Across all languages, child-directed speech is simplified in both obvious and nonobvious ways—whether through overt simplification of words or covert avoidance of complex clauses—even when it sometimes appears that children are being presented with the full complexity of adult speech.

All cultures use baby talk, as we saw in the previous issue in this series, but not all baby talk is the same! In this issue, you’ll see what baby talk (or, more formally, child-directed speech) looks like in the languages of the world!

A lot of people think that baby talk is something only done by American families, a myth we debunked in the previous issue in this series. But baby talk produced by American caregivers does seem to be unique in one respect: American English-speaking parents showed the most extreme changes to their prosody when compared to French, Italian, German, Japanese, and British English (Fernald et al. 1989: 1989). This probably explains why many people think baby talk is a uniquely American phenomenon—because American English speakers do it in an especially pronounced, attention-grabbing way (which is also probably grating to some people). Much of the idea that baby talk is a strange Americanism probably stems from this misperception.

Danish (Indo-European > Germanic) speakers have also been documented using exaggerated intonation in baby talk, but unlike American English, they don’t hyperarticulate the vowels (Cox et al. 2023). Other cultures don’t rely on exaggerated intonation at all. In Quiché Maya (Mayan), for example, baby talk does not use heightened pitch because high pitch is reserved for people of higher social status (Ratner & Pye 1984). The Mi’kmaq and East Cree languages (both in the Algonquian family) are also reported to not use special pitch modifications in child-directed speech (Chee & Henke 2023: 750). And in Cantonese (Sino-Tibetan) baby talk, it is not the intonation that is exaggerated but rather the tones (Xu Rattanasone, Burnham & Reilly 2013).

In the Australian Aboriginal language Warlpiri (Pama–Nyungan), speakers change the pronunciation of vowels (their quality, in technical terms) during baby talk to match the vowels that children themselves are saying (Bundgaard-Nielsen et al. 2023). Some vowels are pronounced further forward in the mouth, while others are pronounced lower in the mouth.

In Japanese baby talk there is no vowel exaggeration, but speakers do emphasize the rhythm of Japanese (Tajima et al. 2013), in which each mora (individual unit of timing) has approximately the same length.1

In addition to changes in prosody and intonation, many languages adjust the pronunciation of individual words in child-directed speech as well. Baby talk in the Cocopah language (Yuman-Cochimí) spoken in Arizona and Mexico has a complex set of conventions where consonants at the beginnings of syllables are replaced in different ways depending on the place and manner of articulation of the consonant in the adult form, and where the stress is. Here are some examples (Crawford 1970: 10):

In fact, the entire consonant inventory of the language is simplified, so that only a subset of the language’s sounds appear in child-directed speech. This is what the consonant inventory of adult-directed speech looks like (Crawford 1970: 11):

And this is what the consonant inventory of child-directed speech looks like (Crawford 1970: 11):

You can see that there are far fewer consonants used in child-directed speech than adult-directed speech.

Similarly, the specialized child-directed vocabulary in Nuuchahnulth (Wakashan) has a reduced consonant inventory compared to adult-directed speech. Here is the inventory for adult-directed speech:

And here is the inventory for child-directed speech:

Reduplication (repetition of sounds, syllables, or entire words) is also a common feature of child-directed speech across languages, such as nana for banana or num‑num for food). Here are examples of reduplication in child-directed speech (CDS) in Inuktitut (Inuit–Yupik–Unangan), compared with their counterparts in adult-directed speech (ADS) (Crago & Allen 1997: 96–97):

What about grammatical differences? Does child-directed speech work differently in languages with elaborate morphology (like polysynthetic languages), which have vastly different grammatical structures than English? Just as an example, here’s a word from the polysynthetic language Yupik (Inuit–Yupik–Unangan):

• Yupik (Inuit–Yupik–Unangan)
  tuntussuqatarniksaitengqiggtuq
tuntu‑ssur‑qatar‑ni‑ksaite‑ngqiggte‑uq
reindeer‑hunt‑FUT‑say‑NEG‑again‑3SG.IND
  ‘He had not yet said again that he was going to hunt reindeer.’
  (Wikipedia: Polysynthetic language)

Do parents perhaps simplify complex grammatical structures like these for their children in the first year or two? After all, children themselves tend to strip words of their inflection when speaking in the polysynthetic languages Severn Ojibwe (a.k.a. Oji-Cree; Algonquian), Inuktitut (Inuit–Yupik–Unangan), and Navajo (Na-Dené) (Chee & Henke 2023: 756), so it would make sense that parents did so too, as another method of accommodating their speech to match the stage of language development of their child.

This isn’t an easy question to answer, precisely because very little research on first language acquisition has been conducted with non–Indo-European languages (Chee & Henke 2023: 742–743). Nonetheless, there is some evidence to suggest that, while caregivers don’t strip otherwise-obligatory affixes from words, they do choose to use simpler (but still grammatical) words. Severn Ojibwe, for example, is about as morphologically complex as it gets, yet child-directed speech in the language tends to use repetition and short constructions, as this example shows:

Ohowe kiniin na.
‘Look at this!’

Awanen aha.
‘Who’s that, eh?’

Awanen aha.
‘Who is that?’

Aacic?
‘Baby?’

You would never know from this example that Severn Ojibwe is a highly morphologically complex language.

🌠 Ergative-Absolutive Marking in Child-Directed Speech

Here’s an especially fascinating aspect of child-directed speech in Inuktitut (Inuit–Yupik–Unangan) which unfortunately requires some robust background in linguistic typology to understand. So if this aside is too technical, just skip it.

Inuktitut doesn’t indicate subjects vs. objects in quite the same way English does. Instead it has something called ergative-absolutive marking: if a clause has two participants (i.e. the verb is transitive), the agent—the doer of the action—gets marked in a special way. This helps listeners determine who is doing what to whom in the clause. That marker of the agent of a transitive clause is called the ergative. If the clause only has one participant (i.e. the verb is intransitive), no special marking is needed, or the participant gets the “default” marking. That non-marking or “default” marking is called the absolutive.

Here’s the cool bit for baby talk: When Inuktitut caregivers use child-directed speech, they almost exclusively use the absolutive (Allen & Crago 1996; Allen 2013: 89). This means that they only use intransitive clauses when talking to children! Since adults nevertheless need to talk about events with two participants, accomplishing this involves a bit of syntactic gymnastics. Adults will use passive verbs or other constructions called antipassives and noun incorporation to express roughly the same event using an intransitive verb. As a result of this, children themselves begin using intransitive clauses before transitive ones, and use intransitives significantly more frequently than transitives overall (Chee & Henke 2023: 757). This same intransitive bias in child speech has also been documented in Navajo, East Cree, and Mohawk; in Navajo, fully 77% of verbs in child speech were intransitive (Chee & Henke 2023: 757).

This is such a fantastic example of how caregivers accommodate their child’s language learning in subtle and subconscious ways. No Inuktitut adult is explicitly thinking, “ergatives are hard, I should only use the absolutive for now”, yet they unwittingly provide an easier entry point into the complexities of Inuktitut grammar by doing so.

Despite the complexity of Native American languages, child-directed speech in these languages isn’t always simplified. Children are often presented with extensively inflected words just like in adult-directed speech. For instance, nouns in East Cree require multiple affixes, and these appear even in child-directed speech, as this example shows:

• East Cree (Algonquian)
  Mâuhî mîn chispituniniuh.
mâu‑hî mîn chi‑spitun‑iniu‑h
DEM‑INAN.PL again 2‑arm‑1PL.INCL‑INAN.PL
  ‘And here are our arms.’
  (Henke 2020: 225)

In Inuktitut, even “baby words” (vocabulary specific to child-directed speech) can still have the full range of complex inflection! Here are examples of the baby words aataaq‑ ‘get hurt’ and aaqqaaq‑ ‘dirt’ in use in child‑directed speech in Inuktitut:

• Inuktitut (Inuit–Yupik–Unangan)
  Aataartaulangasijualunga.
aataaq‑jaulu‑runa‑si‑juq‑aluk‑una
hurt‑PASS‑FUT‑PRES‑PTCP.3SG.SUBJ‑EMPH‑this.one
  ‘That guy is going to get hurt.’
  (Crago, Allen & Pesco 1998: 40)

• Inuktitut (Inuit–Yupik–Unangan)
  Iihiiruluk aaqqaaraaluturumasuuraaluummat.

iihiiq‑guluk
yucky‑bad.little

aaqqaaq‑aluk‑tuq‑guma‑suuq‑aluk‑u‑mmat
dirt‑EMPH‑consume‑want‑HAB‑EMPH‑be‑SUB.3SG.SUBJ
  
  ‘Because the little bugs bit the dirty ones.’
  (Crago, Allen & Pesco 1998: 40)

The result is a somewhat ironic contrast between the “baby word”, which parents presumably use because they find it easier for the child to process, and the full complexity of Inuktitut grammar, which is notably less easy to process!

Lastly, one feature of child-directed speech that has been noted in Native American languages is how fluidly a word can be used as either a noun, verb, or adjective. The Acoma Keres (isolate) word ʔák’aʔák’a can mean ‘drink’, ‘you drink!’, ‘did you drink?’, ‘I want a drink’, ‘I had a drink’, and more (Miller 1965: 112). The Inuktitut root aahaaq‑ can mean ‘to hurt’, ‘thing that causes hurt’, ‘thing that hurts’, and ‘ouch’ (Crago & Allen 1997: 93). This is especially interesting because many Native American languages are known for blurring the distinctions between parts of speech entirely! (Incidentally, this flexibility in parts of speech was actually the focus of my dissertation; Hieber 2021.) So it would seem that this tendency starts early with children. I suspect, however, that Native American languages have been overly-exoticized in this regard, and that if scholars looked at child-directed speech in English they’d find this same flexibility. (Again, this was actually one of the findings from my dissertation: English displays a great deal of part-of-speech flexibility too; this isn’t a phenomenon limited to non–Indo-European languages.)

In sum, then, baby talk can look dramatically different in the world’s languages. Some languages rely heavily on prosodic changes like exaggerated intonation, others on phonological changes like swapping out consonant sounds, and others on entirely distinct sets of “baby words”. Across all languages, child-directed speech is simplified in both obvious and nonobvious ways—whether through overt simplification of words or covert avoidance of complex clauses—even when it sometimes appears that children are being presented with the full complexity of adult speech. Caregivers the world over do a great deal of work—consciously or subconsciously—to accommodate their children’s language-learning process.

In the next issue of this series on the science of baby talk, we’ll look at one of the most controversial claims in child language acquisition research: the Word Gap, which claims that poor households hear up to 30 million fewer words by age 4 than more affluent households. In that issue, we’ll look at how much you really need to be talking to your child.

Be sure to sign up below to get notified when the next issue posts!

Subscribe now

📚 Recommended Reading

📑 References

• Allen, Shanley E. M. & Martha B. Crago. 1996. Early passive acquisition in Inuktitut. Journal of Child Language 23(1). 129–155. https://doi.org/10.1017/S0305000900010126.

• Allen, Shanley E.M. 2013. The acquisition of ergativity in Inuktitut. In Edith L. Bavin & Sabine Stoll (eds.), Trends in Language Acquisition Research, vol. 9, 71–106. Amsterdam: John Benjamins Publishing Company. https://doi.org/10.1075/tilar.9.04all.

• Bundgaard-Nielsen, Rikke L., Carmel O’Shannessy, Yizhou Wang, Alice Nelson, Jessie Bartlett & Vanessa Davis. 2023. Two-part vowel modifications in Child Directed Speech in Warlpiri may enhance child attention to speech and scaffold noun acquisition. Phonetica 80(1–2). 1–42. https://doi.org/10.1515/phon-2022-0039.

• Chee, Melvatha R. & Ryan E. Henke. 2023. Child and child-directed speech in North American languages. In Carmen Dagostino, Marianne Mithun & Keren Rice (eds.), The Languages and Linguistics of Indigenous North America (World of Linguistics 13.2), vol. 2, 741–766. De Gruyter. https://doi.org/10.1515/9783110712742-033.

• Cox, Christopher, Christina Dideriksen, Tamar Keren‐Portnoy, Andreas Roepstorff, Morten H. Christiansen & Riccardo Fusaroli. 2023. Infant‐directed speech does not always involve exaggerated vowel distinctions: Evidence from Danish. Child Development 94(6). 1672–1696. https://doi.org/10.1111/cdev.13950.

• Crago, Martha B. & Shanley E.M. Allen. 1997. Linguistic and cultural aspects of simplicity and complexity in Inuktitut child directed speech. In E. Hughes, M. Hughes & A. Greenhill (eds.), Proceedings of the 21st Annual Boston University Conference on Language Development, 91–102. Cascadilla Press.

• Crago, Martha B, Shanley EM Allen & Diane Pesco. 1998. Issues of complexity in Inuktitut and English child directed speech. In 29th Annual Stanford Child Language Research Forum, 37–46.

• Crawford, James M. 1970. Cocopa baby talk. International Journal of American Linguistics 36(1). 9–13. https://doi.org/10.1086/465083.

• Fernald, Anne, Traute Taeschner, Judy Dunn, Mechthild Papousek, Bénédicte De Boysson-Bardies & Ikuko Fukui. 1989. A cross-language study of prosodic modifications in mothers’ and fathers’ speech to preverbal infants. Journal of Child Language 16(3). 477–501. https://doi.org/10.1017/S0305000900010679.

• Henke, Ryan E. 2020. The first language acquisition of nominal inflection in Northern East Cree: Possessives and nouns. University of Hawaii at Manoa Ph.D. thesis.

• Hieber, Daniel W. 2021. Lexical polyfunctionality in discourse: A quantitative corpus-based approach. Santa Barbara, CA: University of California, Santa Barbara Ph.D. thesis. https://doi.org/10.13140/RG.2.2.11932.00640.

• Kess, Joseph Francis & Anita Copeland Kess. 1986. On Nootka baby talk. International Journal of American Linguistics 52(3). 201–211. https://doi.org/10.1086/466018.

• Miller, Wick R. 1965. Acoma grammar and texts (University of California Publications in Linguistics 40). University of California Press.

• Ratner, Nan Bernstein & Clifton Pye. 1984. Higher pitch in BT is not universal: Acoustic evidence from Quiche Mayan. Journal of Child Language 11(3). 515–522. https://doi.org/10.1017/S0305000900005924.

• Tajima, Keiichi, Kuniyoshi Tanaka, Andrew Martin & Reiko Mazuka. 2013. Is the mora rhythm of Japanese more strongly observed in infant-directed speech than in adult-directed speech? In PRoceedings of Meetings on Acoustics, vol. 19. Montreal, Canada. https://doi.org/10.1121/1.4800508.

• Van Otterloo, Karen. 2011. The Kifuliiru language, Vol. 1: Phonology, tone, and morphological derivation (Publications in Linguistics 146). SIL International.

• Xu Rattanasone, Nan, Denis Burnham & Ronan G. Reilly. 2013. Tone and vowel enhancement in Cantonese infant-directed speech at 3, 6, 9, and 12 months of age. Journal of Phonetics 41(5). 332–343. https://doi.org/10.1016/j.wocn.2013.06.001.

“There is only one thing that no one would ever learn from those ancient songs who did not already know it, and that is that the sky is blue.” ~ Lazarus Geiger

There’s something decidedly weird about the way Homer talks about colors in the Iliad and the Odyssey. Take one of his most famous—and odd—epithets: the “wine-dark sea”. Homer uses this phrase five times in the Iliad and twelve times in the Odyssey. The phrase is so ubiquitous that it is frequently used or alluded to in pop culture even today. But how does wine in any way accurately describe the color of the sea? Surely a comparison with something blue or green or even gray would have made for a better descriptor. Homer was writing in Ancient Greek (Indo-European > Hellenic), and in actuality, the phrase “wine-dark” is already a poetic reinterpretation of the word Homer actually used: οἶνοψ oînops, a compound of οἶνος oînos ‘wine’ + ὄψ óps ‘eye; face’, which would have more literally meant ‘wine-looking’. What about the sea is so wine-like that Homer uses the phrase not one but seventeen times to describe it?

It gets weirder: The only other thing Homer describes using the word oînops is… oxen. (Oxen are, for the record, black, brown, or reddish-brown.) Then there’s the term ἰόεις ióeis ‘violet’, which Homer uses to describe sheep (presumably black sheep), iron, and… the sea. So the sea is wine-looking and violet-like. Got it. Homer also compares the color of Odysseus’ dark hair to the color of a hyacinth, which, in case you didn’t know, looks like this:

📘 Much of this article is based on the discussion of color terms in Guy Deutscher’s excellent book, Through the language glass: Why the world looks different in other languages (Amazon | Bookshop). This also happens to be one of my favorite pop linguistics books. If you enjoy this article, I highly recommend procuring a copy at the links below:

• Amazon

• Bookshop

This is just the beginning of Homer’s chromatic curiosities—oddities that have puzzled scholars for centuries. In 1858, for instance, British politician William Gladstone compiled a comprehensive list of Homer’s erratic use of colors as part of a massive 1,700-page treatise titled Studies on Homer and the Homeric Age. In a chapter tucked away near the end of the book titled “Homer’s perception and use of color”, Gladstone so thoroughly and rigorously documented Homer’s strange use of color terms that the intellectuals of the time were forced to acknowledge that something more was going on with Homer’s color vocabulary than mere poetic license. Gladstone carefully laid out five types of evidence showing that something was amiss in Homer’s descriptions of colors:

First, as we have already seen, Homer often uses the same word to denote colors that are—to modern observers, at least—completely different. Another example of this is his use of the word χλωρός khlōrós, which later just came to mean ‘green’. But Homer applies khlōrós to faces pale with fear, fresh twigs, the olive wood of the club of Cyclops, and… honey. Or take the word κυάνεος kŭắneos, which later came to mean ‘blue’. Homer employs it to describe the eyebrows of Zeus, the hair of Hector, and a dark cloud—never anything blue.

Second, Homer sometimes characterizes the same objects using contradictory color descriptors. Iron, for instance, is described using terms for violet and gray, as well as the word αἴθων aíthōn ‘a thing which blazes and flashes like burnished metal’. aíthōn is otherwise only used to refer to the color of horses, lions, and… oxen. So oxen are wine-looking and blazing like burnished metal. Cool.

Third, Homer uses color terms only sparingly, and not at all in places where we might confidently expect it. He never discusses the color of horses despite including extensive descriptions of them. He never describes the sky as blue; it is starry, broad, great, iron, or copper, but never blue. Poppies, spring flowers, fields of wheat, hills and forests are never described in terms of their colors. In contrast, the producers of the 1939 film The Wonderful Wizard of Oz specifically elected to use the scene involving a vibrant field of red and pink poppies to flaunt their dazzling new Technicolor technology. They used paper flowers to create the effect of a vast field, making it one of the most expensive and visually iconic sequences in movie history. Yet not once does Homer rely on the color of flowers for his vivid imagery.

Fourth, when colors do feature in the Homeric epics, white and black predominate over all other colors. The table below shows how often Homer uses each color term in the Odyssey and the Iliad combined:

Finally, as illustrated by the above table, the size of Homer’s color vocabulary is surprisingly small. There is a complete absence of any terms used to mean ‘blue’, ‘orange’, or ‘pink’, and khlōrós doesn’t appear to mean ‘green’ at this point.

Taken together, all the evidence points to something unquestionably strange about Homer’s use of color. What could possibly be the cause of his uneven, vague, and inconsistent color descriptions?

Why does Homer use colors so strangely?

Homer is often said to have been blind, and some have used this as an explanation for his odd use of color vocabulary, but his blindness is undoubtedly a myth. He may have been prismatically inept, but his visual descriptions are nonetheless vivid. Also, his contemporaries accepted his poetry unquestioningly! The Ancient Greeks found nothing wrong with his monochromatic portrayals of the world. In fact, we find traces of the same chromatic vagueness in other Ancient Greek writers: Pindar (5th C BCE), for example, also wrote about ‘violet-colored hair’.

Gladstone’s proposed explanation for the “faint and indefinite” color descriptions of the Ancient Greeks was that “the organ of colour and its impressions were but partially developed among the Greeks of the heroic age”. He argued that they saw the world primarily in terms of lightness and darkness rather than hue, save for a splash of red. The word erŭthrós ‘red’ was the only color term other than black and white that Homer used to refer only to reddish things (blood, wine, copper). Gladstone argued that the Greeks had been in effect colorblind—a claim made all the more remarkable by the fact that colorblindness was not yet known to the scholarly community in 1858. The reason for this colorblindness, he claimed, was that the Greeks had at the time only just begun to develop artificial dyes. It is only once people are exposed to artificial colors that they evolved the ability to see colors and subsequently name them.

Put differently, the Greeks had not yet conceived of color as an abstract concept, independent of the objects bearing those colors. This isn’t as strange a claim as it might sound: in a similar fashion, it took ancient Sumerians several centuries to conceive of numbers as abstract concepts. Prior to the standardization of their counting system, the Sumerians had different sets of numbers for different commodities. A separate series of numerals was used to count barley and cereal products, malt, barley groats, land area, and calendar time, among others. From today’s standpoint, this proliferation of counting systems seems like an unnecessarily complex deficiency. But given that the primary function of these numeral systems was to tally commodities, having distinct numerals that told you something about which commodity was being counted would have been incredibly useful. This meant, however, that numbers were always tied to the specific concrete entities being counted. It took centuries for the Sumerians to conceptualize numerals in the abstract, and when they did the result was the incredibly sophisticated mathematical tradition of the Babylonians.

Gladstone’s thesis also has some explanatory power. Ignoring Gladstone’s proposed evolutionary mechanism for a moment, if we interpret Homer’s depictions in terms of shades of darkness rather than hue, phrases like ‘violet sheep’ and ‘wine-dark sea’ no longer seem so strange. This hypothesis also explains the total absence of blue: it appears that for Homer, kŭắneos simply referred to any dark color, rather than ‘blue’ specifically. And the use of khlōrós to describe non-green things suddenly makes sense when we realize that its original meaning was ‘young herbage’. As such, it could be used to describe anything fresh, verdant, unripe, or pale—including honey and fearful faces.

But from our modern vantage point, Gladstone’s explanation for why the Ancient Greeks focused on lightness vs. darkness rather than hue seems scientifically antiquated. 2,500 years is not enough time for evolutionary forces to exert sufficient selective pressure on a population that they develop the complexities of color vision. To be fair, however, Gladstone’s magnum opus was published just a few months before Charles Darwin and Alfred Russel Wallace jointly announced a theory of evolution by natural selection in 1858. The prevailing explanation for biological change at the time was Jean-Baptiste Lamarck’s model of inheritance of acquired characteristics. In this view, traits that an organism acquires during its lifetime can be passed on to their offspring. Early giraffes strained to reach the high boughs of a tree, and so their offspring developed long necks. A blacksmith might get strong muscles from their work, and their children would then be stronger as a result.

Thus, to the scholars of the time, an explanation of the Ancient Greeks’ color deficiencies in terms of inheritance of acquired characteristics made perfect sense. The Greeks, it was argued, hadn’t yet developed color terms because they simply didn’t have need of them yet. Once they developed artificial paints and dyes, the Greeks became like those giraffes reaching for the highest leaves, and slowly got better at distinguishing the hues of the rainbow.

Nonetheless, Gladstone’s contemporaries were no less harsh in their reception of his ideas. Prominent classicist John Stuart Blackie trenchantly wrote:

If any man should say that the minstrel was deficient in the organ of colour because he designated the sea by this vague word, I would meet him by saying that the critic is deficient in the organ of poetry.

Yet Gladstone’s thesis turned out to be prescient. Although he misunderstood the mechanism, he was ultimately correct that the color vocabulary of the Greek language—and indeed all languages—shifted its focus over time from shades of darkness to hue. Today we know that Gladstone’s findings actually reflect a universal tendency in how the world’s languages develop color terms over time, and that this color term hierarchy, as it is called, arises due to subtle biological influences on language. In the remainder of this article, I’ll explain what the color term hierarchy is, what causes it, and how it’s realized in the world’s languages.

The color term hierarchy

In the century following Gladstone’s seminal work, scholars came to realize that the Ancient Greeks were not the only ones who were chromatically challenged. About a decade after Gladstone’s Studies on Homer, philologist Lazaraus Geiger gave a plenary lecture at the 1867 Assembly of German Naturalists & Physicians titled “On the Color Sense in Primitive Times and its Evolution”, in which he expanded on Gladstone’s work. He realized that it was not just Homer who struggled with colors, but the authors of all the ancient texts—the Indian Vedic poems (Vedic Sanskrit: Indo-European), the Old Testament (Biblical Hebrew: Semitic), the Icelandic sagas (Old Icelandic: Indo-European > Germanic), and even the Quran (Classical Arabic: Semitic). Biblical Hebrew, for instance, lacks a word for ‘blue’, and the Bible is rife with odd color descriptions: a ‘red horse’ and a ‘red heifer without spot’; faces ‘turned green’ with panic; the feathers of a dove covered with ‘green gold’. As for the Vedic hymns, Geiger says:

These hymns, of more than ten thousand lines, are brimming with descriptions of the heavens. Scarcely any subject is evoked more frequently. The sun and reddening dawn’s play of color, day and night, cloud and lightning, the air and the ether, all these are unfolded before us over and over again, in splendor and vivid fullness. But there is only one thing that no one would ever learn from those ancient songs who did not already know it, and that is that the sky is blue.

Read more

There are few things I enjoy more than a sci-fi story where the characters have to figure out how an alien language works. Why? Because it’s linguistic fieldwork! Xenolinguistics! When writers handle this well, the methods of decipherment adopted by the characters look remarkably like how real-world linguists approach the process of documenting a hitherto undocumented language, and you get fantastic stories like Denis Villeneuve’s movie Arrival, based on Ted Chiang’s short story “Story of Your Life”. And when writers handle this poorly, you get Star Trek’s universal translator—the most boring and implausible linguistic plot device in the history of linguistic plot devices. (Why would you minimize the amount of linguistics in a story? It makes no sense! Even The Hitchhiker’s Guide to the Galaxy takes a paragraph to explain how babelfish work.)

So imagine my delight when I discovered 176 pages into Project Hail Mary that I was about to get another interspecies language documentation project! Jazz hands! 👐 In this article we’ll look at the linguistics of Project Hail Mary and see how it holds up against the real-world science. Since the movie was pretty faithful to the book in terms of linguistics, this should serve as a useful explainer for the movie as well. And if you’re a paying subscriber, don’t miss the line-by-line bonus commentary at the end of this article!

You’re reading Linguistic Discovery, a newsletter about the science and diversity of language—a field known as linguistics. I’m Danny Hieber, a PhD in linguistics who works with indigenous communities to help them document and revitalize their languages. Topics covered in this newsletter include:

• ⚙️ how language works (cognitive linguistics, language change)

• 🌍 grammatical diversity in the world’s languages (typology)

• ℹ️ explainers of terms and concepts in linguistics

• 🗣️ language profiles

• 🗞️ the latest news and research in language and linguistics

• ⭐ linguistic reviews of books and other media

Paid subscribers also get bonus articles and early access to my in-progress book projects!

Subscribe now

How the Eridian language works

In the book, scientist Ryland Grace encounters an alien species he names the Eridians, after their home system, 40 Eridani. Grace has to work closely with one member of species, who he names Rocky, to escape various perilous situations and forestall certain death by means of some hardcore sciencing. He soon discovers that the Eridians have language! 🎶 Amaze! Amaze! 🎶

One of the hard parts of first contact with intelligent alien life—in addition to figuring out everyone’s pronouns (p. 172)—is determining the medium of communication. Do the aliens relay information directly via thought waves, like the Trisolarans in Three-Body Problem? Or via complex visual symbols, like the heptapods in Arrival or Species 10-C in Star Trek: Discovery?

Grace quickly discovers that, like humans, Eridians communicate by propagating sound waves through a medium:

“♩🎝♪♪🎝,” says Rocky.

My jaw drops. Yes, I’m in zero g. It still drops.

There was no pronunciation or inflection of the sounds. Just notes. Like whale song. Except not quite like whale song, because there were several at once. Whale chords, I guess. And he was responding to me. That means he can hear too.

And notably, the sounds were in my range of hearing. Some of the notes were low, some of them high. But definitely audible. That alone is amazing when I think about it. He’s from a different planet, and totally different evolutionary line, but we ended up with compatible sound ranges. (p. 176)

Grace then sets up what documentary linguists would call his first elicitation session, where he uses stimuli or prompts that get the speaker to produce certain target words/phrases. This allows him to both record the words and begin associating them with meanings:

I grab my laptops and launch the waveform-analysis software on one and Excel on the other. I hold up the numeral Ⅰ and point to it. “One,” I say. “One.”

Rocky points to the Ⅰ and says “♪.”

I pause the waveform analyzer and scroll back a few seconds. Rocky’s word for ‘one’ is just two notes played at the same time. There are a bunch of harmonics and resonances in there, too, but the main frequency peaks are just two notes.

I type “one” into the spreadsheet on the other computer and note the relevant frequencies.

I return to the divider and hold up the Ⅴ symbol. “Two,” I say.

“♪,” he says. Another one-syllable word. The oldest words in a language are usually the shortest.

This time, it’s a chord made of four distinct notes. I enter “two” and record the frequencies for that word.

He starts to get excited. I think he knows what I’m up to and it’s got him happy.

I hold up the λ and before I can even speak, he points to it and says, “🎝♪.”

Excellent. Our first two-syllable word. The first syllable has just two notes and the second has five! Rocky can make at least five different notes at the same time. He must have multiple sets of vocal cords or something.

(pp. 199–200, edited for brevity)

Based on this description, Eridian sounds similar to the tonal languages of Earth! In tone languages, the meaning of a word or affix is determined not just by its vowels and consonants, but also the pitch with which you say the word. The classic example is Mandarin Chinese, where the same syllable ma can mean 5 different things depending on the pitch of the word:

1. mā (媽/妈) ‘mother’

2. má (麻/麻) ‘hemp’

3. mǎ (馬/马) ‘horse’

4. mà (罵/骂) ‘scold’

5. ma (嗎/吗) (question marker)

You can think of tone as like the melody with which you say a word (because that’s exactly what it is). You could, if you want, transcribe linguistic tones using musical notation, and linguists have even done so in the past, but this usually isn’t necessary because any given language only has a few tones. Mandarin, as we’ve just seen, has 5 tones, and some languages have been analyzed as having up to 14 tones, but most languages (60%) only make a contrast between two tones, usually high tone (H) vs. low tone (L) or no tone (∅) (Gordon 2016: 219–221). In addition, linguistic tone is based on relative pitch rather than absolute pitch, so the same pitch might be used for a High tone in one context but a Low tone in another.

Below are what the 4 primary tones of Mandarin look like when you plot their change in pitch over time (Wikipedia: Tone (linguistics)). (The fifth tone is a neutral tone whose pitch is mostly determined by the tone of the preceding syllable.)

We can also describe tones in terms of their levels. Mandarin’s 2nd tone, for example, is a 35 tone, while its 1st tone is 55, and so on. The maximum number of distinct tone levels a language can have is 5 (Gordon 2016: 223), a fact which is likely due to perceptual and cognitive limitations on Homo sapiens’ ability to discriminate between different pitches in sequence. Coincidentally, the Eridian language appears to allow a maximum of 5 tone levels as well! The second syllable of the Eridian word for ‘three’ consists of 5 distinct notes, as the book mentions.

Many human tone languages only make simple contrasts between a High (H) tone vs. a Low (L) tone, or High (H) vs. Mid (M) vs. Low (L). Each syllable is either H or L (or M). You don’t have to pay attention to whether the pitch is rising or falling, just whether it’s high, low, or mid (relative to other tones in the utterance). For example, the Gusii language of southwestern Kenya only has High tones and Low tones (Nash 2011):

• éndà HL ‘stomach’ vs. éndá HH ‘louse’

These kinds of languages are often called register tone languages, and they contrast with languages like Mandarin where the pitch can change even within a single syllable. Since in this latter type of language you need to pay attention to rises and falls in the pitch within a syllable, they are called contour tone languages. (In reality, register tone languages also typically involve lots of pitch contours, so the distinction between register and contour tone languages is not entirely principled or clearcut.)

Based on how the author, Andy Weir, transcribes Rocky’s dialogue, it seems as though Eridian has both level tones and contour tones (or rather, level chords and contour chords), parallel to human tone languages. Some syllables consist of a single level chord, which Weir transcribes with a single note, ♩ or ♪, while other syllables consist of a sequence of two chords, transcribed as 🎝—the Eridian equivalent of a contour tone. However, Weir never uses ♫, suggesting that Eridian has falling contour tones but never rising contour tones—and there are some human languages like this too! Kiowa and Jemez are two such examples.

Finally, the fact that Weir transcribes some syllables using a quarter note (♩) and others using an eighth note (♪) suggests that the length of the chord makes a meaningful difference in Eridian. This too parallels human languages! Many languages distinguish between long vowels and short vowels, in the literal sense of long vowels being pronounced for a longer duration than short vowels. Here are some examples of words in Gusii that differ only in their vowel length (indicated by a double vowel; Nash 2011: 40):

• émèrì ‘roots’ vs. émèèrì ‘ship’

• βùtìà ‘conclude’ vs. βùùtìà ‘entertain (a guest)’

Some languages also distinguish between long consonants (called geminate consonants) and short consonants, like Berber (Wikipedia: Gemination):

• ini ‘say’ vs. inni ‘those in question’

• akal ‘earth, soil’ vs. akkal ‘loss’

• imi ‘mouth’ vs. immi ‘mother’

• ifis ‘hyena’ vs. ifiss ‘he was quiet’

So Eridian long vs. short chords parallel human long vs. short vowels and consonants.1

There are however several fundamental differences between Eridian and human tone languages. First, humans can only produce one pitch at a time—at least usually. Some people have learned to produce multiple pitches at once, a technique known as overtone singing, harmonic singing, or throat singing. The Mongolian, Tuvan, Inuit, and Xhosa cultures are all known for this kind of polyphonic singing. But this doesn’t work for everyday language because it erases a lot of the acoustic information we need to discriminate between different consonants/vowels. So as far as human language goes, tonal languages only get 1 tone per word/syllable/vowel/mora (depending on the language). Eridian, by contrast, can use entire musical chords at once! But we know that these chords are, like human tonal languages, based on relative pitch rather than absolute pitch, because of passages like these:

That whole sentence seemed lower in pitch than anything Rocky has ever said before. I select the whole segment in the software’s recording history and bump it up an octave. The octave is a universal thing, not specific to humans. It means doubling the frequency of every note. The computer immediately translates the result. (p. 212)

The second major difference between Eridian and human tone languages is that human tones work in conjunction with consonants, vowels, and semivowels, whereas Eridian works by tones alone. In this respect, Eridian is similar to whistle speech on Earth, where speakers communicate by merely whistling. However, there are no known systems of whistle speech that constitute an entirely independent language. No language is communicated using whistling alone. All whistle communication systems are merely whistled versions of another language. For example, Silbo Gomero is a whistled form of the Spanish used in the Canary Islands.

For tonal languages, the whistled version of a word typically mimics the tonal pattern of the word. For non-tonal languages, the corresponding whistles are based on the acoustic profile of each sound in the language. Whistlers whistle at specific pitches that correspond to each vowel or consonant in the language. For example, in Silbo Gomero the tongue is placed in a similar shape as when pronouncing each spoken sound, giving the whistle a similar pitch. In whistled Turkish, /o/ is whistled at a lower frequency than /a/, and at a different frequency than other vowels, etc.

What kind of language is Eridian, then? It most closely resembles a mix of human contour tone languages, whistle speech, and throat singing, with a distinction between long vs. short chords thrown in. It is, in sum, surprisingly humanlike! But maybe this is to be expected given that both Eridian and human language use the medium of sound. We might expect all sound-based languages in all the galaxies to exhibit some of these same features. So I, for one, laud Andy Weir for creating a fun and surprisingly realistic conlang for the book.

But for a more critical take on the linguistics of Project Hail Mary, check out this interview that Ars Technica did with linguist Betty Birner, professor emerita at Northern Illinois University:

• Project Hail Mary is in theaters—but do the linguistics work? (Ars Technica)

If you want to go deeper into Weir’s treatment of Eridian and how realistic it seems compared to human language, paying subscribers can read my line-by-line commentary of the book after the references below. I made note of any passages that give us insight into how the Eridian language works, along with explanations of the linguistics behind it and how each point compares to human language. Along the way you’ll get to explore some surprising ways that human languages work as well!

📑 References

• Berlin, Brent & Paul Kay. 1969. Basic color terms: Their universality and evolution. University of California Press.

• Cheney, Dorothy L. & Robert M. Seyfarth. 1990. How monkeys see the world: Inside the mind of another species. Chicago: University of Chicago Press.

• Comrie, Bernard. 2013. Numeral bases. In World Atlas of Language Structures Online. Max Planck Institute for Evolutionary Anthropology. http://wals.info/chapter/131.

• Comrie, Bernard. 2021. Typology of numeral systems.

• Gordon, Matthew K. 2016. Phonological typology (Oxford Surveys in Phonology & Phonetics 1). Oxford University Press.

• Harrison, K. David. 2007. When languages die: The extinction of the world’s languages and the erosion of human knowledge. Oxford University Press.

• Hieber, Daniel W. 2016. The cohesive function of prosody in Ékegusií (Kisii) narratives: A functional-typological approach. University of California, Santa Barbara M.A. thesis. http://rgdoi.net/10.13140/RG.2.2.17818.18886. (22 August, 2025).

• Hieber, Daniel W. 2019. The Chitimacha language: A history. In Language in Louisiana: Community and culture (America’s Third Coast Series), 9–27. University Press of Mississippi.

• Mithun, Marianne. 2012. Core argument patterns and deep genetic relations: Hierarchical systems in Northern California. In Pirkko Suihkonen, Bernard Comrie & Valery Solovyev (eds.), Argument structure and grammatical relations: A crosslinguistic typology (Studies in Language Companion Series 126), 257–294. John Benjamins Publishing Company.

• Nash, Carlos M. 2011. Tone in Ekegusii: A description of nominal and verbal tonology. University of California, Santa Barbara Ph.D. thesis.

• Payne, Thomas. 1997. Describing morphosyntax: A guide for field linguists. Cambridge University Press.

• Sterelny, Kim. 2011. From hominins to humans: How sapiens became behaviourally modern. Philosophical Transactions of the Royal Society B: Biological Sciences 366(1566). 809–822. https://doi.org/10.1098/rstb.2010.0301.

• Velupillai, Viveka. 2012. An introduction to linguistic typology. John Benjamins.

The Amazon and Bookshop links on this page are affiliate links, which means that I earn a small commission from Amazon for purchases made through them (at no extra cost to you).

If you’d like to support Linguistic Discovery, purchasing through these links is a great way to do so! I greatly appreciate your support!

Check out my Amazon storefront here.

Check out my Bookshop storefront here.

🌟 Bonus Commentary

Below is my commentary on each passage in Project Hail Mary that tells us something interesting about how the Eridian language works.

Read more

If baby talk helps children learn language better or faster, as we saw in the previous post in this series, it stands to reason that parents would instinctively use it. But if not all parents and cultures use it, how important could it really be?

In this issue of my series on the science of talking to tiny humans, we’ll look at claims that some languages don’t use baby talk and see whether they hold up to scrutiny. Check out the rest of the series at the links below, and sign up to get notified when subsequent issues are published!

Subscribe now

ℹ️ Articles in this Series

• Part 1: Why you should be talking to your infant

• Part 2: What’s the point of baby talk?

• Part 3: Is baby talk good for your child?

• Part 4: Do all cultures use baby talk? [this post]

• Part 5: Baby talk in the languages of the world

• Part 6: How much should you talk to your child? [forthcoming]

• Part 7: What really matters when talking to your child

Is baby talk just a weird American habit?

Many people—some linguists included—think that baby talk is a uniquely Western phenomenon, or that it’s limited to well-educated, white, middle-class parents, or that it’s some weird thing that only Americans do. Here’s an example from a pop linguistics book about child language acquisition from 2006:

If anything, motherese reveals more about parents than children: it is a peculiar cultural phenomenon in modern Western industrialized societies, not part of the universal and biological foundation of language. (Yang 2006: 89)

A Scientific American article from 2017 proclaims that “Parents in a remote Amazon village barely talk to their babies” and that “Ignoring a Western child-rearing practice does not seem to matter for the Tsimané of Bolivia” [emphasis added].

At the outset, we should be suspicious of the idea that only educated white parents participate in an activity that demonstrably improves their children’s language outcomes (as we saw in the previous issue of this series). Such a narrative appeals to implicit ethnocentric, racist, or classist biases, so it’s worth approaching these claims with some healthy skepticism. Indeed, the vast majority of research in child language acquisition has been conducted on precisely this predominantly white, Western demographic (Nielsen et al. 2017), so even if baby talk does exist in other communities (and it does), it’s quite likely that Western scholars wouldn’t have been aware of it (and they weren’t for some time).

Of course, the notion that baby talk is unique to Western culture also appeals to cultural relativists, who are (quite reasonably) cautious of imposing Western perspectives and assumptions on the analysis of other cultures. Yet this too can be an ideologically-motivated position, one based less on scientific caution and more on anti-Western or anti-colonial sentiments. The risk that comes from adopting this position, however, is undue exoticization or romanticization of other cultures. So there’s plenty of opportunity for motivated reasoning all around when it comes to opinions on the universality of baby talk.

For example, widely-reported studies with communities like the Tsimane in the remote Amazon of Bolivia (mentioned by the Scientific American article above; Cristià et al. 2017), the Yucatec Maya in Mexico (Shneidman & Goldin‐Meadow 2012), and the Kaluli in Papua New Guinea (Schieffelin 1990) led some to claim that these and presumably other cultures essentially lack baby talk. In each case, however, it turns out that child-directed speech (CDS) is simply less frequent, occurs later in development, or works differently than it does in Western/American communities. As research into child language acquisition has broadened to include cultures across the world in the last several decades, it’s become increasingly clear that CDS is not limited to white, educated, middle-class families.

You’re reading Linguistic Discovery, a newsletter about the science and diversity of language—a field known as linguistics. I’m Danny Hieber, a PhD in linguistics who works with indigenous communities to help them document and revitalize their languages. Topics covered in this newsletter include:

• ⚙️ how language works (cognitive linguistics, language change)

• 🌍 grammatical diversity in the world’s languages (typology)

• ℹ️ explainers of terms and concepts in linguistics

• 🗣️ language profiles

• 🗞️ the latest news and research in language and linguistics

• ⭐ linguistic reviews of books and other media

Paid subscribers also get bonus articles and early access to my in-progress book projects!

Subscribe now

Baby talk in other languages

Consider the Kaluli people of Papua New Guinea, for example. The Kaluli are sometimes claimed not to have baby talk because young children among the Kaluli are not viewed as having their own intentions. By tradition, however, when infants first use the words for ‘mother’ and ‘breast’, parents very intentionally begin to “teach the child to talk”. They do this by speaking for the child using a special high-pitched, very nasal voice register. Adults spend a lot of time “talking for” children, holding them up to face the addressee and then saying what the child ought to say in that situation, and frequently using the word ɛlɛma ‘say it like this’ (Ochs 1982; Schieffelin 1990). Here’s an example of what this looks like in practice, where a mother is teaching her son, Wanu, how to interact with another child, Binalia (adapted from Ochs & Schieffelin 2009: 306–307):

Mother (to Wanu) Abɛnowo? Ɛlɛma. Whose is it? Say it like that.

Wanu (to Binalia) Abɛnowo? Whose is it?

Mother Gɛnowo? Ɛlɛma. Is it yours? Say it like that.

Wanu Gɛnowo? Is it yours?

Mother Ge oba? Ɛlɛma. Who are you? Say it like that.

Wanu Ge oba? Who are you?

Mother Gi suwo? Ɛlɛma. Did you pick it? Say it like that.

Wanu Gi suwo? Did you pick it?

Mother Ni nuwɛ sukel! Ɛlɛma. My grandmother picked it! Say it like that.

Wanu Ni nuwɛ sukel! My grandmother picked it!

Mother We ni nuwɛ sukel. Ɛlɛma. This my grandmother picked! Say it like that.

Wanu We ni nuwɛ sukel. This my grandmother picked.

This is different from what baby talk looks like in other cultures, but it has the same function of adjusting one’s speech in a way to make it both attention-grabbing and comprehensible to the child.

Another culture sometimes claimed not to use baby talk is that of the Tsimane people of Bolivia. A 2017 study found that Tsimane parents speak to young children for less than one minute every hour (Cristià et al. 2017), which is only about ⅒ the amount of time that U.S. mothers speak to their children of the same age. The graph below shows the average minutes per hour of CDS received by Tsimane children at different age ranges. As you can see, the amount of CDS received stays pretty minimal until about age 4.

The reason for this, tragically, may be the high infant mortality rate in the community. Fully 13% of infants die in their first year, mainly due to infectious diseases (Smith 2017). As a result, parents may try not to be as emotionally attached to their children. Many children aren’t given names until after their first birthday (Smith 2017).

Whatever the cause, this study led Scientific American to proclaim that, “Parents in a remote Amazon village barely talk to their babies”. The researchers also noted (anecdotally), that this relative dearth of child-directed speech seems to have a dilatory effect on child language development, so that children begin to speak later than their U.S. counterparts (Smith 2017).

But if we burrow into the details of the study a bit further, we find that the Scientific American headline lacks some important context (which they do admittedly provide in the article). Firstly, Tsimane children overhear adult speech for about 7 minutes/hour (Smith 2017), which matters because cultures vary as to how much learning happens through direct instruction and modeling versus observation and mimicry. Secondly, sometimes differences in the amount of child-directed speech that infants receive is partly due to differences in overall talkativeness across cultures. Accounting for these differences greatly reduces discrepancies in how much CDS infants hear across cultures (Cristià et al. 2017: 761).

Lastly, we can’t actually compare the results of this study to any others, because they use widely differing methods for measuring the amount of CDS that children hear. This study focuses explicitly on 1-on-1 speech, whereas it’s unclear whether previous studies included overheard speech or speech directed at a group that the child just happened to be part of. It could actually be that CDS rates in other studies and even U.S. households look similar once the same methods are applied. These methodological differences are so significant that it leads the authors to state:

The evidence on who talks to infants is extremely scarce at present. […] Further work employing homogeneous, cross-culturally appropriate methods is needed to more accurately measure the extent of this variation and to make strict comparisons possible. Therefore, we do not attempt direct comparisons. (Cristià et al. 2017: 767)

Moreover, even though children aren’t receiving much directed speech in their early years, they are hearing a great deal of CDS as they move out of infancy. This pattern was also documented among the Yucatec Maya, who hear significantly less child-directed speech than children of the same age in the U.S. (Shneidman & Goldin‐Meadow 2012)

But in their second and third years, Maya children hear an increasingly large amount of CDS, eventually matching that of U.S. children by about 35 months:

This study demonstrates that directed speech doesn’t have to happen within the first year or so to be effective. The authors found that the number of word types used by Maya adults in their child-directed speech at 24 months predicted the children’s vocabulary size a year later. So while CDS is important for child language development, the exact timing of when children receive it seems relatively less important.

Even in those early years, however, children are receiving child-directed speech—just not from adults! Among the Maya, 90% of child-directed speech at age 3 comes from other children (Shneidman & Goldin‐Meadow 2012: 761). By comparison, for Americans only about 10% of CDS comes from other children. This underscores the importance of understanding cross-cultural differences in child language development: communities vary as to which members of the community are most actively involved in child-rearing, and how, and this has effects on the type and amount of child-directed speech that children receive.

Want to learn more about how children’s linguistic capabilities develop over time? Check out this issue of the newsletter about how children learn grammatical rules (and how we know):

Is baby talk universal?

Stepping back and looking more broadly at the body of research on child-directed speech across cultures, the sum of the evidence currently suggests that at least some form of CDS is present in all cultures (Clark 2024: 62–67; Saxton 2017: 112–114). Exaggerated intonation, for instance, has been documented in English, German, Hebrew, Italian, Japanese, Russian, Spanish, Swedish, Luo (a Nilotic language of East Africa), Severn Ojibwe (a Native American language of Ontario and Manitoba also known as Oji-Cree; Upper 1993: 121), and Nuuchahnulth (a Native American language of British Columbia; Kess & Kess 1986: 203) (Burnham, Kitamura & Vollmer-Conna 2002: 1435; Soderstrom 2007). This feature abounds especially during the child’s first year, and so might be considered special to infant-directed speech (IDS) specifically rather than child-directed speech (CDS) more broadly. One experiment also showed that people from 187 countries could recognize IDS even when it wasn’t in a language they knew (Hilton et al. 2022).

Repetitions and recasts—saying words back to the child as a model of correct pronunciation—also appear to be universal features of CDS, and are documented for Danish, English, French, Hebrew, Japanese, K’iche’ Maya (Guatemala), Korean, Mandarin Chinese, Manus (New Guinea), Persian (Iran, Afghanistan, Tajikistan), Samoan (Samoan Islands), and Warlpiri (Australia) (Saxton 2017: 113). And just like English caregivers enunciate individual sounds (as discussed in a previous issue of this series), Hidatsa caregivers have also been documented slowing and emphasizing their speech, exaggerating certain sounds, and fully pronouncing consonant clusters that would ordinarily be reduced or simplified in adult speech (Voegelin & Robinett 1954).

Even people who claim they don’t engage in baby talk actually do. A study of the African American community of Trackton, South Carolina (Heath 1983) found that adults consider the idea of modifying their speech to talk to infants irrational, yet they engage in the same kinds of structured repetitions and recasts as documented in other communities (Saxton 2017: 113). Another study found that 18 out of the 82 Kuwaiti parents who were interviewed insisted they made no modifications to their speech when speaking to their children. Yet every one of those 18 were observed to use typical features of CDS with their children (Haggan 2002). A similar study conducted in the city of Babol, Iran, found that parents generally deny correcting their children’s speech errors even though most parents actually did so (Birjandi & Nasrolahi 2012).

Even if parents don’t engage in CDS with their children, parents are not the only caregivers who children receive linguistic input from. Older siblings and other family members play a large role in child language development in many communities. Even children as young as 4 years old will modify their speech when talking to toddlers! (Shatz & Gelman 1973; Weppelman et al. 2003)

All this leads prominent child language researcher Eve Clark to conclude that “adult speech with young children appears to play a critical role in early acquisition, even in communities where adults may talk rather less with their children than in many middle-class Western societies.” (Clark 2024: 66). At present, then, there is little reason to suspect that CDS is not a universal feature of the world’s cultures. It may not be strictly necessary, but all communities seem to use it regardless.

What’s equally clear, however, is that CDS can look very different across languages! In the next issue we’ll look at some of the neat ways that baby talk varies in the languages of the world. Sign up below to be notified when the next installment in this series is published!

Subscribe now

ℹ️ Articles in this Series

• Part 1: Why you should be talking to your infant

• Part 2: What’s the point of baby talk?

• Part 3: Is baby talk good for your child?

• Part 4: Do all cultures use baby talk? [this post]

• Part 5: Baby talk in the languages of the world

• Part 6: How much should you talk to your child? [forthcoming]

• Part 7: What really matters when talking to your child

📚 Recommended Reading

First language acquisition

Understanding child language acquisition

📑 References

• Birjandi, Parviz & Atefeh Nasrolahi. 2012. Negative evidence in iranian children: How do Iranian parents correct their children’s errors? Journal of Language Teaching and Research 3(4). 700–706. https://doi.org/10.4304/jltr.3.4.700-706.

• Burnham, Denis, Christine Kitamura & Uté Vollmer-Conna. 2002. What’s new, pussycat? On talking to babies and animals. Science 296(5572). 1435–1435. https://doi.org/10.1126/science.1069587.

• Clark, Eve V. 2024. First language acquisition. 4th edn. Cambridge University Press. https://doi.org/10.1017/9781009294485.

• Cristià, Alejandrina, Emmanuel Dupoux, Michael Gurven & Jonathan Stieglitz. 2017. Child‐directed speech is infrequent in a forager‐farmer population: A time allocation study. Child Development 90(3). 759–773. https://doi.org/10.1111/cdev.12974.

• Haggan, Madeline. 2002. Self-reports and self-delusion regarding the use of Motherese: Implications from Kuwaiti adults. Language Sciences 24(1). 17–28. https://doi.org/10.1016/S0388-0001(00)00044-9.

• Heath, Shirley Brice. 1983. Ways with words: Language, life, and work in communities and classrooms. Cambridge University Press.

• Hilton, Courtney B., Cody J. Moser, Mila Bertolo, Harry Lee-Rubin, Dorsa Amir, Constance M. Bainbridge, Jan Simson, et al. 2022. Acoustic regularities in infant-directed speech and song across cultures. Nature Human Behaviour 6(11). 1545–1556. https://doi.org/10.1038/s41562-022-01410-x.

• Kess, Joseph Francis & Anita Copeland Kess. 1986. On Nootka baby talk. International Journal of American Linguistics 52(3). 201–211. https://doi.org/10.1086/466018.

• Nielsen, Mark, Daniel Haun, Joscha Kärtner & Cristine H. Legare. 2017. The persistent sampling bias in developmental psychology: A call to action. Journal of Experimental Child Psychology 162. 31–38. https://doi.org/10.1016/j.jecp.2017.04.017.

• Ochs, Elinor. 1982. Talking to children in Western Samoa. Language in Society 11(1). 77–104. https://doi.org/10.1017/S0047404500009040.

• Ochs, Elinor & Bambi B. Schieffelin. 2009. Language acquisition and socialization: Three developmental stories and their implications. In Alessandro Duranti (ed.), Linguistic anthropology: A reader (Blackwell Anthologies in Social & Cultural Anthropology 1), 296–328. 2nd edn. Malden, MA: Wiley-Blackwell.

• Saxton, Matthew. 2017. Child language: Acquisition and development. 2nd edn. SAGE.

• Schieffelin, Bambi B. 1990. The give and take of everyday life: Language socialization of Kaluli children (Studies in the Social and Cultural Foundations of Language 9). Cambridge University Press.

• Shatz, Marilyn & Rochel Gelman. 1973. The development of communication skills: Modifications in the speech of young children as a function of listener. Monographs of the Society for Research in Child Development. Oxford University Press 38(5). https://doi.org/10.2307/1165783.

• Shneidman, Laura A. & Susan Goldin‐Meadow. 2012. Language input and acquisition in a Mayan village: How important is directed speech? Developmental Science 15(5). 659–673. https://doi.org/10.1111/j.1467-7687.2012.01168.x.

• Upper, Mary. 1993. Lessons from language acquisition: Reports on a study of Oji-Cree first language learning in the home. In Jo-Ann Archibald & Sheena Selkirk (eds.), Selected papers from the 1988 and 1990 Mokakit conferences: Establishing pathways to excellence in First Nations Education. Vancouver, BC: Mokaki Indian Education Research Association.

• Voegelin, C. F. & Florence M. Robinett. 1954. “Mother Language” in Hidatsa. International Journal of American Linguistics 20(1). 65–70. https://doi.org/10.1086/464252.

• Weppelman, Tammy L, Angela Bostow, Ryan Schiffer, Evelyn Elbert-Perez & Rochelle S Newman. 2003. Children’s use of the prosodic characteristics of infant-directed speech. Language & Communication 23(1). 63–80. https://doi.org/10.1016/S0271-5309(01)00023-4.

• Yang, Charles D. 2006. The infinite gift: How children learn and unlearn the languages of the world. Scribner.

The Amazon and Bookshop links on this page are affiliate links, which means that I earn a small commission from Amazon for purchases made through them (at no extra cost to you).

If you’d like to support Linguistic Discovery, purchasing through these links is a great way to do so! I greatly appreciate your support!

Check out my Amazon storefront here.

Check out my Bookshop storefront here.

Nope, but you’d be forgiven for reaching that conclusion after reading these headlines reporting on research published in the Proceedings of the National Academy of Sciences last week:

Archaeologists may have discovered the oldest form of writing—ever (Popular Mechanics)

Stone Age symbols may push back the earliest form of writing (New Scientist)

Ancient art could hold clues to the origins of written language (Scientific American)

Ancient artifacts hint at earliest protowriting (Science)

And here’s how the New Scientist article opens:

Stone Age people 40,000 years ago used a simple form of writing comparable in complexity to the earliest stages of the world’s first writing system, cuneiform, according to a study of mysterious signs engraved on figurines and other artefacts found in Germany. If confirmed, this pushes back the emergence of a proto-writing system by more than 30,000 years. (New Scientist)

Big if true. In today’s issue we’ll look at what the study found, why it doesn’t push back the date of the earliest writing by tens of thousands of years, and why the findings are actually way cooler, giving us insight into the very nature of what it is to be human.

Background: Behaviorally modern humans in Europe

The earliest groups of Homo sapiens to arrive in Europe appear ca. 45,000 years ago (ya), and with them an artistic explosion in the form of cave paintings, petroglyphs, carvings, and engravings on bone or ivory, and a marked increase in the overall diversity of the types of artifacts associated with human activity.

This archaeological period is known as the Upper Paleolithic (Late Stone Age), and is typically associated with the emergence of behavioral modernity—when humans begin to exhibit abstract and symbolic thought, advance planning, and social learning.1 During this era, we see a dramatic increase in figurines of people and animals as well as objects adorned with sequences of geometric signs like lines, crosses, and dots. This archaeological industry associated with the first modern humans in Europe is known as the Aurignacian complex.

Linguist Christian Bentz and archaeologist Ewa Dutkiewicz were especially curious about the sequences of geometric marks: Were they just decorative, or did they have specific meanings? The fact that sequences of geometric signs recur so frequently and systematically in the archaeological record of this period suggests that they had some sort of meaning, maybe as tallies of kill counts, or marks on a lunar calendar, etc. For example, the Adorant figurine shown above, carved from mammoth ivory, depicts a lion-human, and is engraved with dots and notches in rows of 13 or 12. These engravings could well be calendric observations made to keep track of the passing of time. In fact, another study from 2023 argues that sequences of lines, dots, and Y shapes frequently found on European cave paintings of animals from the same time period were used to record information about the mating and birthing cycles of commonly-hunted animals (a phenological calendar).

So it’s entirely plausible that the sequences of marks in Aurignacian artifacts have some type of symbolic meaning, but how could we ever possibly know?

Reporting

• Could these cave markings be the earliest form of writing? New research proposes that symbols in 20,000-year-old cave drawings are a proto-writing system, but not all scientists are convinced (Smithsonian Magazine)

• 20,000-year-old cave painting ‘dots’ are the earliest written language, study claims. But not everyone agrees. (Live Science)

You’re reading Linguistic Discovery, a newsletter about the science and diversity of language—a field known as linguistics. I’m Danny Hieber, a PhD in linguistics who works with indigenous communities to help them document and revitalize their languages. Topics covered in this newsletter include:

• ⚙️ how language works (cognitive linguistics, language change)

• 🌍 grammatical diversity in the world’s languages (typology)

• ℹ️ explainers of terms and concepts in linguistics

• 🗣️ language profiles

• 🗞️ the latest news and research in language and linguistics

• ⭐ linguistic reviews of books and other media

Paid subscribers also get bonus articles and early access to my in-progress book projects!

Read more

Now there is another one. 🐤🐤

There are two of them. There are two _____.

You’ve just participated in one of the most famous linguistics experiments of all time, called The Wug Test. In today’s issue, we’ll see just how much this tiny experiment can teach us about the way children acquire language and how language itself works.

You’re reading Linguistic Discovery, a newsletter about the science and diversity of language—a field known as linguistics. I’m Danny Hieber, a PhD in linguistics who works with indigenous communities to help them document and revitalize their languages. Topics covered in this newsletter include:

• ⚙️ how language works (cognitive linguistics, language change)

• 🌍 grammatical diversity in the world’s languages (typology)

• ℹ️ explainers of terms and concepts in linguistics

• 🗣️ language profiles

• 🗞️ the latest news and research in language and linguistics

• ⭐ linguistic reviews of books and other media

Paid subscribers also get bonus articles and early access to my in-progress book projects!

Subscribe now

How do children learn grammatical rules?

The creator of the Wug Test, Jean Berko Gleason, wanted to understand how children acquired the rules of English grammar. For example, most English nouns form their plural in the regular way, by adding the ‑s suffix:

• cat → cats

• dog → dogs

• house → houses

But English also has plenty of irregular plurals:

• man → men

• mouse → mice

• sheep → sheep

Regular plurals comport well with the perspective that children learn a rule STEM + ‑s = PLURAL, but irregular plurals make it seem like children have to memorize each plural individually. Gleason wanted to know which strategy children were using when they learned different grammatical affixes like plural ‑s, past tense ‑ed, and progressive ‑ing—what linguists call the different inflections of a word. Do they simply memorize each form on a case-by-case basis, or do they learn a more abstract rule that they can then apply to new words?

Complicating matters, the task facing children when they learn new inflections is actually much more challenging than simply adding an ‑s or ‑ed to the ends of words. In many cases, English uses different pronunciations of an affix depending on the sound around it. In fact, there are actually three different pronunciations of each of the plural ‑s, the possessive ‑s, and the past tense ‑ed, depending on the sound at the end of the word they attach to.

Take plural ‑s to start: If a noun ends in what linguists call a voiceless consonant—a sound where your vocal folds aren’t vibrating, any of the sounds /f, k, p, t, θ/—then the plural is simply pronounced /s/. So cat /kæt/ becomes cats /kæts/, cup /kʌp/ becomes cups /kʌps/, and so on:

• cat /kæt/ → cats /kæts/

• cup /kʌp/ → cups /kʌps/

• month /mʌnθ/ → months /mʌnθs/

ℹ️ Here I’m writing words as they are pronounced, not as they are spelled, so you can see these pronunciation rules at work. To do so, I use the International Phonetic Alphabet (IPA). When a word/affix is written in the IPA in this article, it’ll be between /slashes/. Words are transcribed using a General American accent, so you might pronounce these words differently depending on your own accent.

However, if the word ends in a voiced sound—where your vocal folds are vibrating—then the plural ‑s is pronounced /z/. Voiced sounds in English include:

• vowels

• semivowels: /w, j/

• voiced consonants: /b, d, ð, g, l, r, m, n, ŋ, v/

So you say dogs /dɑgz/ not */dɑgs/, beds /bɛdz/ not */bɛds/, and cans /kænz/ not */kæns/.

ℹ️ An asterisk * is used to indicate an incorrect or ungrammatical form.

• dog /dɑg/ → dogs /dɑgz/

• bed /bɛd/ → beds /bɛdz/

• can /kæn/ → cans /kænz/

Notice that these plural forms end in /z/, even though the word is spelled with an ⟨s⟩. English spelling is insensitive to this particular pronunciation rule (because we apply it subconsciously). If you’re having trouble convincing yourself there’s actually a /z/ there, here are a few contrasting pairs of words (minimal pairs) that show what happens if you pronounce the plural as /s/ instead. You get an entirely different word! These should convince you that the plural is in fact pronounced as /z/, distinct from /s/.

Finally, if the word ends in one of English’s sibilant sounds—/s, z, ʃ, ʒ, tʃ, dʒ/—then the plural is pronounced /ɪz/. English inserts a separator vowel (called an epenthetic vowel) to keep the two sibilants apart (otherwise they’d be unpronounceable). So house /haᶷs/ becomes /haᶷsɪz/, not /haᶷss/; maze /meᶦz/ becomes /meᶦzɪz/, not /meᶦzs/, etc.

• house /haᶷs/ → houses /haᶷsɪz/

• maze /meᶦz/ → mazes /meᶦzɪz/

• glass /glæs/ → glasses /glæsɪz/

You can think of these three different forms of the plural ‑s as context-conditioned pronunciation variants of the same suffix. The technical term for this type of variant is an allomorph.

Suddenly “add an ‑s to make it plural” doesn’t seem so straightforward after all!

Thankfully, the pronunciation rules for the possessive ‑’s are exactly the same as for plural ‑s:

• cat /kæt/ → cat’s /kæts/

• cup /kʌp/ → cup’s /kʌps/

• month /mʌnθ/ → month’s /mʌnθs/

• dog /dɑg/ → dog’s /dɑgz/

• bed /bɛd/ → bed’s /bɛdz/

• can /kæn/ → can’s /kænz/

• house /haᶷs/ → house’s /haᶷsɪz/

• maze /meᶦz/ → maze’s /meᶦzɪz/

• glass /glæs/ → glass’s /glæsɪz/

These possessives may be spelled differently from their plural forms, but they are pronounced the same.

The rules for the pronunciation of different allomorphs of the past tense ‑ed parallel those of the plural and the possessive. After voiceless consonants, ‑ed is pronounced as /t/:

• look /lʊk/ → looked /lʊkt/

• ask /æsk/ → asked /æskt/

• help /hɛlp/ → helped /hɛlpt/

After voiced sounds, ‑ed is pronounced as /d/:

• use /juz/ → used /juzd/

• call /kɑl/ → called /kɑld/

• try /traᶦ/ → tried /traᶦd/

But after a similar sound—an alveolar stop /t d/—an epenthetic vowel is inserted, so that ‑ed is pronounced /ɪd/:

• want /wɑnt/ → wanted /wɑntɪd/

• need /nid/ → needed /nidɪd/

• start /stɑrt/ → started /stɑrtɪd/

We can generalize the rules for all three of these suffixes in the following way:

1. If the word ends in a similar sound as the suffix, insert an epenthetic vowel to separate the two sounds:

   • sibilant + ‑s → /ɪz/

   • alveolar stop + ‑ed → /ɪd/

2. Otherwise, use the form of the suffix that matches the voicing of the preceding consonant:

   • voiceless: /s/ or /t/

   • voiced: /z/ or /d/

Yet even this “simplification” is a complex, highly abstract rule! English spelling often obscures the intricacies of these pronunciation rules, but all native English speakers are nonetheless aware of these rules at a subconscious level.

But are children? And if so, when do they become aware of them? Do they learn all the allomorphs of an affix at once? Do they memorize individual forms, or do they learn the abstract rules for each of these forms? Obviously children learn the rules eventually, but what happens while they’re learning the language?

When Gleason first conducted the Wug Test in 1958, linguists didn’t yet have answers to these questions. In fact, Gleason’s study can perhaps be considered the first ever scientific experiment in child language research (Saxton 2017: 12). That’s only part of what has given the study its enduring fame, however. In addition to furnishing the field of linguistics with an unofficial mascot in the adorable wug (you can grab your own wug merch here), the experiment’s astute design provided the first method in linguistics for interrogating the internal grammatical knowledge of children, and it was one of the first studies to offer experimental evidence that children regularize irregular forms (Clark 2024: 225). Let’s look at how the experiment worked and what Gleason discovered.

How the Wug Test works

Gleason reasoned that if children learned plurals and other inflections by rote, they shouldn’t make inflectional errors. They would say cats and men because that’s what the adults say. But if children are learning patterns and applying them productively, they should sometimes overgeneralize, producing forms like these:

• Past Tense ‑ed

  • fall → *falled (fell)

  • go → *goed (went)

  • break → *broked (broke)

  • sing → *singed1 (sang)

• Plural ‑s

  • man → *mans (men)

  • foot → *foots (feet)

  • tooth → *teeths (teeth)

  • person → *peoples (people)

To find out what children were actually doing, Gleason created a number of nonsense words to describe colorful pictures of fantastical creatures (including the wug) and people performing novel actions. Below are a few of the original elicitation stimuli she used. The first four are designed to test for plural nouns, and the last two are designed to test for past tense verbs.

Gleason also created cards to test for the progressive ‑ing, the third person singular ‑s, and the possessive ‑’s and plural possessive ‑s’, but most items focused on plural ‑s and past tense ‑ed.

This experimental format worked particularly well with young children because it mirrors picture books, which they all had experience with. The nonsense words were also carefully built to obey English sound patterns. In fact, they were so plausible that Gleason writes:

It was, moreover, evident that a great number of these children thought they were being taught new English words. It was not uncommon for a child to repeat the nonsense word immediately upon hearing it and before being asked any questions. Often, for example, when the experimenter said “This is a gutch”, the child repeated, “Gutch”. Answers were willingly, and often insistently, given. (p. 3–4)

The words were designed to test whether children had internalized the various pronunciation rules explained above. Some ended in voiceless consonants, others in voiced sounds, others in sibilants, and others in alveolar stops. If the child really knew these rules, they should be able to apply them to new words to produce the right allomorph: /wʌgz/ for wugs, /rɪkt/ for ricked, and so on. Gleason also included a few real words to compare children’s performance on familiar words versus made-up ones. These came later in the experiment so that they would not prime the children’s responses for the nonsense words. For example, glass only appeared several turns after tass.

Before running the experiment with children, Gleason tested the materials with twelve adults, and used their answers as the standard of correctness. A child’s response was considered “correct” if it matched any adult response. The experiment was conducted with two groups of children in the U.S.: one group in preschool (where students are between 4 and 5 years old) and one group in first grade (where children are between 5.5 and 7 years old).

Results I: How children learn plural nouns

What did we learn from the Wug Test? Table 2 lays out the raw results, which we’ll break down in a moment. It shows the percentage of correct answers given for each test item in each age group.

Let’s start with plural nouns. Table 3 combines the data from the two age groups for plurals.

The word bik was also used in a pretest with the children, and Gleason reports that they used the correct form of bik with the same degree of accuracy as they did for wug.

The first thing to notice is the huge spread in accuracy across items—from 91% for glasses to 28% for nizzes. What explains this disparity? First, children have little trouble with the simpler allomorphs /‑s/ and /‑z/, even for novel words. Any difficulty they do have vanishes by first grade, as Table 2 shows. So at the outset we have strong evidence that children have in fact internalized an abstract rule about the formation of English plurals by age 7—already a very cool and (at the time) groundbreaking result.

Yet even among the /‑s/ and /‑z/ forms, children’s accuracy varies. Why do they perform so well with bik and wug but relatively worse on cra? Gleason posits that this is due to a quirk of the English sound system: Every language has rules about which sounds are allowed to appear next to each other and how they can be arranged within syllables and words. These are called the phonotactics or phonotactic rules for a language. One phonotactic rule of English is certain consonant clusters are disallowed at the ends of syllables. In particular, sibilants must match the voicing of the previous consonant. This means that the sequences */kz/ and */gs/ never occur—which coincidentally rules out forms like */wʌgs/ and */bikz/.

Because of this phonotactic rule, bik and wug are not actually the decisive test cases they seem to be. Correctly using /‑s/ with bik and /‑z/ with wug isn’t technically enough to demonstrate that children know the pronunciation rules for plural nouns. They might just be following a more general pattern about how English sounds can be arranged. The real test case, therefore, is nouns that end in vowels, the semivowels /w j/, or the liquid consonants /l r/. That’s precisely what lun, tor, and cra are designed to test. The fact that children perform well with these nouns shows definitively that they have internalized the pronunciation rules for the /‑s/ and /‑z/ variants of the English plural. And children’s mastery of these forms solidifies between preschool and first grade.

At the same time, however, children did not perform as well on lun, tor, and cra as they did on bik and wug. Why? It may be that children do better when plural form also aligns with English’s natural sound patterns. But for lun, tor, and cra, there is no such general pattern to follow. Children are forced to choose between /‑s/ and /‑z/ based on some other principle. The children who missed these items may have learned the general phonotactic rule about English but not the pronunciation rules for the English plural. What I find especially cool about this is that it demonstrates how two individuals can produce the same surface forms but have different underlying grammars. Linguists call this grammatical isomorphism.

Next, notice how poorly children performed with sibilant-final nouns—tass, gutch, kazh, nizz. The one exception was the word they already knew, glass. They performed better with glass than any of the nonsense words, but they struggled to extend /‑ɪz/ to new words, even by age 7. At first glance, this looks like rote learning: children produce glasses because they’ve memorized it; they haven’t learned the rule for /‑ɪz/ yet. But a closer look suggests something much more advanced (and cooler!) is going on. To understand what, we need to understand the difference between rules and schemas.

Rules vs. Schemas

What’s striking is not that children got these items wrong, but how confidently they got them wrong:

It must be noted, however, that in these items, the children delivered the wrong form with a great deal of conviction: 62% of them said “one tass, two tass” as if there were no question that the plural of tass should and must be tass. (p. 163)

Contrast this with the children’s behavior when they were uncertain:

Every child interviewed understood what was being asked of [them]. […] In some instances they pronounced the inflexional endings they had added with exaggerated care, so that it was obvious that they understood the problem and wanted no mistake made about their solution. Sometimes, they said “That’s a hard one,” and pondered a while before answering, or answered with one form and then corrected themselves. The answers were not always right so far as English is concerned; but they were consistent and orderly answers, and they demonstrated that there can be no doubt that children in this age range operate with clearly delimited morphological rules.

Elsewhere, Gleason notes that the most common response when children were uncertain was simple silence. So why the enthusiasm with sibilant-final nouns? It seems like children actually have learned a rule—just not the correct one. But what rule?

The rule these children have internalized isn’t, “Add an ‑s to make a noun plural,” it’s “Nouns that end in sibilants are plural.” When they hear gutch or tass, they assume it’s already plural because it ends in a sibilant. The difference between these two formulations is that the former is a rule-based approach with inputs and outputs (STEM + ‑s ⇒ PLURAL), and the latter is a schematic approach that checks whether the word matches a template or pattern. A rule-based approach suggests that children pay attention to the source or input form and then add inflectional endings to it, while a schematic approach suggests that children pay attention to the goal or target form and check whether the word they’re using fits the schema. Subsequent research supports the schematic perspective. Children do in fact rely on schemas rather than rules in their acquisition of inflection (Clark 2024: 239–240).

Children’s performance with heaf is also revealing in this regard: While 42% of adults gave heaves as the plural of heaf, only 3 out of 89 children did. Most children said heafs. This shows that children have not yet acquired the irregular /f/ ~ /v/ alternation, as in leaf ~ leaves. But more relevantly to the point about schemas, 9 children (10%) produced the singular heaf, and 4 used /‑ɪz/ as though /f/ were a sibilant. While /f/ is not actually a sibilant, it is perceptually very similar. Children who said heaf or heafes were treating /f/ as a sibilant, and either correctly applied the /‑ɪz/ suffix like any other sibilant, or matched it to the sibilant schema and treated it as already plural. What incredible pattern-matching and analogical reasoning skills these kids have!

Results II: Past tense and tricksy possessives

Children’s difficulty with the /‑ɪz/ form of the possessive ‑’s mirrored their difficulty with the /‑ɪz/ form of the plural ‑s, as you can see in Table 4:

This is no surprise, since the possessive ‑’s follows the same pronunciation rules as plural ‑s.

For plural possessives ‑s’, however, children didn’t add anything. They simply repeated the singular. While this is traditionally correct for plural possessives (e.g. the possessive of girls is girls’, and both words are pronounced the same: /ɡɝlz/), the children were probably producing the correct form for the wrong reason: they treated any noun ending in a sibilant as already possessive (and plural). Same output, but different underlying grammar.

😄 Amusingly, it was actually the adults who had trouble with plural possessives: 33% of them said wugses /‑zɪz/ and bikses /‑sɪz/, although none said nizzeses /‑ɪzɪz/. Gleason believes this is due to influence from proper names like “the Joneses” and states, “No adult would say that if two dogs own hats, they are the *dogses /‑zɪz/ hats.”

But I get the impression that now, 70 years later, people do in fact say this, at least when they’re trying to enunciate or clarify a plural possessive. And there’s certainly been an increase in the jocular use of this form thanks to Gollum in the Lord of the Rings movies. But even setting aside jocular uses, it seems to me that the use of /‑ɪz/ for plural possessives is actually starting to gain some acceptance. Perhaps in a hundred years we’ll all be saying “the girlses books” instead of “the girls’ books”.

As for the past tense ‑ed, children could easily handle the /‑t/ and /‑d/ variants, but not the /‑ɪd/ variant. Like with /‑ɪz/, they could correctly apply the /‑ɪd/ version to the real word melt → melted, but not the nonsense words mot and bod. And also like /‑ɪz/, their performance with /‑ɪd/ improved with time (14% of younger children correctly produced bodded as compared to 31% of older children), but overall their performance for this variant was dismal. Here again it is likely that children internalize a schema “verbs that end in an alveolar stop /t d/ are past tense” rather than a rule “add ‑ed to the stem”.

Another pattern that emerges from the past tense data is that children do not create irregular past tense forms like bing → bang or gling → glung. Virtually every English verb whose stem ends in ‑ing has an irregular past:

• sing → sang

• ring → rang

• cling → clung

So it’s reasonable to expect that children might pick up on this pattern too, as common as it is. Yet while 50% of adults use an irregular past for the nonsense words bing and gling, only one child did so for bing and one for gling. Remember that children likewise do not form an irregular plural heaves from heaf, even though adults do. Taken together, these results suggest that children at this age have not yet begun to internalize the handful of patterns governing irregular inflections—they’re still working on the regular ones!

What grammatical words/affixes do children learn first?

Overall, children’s performance with past tense verbs parallels their performance with plurals, with one notable difference: they do better with plurals than past tense. And that’s not surprising, because the plural ‑s suffix is generally acquired before the past tense ‑ed suffix! (Brown 1973). Table 5 shows the order of acquisition for 14 grammatical words and affixes in English.

(Spock apparently never made it past stage 12, since he doesn’t use contractions. 🖖🏼)

The three factors which most strongly influence this order are semantic complexity, formal complexity, and frequency (Clark 2024: 227). Semantic complexity refers to the number of meanings conveyed by a word/affix. The more meanings a word encodes, the longer children take to learn it. Notice how, in Table 5, the further down the list a word is, the more meanings it encodes.

Formal complexity refers to the number of forms that a word/affix has for the same function. The more forms there are, the more “moving parts” children have to learn. In Dutch, for example, there are two totally distinct plurals, ‑s and ‑en (which English has a remnant of in brethren), and the plural ending you use depends on the specific noun:

• Nouns that take the ‑en plural

  • boek ‘book’ → boeken ‘books’

  • kat ‘cat’ → katten ‘cats’

• Nouns that take the ‑s plural

  • tafel ‘table’ → tafels ‘tables’

  • dokter ‘doctor’ → dokters ‘doctors’

Even more complex is Swahili, where the plural prefix depends on the noun class, of which there are many. (Think of noun classes like grammatical gender, but situated around other semantic categories than masculine/feminine/neuter. Possible categories include humans, plants, body parts, man-made objects, animals, etc. The semantic basis of noun classes varies by language.) So the plural in Swahili is not one affix, but rather a whole set of affixes:

As you can see, Swahili plural nouns are quite complex, and children generally take longer to master all their nuances than English-speaking children do to master the comparatively simple ‑s suffix and its variants.

Finally, the more frequently a child is exposed to a form, the earlier they typically acquire it. This is one reason child-directed speech (“baby talk”) can have such an outsized effect on language acquisition. I discuss these factors in more detail in this article on baby talk.

Words or rules?

So what are children actually learning: productive rules or a big catalog of memorized forms? One popular textbook on child language acquisition frames the problem this way:

In principle, a child could get by in life without any knowledge of [grammatical affixes]. All we would need is a powerful long-term memory storage system, something which humans do, in fact, possess. Using a sufficiently extensive memory, we could learn every word form as it stands, with no reference to or acknowledgement of any internal morphological structure. On this model of learning, the child would treat words like parrots, believing and wanted as indivisible wholes. The child would never need to analyse words into their constituent parts (e.g., parrot + ‑s). (Saxton 2017: 178)

There is in fact good evidence that this is how children start out. They begin as whole-word learners, memorizing specific inflections for specific uses, and only later generalize the pattern to other words (Saxton 2017: 178; Evans 2019: 154; Zukowski & Bernstein Ratner 2022: 155–156; Clark 2024: 228). This means that children may produce adult-like forms long before they learn a productive rule. Children acquiring Hungarian, for instance, produce correctly-inflected words even during the earliest one-word stage of language (MacWhinney 1985). Initially, there are no signs that the inflectional affixes are productive. Children only produce inflected forms that they’ve already heard from their parents (Tomasello & Brooks 1998).

Children also produce inflected forms before they understand them. In one experiment, toddlers were presented with pictures of 1 novel object on one side and 2 novel objects on the other, and were told “Look at the X!”, where X would be a nonsense word either in the singular or plural, e.g. “Look at the jeels!” If children understand plural ‑s, they should look reliably at the correct picture. But children as old as 24 months were insensitive to the plural marker. They’d stare at either picture with equal likelihood. Only around 30 months did they begin to look consistently at the picture that matched the singular or plural form, even though they correctly produce plural markers well before 30 months! (Jolly & Plunkett 2008)

Another strategy children rely on, especially in highly inflected languages with lots of noun cases or verb conjugations, is to pick one form as the default—e.g., the infinitive form of the verb—and use it everywhere. So while the word itself may be well-formed, its use in context is often ungrammatical. But because, as we saw earlier, children tend to acquire the most frequent forms, this strategy may actually mean that children are getting the form right in the majority of cases!

This is another example of children using grammatical assumptions that approximate—but don’t quite match—adult grammar, like the plural and past tense schemas we saw earlier. Since numerous equivalent (isomorphic) grammars could produce the same surface-level output, children are faced with the unenviable task of inferring exactly which grammatical rules will best result in language that matches that of their caregivers. What this means in practice is that children constantly (and subconsciously) massage their grammatical assumptions in order to more and more closely approximate what they hear, until their speech converges on the adult target. Of course, sometimes they land on slightly different rules than their parents did, with the result that language changes with each generation, and now we have things like lonely transitives and costed as a regular past tense instead of cost. That’s language change for ya 🤷🏼

So do children learn abstract grammatical rules? Yes, eventually. They learn the most general rules first (e.g. add an /‑s/ to make a word plural) and only later do they master more specific rules that only apply in certain contexts (e.g. add an /‑ɪz/ only if the word ends in a sibilant). But evidence suggests that children aren’t so much acquiring rules as they are schemas or templates, and this explains their poor performance with the /‑ɪz/ form of the plural and the /‑ɪd/ form of the past. And for the children who do not answer items on the Wug Test correctly but nonetheless do produce correct plurals for real words, it is safe to say that they have not yet learned the abstract schema. They are still memorizing words on an item-by-item basis. It is only after sufficient exposure to the pattern/template that they learn to generalize.

Of course, children eventually do become sensitive to the internal structure of words and master all the grammatical rules of their language, and to this day one of the best demonstrations of that fact is the Wug Test.

📚 Recommended Reading

Gleason’s original 1958 article (published under her maiden name, Berko) is available via open access here:

• Berko, Jean. 1958. The child’s learning of English morphology. Word 14(2–3). 150–177. https://doi.org/10.1080/00437956.1958.11659661.

I have an entire series of articles on the science of baby talk here:

👶 The science of baby talk

• Part 1: Why you should be talking to your infant

• Part 2: What’s the point of baby talk?

• Part 3: Is baby talk good for your child?

• Part 4: Do all cultures use baby talk? [forthcoming]

• Part 5: Baby talk in the languages of the world

• Part 6: How much should you talk to your child?

• Part 7: What really matters when talking to your child

This is (in my opinion) the best introduction to child language acquisition:

Here are two popular YouTube videos about Gleason’s seminal contribution to the field of linguistics:

📑 References

• Berko, Jean. 1958. The child’s learning of English morphology. Word 14(2–3). 150–177. https://doi.org/10.1080/00437956.1958.11659661.

• Brown, Roger. 1973. A first language: The early stages. Harvard University Press.

• Clark, Eve V. 2024. First language acquisition. 4th edn. Cambridge University Press. https://doi.org/10.1017/9781009294485.

• Evans, Vyvyan. 2019. Cognitive linguistics: A complete guide. 2nd edn. Edinburgh University Press.

• Jolly, Helen R. & Kim Plunkett. 2008. Inflectional bootstrapping in 2-year-olds. Language & Speech 51(1–2). 45–59. https://doi.org/10.1177/00238309080510010401.

• MacWhinney, Brian. 1985. Hungarian language acquisition as an exemplification of a general model of grammatical development. In The crosslinguistic study of language acquisition. Psychology Press.

• Saxton, Matthew. 2017. Child language: Acquisition and development. 2nd edn. SAGE.

• Tomasello, Michael & Patricia J. Brooks. 1998. Young children’s earliest transitive and intransitive constructions. Cognitive Linguistics 9(4). 379–396. https://doi.org/10.1515/cogl.1998.9.4.379.

• Zukowski, Andrea & Nan Bernstein Ratner. 2022. Putting words together: Comprehension and production of morphology and syntax in the preschool years. In Jean Berko Gleason & Nan Bernstein Ratner (eds.), The development of language. 10th edn. Plural Publishing.

The Amazon and Bookshop.org links on this page are affiliate links, which means that I earn a small commission from Amazon for purchases made through them (at no extra cost to you).

If you’d like to support Linguistic Discovery, purchasing through these links is a great way to do so! I greatly appreciate your support!

Check out my Amazon storefront here.

Check out my Bookshop storefront here.

There’s a difference between “parentese” and baby talk! Parentese is used in conjunction with English (or whatever your native tongue is). Gibberish generally isn’t helpful in developing language skills. (link)

“Motherese” is pervasive across cultures and may be something we’ve evolved to do. It would seem odd to me that we do it as we do if there weren’t some evolutionary advantage. (link)

I never bothered with baby talk with all of the bubs I’ve cared for, including my 2. I couldn’t see the point. Not sure if this was the reason, but my siblings & my children & now my grandchildren, all have an extensive & varied vocabulary & excellent command of the English language. (link)

It seems to me that baby talking to them teaches them a particular pitch, prosody, and vocabulary that they’ll then have to unlearn whenever the parent decides to start talking “normally” to them. I have four children and while one is just barely past a year old, the others are 2.5, 4, and 13, and they have all three spoken very well for their ages—we never allowed baby talk around them. Perhaps a case of “correlation but not causation,” but I don’t believe so. (link)

It depends on what you mean? Short sentences and sweet nicknames? Fine. Making up silly words, speaking incorrectly or saying words wrong on purpose to be cute? Bad. (link)

Good! When kids are very little, using the two-word sentences they say can help with comprehension. Not saying to do baby talk exclusively, but a well placed “find shoes?” or “want snuggle?” can streamline understanding and effective communication (link)

You are raising adults! Go ahead and baby talk if you have speech therapy money (link)

What do you mean by baby talk? Slow, sing-song intonation? Great! Adults deliberately using incorrect grammar like “me want cookie” when speaking to babies? Bad! And also fuckin weird! Nursery words? No idea but I’ve never met a six year old who doesn’t know that “choo choo” and “train” are the same thing so I doubt it’s a big deal and I have had fights in person about it when someone other than the kid’s parents tried to tell me not to say “bunny rabbit.” (link)

I always talked my kids with an adult voice. Why should they have to learn a second vocabulary and grammar? (link)

It’s not beneficial. Infants are apers. They mimic their parents and learn. Talk to them like people, so they learn to speak like people. (link)

This question got a lot more responses than I was expecting! According to a poll I posted on Instagram and Threads with several hundred responses on each platform, the overall bias was in favor of baby talk:

Poll: Is baby talk good or bad for your child?

This is an issue that people are divided and passionate about! So which take is correct? Today we’ll consider the advantages and disadvantages of baby talk in this third installment in my series on the science of talking to babies. Be sure to subscribe below to get notified when the other issues in this series go live!

Subscribe now

ℹ️ Articles in this Series

• Part 1: Why you should be talking to your infant

• Part 2: What’s the point of baby talk?

• Part 3: Is baby talk good for your child? [this issue]

• Part 4: Do all cultures use baby talk?

• Part 5: Baby talk in the languages of the world

• Part 6: How much should you talk to your child? [forthcoming]

• Part 7: What really matters when talking to your child

Does baby talk even work?

A 2006 book about child language acquisition states:

[Being told] “Your instinct is supported by science” is, of course, just the kind of thing that parents like to hear. But let’s get the facts right. Preference for motherese does not establish its necessity, or even its usefulness. […] For all we know, motherese is neither good nor bad for child language learning: it is simply irrelevant. Language learning is remarkably resilient. All you need to do is to talk to children—baby-talk or otherwise—and biology will take care of rest. (Yang 2006: 101–103)

While the second part of this claim is undeniably true—your children will learn language proficiently as long as you speak to them regularly—the subsequent two decades of research since the book was published have shown a number of benefits from using baby talk (also called infant-directed speech (IDS)). The perspective represented in that book stems from a time when the majority of linguists believed that grammar is genetically endowed—that we have a “language instinct”, to use Steven Pinker’s phrase—a position called linguistic nativism. Because grammar is hardwired into us, the thinking goes, it doesn’t really matter how much or what kind of language children are exposed to. Children merely need to be exposed to a language, and then their language instinct will kick in and apply those innate grammatical rules to whatever language it is they’re hearing (or seeing, in the case of infants exposed to sign language). That innate grammar—called Universal Grammar—is thought to help children make sense of the messy, overwhelming, inconsistent language input they encounter in the world.

Linguist Noam Chomsky famously claimed that without an innate grammar, it was impossible for children to learn grammatical rules as proficiently as they do. He posited that the amount of linguistic data that children receive in infancy is simply insufficient for them to work out the complex rules of language—there exists a Poverty of the Stimulus, to use his term.

For linguistic nativists, whether adults use baby talk or regular speech with children is irrelevant, because Universal Grammar will impose the same innate grammatical rules regardless. However, advances in our understanding of cognition in the last several decades have inclined more and more researchers away from the nativist position. Many, perhaps the majority, of linguists now believe that children have all the cognitive tools they need to make sense of the messy language input they receive in infancy, without having to posit an innate grammar mechanism. This approach to language is typically referred to as cognitive linguistics due to its emphasis on aligning linguistic theory with insights from the rest of cognitive science.

One of the many pieces of evidence supporting the cognitive approach is the fact that baby talk does in fact have significant effects on a child’s linguistic proficiency. In the last three decades especially, a growing body of research has demonstrated that the amount and type of linguistic input children receive really does affect their knowledge of language.

For example, children tend to master the most frequent elements of their language first. Morphemes that are used with high frequency on many different word stems in adult speech, such as the plural ‑s in English, typically appear early in the child’s speech (Clancy 2018: 352). Children of mothers who use more yes/no questions in their speech acquire auxiliary verbs earlier (Newport, Gleitman & Gleitman 1977 ). And the frequency of a given word in adult speech, along with the number of uses in one-word utterances and the length of the word, predicts children’s comprehension and production of those words at 15 months (Clark 2024: 38).

Another great example of the importance of frequency in language acquisition is how children learn Peninsular (Spain) Spanish vs. Mexican Spanish: In Peninsular Spanish, the present perfect tense (e.g. he comido ‘I have eaten’) is more common than it is in Mexican Spanish. As a result, the first tense contrast that Spanish children learn is the present tense vs. the present perfect tense, whereas the first tense contrast that Mexican children learn is the present tense vs. the past preterite tense (e.g. comí ‘I ate’) (Chee et al. 2023: 381). Similarly, in Mexican Spanish the plural /s/ is almost always expressed, but in Chilean Spanish the /s/ is unpronounced in certain contexts. As a result, Chilean children take longer to master the formation of plurals than Mexican children do (Chee et al. 2023: 381).

Want to learn more about how children’s linguistic capabilities develop over time? Check out this issue of the newsletter about how children learn grammatical rules (and how we know):

The type of input children receive matters too: one study found that the greater the proportion of commands that a caregiver used with their children (as opposed to questions), the less complex the children’s verb phrases and noun phrases were later on. The same researchers also found that more frequent use of words that rely on context (deixis), such as this, that, then, you, here, them, etc., correlates with vocabulary growth and children’s later development of noun phrases. (Newport, Gleitman & Gleitman 1977)

All this shows us that the amount and type of input that children receive does matter, contra the claims of the linguistic nativists. But what about baby talk specifically? We turn to this in the next section.

Is baby talk good for your child?

The main way that baby talk helps children learn language is by making it easier for children to understand what’s being said to them, through slower, exaggerated speech, simpler syntactic constructions, and targeted choice of vocabulary. This was the main point of Part 2 of this series on baby talk, so you can read all about the details of those features of baby talk there. But that’s only part of how baby talk benefits children.

For starters, it’s well established that babies prefer listening to infant-directed speech (IDS)—and particularly IDS with exaggerated prosody—over other types of speech (Cooper & Aslin 1990; Cooper 1993; Hernik & Broesch 2019; The ManyBabies Consortium et al. 2020). This matters for language learning because it helps get the child’s attention! Infants remember individuals who address them in IDS better than those who don’t, and look longer at them. This in turn makes the child more interesting to the adult, who is then more likely to continue using IDS with them! (Schachner & Hannon 2011) One study shows that infants who are good at observing their caregiver’s mouth and following their gaze acquire vocabulary especially quickly (Tenenbaum et al. 2015).

However, the preference for IDS over adult-directed speech starts to wane around the child’s first birthday (Ibbotson 2022: 66), which may be why adults use IDS most prominently during the child’s first year but less so later (Saxton 2017: 88). This is another great example of how adults subconsciously adjust their speech to the needs and preferences of the child over time, as I emphasized in Part 2 of this series.

The exaggerated prosody and pronunciation used in baby talk also have positive effects for children’s language acquisition: infants whose parents hyperarticulate speech sounds perform better at discriminating between consonants in their parents’ language (Liu, Kuhl & Tsao 2003), and exaggerated vowels and slower speaking rates help infants recognize words better (Song, Demuth & Morgan 2010).

Infants who were receiving a greater amount of IDS at 18 months also had larger expressive vocabularies by 25 months (Weisleder & Fernald 2013). This was specifically correlated with IDS, and not speech simply overheard by the child. Social interaction is key to the language-learning process. Use of IDS with an infant also predicts greater lexical diversity, length of utterances, and frequency of conversational turn-taking in the child’s speech at age 5 (Ferjan Ramírez et al. 2024). (It’s important to point out here that this is just a correlation. Other aspects of the child’s environment likely contribute to their conversational style, vocabulary size, and utterance lengths as well.)

We’ll explore other benefits of IDS in a later installment in this series, but suffice it to say for now that baby talk is basically all benefits and no drawbacks. So if you’re a caregiver worried about the detrimental effects of baby talk on your infant, rest easy—you’re simply making the incredibly complex task of language learning easier for them.

Nonetheless, baby talk isn’t strictly necessary for successful child language acquisition. This raises some fascinating questions: Is baby talk universal? Do all parents use it? Do all cultures use it? In the next issue, we’ll look at baby talk in the languages of the world! Be sure to subscribe to the newsletter if you haven’t already in order to receive future issues in this series.

Subscribe now

ℹ️ Articles in this Series

• Part 1: Why you should be talking to your infant

• Part 2: What’s the point of baby talk?

• Part 3: Is baby talk good for your child? [this issue]

• Part 4: Do all cultures use baby talk?

• Part 5: Baby talk in the languages of the world

• Part 6: How much should you talk to your child? [forthcoming]

• Part 7: What really matters when talking to your child

📚 Recommended Reading

How babies talk: The magic and mystery of language in the first three years of life

Understanding child language acquisition

📑 References

• Chee, Melvatha R., Frances V. Jones, Jill P. Morford & Naomi L. Shin. 2023. Usage-based approaches to child language development: Insights from studies of Navajo, ASL, and Spanish. In Manuel Díaz-Campos & Sonia Balasch (eds.), The handbook of usage-based linguistics (Blackwell Handbooks in Linguistics), 379–392. Wiley-Blackwell.

• Clancy, Patricia M. 2018. First language acquisition. In Carol Genetti (ed.), How languages work: An introduction to language and linguistics. 2nd edn. Cambridge University Press. https://doi.org/10.1017/9781108553988.

• Clark, Eve V. 2024. First language acquisition. 4th edn. Cambridge University Press. https://doi.org/10.1017/9781009294485.

• Cooper, R. P. 1993. The effect of prosody on young infants’ speech perception. Advances in Infancy Research 8. 137–167.

• Cooper, Robin Panneton & Richard N. Aslin. 1990. Preference for infant-directed speech in the first month after birth. Child Development 61(5). 1584–1595. https://doi.org/10.2307/1130766.

• Ferjan Ramírez, Naja, Yael Weiss, Kaveri K. Sheth & Patricia K. Kuhl. 2024. Parentese in infancy predicts 5-year language complexity and conversational turns. Journal of Child Language 51(2). 359–384. https://doi.org/10.1017/S0305000923000077.

• Hernik, Mikołaj & Tanya Broesch. 2019. Infant gaze following depends on communicative signals: An eye‐tracking study of 5‐ to 7‐month‐olds in Vanuatu. Developmental Science 22(4). e12779. https://doi.org/10.1111/desc.12779.

• Ibbotson, Paul. 2022. Language acquisition: The basics (The Basics Series). Routledge. https://doi.org/10.4324/9781003156536.

• Liu, Huei‐Mei, Patricia K. Kuhl & Feng‐Ming Tsao. 2003. An association between mothers’ speech clarity and infants’ speech discrimination skills. Developmental Science 6(3). https://doi.org/10.1111/1467-7687.00275.

• Newport, Elissa L., Henry Gleitman & Lila R. Gleitman. 1977. Mother, I’d rather do it myself: Some effects and non-effects of maternal speech style. In Catherine E. Snow & Charles A. Ferguson (eds.), Talking to children: language input and acquisition, 101–149. Cambridge University Press.

• Saxton, Matthew. 2017. Child language: Acquisition and development. 2nd edn. SAGE.

• Schachner, Adena & Erin E. Hannon. 2011. Infant-directed speech drives social preferences in 5-month-old infants. Developmental Psychology 47(1). 19–25. https://doi.org/10.1037/a0020740.

• Song, Jae Yung, Katherine Demuth & James Morgan. 2010. Effects of the acoustic properties of infant-directed speech on infant word recognition. The Journal of the Acoustical Society of America 128(1). 389–400. https://doi.org/10.1121/1.3419786.

• Tenenbaum, Elena J., David M. Sobel, Stephen J. Sheinkopf, Bertram F. Malle & James L. Morgan. 2015. Attention to the mouth and gaze following in infancy predict language development. Journal of Child Language 42(6). 1173–1190. https://doi.org/10.1017/S0305000914000725.

• The ManyBabies Consortium, Michael C. Frank, Katherine Jane Alcock, Natalia Arias-Trejo, Gisa Aschersleben, Dare Baldwin, Stéphanie Barbu, et al. 2020. Quantifying sources of variability in infancy research using the infant-directed-speech preference. Advances in Methods and Practices in Psychological Science 3(1). 24–52. https://doi.org/10.1177/2515245919900809.

• Weisleder, Adriana & Anne Fernald. 2013. Talking to children matters: Early language experience strengthens processing and builds vocabulary. Psychological Science 24(11). 2143–2152. https://doi.org/10.1177/0956797613488145.

• Yang, Charles D. 2006. The infinite gift: How children learn and unlearn the languages of the world. Scribner.

The Amazon and Bookshop.org links on this page are affiliate links, which means that I earn a small commission from Amazon for purchases made through them (at no extra cost to you).

If you’d like to support Linguistic Discovery, purchasing through these links is a great way to do so! I greatly appreciate your support!

Check out my Amazon storefront here.

Check out my Bookshop storefront here.

Prefer a video version of this post? Watch here:

For those not familiar with Proto-Indo-European: it’s the hypothetical, reconstructed ancestor language from which all modern Indo-European languages descend and was spoken around 4000 BCE. There are an estimated 454 living Indo-European languages in the family, including English, Hindi, Spanish, French, Russian, Portuguese, German, Persian, and many others.

🔠 Abbreviations Used in this Article

• DE: Dutch

• L: Latin

• ME: Modern English

• OE: Old English

• PG: Proto-Germanic

• PIE: Proto-Indo-European

The first reflex (descendant) of *leubʰ- in this list comes via Latin. In the Italic branch of Indo-European, *leubʰ- developed into the Latin word libēre ‘to please’ and the derived noun libīdō ‘desire’. As you can guess, that noun was then borrowed into English at the turn of the 1900s as the word libido.

PIE *leubʰ- ‘love, care, desire’ → L libēre ‘to please’ → L libīdō ‘desire’ ⇢ ME libido

ℹ️ The asterisk * before a word indicates that it has been reconstructed using the comparative method in historical linguistics. The word was never actually attested in any written source. Instead, it is a hypothetical reconstruction based on available evidence and the many techniques linguists have at their disposal for winding back the clock on language history.

If an entire parent language has little or no extant written texts and has been reconstructed using these techniques, it has the prefix “Proto-”. Proto-Indo-European is the reconstructed parent language of all the modern Indo-European languages, while Proto-Germanic is the reconstructed parent language of all the modern Germanic languages (and is itself a child language descended from Proto-Indo-European).

In the Germanic branch of the Indo-European family, however, the word took on a new meaning: Proto-Germanic *laubō ‘permission’. This is a great example of how the meanings of words can sometimes shift from one stage of an action to another, or from one concept to something closely related. In this case, if you desire something, you may be looking for permission to do or acquire it, and so the meaning of the word expanded to include not just ‘desire’ but also ‘permission’. This word then evolved into the Old English lēaf, which became Modern English leave, as in, “to take one’s leave”. (This sense of leave is unrelated to the sense of ‘to go away’; the two words just happen to be homonyms.)

PIE *leubʰ- ‘love, care, desire’ → PG *laubō ‘permission’ → OE lēaf ‘leave’ → ME leave

You’re reading Linguistic Discovery, a newsletter about the science and diversity of language—a field known as linguistics. I’m Danny Hieber, a PhD in linguistics who works with indigenous communities to help them document and revitalize their languages. Topics covered in this newsletter include:

• ⚙️ how language works (cognitive linguistics, language change)

• 🌍 grammatical diversity in the world’s languages (typology)

• ℹ️ explainers of terms and concepts in linguistics

• 🗣️ language profiles

• 🗞️ the latest news and research in language and linguistics

• ⭐ linguistic reviews of books and other media

Paid subscribers also get bonus articles and early access to my in-progress book projects!

Subscribe now

There was also a prefixed form of that Old English word: belēfan ‘to give leave or permission to’, or perhaps literally ‘to hold dear’ (possibly retaining some of the original meaning from Proto-Indo-European), and that led to the Modern English believe.

PIE *leubʰ- ‘love, care, desire’ → PG *laubō ‘permission’ → OE belēfan ‘believe’ → ME believe

The Proto-Germanic word *laubō ‘permission’ was also inherited by Dutch and appears in the prefixed word verlof ‘permission’, literally ‘for-leave’. In the 1600s the Dutch word was borrowed into English and the final /f/ sound was lost, yielding Modern English furlough.

PIE *leubʰ- ‘love, care, desire’ → PG *laubō ‘permission’ → D verlof ‘permission’ ⇢ ME furlough

Lastly, that Proto-Indo-European root *leubʰ- didn’t only change its meaning to ‘permission’ in Proto-Germanic. It also kept its original meaning ‘love’ in the form *lubō. In Old English *lubō became lufu, and that evolved into the word love.

PIE *leubʰ- ‘love, care, desire’ → PG *lubō ‘love’ → OE lufu ‘love’ ⇢ ME love

You can see the complete etymological flowchart for all these word histories below.

Happy Valentine’s Day! 💜

📖 Recommended Reading

📑 Sources

• The Oxford dictionary of word origins

• The Chambers dictionary of etymology

• The American Heritage dictionary of Indo-European roots

The Amazon and Bookshop.org links on this page are affiliate links, which means that I earn a small commission from Amazon for purchases made through them (at no extra cost to you).

If you’d like to support Linguistic Discovery, purchasing through these links is a great way to do so! I greatly appreciate your support!

Check out my Amazon storefront here.

Check out my Bookshop storefront here.

ℹ️ Articles in this Series

• Part 1: Libfixes: When word parts go rogue

• Part 2: The science of libfixes: How libfixes work, where they come from, and what they teach us about language and the mind [this issue]

Libfixes vs. blends

Like all newer technical terms, the exact definition of a libfix is still contentious, but we can nonetheless delineate the concept a bit more precisely.

First off, libfixes are more than just simple rebracketings. What do I mean by this? The English words apron, newt, and umpire were originally more like napron, ewt, and numpire, but the boundary between the indefinite article a/an and the word was reanalyzed in a process of rebracketing:

• a napron → an apron

• an ewte → a newt

• a noumpere → an umpire

A similar process happened with the Romanian adjectival suffix ‑os ‘abounding in …’. It was originally used like so:

• dealuri ‘hills’ → deluros ‘hilly’

• colțuri ‘corners’ → colțuros ‘edgy’

Speakers saw the sequence ‑uros frequently enough that they took it as the adjectival suffix, and now it can be used to form adjectives like buburos ‘pimply’ (bubă ‘pimple’ + ‑uros) (Vasileanu & Niculescu-Gorpin 2025: 244).

Neither of the above sets of examples are considered libfixes. They merely adjust the boundaries of existing words/affixes rather than create new ones. While libfixes do always involve some type of “re”bracketing (more on this later), rebracketing alone isn’t what defines them.

Secondly, existing affixes used on new words with their original meaning intact also do not count as libfixes (Vasileanu & Niculescu-Gorpin 2025: 244). For example, ‑cracy, from the Ancient Greek suffix ‑κρατία ‑kratía ‘rule, government’, has been used playfully/critically in recent years to form new words such as idiocracy, shopocracy, kakistocracy. Likewise, we now have rizzology to accompany older uses of ‑ology ‘the study of a particular subject’. But no new affix has been created in these cases.

However, libfixes do include cases of existing affixes that have undergone significant changes in meaning. For example, ‑verse is generally considered a libfix because it now refers to ‘community of X’ rather than its originally meaning ‘turned’ (uni‑verse etymologically meant ‘turned into one’). This is also the case for many of the prefixal libfixes listed above:

• cyber‑ (cybercensorship, cyberbully) counts as a libfix because it means ‘relating to the internet or computers generally’, and no longer has any connection to the original Greek κυβερνήτης kŭbernḗtēs ‘steersman’. cyber‑ is simply a blending of cybernetic.

• petro‑ (petrodollar, petrostate) counts as a libfix because it isn’t based on the original Latin or Greek roots petra or πέτρα pétra ‘stone, rock’, but instead comes blends involving petroleum.

• crypto‑ (cryptocoin, cryptoverse) counts as a libfix because it’s based on the term cryptocurrency and refers to the ‘currency’ part of its source word rather than ‘cryptography’ part.

This connection between libfixes and blends turns out to be key to their definition, as we’ll see in a bit.

You’re reading Linguistic Discovery, a newsletter about the science and diversity of language—a field known as linguistics. I’m Danny Hieber, a PhD in linguistics who works with indigenous communities to help them document and revitalize their languages. Topics covered in this newsletter include:

• ⚙️ how language works (cognitive linguistics, language change)

• 🌍 grammatical diversity in the world’s languages (typology)

• ℹ️ explainers of terms and concepts in linguistics

• 🗣️ language profiles

• 🗞️ the latest news and research in language and linguistics

• ⭐ linguistic reviews of books and other media

Paid subscribers also get bonus articles and early access to my in-progress book projects!

Subscribe now

The phonology of libfixes

The boundaries of libfixes are often fuzzy. Consider two different versions of the ‑geddon libfix: ‑geddon vs. ‑mageddon:

• ‑geddon: cybergeddon

• ‑mageddon: gaymaggedon

Or ‑holic vs. ‑aholic/‑oholic:

• sugarholic

• workaholic

The reason for this variation is that libfixes like to preserve the syllable structure and stress pattern (the prosody) of their original source word (Vasileanu & Niculescu-Gorpin 2025: 246). For example, most (but not all) ‑tainment derivatives are 4-syllable words stressed on the penultimate (second-to-last) syllable, just like entertainment:

• e.du.ˈtain.ment

• in.fo.ˈtain.ment

• sci.en.ˈtain.ment

ℹ️ How to read linguistics

If any of the abbreviations or notations in this article are unfamiliar to you, check out the “How to read linguistics” page. It lists all the abbreviations, conventions, and notations used in Linguistic Discovery content.

In the case of ‑geddon, the speaker chooses whichever version of the libfix—‑geddon or ‑mageddon—best preserves the prosodic template of the original word armageddon. When ‑geddon attaches to a word with 2 or more syllables, the simple ‑geddon version suffices to fill out the template:

• cy.ber.ged.don

• ba.na.na.ged.don

But when ‑geddon attaches to a 1-syllable word, speakers recruit more material from the original armageddon to fill in the rest of the template, usually ‑mageddon but sometimes just ‑aggedon:

• car.ma.ged.don

• snow.ma.ged.don

• porn.a.ged.don

Another great example of this template-filling process is the word Whoniverse ‘all the fictional media related to the television program Dr. Who’. Almost always, the libfix ‑verse appears as ‑verse, but in this case it recruits the ‑ni‑ of universe as well, because the resulting derivation Whoniverse almost exactly matches the prosodic template of the original word, universe (just with an extra /h/ sound at the beginning).

Of course, sometimes the prosodic templates of the original word and the new base word match so closely that it’s impossible to tell where the boundary lies:

• Karmageddon (a 2011 movie) ← karma + ???

It is this preference for preserving the prosodic template of the source word that motivates the insertion of a dummy vowel (called an epenthetic vowel) in many libfixes:

• work‑a‑holic

• hack‑a‑thon

• crunch‑a‑licious

• blog‑o‑sphere

In fact, most libfixes are semantically transparent precisely because they preserve enough of the original word and its prosodic template that you know the original word being referred to. Some researchers actually include this in their definition of a libfix: “word parts emerging in blends that are analogically used in coining new words, while still preserving the connection with their source-words” (Vasileanu & Niculescu-Gorpin 2025: 241). So even if you’ve never encountered the word snowmageddon—or the libfix ‑geddon in general—you can still guess its meaning because enough of the original word armageddon is preserved, either directly in its consonants and vowels, or indirectly in its overall prosodic profile.

Given this behavior, it may be more useful to think of libfixes not so much as a new type of affix, but rather as a new template that can be used to create new blends.

Lastly, let’s look at what happens to libfixes when there’s a mismatch between their spelling and their pronunciation. Take the libfix ‑splain as an example. It originates as a blend with the word explain, but it’s spelled ‑splain not ‑xplain. The reason for this is because the letter ⟨x⟩ in the Roman alphabet typically represents a sequence of two sounds /ks/, and the syllable boundary in explain just so happens to fall right between those two sounds—right in the middle of the ⟨x⟩, so to speak. If I were to write explain phonemically (according to its sounds rather than its spelling), it’d be /ɛk.spleᶦn/. The part of the word that became the libfix was only the second syllable, /spleᶦn/, so we write it as ‑splain.

What all this demonstrates is that speakers create blends and libfixes based on their sounds rather than their spelling—and do so without even realizing it! In fact, all the phonological behaviors in this section illustrate just how much unconscious knowledge speakers have about sound patterns in their own language. We apply these moderately complex rules about prosodic profiles and syllable boundaries with no awareness that that’s what we’re doing, and I think that’s pretty remarkable.

Where do libfixes come from?

All libfixes start out as parts of a blend. A blend is a word formed by combining parts of existing words (Bauer 2004: 22). Canonical examples in English are smog (smoke + fog) and motel (motor + hotel). Blends were also famously called portmanteau words by Lewis Carroll, who coined the term to describe the words of the Jabberwocky poem in Through the Looking Glass (1871), like slithy (slimy + lithe) and mimsy (miserable + flimsy):

You see it’s like a portmanteau—there are two meanings packed up into one word.

~ Humpty Dumpty explaining words in the Jabberwocky poem to Alice.

In the English of the time, a portmanteau was a suitcase that opened into two equal sections, from the French porte‑manteau (porter ‘to carry’ + manteau ‘cloak’) (Etymonline: portmanteau). In this article, however, I’ll continue to refer to them as blends.

The chopped up bits of words that arise from blends are often called splinters. Every libfix starts out life as a splinter. Very occasionally, splinters can even transcend the affix stage and become fully independent words. This is what happened with burger: The word burger comes, of course, from hamburger, but the internal boundaries of that word were not originally ham‑burger. Instead the word is a reference to the city of Hamburg—a “Hamburger sandwich”, as it was originally called. But the presence of ham in that word motivated English speakers to reanalyze its internal boundaries as ham‑burger—a process called analogical reanalysis, etymological reinterpretation, or most commonly, folk etymology, where speakers reinterpret the internal pieces of a word to match other words or affixes they’re already familiar with. This reanalysis then made it possible to use hamburger in a number of blends: cheeseburger, veggieburger, fishburger, nothingburger, etc. The new libfix ‑burger was born.

In addition to its status as a libfix, burger also became an independent word. This happened early in the libfixation process: cheeseburger is first documented in 1938 (and is the earliest ‑burger blend I’m aware of), but burger appears by itself as an independent word just a year later (Etymonline: burger). Because of its early appearance, many linguists analyze burger as simply a clipping (shortening) of hamburger rather than a bona fide libfix.

You should be starting to notice a pattern at this point: libfixes always follow the same diachronic (historical) trajectory (Vasileanu & Niculescu-Gorpin 2025: 245):

1. [optional] Speakers reanalyze an existing word as having different internal boundaries than it did historically (analogical reanalysis / etymological reinterpretation / folk etymology). This is what happened with the ‑burger of hamburger, and maybe the ‑copter of helicopter and the ‑ol of alcohol. This step isn’t necessary, but it can strengthen the perception of an internal boundary at the place where blending later happens (Step 2).

2. Two words are blended (e.g. workaholic, cheeseburger, Brexit).

3. The new blend becomes popular.

4. The popularity of the original blend serves as a template by which subsequent blends are formed: shopaholic, veggieburger, Grexit). The splinter from the original blend (‑holic, ‑burger, ‑exit) can now be reused with a variety of words.

The key point in all this is that libfixes are always based on an initial blend, not the words that originally made up the blend or their etymologies. chocoholic and twitterholic were formed by analogy to workaholic, not alcoholic. Tipflation and shrinkflation were formed by analogy to stagflation, not inflation.

This better explains why forms like crypto‑, cyber‑, and petro‑ count as libfixes but ones like ‑cracy and ‑ology don’t. Because every libfix is based on a blend and relies on that blend for its meaning, the original pieces of that blend and their etymologies are irrelevant. crypto‑ is based on cryptocurrency, not cryptography; cyber‑ is based on cybernetics, not the Greek κυβερνήτης kŭbernḗtēs ‘steersman’; ‑gate is based on Watergate, not the original word gate. By contrast, new uses of ‑cracy and ‑ology are always based the same original ‑κρατία ‑kratía and ‑λογῐ́ᾱ ‑logĭ́ā suffixes that they’ve always been.

The cognitive linguistics of libfixes

From a cognitive perspective, it makes perfect sense that libfixes refer to the meaning of their source blend and not the meanings or etymologies of the parts of the source blend, because our brains don’t generally pay attention to the internal structure of words. You learn Watergate as a holistic unit, a single proper noun, not as two pieces, water + gate. When the libfix ‑gate hearkens back to the meaning of Watergate, that holistic meaning is the one that’s being referenced. Same with crypto‑: the libfix relies on the entire word cryptocurrency for its meaning, not its individual parts. And since the word cryptocurrency refers to a type of currency and not a type of cryptography, it is the currency meaning that gets highlighted every time crypto‑ forms a new word. A cryptocoin is any specific cryptocurrency; the cryptoverse is the world of cryptocurrency; a cryptobro is a bro who’s overzealous about cryptocurrency; and so on.

This also explains why the libfix ‑doodle exists as such rather than as *‑oodle. The blend that originally made the form popular was Labradoodle, whose splinters were Labrad‑ (< Labrador) and ‑oodle (< poodle). But it didn’t matter to speakers what the original splinters were, because speakers don’t pay attention to the internal structure of words! ‑doodle is based on an analogy to the entire word Labradoodle, not its parts. All that mattered was the prosodic profile of the whole word, Labradoodle. When speakers used Labradoodle as a template for other blends, it felt more natural to begin the libfix at the syllable boundary, as ‑doodle, than awkwardly in the middle of a syllable, as *‑oodle, even though *‑oodle was the original splinter. Labradoodle basically underwent rebracketing before spawning the libfix ‑doodle.

Even if our brains were capable of breaking down every word into its pieces every time we heard it (an implausibly cognitively demanding task), we simply don’t have the option. Unless you’re a lover of etymology (and even if you are), you probably aren’t aware of the historical meaning of the vast majority of words and their parts.

As a result, blends often give rise to libfixes that look exactly the same as a word or affix that was part of the original blend. But due to the libfixation process, these are now two totally distinct affixes! The Greek prefix crypto‑ is not the same entity as the libfix crypto‑; the English noun gate is not the same entity as the libfix ‑gate. They are splinters of blends that just so happen to have the exact same boundaries as the original parts of the blend.

This is what I meant earlier when I said that libfixes always involve “re”bracketing: two words are blended together in a way that creates a new word-internal boundary. Sometimes those new internal boundaries are different than they were before (helico‑pter → gyro‑copter, or al‑kuḥūl → work‑a‑holic), but sometimes they coincidentally happen to stay in the same spot (e.g. Water‑gate → nipple‑gate or crypto‑graphy → crypto‑currency). Nonetheless, a reanalysis always takes place.

What all this teaches us about language is that we as language-users care more about patterns than structure per se. Your brain doesn’t care where the internal boundaries of a word are—all it cares about is that ‑doodle has the same prosodic profile as poodle; hence bernadoodle instead of *berna‑oodle or *bernoodle. Your brain doesn’t care that ‑gate has nothing to do with actual gates, or that ‑aholic chops the word alcohol in two. Once it spots a pattern, it runs with it—even if that pattern is etymological nonsense.

Are libfixes just slang?

You probably noticed that the majority of libfixes convey a sense of either jocularity or pejoration (Gorman 2013)—basically a kind of wordplay or slang. Some scholars relegate libfixes to the margins of grammar or discount them entirely because of this (Vasileanu & Niculescu-Gorpin 2025: 241). These scholars view libfixes as mere acts of sporadic creativity rather than stable, rule-governed, productive affixes.

In reality, creativity and productivity sit on either end of a continuum. On the creative side of the continuum are the many, many, many nonce words (one-off creations or occasionalisms) coined for a specific context that never garner uptake outside that context. They are essentially wordplay. For example, it’s quite common for specific libfixes to be popular but still have none of their derivatives gain traction. ‑core and ‑ception are good examples of this: it’s easy to find one-off examples like momcore or dogception all over the internet because they were funny and made sense in the specific context they were used for, but outside that context they have no uptake. The libfix ‑ception comes from the name of the movie Inception, in which characters create alternate realities within a person’s mind, each reality embedded within the last. The ‑ception libfix now refers to any type of recursion like this. Thus the word dogception is incredibly useful for the TikTok video of a dog barking at a video of a dog barking at a video of a dog, but not a particularly helpful word otherwise. The vast majority of neologisms are like this: mere twinkles in the linguistic starscape—a brief flash and then they’re gone.

But occasionally conditions are propitious for broader adoption. If the new construction catches the right cultural wave, or maybe that playful one-off use becomes a viral TikTok, it can become ensconced within a particular linguistic community. One such use of the otherwise ephemeral ‑core libfix that seems to have reached this point is cottagecore, an internet aesthetic and subculture concerned with an idealized rural lifestyle. The word now has a stable meaning outside its individual contexts of use. Similarly, with sufficient repeated use, individual libfixes may even become extremely productive for the entire language community, like ‑holic, ‑gate, and ‑licious have.

The single best example of this is probably ‑exit, which began life in the blend Brexit (British + exit), referring to the withdrawal of the United Kingdom (UK) from the European Union (EU). It first appeared in a blog post from 2012, where it seemed to be little more than witty word play. But the topic was so widely discussed, and term so widely adopted, that it swiftly spawned numerous derivatives, such as terms referring to other countries leaving the EU:

• Frexit (France)

• Grexit (Greece)

• Itexit (Italy)

• Spexit (Spain)

And the possibility of Scotland leaving the UK in response to Brexit:

• Scexit (Scottish exit)

And then the possibility of US states seceding:

• Calexit (California)

• Texit (Texas)

And then use in popular media:

• Bakexit (see also its use on X here)

Finally, the blend Brexit itself started to be used in other blends! Blends of blends!

• regrexit ‘regret over Brexit’

• Lexit ‘left-wing supporter of Brexit’

And not just the ending, but the beginning of Brexit was coopted for blends as well:

• Braccident

• Breferendum

• Bregret

• Bremain

• Bremorse

• Brextension

• Brexistential crisis

• Brexpert

• in-Bretween

• Brentry

We know these are blends of blends because, unlike the original libfix ‑exit which meant ‘end association with X’, these mean ‘X in relation to Brexit’. This is the meaning you’d expect from blending Brexit: regrexit means ‘regret about Brexit’ because it’s a blend of regret + Brexit rather than regret + ‑exit; Brexpert means ‘expert on Brexit’ because it’s a blend of Brexit + expert rather than British + expert.

This reblending is really surprising (and cool) because blended blends are naturally quite confusing, especially in this case since the boundaries between the splinters are exactly the same in the first blend and the reblend. In any case, its reblendability is a beautiful illustration of just how deeply ensconced Brexit has become in the British lexicon—a “real” word if ever there was one. Fittingly, the Collins English Dictionary selected Brexit as the word of the year for 2016. Brexit thus makes for a great illustration of how a new expression can go from a playful one-time creation to a stable term with a predictable meaning in relatively short order, sliding all the way from one end of the creativity ↔ productivity scale to the other.

This process of graduating from creative → productive has as its keystone one crucial fact: speakers would not take up a new expression if it didn’t utilize existing words and rules of the language in the first place. Neologisms like libfixes don’t violate or break the grammar of a language—they expand on it. Libfixes work precisely because existing popular blends provide them with lots of affordances: a speaker hears chocoholic for the first time and can (unconsciously) rely on workaholic and shopaholic and the general principles of how libfixes work to figure out that it means ‘person addicted to chocolate’. Far from being “mere slang”, libfixes are in fact illustrative of one of the most foundational processes by which language—and the human mind in general—operates: analogy. And in my opinion, they demonstrate that we humans are pretty damn good at it.

I hope this article has given you lots of good affixana to share with your acquaintances. I encourage you to share your favorite libfix from this article and tag me on social media! Thanks for joining me in this deep-dive into the linguaverse!

Share

🙏 Credits

This issue of the Linguistic Discovery newsletter was edited by Amy Treber. If you’re looking for professional copyediting services, email Amy at amytreberedits@gmail.com.

The final responsibility for any mistakes or omissions is of course still wholly my own.

📑 References

• Bauer, Laurie. 2004. A glossary of morphology. Georgetown University Press.

• Gorman, Kyle. 2013. Defining libfixes. Wellformedness. https://www.wellformedness.com/blog/defining-libfixes/. (20 January, 2026).

• Vasileanu, Monica & Anabella-Gloria Niculescu-Gorpin. 2025. Romanian libfixes in the making. In Sabine Arndt-Lappe & Natalia Filatkina (eds.), Dynamics at the lexicon-syntax interface (Formulaic Language 6), 241–266. De Gruyter. https://doi.org/10.1515/9783111321905-009.

The Amazon and Bookshop.org links on this page are affiliate links, which means that I earn a small commission from Amazon for purchases made through them (at no extra cost to you).

If you’d like to support Linguistic Discovery, purchasing through these links is a great way to do so! I greatly appreciate your support!

Check out my Amazon storefront here.

Check out my Bookshop storefront here.

There is no empirical evidence suggesting that baby talk in any way hinders child language development—quite the opposite, in fact.

What exactly is baby talk and how is it different from the way adults regularly speak? More importantly, why do adults use it with children?

In this second issue of my special series on the science of baby talk, we’ll look at all the surprisingly sophisticated strategies and tactics that parents use to make their speech more accessible to their children.

ℹ️ Articles in this Series

• Part 1: Why you should be talking to your infant

• Part 2: Why do people use baby talk? [this issue]

• Part 3: Is baby talk good for your child?

• Part 4: Do all cultures use baby talk?

• Part 5: Baby talk in the languages of the world

• Part 6: How much should you talk to your child? [forthcoming]

• Part 7: What really matters when talking to your child

The technical term for baby talk is infant-directed speech (IDS) or child-directed speech (CDS), depending on the focus (younger children or children of any age). Baby talk is a way of speaking used specifically during the process of socializing young children—teaching them how to be competent members of society, which includes understanding how to use the language of the community. This makes baby talk a special register—a way of speaking associated with specific social situations or specialized activities of particular groups. Baby talk used to be called motherese, but scholars now eschew this term to avoid gender stereotyping. More recently, as researchers have realized the important role that caregivers other than parents often play in child language development, the term caregiver speech has emerged as a more broadly-encompassing expression.

In this and the other articles in this series I’ll simply use baby talk unless I want to be more specific.

Some people also distinguish between baby talk and parentese, the idea being that baby talk uses cutesy, shortened, or nonsense words (nana for banana, num-num for food) whereas parentese doesn’t. Those who make this distinction typically claim that baby talk is bad for your child but parentese is good for them. They assert that using nonsense words and incorrect grammar with children hinders their language development. This idea is, to put it frankly, total bullshit. There is no empirical evidence suggesting that baby talk in any way hinders child language development—quite the opposite, in fact, and we’ll look at the cool ways that baby talk actually abets your child’s language learning in the following issues in this series. The idea that baby talk is harmful for your child’s language development is simply a holdover from the stigma against any type of infant-directed speech in general (and probably also due to the fact that many adults find it annoying to listen to).

In reality, using simplified pronunciations and grammatical structures may actually do a better job of meeting infants where they’re at in their language-learning journey: children progress from extremely simple syllable structures and grammatical structures to more complex ones over time (e.g. only consonant-vowel (CV) syllables [Chee & Henke 2023: 752], and only 2-word combinations with no function words). (You can listen to a timelapse of this happening with a single child in this fantastic TED Talk.) Consider, for example, some ways that infants simplify words early in their language acquisition process (Ibbotson 2022: 145–146):

• reduplication: the complete or partial repetition of a syllable, replacing other syllables

  • bottle → bobo

  • dummy → dudu

• consonant harmony: adjusting the consonants in a word so that they are more similar to (or exactly match) other consonants in the word

  • cat → tat

  • dog → dod

• context-sensitive voicing: voicing all the consonants in certain positions, such as at the start of words

  • pat → bat

  • pig → big

• final devoicing: devoicing all consonants at the ends of words

  • pad → pat

  • bag → bak

• fronting: articulating sounds further forward in the mouth

  • car → tar

  • goat → doat

  • pig → pid

• final consonant deletion: deleting the consonant at the end of a word or syllable

  (Notice how this change results in CV syllables.)

  • dog → do

  • mum → mu

• cluster reduction: reducing or simplifying consonant clusters to a single consonant

  • plane → pane

  • flag → fag

  • spoon → poon

  • splash → plash

• weak syllable deletion: dropping the unstressed syllable entirely

  • banana → nana

• stopping: pronouncing fricatives as stops (plosives)

  • van → pan

  • jump → dump

  • sun → tun

  • that → dat

• gliding: pronouncing liquid consonants as glides, so that /r/ → /w/ and /l/ → /j, w/

  • red → wed

  • yellow → yeyow

• metathesis: switching the order of sounds

  • elephant → efalent

English is hardly alone in this. Native American children learning a variety of indigenous languages have also been reported to change, replace, simplify, or delete syllables, as shown in the table below (Chee & Henke 2023: 753).

Reduplication is also quite common in child speech around the world. Here are examples from several Native American languages (Chee & Henke 2023: 753):

These words sound a lot like baby talk, don’t they? It’s no coincidence that caregivers make phonotactic simplifications when talking to infants, because baby talk is really about linguistic accommodation—meeting children where they’re at in their language development by making communication as clear and understandable as possible. Parents naturally and intuitively adjust their speech to their child’s linguistic capabilities and preferences as the child learns (Kitamura & Lam 2009). They’ll often use the same pronunciations in baby talk that children are capable of producing themselves. As children master increasingly more sophisticated sounds and sound combinations, parents gradually raise their expectations of their children’s phonetic abilities, and normalize their own pronunciations in baby talk—all without even realizing they’re doing it! As another example of this linguistic accommodation, IDS tends to be more emotional at three months, more approving at six months, and more directive (”yes, look at the doggie”) at nine months (Kitamura & Burnham 2003). All this accommodation makes it easier for the child to sift through the enormous amount of linguistic input they receive and generalize grammatical patterns from it.

Want to learn more about how children’s linguistic capabilities develop over time? Check out this issue of the newsletter about how children learn grammatical rules (and how we know):

Why do people use baby talk?

When speaking to young children, one problem that caregivers have to solve is getting children to realize that speech is being directed towards them in the first place. These attention-getting (and holding) devices are a major factor shaping baby talk (Dawson, Hernandez & Shain 2022: 351). The most salient tactic that caregivers use to capture and retain young children’s attention is exaggerated prosody—higher pitch and a broader pitch range (Soderstrom et al. 2008).

Prosody is the set of phonetic and phonological cues that speakers use to structure their speech (Hieber 2016). It includes things like intonation, rhythm and timing, voice quality, pauses—anything that helps signal where one stretch of speech (syllables, words, clauses, etc.) ends and another begins, and how those sections of speech relate to one another. For example, a recent study shows that consonants at the beginnings of words are systematically lengthened across languages (Blum et al. 2024), which helps listeners determine where one word ends and the next begins.

In baby talk, adults generally use a wider pitch range and higher overall pitch—“great swooping curves of sound over an extended pitch range” (Saxton 2017: 88; Clark 2024: 40). Their speech tends to be slower, with lengthened syllables, longer pauses, and fewer disfluencies (hesitations and restarts) (Saxton 2017: 88). Baby-talking adults also tend to place object words at the ends of sentences and pronounce them fairly loudly, thus giving them special prominence (Messer 1981). Very young children especially are strongly attracted to this type of exaggerated prosody, so it serves as an excellent attention-getting device. At the same time, you can see how exaggerating these types of cues would be useful for infants trying to figure out how their language works.

Deaf parents of deaf children have additional attention-getting devices at their disposal: signs are often made in the visual field of the child rather than in their typical position near the signer. The two-handed ASL sign RABBIT, for example, is normally signed with both hands on either side of the head, but can be articulated in the child’s visual field, such as close to a picture of a rabbit in a book, or on the head of the child (Baker et al. 2016: 56). Deaf adults make sure that they sign to their deaf child when the child is actually looking at them, but they can also gain the child’s attention by tapping the child or waving their hand in the child’s visual field.

You’re reading Linguistic Discovery, a newsletter about the science and diversity of language—a field known as linguistics. I’m Danny Hieber, a Ph.D. in linguistics who works with indigenous communities to help them document and revitalize their languages. Topics covered in this newsletter include:

• ⚙️ how language works (cognitive linguistics, language change)

• 🌍 grammatical diversity in the world’s languages (typology)

• ℹ️ explainers of terms and concepts in linguistics

• 🗣️ language profiles

• 🗞️ the latest news and research in language and linguistics

• ⭐ linguistic reviews of books and other media

Paid subscribers also get bonus articles and early access to my in-progress book projects!

Subscribe now

Another strategy for retaining a child’s attention is by attending to things that the child is already paying attention to. Adults generally follow up on topics introduced by children (whether verbally or whether with their gaze, pointing, or other vocalizations), so much so that children set the agenda much more often than not. One study found that a particular child introduced about 20 new topics per hour, whereas her mother only introduced 5. And the mother followed up on the topics her daughter introduced (Moerk 1983). This responsiveness has been directly linked to later vocabulary size, but only when the response is linked to the child’s vocalization or babble (Clark 2024: 49).

Another factor that plays a strong influence on baby talk is topic choice—adults choose topics that maximize the likelihood that the child will understand what is being said. This results in a smaller range of words (Ibbotson 2022: 57), and words that focus on the here-and-now, rather than concepts removed in time or space. CDS thus includes a higher frequency of concrete terms like cup, juice, and tree than abstract concepts like kindness, beauty, and justice (Saxton 2017: 89). Five topics in particular dominate infant-directed speech (Ferguson 1977; Ferguson & Debose 1977; Dawson, Hernandez & Shain 2022: 351):

1. members of the family: mommy, daddy

2. animals, games, and toys: peek-a-boo, choo-choo ‘train’

3. parts of the body / bodily functions and routines: wee-wee ‘urinate’, night-night ‘go to sleep’

4. food

5. clothing

This trend holds across spoken and signed languages. Work with the Comanche, East Cree, Nuuchahnulth, and Twana languages, for instance, has found that common categories for specialized CDS vocabulary include kinship terms, body parts and bodily functions, and everyday actions, animals, and objects (Chee & Henke 2023: 748), and work with sign languages shows that the most frequently-discussed semantic domains are the same ones listed above (Fenson et al. 1994).

Linguist Madeline Beekman talks more about the functions of baby talk on her Substack:

Madeleine's Substack

Baby Talk

Have you ever wondered why almost everyone speaks differently to a little child? Our voice sounds higher, our pitch becomes more variable, we speak slower and more deliberate. Changes in the way we speak when interacting with a baby or toddler is so widespread we even have a special name for it: infant-directed speech. But most of us probably know it as…

Read more

2 months ago · 5 likes · Madeleine Beekman

What are the features of baby talk?

Let’s look at some other ways caregivers adjust their speech to infants. Not all caregivers engage in all of these strategies, and the details vary by individual and by culture (more on that in Part 4 of this series), but these are recurring patterns observed by child language acquisition researchers.

Generally speaking, baby talk is characterized by shorter sentences, simpler words, more repetition, and slower speech. This is true for both spoken and signed languages: adults articulate their signs more slowly with young children, making it easier for the child to see their signs (Baker et al. 2016: 56). In spoken language, individual sounds are pronounced more clearly than in adult-directed speech (ADS) (Dilley et al. 2014), whether vowels (Burnham, Kitamura & Vollmer-Conna 2002) or consonants (Cristià 2010). Sounds are also more evenly timed in IDS than ADS (Payne et al. 2010). In normal adult speech, vowels and consonants are pronounced longer or shorter depending on the sounds around them; in IDS, by contrast, vowels and consonants are given more similar lengths.

Caregivers also use vocabulary unique to the CDS register, such as tum-tum ‘tummy’, din-din ‘dinner’, and go bye-bye ‘leave’. Usually this child-directed vocabulary is based on words in the adult lexicon, but with simplification, sound substitution, or other sound changes. Here are similar examples from the Nuuchahnulth language (Kess & Kess 1986: 209):

However, the CDS register can include words that are completely unrelated to their ADS counterparts, such as how the word boo-boo isn’t related to any word like injury or wound. Here are similar cases from Nuuchahnulth, where the CDS form is unrelated to the ADS form (Kess & Kess 1986: 209):

Words used in CDS are also:

• short, containing fewer sounds (phonemes) on average than words used with adults

• higher frequency, i.e. words that adults use the most often in everyday speech

• more predictable, in that they contain very common sound sequences, e.g. the sequence /ði/ in the word the is a much more common sequence than the reverse, /ið/, as in seethe, teethe, or wreathe.

• more similar to each other, meaning that there are a higher number of words that sound similar to any given word

  (Jones et al. 2023)

This last feature results in a large number of pairs of words that differ only by a single sound, e.g. pat vs. bat, called minimal pairs. Such minimal pairs are extremely useful for children learning to differentiate sounds in a language.

In terms of syntax, words are also more likely to be delivered in isolation in IDS (Ibbotson 2022: 57), and there are fewer instances of both negation and complex clauses (Sachs, Brown & Salerno 1976). Parents often leave out function words and grammatical endings (Dawson, Hernandez & Shain 2022: 353), and use entire noun phrases instead of pronouns. Because pronouns stand in for a noun phrase and/or refer back to them, understanding them requires the ability to track topics over the course of a conversation or text, making them difficult for children to master.

Subjects of sentences in IDS have a strong tendency to be agents (the person doing the action) (Rondal & Cession 1990), even though this is just one of the very many functions that subjects have in English. For example, the subject of a passive sentence is a patient (person affected by the action) rather than an agent, and the subject of a meteorological verb like to rain is a meaningless dummy subject it (an expletive or pleonastic subject), e.g. it is raining. In fact, the notion of a grammatical subject is so complex that children don’t demonstrate full control over the category until age 9! (Braine et al. 1993) Limiting the use of subjects to just agents when children are young helps ease infants into grasping this multifaceted construction.

Deaf caregivers have another means of simplifying the syntax of baby talk. In sign languages, facial expressions, mouth movements, or body movements can also be part of the grammar or lexicon of the language. For instance, it is common for sign languages to use facial expressions like raised eyebrows to mark questions (much like many spoken languages use rising intonation to mark questions). Such gestures can also be part of the form of individual signs. For example, the sign for SEARCH in Sign Language of the Netherlands is produced with the eyebrows lowered, and the sign for BE-PRESENT in British Sign Language and Sign Language of the Netherlands is produced with a mouth movement sh, even though that sound is not part of either the corresponding Dutch or English word (Baker et al. 2016: 4). Signers would be surprised or confused about these signs if they did not include these facial gestures.

This group of linguistic gestures and facial expressions made in addition to hand movements are called non-manual features/components/gestures. They can be difficult for a child to learn because the face is also used to display emotion, so the child has to learn to distinguish when a facial expression or gesture is being used linguistically versus emotively / gesturally. So it should come as no surprise that one of the ways deaf mothers accommodate their child’s learning process is by avoiding the use of non-manual markers in a grammatical way until children are between 2;0 and 2;6 (Baker et al. 2016: 60).

ℹ️ In first language acquisition research, age is written as years;months, so that “3;9” means ‘3 years, 9 months old’.

Other ways that adults assist children in language learning are even more subtle. In his brilliant TED Talk summarizing his equally brilliant research, Deb Roy shows how anytime his son started using a new word, it coincided with him and his wife subconsciously shortening any utterances containing that word when they spoke to their son (Roy et al. 2015). This unintentionally gave their son focused, targeted practice with that word.

Finally, CDS utterances are also remarkably well-formed. Normal adult conversation is riddled with errors and restarts. We tend not to notice these errors because they’re a natural part of conversation, but if you look at the raw transcripts of conversations, you quickly realize how remarkable it is we’re ever able to make sense of anything we say to each other. But IDS is almost completely devoid of both disfluencies (false starts, mispronunciations, hesitations) and grammatical errors. The table below shows just how few disfluencies occurred, on average, per 100 words in CDS vs. ADS (Broen 1972: 11). Another survey found just one grammatical error in a corpus of 1,500 utterances (Newport, Gleitman & Gleitman 1977).

Taken together, it is clear that the modifications adults make in IDS are done for the benefit of the child’s language development. Caregivers continuously try to determine their infant’s state of knowledge—a skill called theory of mind—and knowingly or unknowingly adjust their speech to accommodate that state of knowledge.

🌠 The fact that baby talk is all about theory of mind probably also explains why other primates don’t do it. A recent study showed that the other great ape species directed almost none of their vocalizations towards their infants (Wegdell et al. 2025). This is likely because non-human primates (with some borderline exceptions like Kanzi, as I discuss in this newsletter issue), generally don’t exhibit much in the way of theory of mind. For humans, on the other hand, theory of mind is absolutely central to all our social interactions.

Yet even though the intent (conscious or subconscious) of baby talk is to facilitate child language development, this doesn’t mean it’s necessary—or even useful. All developmentally normal children follow roughly the same stages of language development regardless of what percentage of the speech they hear is baby talk (although these stages vary by language, and the exact timing varies by individual). So what’s the point of baby talk? I’ve already hinted at its potential benefits in this series, but in the next issue of this series we’ll look more extensively at the evidence that IDS actually improves child language acquisition. Be sure to subscribe to receive the rest of the issues in the series!

Subscribe now

ℹ️ Articles in this Series

• Part 1: Why you should be talking to your infant

• Part 2: Why do people use baby talk? [this issue]

• Part 3: Is baby talk good for your child?

• Part 4: Do all cultures use baby talk?

• Part 5: Baby talk in the languages of the world

• Part 6: How much should you talk to your child?

• Part 7: What really matters when talking to your child

📚 Recommended Reading

“The birth of a word”

MIT researcher Deb Roy wanted to understand how his infant son learned language—so he wired up his house with video cameras to catch every moment (with exceptions) of his son’s life, then parsed 90,000 hours of home video to watch gaaaa slowly turn into water. Astonishing, data-rich research with deep implications for how we learn.

HELLO Lab Presents

The Hearing Experience & Language Learning Outcomes (HELLO) Lab at the University of Connecticut has a great series of YouTube Videos about child language acquisition for parents.

• HELLO Lab on YouTube

How babies talk: The magic and mystery of language in the first three years of life

📑 References

• Baker, Anne, Beppie Van Den Bogaerde, Roland Pfau & Trude Schermer (eds.). 2016. The linguistics of sign languages: An introduction. John Benjamins. https://doi.org/10.1075/z.199.

• Blum, Frederic, Ludger Paschen, Robert Forkel, Susanne Fuchs & Frank Seifart. 2024. Consonant lengthening marks the beginning of words across a diverse sample of languages. Nature Human Behaviour 8(11). 2127–2138. https://doi.org/10.1038/s41562-024-01988-4.

• Braine, Martin D.S., Patricia J. Brooks, Nelson Cowan, Mark C. Samuels & Catherine Tamis-LeMonda. 1993. The development of categories at the semantics/syntax interface. Cognitive Development 8(4). 465–494. https://doi.org/10.1016/S0885-2014(05)80005-X.

• Broen, Patricia Ann. 1972. The verbal environment of the language-learning child (ASHA Monographs 17).

• Burnham, Denis, Christine Kitamura & Uté Vollmer-Conna. 2002. What’s new, pussycat? On talking to babies and animals. Science 296(5572). 1435–1435. https://doi.org/10.1126/science.1069587.

• Chee, Melvatha R. & Ryan E. Henke. 2023. Child and child-directed speech in North American languages. In Carmen Dagostino, Marianne Mithun & Keren Rice (eds.), The Languages and Linguistics of Indigenous North America (World of Linguistics 13.2), vol. 2, 741–766. De Gruyter. https://doi.org/10.1515/9783110712742-033.

• Clark, Eve V. 2024. First language acquisition. 4th edn. Cambridge University Press. https://doi.org/10.1017/9781009294485.

• Cristià, Alejandrina. 2010. Phonetic enhancement of sibilants in infant-directed speech. The Journal of the Acoustical Society of America 128(1). 424–434. https://doi.org/10.1121/1.3436529.

• Dawson, Hope, Antonio Hernandez & Cory Shain (eds.). 2022. Language files: Materials for an introduction to language and linguistics. 13th edn. The Ohio State University Press.

• Dilley, Laura C., Amanda L. Millett, J. Devin Mcauley & Tonya R. Bergeson. 2014. Phonetic variation in consonants in infant-directed and adult-directed speech: the case of regressive place assimilation in word-final alveolar stops. Journal of Child Language 41(1). 155–175. https://doi.org/10.1017/S0305000912000670.

• Fenson, Larry, Philip S. Dale, J. Steven Reznick, Elizabeth Bates, Donna J. Thal & Stephen J. Pethick. 1994. Variability in early communicative development (Monographs of the Society for Research in Child Development 242). Vol. 59. https://doi.org/10.2307/1166093.

• Ferguson, Charles A. 1977. Baby talk as a simplified register. In Catherine E. Snow & Charles A. Ferguson (eds.), Talking to children: language input and acquisition. Cambridge University Press.

• Ferguson, Charles A. & C. E. Debose. 1977. Simplified registers, broken language, and pidginization. In A. Valdman (ed.), Pidgin and creole linguistics. Indiana University Press.

• Fernald, Anne, Traute Taeschner, Judy Dunn, Mechthild Papousek, Bénédicte De Boysson-Bardies & Ikuko Fukui. 1989. A cross-language study of prosodic modifications in mothers’ and fathers’ speech to preverbal infants. Journal of Child Language 16(3). 477–501. https://doi.org/10.1017/S0305000900010679.

• Hieber, Daniel W. 2016. The cohesive function of prosody in Ékegusií (Kisii) narratives: A functional-typological approach. University of California, Santa Barbara M.A. thesis. http://rgdoi.net/10.13140/RG.2.2.17818.18886. (22 August, 2025).

• Ibbotson, Paul. 2022. Language acquisition: The basics (The Basics Series). Routledge. https://doi.org/10.4324/9781003156536.

• Jones, Gary, Francesco Cabiddu, Doug J. K. Barrett, Antonio Castro & Bethany Lee. 2023. How the characteristics of words in child-directed speech differ from adult-directed speech to influence children’s productive vocabularies. First Language 43(3). 253–282. https://doi.org/10.1177/01427237221150070.

• Kess, Joseph Francis & Anita Copeland Kess. 1986. On Nootka baby talk. International Journal of American Linguistics 52(3). 201–211. https://doi.org/10.1086/466018.

• Kitamura, Christine & Denis Burnham. 2003. Pitch and communicative intent in mother’s speech: adjustments for age and sex in the first year. Infancy 4(1). 85–110. https://doi.org/10.1207/S15327078IN0401_5.

• Kitamura, Christine & Christa Lam. 2009. Age‐specific preferences for infant‐directed affective intent. Infancy 14(1). 77–100. https://doi.org/10.1080/15250000802569777.

• Moerk, Ernst L. 1983. The mother of Eve—as a first language teacher (Monographs on Infancy). Ablex.

• Newport, Elissa L., Henry Gleitman & Lila R. Gleitman. 1977. Mother, I’d rather do it myself: Some effects and non-effects of maternal speech style. In Catherine E. Snow & Charles A. Ferguson (eds.), Talking to children: language input and acquisition, 101–149. Cambridge University Press.

• Payne, Elinor, Brechtje Post, Lluïsa Astruc, Pilar Prieto & Maria del Mar Vanrell. 2010. Rhythmic modification in child directed speech. In Michela Russo (ed.), Prosodic universals: Comparative studies in rhythmic modeling and rhythm typology, 147–184. Aracne Editrice.

• Rondal, Jean A. & Anne Cession. 1990. Input evidence regarding the semantic bootstrapping hypothesis. Journal of Child Language 17(3). 711–717. https://doi.org/10.1017/S0305000900010965.

• Roy, Brandon C., Michael C. Frank, Philip DeCamp, Matthew Miller & Deb Roy. 2015. Predicting the birth of a spoken word. Proceedings of the National Academy of Sciences 112(41). 12663–12668. https://doi.org/10.1073/pnas.1419773112.

• Sachs, J., R. Brown & R. Salerno. 1976. Adult’s speech to children. In W. von Raffler Engel & Y. Lebrun (eds.), Baby talk and infant speech (Neurolinguistics 5). Peter de Ridder Press.

• Saxton, Matthew. 2017. Child language: Acquisition and development. 2nd edn. SAGE.

• Soderstrom, Melanie. 2007. Beyond babytalk: Re-evaluating the nature and content of speech input to preverbal infants. Developmental Review 27(4). 501–532. https://doi.org/10.1016/j.dr.2007.06.002.

• Wegdell, Franziska, Caroline Fryns, Johanna Schick, Lara Nellissen, Marion Laporte, Martin Surbeck, Maria A. Van Noordwijk, et al. 2025. The evolution of infant-directed communication: Comparing vocal input across all great apes. Science Advances 11(26). eadt7718. https://doi.org/10.1126/sciadv.adt7718.

The Amazon and Bookshop.org links on this page are affiliate links, which means that I earn a small commission from Amazon for purchases made through them (at no extra cost to you).

If you’d like to support Linguistic Discovery, purchasing through these links is a great way to do so! I greatly appreciate your support!

Check out my Amazon storefront here.

Check out my Bookshop storefront here.

Prefer a video version of this post? Watch here:

However, February is one of those words which underwent etymological respelling, where scholars during the Renaissance changed the spelling of the word to match its original Classical Latin or Greek spelling, even though the word was no longer pronounced that way. If you’ve been following the newsletter for a while, you might have read about another case of etymological respelling as well—herb!

Here are some other examples:

• aisle

• debt

• doubt

• indict

• island

• plumber

• receipt

• salmon

• subtle

The silent consonants in those examples were pronounced in Latin but lost by the time those words got to English. Then they were later respelled to more closely resemble the original Latin.

What did the English call February before they borrowed the word from Old French? The word that roughly corresponded to that same time of year was solmonað, but scholars are unsure of its meaning. Possible meanings are ‘Mud Month’ (from sol ‘mud, wet sand’), or (according to Bede) ‘Cake Month’ because this was purportedly the time when Anglo-Saxons would offer cakes to their gods. (Given England’s weather at that time of year, I find the ‘Mud Month’ to be the most plausible.)

Here’s the etymological flowchart for February:

📚 Recommended Reading

📑 Sources

• Etymonline: February

• Wiktionary: February

The Amazon and Bookshop.org links on this page are affiliate links, which means that I earn a small commission from Amazon for purchases made through them (at no extra cost to you).

If you’d like to support Linguistic Discovery, purchasing through these links is a great way to do so! I greatly appreciate your support!

Check out my Amazon storefront here.

Check out my Bookshop storefront here.

ℹ️ Articles in this Series

• Part 1: Libfixes: When word parts go rogue [this issue]

• Part 2: The science of libfixes: How libfixes work, where they come from, and what they teach us about language and the mind

You’re reading Linguistic Discovery, a newsletter about the science and diversity of language—a field known as linguistics. I’m Danny Hieber, a Ph.D. in linguistics who works with indigenous communities to help them document and revitalize their languages. Topics covered in this newsletter include:

• ⚙️ how language works (cognitive linguistics, language change)

• 🌍 grammatical diversity in the world’s languages (typology)

• ℹ️ explainers of terms and concepts in linguistics

• 🗣️ language profiles

• 🗞️ the latest news and research in language and linguistics

• ⭐ linguistic reviews of books and other media

Paid subscribers also get bonus articles and early access to my in-progress book projects!

Subscribe now

⚠️ Content Warning

This article contains a number of sexually explicit and derogatory terms (but not pictures or videos). These terms are included here solely for educational purposes and are merely discussed, not used with their normal pragmatic intent. (See the Use-Mention Distinction for more on this difference.) Nonetheless, you may want to screen this article before sharing it with kids, and/or exercise judgement if viewing this at work.

No Longer Bound: The Libfix Revolution

The ‑holic in chocoholic and shopaholic and workaholic didn’t start out free. Its path to liberation was long and circuitous: It began its life in the depths of history as the Akkadian word guḫlum, referring to a lustrous gray mineral we today call stibnite. Traditionally, stibnite was reduced to a cosmetic powder called kohl that was then used to darken the skin around the eyes. Thus the word guḫlum could refer to either the mineral or the cosmetic powder. Aramaic later borrowed this word as kuḥlā, and Arabic then borrowed it from Aramaic as the word كُحْل kuḥl sometime during the Old Arabic period (c. 1st millennium BCE) (Wiktionary: كحل).

The Arab conquest of the Iberian Peninsula by the Umayyad Caliphate brought the word with it in the form of Andalusian Arabic كُحُول kuḥūl (Wiktionary: alcohol). This period was the beginning of the Islamic Golden Age as well as the Islamic–European knowledge exchange, when the Islamic world contributed significantly to the intellectual advancement of Europe. These contributions came in many forms, including Latin translations of previously lost Ancient Greek texts (like Aristotle), and numerous Arabic texts in astronomy, mathematics, science, and medicine. Many technical terms in English originate from words borrowed from Arabic during this period: algebra, algorithm, chemistry, cipher, zero, zenith, and many others.

Among those words was alcohol, the Medieval Latin rendering of the Arabic phrase اَلْكُحُول al-kuḥūl ‘the kohl’. al‑ is the definite article ‘the’ in Arabic, but many European scholars mistook it as part of the following word, which is why so many Arabic borrowings in European languages begin with a‑ or al‑ (e.g. Spanish azúcar ‘sugar’ and algodón ‘cotton’). In Medieval Latin alcohol initially referred only to kohl, the cosmetic powder, but soon it took on a more generic meaning, referring to any powder obtained from titrating a material. This was the meaning it had when it was subsequently borrowed into Spanish and French.

English got ahold of the word in the 1540s from either Spanish, French, or Medieval Latin, and it quickly expanded in meaning to not just powders but distillates of liquids as well. By the 1670s, alcohol could refer to any sublimated substance, or metaphorically to the “essence” or “spirit” of anything. The first use of the modern sense of the word, referring to intoxicating beverages, appears in 1753 in the phrase “alcohol of wine”. (Barnhart 1988: 22–23; Etymonline: alcohol)

‑holic entered its captivity in 1790 when the word alcoholic first appears in the record. At the time, the word simply meant ‘of or pertaining to alcohol’. The meaning ‘caused by drunkenness’ is attested in 1872, and finally the modern sense of ‘habitually drunk’ in 1910. It was at this point that a fortuitous coincidence occurred: just a few years prior, the International Conference on Chemical Nomenclature resolved to name alcohols using a shortened form of the word alcohol as a suffix, ‑ol, coining terms like ethanol (ethyl alcohol), methanol (methyl alcohol), and phenol (benzine alcohol) (Wikipedia: Ethanol). (Those coinages also include an infix ‑an‑ meaning ‘a single bond’.) Suddenly, ‑ol was a productive suffix that could be used to form new words at will, and in this suffix ‑holic received its first glimpse of freedom. If ‑ol could run wild, attaching itself to any word it pleased, why couldn’t the adjective form ‑(h)olic?

The fateful day came in 1965, when an unwitting English speaker, perhaps influenced by words like ethanol into thinking that alcoholic too contained the suffix ‑ol, rent the word alcohol asunder, cast down the beginning alco‑, and in a flash of creative insight coined a strange new frankenword: sugarholic.

After centuries of captivity, ‑holic seized upon its chance and was at last free. Within the year it had formed an alliance with food to create foodoholic, and in 1968 it achieved its masterstroke: workaholic. This confederation proved to be unstoppable, rocketing ‑holic to such heights that it would never have to worry about lexical captivity again:

With the model of workaholic deeply entrenched and popularized, other words soon fell into line: golf and chocolate succumbed quickly as golfaholic and chocoholic in 1971, then shopping as shopaholic in 1984 (Etymonline: -aholic). Today there are entire Wikipedia pages filled with ‑holic offspring, a testament to its ascendancy.

What are libfixes?

The etymological history in the previous section illustrates several common types of language change:

• borrowing words between languages, which often involves adapting the pronunciation of the word to the new language (termed phonological adaptation), e.g. Andalusian Arabic al-kuḥūl → Medieval Latin alcohol, where Latin replaced the foreign ⟨ḥ⟩ (voiceless pharyngeal fricative /ħ/) sound with native /h/.

• rebracketing, where the boundaries of words or affixes are reanalyzed. In the case of al-kuḥūl becoming alcohol, the boundary was lost entirely, a process called juncture loss.

• blending, combining parts of two or more words, such as ethyl + alcohol → ethanol.

• libfixation, the focus of today’s article, when parts of words escape their parent words and become new affixes.

ℹ️ How to read linguistics

If any of the abbreviations or notations in this article are unfamiliar to you, check out the “How to read linguistics” page. It lists all the abbreviations, conventions, and notations used in Linguistic Discovery content.

The term libfix was first coined by linguist Arnold Zwicky on his blog in 2010 and defined as follows:

The “liberation” of parts of words […] to yield word-forming elements that are semantically like the elements of compounds but are affix-like in that they are typically bound. (Arnold Zwicky’s blog: Libfixes)

The word itself is a blend of liberated affix, and follows the pattern of other affix names in linguistics: prefix, suffix, circumfix, infix, and others. Other terms used to describe libfixes include fragment, secreted affix (gross), combining form, and splinter (see Delhem & Marty 2024: 5 for sources).

The best-known libfix is ‑gate, originally part of Watergate, the name of a hotel where an infamous political scandal involving former US President Richard Nixon occurred in the early 1970s. The episode quickly became known as “the Watergate scandal”, and soon commentators were using that phrase as a template for creating other neologisms like Volgagate (in a 1973 article in National Lampoon; OED: ‑gate). New York Times columnist William Safire popularized the construction soon after, writing about Vietgate in September 1974, and then applying the pattern with abandon: Billygate, Briefingate, Contragate, Debategate, Lancegate, Nannygate, Raidergate, and many others. His intention may have been to “minimize the relative importance of the crimes committed by his former boss with this silliness” (source), but the inadvertent effect of his sarcastic use of ‑gate was to ensconce the suffix in public consciousness. Today ‑gate is used for all manner of scandals and controversies, with a giant Wikipedia list dedicated to its appearances in English. Some examples from recent memory include gamergate, nipplegate, emailgate, and poopgate (which will forever be part of local Chicago lore—but don’t let it deter you from doing the fantastic architectural boat tour!).

Some other well-known libfixes include:

• ‑cation ‘a specific type of vacation’ ← vacation

  • girlcation ‘vacation with girl friends’

  • kidcation ‘vacation without kids’

  • staycation ‘vacation at home’

• ‑copter ‘having a spinner rotor for flight’ ← helicopter

  • gyrocopter

  • quadcopter

• ‑dar ‘ability to perceive X’ ← radar

  • gaydar

  • humordar

  • jewdar

  • sarcasmdar

• ‑doodle ‘cross with a poodle’ ← labradoodle

  • bernadoodle

  • goldendoodle

• ‑erati ‘the elite within a group’ ← illuminati

  • literati ‘the educated/literate elite’

  • digerati ‘internet elite’

  • glitterati ‘famous, wealthy, and attractive people’

  • twitterati ‘Twitter elite’

  • wokerati ‘extremely woke people’

• ‑flation ‘growth in X’ ← inflation

  • tipflation ‘growth in the size of tips’

  • shrinkflation ‘increase in the reduction of the size of goods’

  • stagflation ‘increase in stagnation’ / ‘stagnation + inflation’

• ‑gasm ‘explosive or pleasurable experience’ ← orgasm

  • braingasm ‘intense intellectual stimulation’

  • eargasm ‘climax of musical excitement’

  • foodgasm ‘intense enjoyment of a food’

  • nerdgasm ‘intense excitement at nerdy things/topics’

• -iversary ‘yearly recognition of X’ OR ‘recognizing an event every X intervals’ ← anniversary

  • blogiversary ‘anniversary of a blog’

  • adoptiversary ‘anniversary of an adoption’

  • weekiversary ‘weekly recognition of an event’

  • monthiversary ‘monthly recognition of an event’

• ‑licious ‘appealing in reference to X’ ← delicious

  • babelicious

  • bubblelicious ‘appealing bubbles’ (gum, boba tea)

  • bootylicious

  • Fergalicious ‘the sexual appeal of music artist Fergie’

• ‑(ma)geddon ‘disastrous or cataclysmic event’ ← armaggedon

  • carmaggedon ‘incredible amount of traffic; large pileup’

  • snowmaggedon

  • toemaggedon

• ‑nomics ‘pseudo-scientific; economic policy of X’ ← economics

  • Freakonomics

  • Reaganomics

  • hope-a-nomics

• ‑preneur ‘type of entrepreneur’ ← entrepreneur

  • kidpreneur

  • mompreneur

  • solopreneur

• ‑pocalypse ‘a catastrophic event relating to or caused by X’ ← apocalypse

  • beepocalypse

  • snowpocalypse

• ‑scape ‘environment of X’ ← landscape

  • mindscape

  • seascape

  • soundscape

• ‑splain ‘explain condescendingly or to one who is more knowledgeable in an area’ ← explain

  • ablesplain ‘to explain a disability-related topic without knowledge of it’

  • goysplain ‘to explain a Judaism-related topic to a Jewish person’

  • mansplain ‘to explain condescendingly to a female listener’

  • whitesplain ‘to explain a race-related topic to a non-white person’

• ‑tainment ‘type of entertainment’ ← entertainment

  • edutainment

  • familytainment

  • gastrotainment / eatertainment

  • infotainment

  • scientainment

• ‑tarian ‘one who eats X’ ← vegetarian

  • flexitarian

  • pescetarian

  • pizzatarian

• ‑thon ‘event characterized by intense effort and/or long duration ← marathon

  • boreathon

  • hackathon

  • excuseathon

  • knitathon

  • readathon

  • telethon

  • walkathon

• ‑Tok ‘TikTok subcommunity’ ← TikTok

  • BookTok

  • CleanTok

  • DanceTok

  • MusicTok

  • PlantTok

  • TechTok

• ‑ussy ‘resemblance to a vulva or vagina’ ← pussy

  • bussy ‘boy pussy’ or ‘butt pussy’ (anus)

  • thrussy ‘throat pussy’

  • donutussy ‘donut hole’ (the physical hole)

  • baristaussy ‘barista pussy’ (source)

    • Swarthmore College linguistics professor Kirby Conrad explained the use of the ‑ussy libfix by saying, “So, if you’re a really dedicated coffee artist and you really put your whole baristaussy into this latte, I’m expecting some killer latte art.” Incidentally, this is also a great example of the ass camouflage construction at work, where the body part stands in metonymically for the entire person.

    • ‑ussy was actually selected as the Word of the Year by the American Dialect Society for 2022.

• ‑verse ‘the community of X’ ← universe

  • cryptoverse

  • Twitterverse

• ‑zilla ‘biggest/baddest/meanest/nastiest of its type’ ← Godzilla

  • dogzilla

  • bridezilla

  • momzilla

Technologies seem to be an especially fertile ground for libfixes (or blends in general) (O’Dell 2016):

• blog-based

  • blogosphere

  • bloggerati ‘the blogger elite’

• Twitter-based

  • twitterati

  • twittersphere

  • twitterpated ‘making excessive use of Twitter to tell the world about the minutiae of one’s life’

  • twittercide ‘deleting one’s Twitter account’

  • twittercation ‘a break from using Twitter’

  • twitterhoea ‘sending too many tweets in a short span of time’

  • twisticuffs ‘strong disagreement between tweeters’

  • twitterholic

Libfixes don’t have to be suffixes either; some are prefixes:

• alt‑ ‘outside the mainstream’ ← alternative

  • altcoin

  • alt-history

  • alt-right

  • alt-rock

• crypto‑ ‘related to cryptocurrency’ ← cryptography

  • cryptobro

  • cryptocoin

  • cryptocurrency

  • cryptoverse

• franken‑ ‘human-altered interference with nature’ ← Frankenstein

  • frankenfood

  • frankenplant

  • frankenword

• heli‑ ‘type of helicopter’ ← helicopter

  • helipad

  • heliport

  • helidrome

• hyper‑ ‘relating to hypertext’ ← hypertext

  • hyperlink

  • hypermedia

• petro‑ ‘relating to petroleum’ ← petroleum

  • petrodollar

  • petrochemical

  • petrocurrency

  • petrostate

All this is just a small sampling of libfixes. There are many many more. This sampling doesn’t include Covid- and pandemic-related terms (quarantini, quarantigue, maskne, maskulinity, maskhole, covidiot, moronovirus, Faucism, Magat, or Spanish pandejo ← pandemic + pendejo ‘asshole’ [Berry 2022]) because it would be too exhausting to catalog them all. Something about that time period—perhaps the fact that everyone was chronically online, but more likely because there were so many new concepts and experiences that needed naming—led to a shocking proliferation of new blends and libfixes.

The big list of examples above notwithstanding, libfixation seems to be a relatively novel word-formation strategy, at least in English. By any measure of productivity (how readily a construction can be applied to new words or contexts), the most common English libfixes have become more productive over time. The charts below show a few different measures of productivity for the libfixes ‑holic, ‑licious, and ‑nomics, as well as the traditional suffixes ‑dom and ‑hood for comparison.

Given how closely libfixes are associated with the tech and internet spaces, their rise to prominence was probably spurred by the rise of computing and internet terminology, in much the same way that scientific terminology gave rise to an increase in acronyms and initialisms as a word-coinage strategy in the early 20th century. (Before the 1990s, acronyms and initialisms simply weren’t used to create new words. shit does not come from ship high in transit; news does not come from north-south-east-west; and so on. A good rule of thumb is: if someone tells you that a word was originally an acronym, no it wasn’t.)

Regardless, libfixes have now become so common that other languages in contact with English are starting to adopt the strategy for new words (Vasileanu & Niculescu-Gorpin 2025). So let’s look at libfixes in other languages too.

Libfixes in languages other than English

Of course, other languages have their own native libfixes. Just because English has influenced other languages in a way that makes libfixes more common doesn’t mean that libfixes weren’t already present in those languages. Some examples:

• Italian: ‑opoli ‘‑ville’, referring to a scandal ← Tangentopoli ‘Bribesville’ (Wikipedia: Libfix)

  • Bancopoli

  • Calciopoli

• Brazilian Portuguese: ‑ão ‘scandal’ ← Mensalão ‘Monthly payment–gate’ (Wikipedia: Libfix)

  • Petrolão

  • Metrolão

  • Trensalão

• Dutch (Hamans 2021)

  • ‑talië ‘Italy’ (but used generically to mean ‘country’

    • vertalen ‘interpret’ → Vertalië ‘country of interpreters’

    • hospitaal ‘hospital’ → hospitalië ‘hospital as long-term residence’

    • kapitaal ‘capital’ → Kapitalië ‘country where capital rules’

    • betalen ‘pay’ → Betalië ‘country where one has to pay for everything’

• Romanian: ‑izdă ‘vulva’ (offensive/sarcastic) ← pizdă ‘vulva’ [slang] (Vasileanu & Niculescu-Gorpin 2025)

  • feminizdă ‘female feminist’ [offensive] ← feministă ‘female feminist’

  • Merkelizdă ‘Merkel’ [offensive] ← (Angela) Merkel

• Spanish: pandejo ‘covidiot’ ← pandemic + pendejo ‘asshole’

It’s difficult to know, however, how much the origin of the Italian and Portuguese libfixes can be attributed to influence from English ‑gate, especially given that ‑gate is by far the most popular English libfix, and also the one most likely to appear in international news outlets because of its political nature. Perhaps those two libfixes should be considered calques (loan translations) rather than truly native constructions. I leave it to other linguists to sort that one out. (Anybody need a good thesis topic?)

The Romanian libfix, on the other hand, likely isn’t a calque of English ‑ussy, since ‑ussy is very recent and data on ‑izdă comes from a corpus spanning the last 15 years (Vasileanu & Niculescu-Gorpin 2025). ‑izdă also only attaches to animate nouns, and refers derogatorily to the entire person rather than to a specific body part, unlike ‑ussy which may attach to inanimate nouns, refers specifically to the body part, and also (surprisingly) is typically used in a positive and playful manner rather than derogatorily.

Despite pizdă being a feminine noun, a few masculine versions of the libfix have been back-formed from the feminine ones too:

• ziarizd ‘male journalist’ [offensive] ← ziarizdă ‘female journalist’ [offensive] ← ziaristă ‘female journalist’

• minimalizd ‘male adept of minimalism’ [offensive]

Notably, no use of the feminine version, minimalizdă, has been documented, suggesting the emergence of an independent masculine version of the suffix.

Because scholarly awareness of libfixes is still so recent, there haven’t yet been good studies of them in other languages. There have however been a few studies which examine the influence of English libfixes on other languages. Other languages can outright borrow English libfixes and attach them to native words (with minor spelling changes and phonological adaptation), as shown in the examples below.

ℹ️ If you know of libfixes in other languages, please tell me about them in this form. I’ll be sure to share any responses I get with the community of research linguists so we can learn more about how libfixes work in the world’s languages!

• Polish Rywingate ‘political scandal caused by Lech Rywin’ teleholik ‘teleholic’ (Konieczna 2012)

• Romanian Băsescugate ‘scandal involving former Romanian president Băsescu’ SRIgate ‘scandal involving the SRI [Romanian intelligence service]’ romgate ‘scandal involving Rome’

  curcubeoholic ‘addicted to rainbows’ ← curcubeu ‘rainbow’ ihtioholic ‘addicted to fish’ (literally ‘ichthyoholic) ← ihtio‑ ‘ichthyo‑’ călătoholic ‘addicted to traveling’ ← călători ‘to travel’

  veveritzilla ‘huge squirrel’ ← veveriță ‘squirrel’ șnitzilla ‘huge schnitzel’ ← șnițel ‘schnitzel’

  varaton ‘long vacation session during the summer’ ← vară ‘summer’

  sarmaghedon ‘a meal where one eats too many meat rolls’ ← sarma ‘meat rolls’ pornaghedon ‘a sudden end to a politician’s career due to video recordings of him having sex’ ← porno ‘porn’

  (Vasileanu & Niculescu-Gorpin 2025)

Italian has also borrowed ‑gate and ‑nomics, and Bulgarian has borrowed ‑holic (Vasileanu & Niculescu-Gorpin 2025: 248).

As mentioned above, languages can also borrow entire word-formation strategies like libfixation without borrowing specific words or libfixes. One study documents the increased use of clipping (truncation/shortening of words) in Polish and noun-noun compounding in both French and Polish due to influence from English (Renner 2018). Another showed that Dutch libfixing has increased under influence from English (Hamans 2021), and a third showed that blends (defined below) have increased in Bulgarian, Italian, and Dutch (see Vasileanu & Niculescu-Gorpin 2025: 248 for sources). In Romanian, pervasive influence from English has led blending to become the default strategy for naming hybrid objects in younger generations. In one experiment, subjects were asked to name, using only a single word, some imaginary hybrid entities. Fully 68.31% of the words they created for these hybrid entities were blends, and only 12.22% were compounds, which was unexpected because compounding has been a major word-formation process in Romanian but blending was not (until recently). (Vasileanu, Niculescu-Gorpin & Radu-Bejenaru 2024)

I also see the influence of English word-formation processes all the time in my language revitalization work: Many Native American communities are working to revitalize their endangered languages, and part of this process involves coining new words for things and activities in the community. However, because these languages are endangered, many of the members within these communities speak English rather than their heritage language. And English speakers love to use noun compounding when coining words. This is unfortunate, because most Native American languages are extremely verb-oriented: traditionally, nouns were derived from complex descriptive verbs whenever possible.

As an example, the traditional Chitimacha word for ‘writing utensil’ (pen, pencil) is naakxtempuyna, which literally means ‘they habitually write with it’, from the verb naakxte‑ ‘write’ (literally ‘to paper’). Similarly, ‘bridge’ is pamtuyna ‘they habitually cross it’, from the verb pamte‑ ‘cross’. But when the Chitimacha tribe needed a word for ‘net’ (in sports), they chose to create the compound noun cii tuukun, literally ‘rope bag’ (ciq ‘rope’ + tuukun ‘bag’), rather than something like *dzehtuyna ‘they usually catch/trap with it’, from the verb dzeht‑ ‘to trap/catch …; to make a play in stickball’. Since I began working with the Chitimacha tribe in 2008, I have gently encouraged them to reach for verbs before nouns when coining new words. Nonetheless, years of English influence have resulted in a significantly higher proportion of compound nouns in the lexicon than traditional Chitimacha had. Shifts in grammatical tendencies like this are common in endangered languages (Palosaari & Campbell 2011).

In the second half of this series, we’ll explore the science behind libfixes: the sound patterns they follow, where they come from, how your brain processes them, and what they teach us about how language changes. If you haven’t subscribed already, be sure to do so to receive Part 2!

Subscribe now

ℹ️ Articles in this Series

• Part 1: Libfixes: When word parts go rogue [this issue]

• Part 2: The science of libfixes: How libfixes work, where they come from, and what they teach us about language and the mind

📑 References

• Barnhart, Robert K. (ed.). 1988. Chambers dictionary of etymology. Chambers.

• Berry, Jack. 2022. Neologasming: Topics in Modern English word formation. Term paper.

• Delhem, Romain & Caroline Marty. 2024. Measuring the correlation of productivity rates for English libfixes. Presented at the Unraveling linguistic productivity: Insights into usage, processing and variability.

• Hamans, Camiel. 2021. A lesson for covidiots about some contact induced borrowing of American English morphological processes into Dutch. Studia Anglica Posnaniensia 56(s1). 659–691. https://doi.org/10.2478/stap-2021-0009.

• Konieczna, Ewa. 2012. Lexical blending in Polish: A result of the internationalisation of Slavic languages. In Vincent Renner, François Maniez & Pierre Arnaud (eds.), Cross-Disciplinary Perspectives on Lexical Blending (Trends in Linguistics: Studies & Monographs 252), 51–74. De Gruyter. https://doi.org/10.1515/9783110289572.51.

• O’Dell, Felicity. 2016. Creating new words: affixation in neologisms. ELT Journal 70(1). 94–99. https://doi.org/10.1093/elt/ccv054.

• Renner, Vincent. 2018. French and English lexical blends in contrast. Languages in Contrast 19(1). 27–47. https://doi.org/10.1075/lic.16020.ren.

• Vasileanu, Monica & Anabella-Gloria Niculescu-Gorpin. 2025. Romanian libfixes in the making. In Sabine Arndt-Lappe & Natalia Filatkina (eds.), Dynamics at the lexicon-syntax interface (Formulaic Language 6), 241–266. De Gruyter. https://doi.org/10.1515/9783111321905-009.

• Vasileanu, Monica, Anabella-Gloria Niculescu-Gorpin & Cristina-Andreea Radu-Bejenaru. 2024. Keep calm and carry on blending: Experimental insights into Romanian lexical blending. Word Structure 17(1–2). 56–90. https://doi.org/10.3366/word.2024.0236.

The Amazon and Bookshop.org links on this page are affiliate links, which means that I earn a small commission from Amazon for purchases made through them (at no extra cost to you).

If you’d like to support Linguistic Discovery, purchasing through these links is a great way to do so! I greatly appreciate your support!

Check out my Amazon storefront here.

Check out my Bookshop storefront here.

The word infant comes from the Latin “īnfāntem” ‘unable to speak’.

A TikTok mom recently went viral because she asked:

Does anybody else not talk to their baby? My daughter is 11 months old, and she can’t talk so I don’t talk to her. Is that weird? Should I be just saying things so she can learn how to talk? I make noises, I do funny faces, I dance, but I don’t converse.

@slrlsharLike how ? But you sitting on your phone talking to strangers @mak_and_momlife …..##resilientjenkins##badmom##kidsoftiktok##fyp##viral##mom##needtoknow

Tiktok failed to load.

Enable 3rd party cookies or use another browser

Unfortunately this mom was so thoroughly excoriated in the comments and in numerous response videos that she took the original video down.

But this isn’t a dumb question!

If you aren’t familiar with research on child language acquisition, it’s entirely reasonable to wonder whether baby talk is bad for your child, or whether it’s pointless to speak to children before they begin speaking themselves. The word infant itself even comes from the Latin īnfāntem ‘unable to speak’. In some cultures parents rarely talk to infants, and those children learn their native languages just fine. How important can talking to your newborn really be if not everybody does it?

Well, pretty important actually. Even those purportedly low-conversation parents in other cultures actually talk to their kids quite a bit, just in different ways and on different timetables. (We’ll learn more about that in Parts 4 & 5 [forthcoming] of this series.)

On the other end of the spectrum, many parents have been terrified into thinking that they’re not speaking to their children enough because of an infamous 1995 study showing that children from lower socioeconomic backgrounds hear up to 30 million fewer words by age 4 than their high-income peers, and that this gap correlates with reduced vocabulary development and educational outcomes.1

But as we’ll see in Part 6 of this series, the sheer amount of exposure kids have to language doesn’t actually make much difference for their language development. Those reported differences in vocabulary size are a) questionable, and b) have more to do with other socioeconomic factors than hitting a daily word count with your toddler.

So in this series, we’ll look at what the research has to say about speaking to your infant, answering the following questions along the way:

ℹ️ Articles in this Series

• Part 1: Why you should be talking to your infant (this issue)

• Part 2: What’s the point of baby talk?

• Part 3: Is baby talk good for your child?

• Part 4: Do all cultures use baby talk?

• Part 5: Baby talk in the languages of the world

• Part 6: How much should you talk to your child? [forthcoming]

• Part 7: What really matters when talking to your child

Subscribe to get notified when the other issues in the series are published!

You’re reading Linguistic Discovery, a newsletter about the science and diversity of language—a field known as linguistics. I’m Danny Hieber, a Ph.D. in linguistics who works with indigenous communities to help them document and revitalize their languages. covered Topics in this newsletter include:

• ⚙️ how language works (cognitive linguistics, language change)

• 🌍 grammatical diversity in the world’s languages (typology)

• ℹ️ explainers of terms and concepts in linguistics

• 🗣️ language profiles

• 🗞️ the latest news and research in language and linguistics

• ⭐ linguistic reviews of books and other media

Paid subscribers also get bonus articles and early access to my in-progress book projects!

Subscribe now

Why you should be talking to your newborn

Should you be talking to your newborn child? The short answer is decidedly yes. Infants need some type of linguistic input to be able to learn a language—it’s not as though they’re born with an innate knowledge of English or any other language. Instead, they’re born with an incredibly advanced cognitive toolkit that allows them to learn any language they’re exposed to within just a few years.

But that exposure is crucial. We know from cases of extreme childhood neglect that if children aren’t sufficiently exposed to language before puberty (a timeframe known as the critical period or sensitive period in linguistics and developmental psychology), language acquisition becomes significantly more difficult and ultimately less successful. The most infamous and heart-wrenching case is that of Genie, who was confined in a dark room with minimal contact from about 20 months of age until she was almost 14 years old.2

When Genie was finally discovered she was extremely malnourished. Physically, she looked like a 6- or 7-year-old, weighed 59 pounds, had difficulty standing and walking, and couldn’t chew solid food. She never spoke, and seemed to only understand a few words at best. As she was treated and given a healthier cognitive and social environment, her rehabilitation quickly triggered the onset of puberty. Her cognitive development reached that of a 6- to 8-year-old by the following year. After 5 months she began to use single words, and her vocabulary grew quickly. She started to combine two, then three, then four words (1), and after two years she produced apparent cases of embedded clauses (2):

1. want milk
   Genie love Curtiss

   big elephant long trunk

2. ask [go shopping]

   tell [door lock]

However, Genie never mastered the complexities of English morphology and syntax. All grammatical words and morphology were missing.

Want to learn more about how children’s linguistic capabilities develop over time? Check out this issue of the newsletter about how children learn grammatical rules (and how we know):

Less lurid but still unfortunate cases have also been documented for profoundly deaf children who were not exposed to sign language in childhood. Because 90–95% of deaf children are born to hearing parents,3 it occasionally happens that those parents fail to realize that their child is deaf, so the child doesn’t receive any useful linguistic input until much later—usually not until they start school. These children are called late learners. Many late learners will nonetheless spontaneously develop a manual communication system using a combination of gestures they observe in their surroundings and signs they invent. This creates an idiosyncratic form of communication called homesign.4 Homesign is generally not a fully-developed language, but rather is more like a pidgin: an extremely limited lexicon with no grammatical constructions.

Unfortunately, late learners suffer lifelong consequences from their delayed acquisition. Deaf children who began learning a sign language at age 5 were significantly less fluent than native signers (who are exposed to sign language from birth) even after 30 years of daily use. While able to fully express themselves, their use of complex morphology and syntax is systematically different from that of deaf children who learned sign from birth.5 Deaf children who weren’t exposed to sign language until age 12 or later were profoundly hindered in their ability to learn sign.6 In one case of three deaf children who were not exposed to language of any kind until they were about 14 years old, by two years later they were acquiring vocabulary on a timeline similar to typical toddlers, with many words for concrete objects and actions but few grammatical words like prepositions.7 In another case, a deaf boy was raised in a rural area with little formal education before age 12, and no access to the deaf community. At age 15, he was fitted with hearing aids and began learning Spanish, but he struggled to learn grammatical constructions like pronouns and the singular-plural distinction.8 A final case is Chelsea, a deaf woman who did not start learning language until her early thirties, who showed similar results to Genie.9

So we’ve established that children need linguistic input as early in childhood as possible in order to fully master language. But does any kind of input work? Could you simply plop your child on the couch to watch some reality TV? If only it were that easy! It’s well established that children cannot learn language from media alone. In fact, a 2024 study even found that the greater the total screen time of children aged 2;5 to 4;0 on weekends, the worse their vocabulary and grammar skills.10 Children need social cues to help direct their attention and start associating social meaning with linguistic meaning. One salient illustration of this is the case of Jim, a hearing child born to deaf parents whose only exposure to English was via (a great deal of) television, and whose parents did not generally speak or sign to him.11 By age 3;9, Jim’s language use was decidedly abnormal for children his age:

• This is how plane

• I want that make

• House chimney my house chimney

Jim could learn new vocabulary, but was limited in his ability to form grammatical sentences.

ℹ️ In first language acquisition research, age is written as years;months, so that “3;9” means ‘3 years, 9 months old’.

ℹ️ If you’re concerned that your child’s language development might be abnormal, here’s a summary of developmental milestones you can expect:

• Age-Appropriate Speech and Language Milestones (Stanford Medicine Children’s Health)

Other studies12 have shown that children can’t learn new words from television before the age of 2.13 After age 2, it’s possible for children to learn some vocabulary from television, but only from shows designed specifically for children, and only well-designed ones.14 (Children who watched Teletubbies did not show significant vocabulary development, while children who watched Dora the Explorer, Blue’s Clues, or Dragon’s Tales showed greater vocabulary gains.15) One study showed that children only learned new vocabulary from video when an adult facilitated their learning with concomitant social interaction,16 again showing the importance of socially-directed cues for child language learning. By comparison, one study showed that children aged 21–27 months learned new vocabulary from a shared book reading activity, but none from TV viewing.17

Linguist Noam Chomsky actually once claimed explicitly that “a child may pick up a large part of his vocabulary and ‘feel’ for sentence structure from television”.18 Chomsky believes that mere exposure is sufficient for language acquisition, because he thinks that some part of grammar is innate rather than learned, a position known as linguistic nativism. But as Saxton (2017: 131) trenchantly states, “Chomsky’s early assertions about language acquisition are the result of armchair speculation, not empirical enquiry.” In reality, the amount of speech that children overhear does not predict vocabulary size, whereas the amount of speech specifically targeted at them does.19 The body of first language acquisition research has shown that targeted social interaction is crucial for child language development.

So we’ve established that some type of interactive linguistic input is necessary for children to learn language. But what kind of language works best? Baby talk? Normal adult speech? We’ll look at this question in the next two issues of the Linguistic Discovery newsletter! If you haven’t already subscribed, be sure to do so to receive the other issues in this series.

ℹ️ Articles in this Series

• Part 1: Why you should be talking to your infant (this issue)

• Part 2: What’s the point of baby talk?

• Part 3: Is baby talk good for your child?

• Part 4: Do all cultures use baby talk?

• Part 5: Baby talk in the languages of the world

• Part 6: How much should you talk to your child? [forthcoming]

• Part 7: What really matters when talking to your child

Subscribe to get notified when the other issues in the series are published!

Subscribe now

📚 Recommended Reading

“The birth of a word”

MIT researcher Deb Roy wanted to understand how his infant son learned language—so he wired up his house with video cameras to catch every moment (with exceptions) of his son’s life, then parsed 90,000 hours of home video to watch gaaaa slowly turn into water. Astonishing, data-rich research with deep implications for how we learn.

HELLO Lab Presents

The Hearing Experience & Language Learning Outcomes (HELLO) Lab at the University of Connecticut has a great series of YouTube Videos about child language acquisition for parents.

• HELLO Lab on YouTube

How babies talk: The magic and mystery of language in the first three years of life

📑 References

• Anderson, Daniel R. & Tiffany A. Pempek. 2005. Television and very young children. American Behavioral Scientist 48(5). 505–522. https://doi.org/10.1177/0002764204271506.

• Baker, Anne, Beppie Van Den Bogaerde, Roland Pfau & Trude Schermer (eds.). 2016. The linguistics of sign languages: An introduction. John Benjamins. https://doi.org/10.1075/z.199.

• Barr, Rachel & Nancy Wyss. 2008. Reenactment of televised content by 2-year olds: Toddlers use language learned from television to solve a difficult imitation problem. Infant Behavior & Development 31(4). 696–703. https://doi.org/10.1016/j.infbeh.2008.04.006.

• Chomsky, Noam. 1959. Review of “Verbal behavior” by B.F. Skinner. Language 35(1). 26–58. https://doi.org/10.2307/411334.

• Curtiss, Susan. 1988. Abnormal language acquisition and the modularity of language. In Frederick J. Newmeyer (ed.), Linguistic theory: extensions and implications (Linguistics: The Cambridge Survey), vol. 2. Cambridge University Press.

• Hart, Betty & Todd R. Risley. 1995. Meaningful differences in the everyday experience of young American children. Baltimore: P. H. Brookes.

• Krcmar, Marina, Gernard Grela & Kirsten Lin. 2007. Can toddlers learn vocabulary from television? An experimental approach. Media Psychology 10. 41–63. https://doi.org/10.108/15213260701300931.

• Kuhl, Patricia K., Feng-Ming Tsao & Huei-Mei Liu. 2003. Foreign-language experience in infancy: Effects of short-term exposure and social interaction on phonetic learning. Proceedings of the National Academy of Sciences 100(15). 9096–9101. https://doi.org/10.1073/pnas.1532872100.

• Linebarger, Deborah L. & Dale Walker. 2005. Infants’ and toddlers’ television viewing and language outcomes. American Behavioral Scientist 48(5). 624–645. https://doi.org/10.1177/0002764204271505.

• Lust, Barbara. 2006. Child language: Acquisition and growth (Cambridge Textbooks in Linguistics). Cambridge University Press.

• Mumme, Donna L. & Anne Fernald. 2003. The infant as onlooker: Learning from emotional reactions observed in a television scenario. Child Development 74(1). 221–237. https://doi.org/10.1111/1467-8624.00532.

• Newport, Elissa L. 1990. Maturational constraints on language learning. Cognitive Science 14(1). 11–28. https://doi.org/10.1207/s15516709cog1401_2.

• O’Doherty, Katherine, Georgene L. Troseth, Priya M. Shimpi, Elizabeth Goldenberg, Nameera Akhtar & Megan M. Saylor. 2011. Third-party social interaction and word learning from video: Social interaction and video. Child Development 82(3). 902–915. https://doi.org/10.1111/j.1467-8624.2011.01579.x.

• Patterson, Janet L. 2002. Relationships of expressive vocabulary to frequency of reading and television experience among bilingual toddlers. Applied Psycholinguistics 23(4). 493–508. https://doi.org/10.1017/S0142716402004010.

• Ramírez, Naja Ferjan, Amy M. Lieberman & Rachel I. Mayberry. 2013. The initial stages of first-language acquisition begun in adolescence: when late looks early. Journal of Child Language 40(2). 391–414. https://doi.org/10.1017/S0305000911000535.

• Rice, Mabel L., Aletha C. Huston, Rosemarie Truglio & John C. Wright. 1990. Words from “Sesame Street”: Learning vocabulary while viewing. Developmental Psychology 26(3). 421–428. https://doi.org/10.1037/0012-1649.26.3.421.

• Rice, Mabel L. & Linda Woodsmall. 1988. Lessons from television: Children’s word learning when viewing. Child Development 59(2). 420. https://doi.org/10.2307/1130321.

• Sachs, Jacqueline, Barbara Bard & Marie L. Johnson. 1981. Language learning with restricted input: Case studies of two hearing children of deaf parents. Applied Psycholinguistics 2(1). 33–54. https://doi.org/10.1017/S0142716400000643.

• Sandler, Wendy & Diane Lillo-Martin. 2017. Sign languages. In Mark Aronoff & Janie Rees-Miller (eds.), The handbook of linguistics (Blackwell Handbooks in Linguistics), 371–396. 2nd edn. Wiley-Blackwell.

• Saxton, Matthew. 2017. Child language: Acquisition and development. 2nd edn. SAGE.

• Shneidman, Laura A., Michelle E. Arroyo, Susan C. Levine & Susan Goldin-Meadow. 2013. What counts as effective input for word learning? Journal of Child Language 40(3). 672–686. https://doi.org/10.1017/S0305000912000141.

• Snow, C. E., A. Arlman-Rupp, Y. Hassing, J. Jobse, J. Joosten & J. Vorster. 1976. Mothers’ speech in three social classes. Journal of Psycholinguistic Research 5(1). 1–20. https://doi.org/10.1007/BF01067944.

• Tulviste, Tiia & Jaan Tulviste. 2024. Weekend screen use of parents and children associates with child language skills. Frontiers in Developmental Psychology 2. 1404235. https://doi.org/10.3389/fdpys.2024.1404235.

• Uchikoshi, Yuuko. 2005. Narrative development in bilingual kindergarteners: Can Arthur help? Developmental Psychology 41(3). 464–478. https://doi.org/10.1037/0012-1649.41.3.464.

The Amazon and Bookshop.org links on this page are affiliate links, which means that I earn a small commission from Amazon for purchases made through them (at no extra cost to you).

If you’d like to support Linguistic Discovery, purchasing through these links is a great way to do so! I greatly appreciate your support!

Check out my Amazon storefront here.

Check out my Bookshop storefront here.

To some of you, this may seem like an odd question. If so, that’s probably because you don’t pronounce the ⟨g⟩ in longevity twice.

You see, there are two common pronunciations of longevity: one where the first syllable sounds like lawn, and one where that syllable sounds like long. In the International Phonetic Alphabet (IPA), for a General American dialect, those two pronunciations might be transcribed like this:

• When the first syllable sounds like lawn, ⟨g⟩ is pronounced once: /lɑnˈd͡ʒɛv.ɪ.ti/

• When the first syllable sounds like long, ⟨g⟩ is pronounced twice: /lɑŋˈd͡ʒɛv.ɪ.ti/

There is of course a lot of additional variation in how this word is pronounced across various dialects, but the relevant difference here is whether the nasal consonant before the ⟨g⟩ is pronounced as /n/ (like at the end of the word sin) or /ŋ/ (like at the end of the word sing). That /ŋ/ sound is called eng (or engma) in linguistics. From what I can tell, the use of eng in longevity is more common in North American English, Australian English, and New Zealand English than it is in British English, but both pronunciations can be heard anywhere in the core Anglosphere.

The eng pronunciation also seems to be a more recent development, based solely on the fact that major dictionaries (the OED, Merriam-Webster, Dictionary.com) don’t even list that pronunciation yet. Wiktionary, however, does list this variant (which is an excellent example of how good Wiktionary’s descriptive coverage has become, thanks to crowdsourcing its data).

Among those who do pronounce longevity with an eng, however, the mystery of why the ⟨g⟩ gets pronounced twice is deeply puzzling, at least if the number and view count of online videos about this question are any indication:

@iambrianfinkHow do YOU pronounce it? 😤 #english #weirdwords #longevity #brianfink #fyp

Tiktok failed to load.

Enable 3rd party cookies or use another browser

It even caught the attention of some popular online educators!

@jasonkpargin @jasonkpargin ♬ Elevator Music - Bohoman

Tiktok failed to load.

Enable 3rd party cookies or use another browser

Of course, videos like these tend to spark backlash in the comments from people who either insist that there’s only one correct pronunciation of a controversial word, or claim to have never heard the alternate pronunciation.

I say “claim” because few people are actually so linguistically isolated that they’ve genuinely never heard common alternate pronunciations of everyday words. If they’re globally connected enough to be commenting on language videos on the internet, they’ve likely encountered those other pronunciations but simply weren’t paying attention. This isn’t a criticism of how observant people are. Your brain ignores a great deal of the linguistic variation it encounters because it’s focused not just on the sounds you hear but on identifying the words being used and the message being conveyed. Our brains are incredible categorization machines, capable of disregarding incidental differences between stimuli in order to identify instances of the same abstract type. Even widely variant pronunciations of a word still get categorized as instances of the “same” word by your brain. We would not be able to do language with each other if our brains weren’t capable of ignoring these interpersonal disparities.

Ironically, it’s precisely our incredible linguistic abilities and our brains’ ability to filter out semantically irrelevant differences which make people so upset when those differences are pointed out to them. It makes for some great internet ragebait, as the above videos show.

Anyway, I digress. Now that we’ve established that both /lɑnˈd͡ʒɛv.ɪ.ti/ and /lɑŋˈd͡ʒɛv.ɪ.ti/ are in fact common pronunciations of longevity, let’s talk about how it got to be pronounced that way, and how its spelling is actually playing a neat mental trick on you.

You’re reading Linguistic Discovery, a newsletter about the science and diversity of language—a field known as linguistics. I’m Danny Hieber, a Ph.D. in linguistics who works with indigenous communities to help them document and revitalize their languages. Articles in this newsletter cover:

• ⚙️ how language works (cognitive linguistics, language change)

• 🌍 grammatical diversity in the world’s languages (typology)

• ℹ️ explainers of terms and concepts in linguistics

• 🗣️ language profiles

• 🗞️ the latest news and research in language and linguistics

• ⭐ linguistic reviews of books and other media

Paid subscribers also get bonus articles and early access to my in-progress book projects!

Subscribe now

The spelling of longevity is messing with your head

You might be tempted at first to explain the eng pronunciation by saying that the ⟨g⟩ is shared by two different syllables—it’s part of both the first syllable, long-, and the second syllable, -gev-. And you’d be clever for suggesting this, because this is in fact something that happens occasionally in English. Linguists often analyze words like dinner and donor as containing an ambisyllabic consonant—a consonant sound that is part of two syllables at once. (Note that we’re talking about sounds here, not letters.) However, the ⟨g⟩ in longevity isn’t one of those cases. The syllable boundary around that ⟨g⟩ is actually quite clear: the first syllable is /lɑŋ/ or /lɑn/ depending on your pronunciation, and the second syllable is /d͡ʒɛv/.

🌟 Bonus Content: How do we know where syllable boundaries are?

What are the syllables in donor, hurry, and lemon? Does thinking about that make your head hurt? If so, you’re not alone! Linguists puzzle over this same problem. If you’re a paying subscriber, stick around til the end of the article for some bonus content about ambisyllabic consonants and how linguists think about syllable boundaries.

In order to understand what’s happening with longevity, you need to make a mental distinction between how the word is spelled and how it’s pronounced:

• spelling: ⟨longevity⟩

• pronunciation: /lɑŋˈd͡ʒɛv.ɪ.ti/ (eng pronunciation)

You’ll notice that the spelled version of longevity has ⟨angle brackets⟩ around it, while the pronunciation version has /slashes/ around it. Linguists use these conventions to clarify when we are talking about the written representation of a word (its spelling) versus when we are talking about the spoken representation of a word (its pronunciation).

If you pay attention to the pronunciation of longevity rather than its spelling, you might notice something odd given the subject of this article: there’s no /g/! Even though longevity is spelled with one ⟨g⟩ letter, it doesn’t have a /g/ sound when you write it phonetically. That’s because in the International Phonetic Alphabet, the symbol /g/ always refers to the “hard” /g/ as in go or good. Since longevity doesn’t have a “hard” /g/ in the pronunciation, there’s no /g/ in its phonetic transcription.1

In English spelling conventions—English’s orthography, if you prefer the technical term—the letter ⟨g⟩ is commonly used to write a few different sounds:

And certain combinations of letters (known as digraphs) with ⟨g⟩ also commonly represent certain sounds:

In the word longevity, the letter ⟨g⟩ is being used to represent two of these “g sounds” at once: the eng sound /ŋ/ and the “soft g” sound /d͡ʒ/. Even though neither of these is technically a /g/ sound in the sense of being a “hard g”, that’s why you perceive there to be two ⟨g⟩’s in the word—it contains two different sounds commonly associated with the letter ⟨g⟩ in English.

You may have also just noticed that the word English itself has this same quirk: the ⟨g⟩ represents both /ŋ/ and /g/ at the same time (though the story behind why that happens in the word English is actually different from why it happens in longevity, even though the result is similar).

How did this dual use of ⟨g⟩ come about? Which sound did that ⟨g⟩ originally represent? Why isn’t it spelled ⟨longgevity⟩ instead? To understand how we got here, we need to trace longevity‘s journey from Classical Latin to modern English.

The history of longevity

The word longevity was first documented in English in 1569, making it one of the many words borrowed from Latin during the Renaissance. The Latin word it came from was longaevitās ‘long life’, which itself is a combination of longus ‘long’ + aevum ‘time, eternity; age, generation’ + -tās, an abstract noun suffix indicating states of being.

It’s important to note that not all sounds are meaningful in all languages. For example, Latin didn’t make a meaningful contrast between /n/ and /ŋ/ like English does. You can find pairs of words in English like sin /sɪn/ vs. sing /sɪŋ/ which prove that /n/ and /ŋ/ are meaningfully distinct sounds in the language, but you can’t find such pairs in Latin. Pairs of words like this which differ by only a single sound are called minimal pairs, and they are how linguists determine which sounds are distinct and meaningful in a language. In English, you have to be able to recognize the difference between /n/ and /ŋ/ if you want to be able to distinguish sin from sing or kin from king. This is what I mean what I say that there’s a “meaningful contrast” between /n/ and /ŋ/ in English. When a language does make a meaningful contrast between two sounds like this, we call them distinct phonemes.

Latin had a phoneme /n/, written as ⟨N⟩, but not a phoneme /ŋ/. Instead, /n/ could be pronounced as either [n] or [ŋ] depending on the sounds around it. So while Latin speakers pronounced an eng [ŋ] in many words, they didn’t perceive it as a distinct sound; they perceived it as just a variation of /n/. In particular, the [ŋ] variant was only used before other velar consonants like /g, k/. (When a phoneme’s pronunciation changes to be more like the sounds around it in this way, the process is called assimilation.)

In technical terms, we’d say that the [ŋ] pronunciation is an allophone of the phoneme /n/. An allophone is simply a pronunciation variant of a particular phoneme, usually conditioned by the sounds around it. In English, for instance, the phoneme /t/ can be pronounced [t] (plain), [tʰ] (aspirated), [ʔ] (glottalized), [t̚] (unreleased), and [ɾ] (flapped). Each of these pronunciations is a different allophone of /t/, and each has its own rules about when it’s used. The aspirated [tʰ], for example, is generally the pronunciation used at the beginnings of syllables. But if you’re a native English speaker, you just hear all these different sounds as /t/. [t, tʰ, ʔ, t̚, ɾ] are not meaningfully distinct in English, just like [n] and [ŋ] were not meaningfully distinct sounds in Latin.

Notice the additional transcription convention I’ve introduced here too: when a sound is a distinct phoneme in a language, we write it between /slashes/; but when a sound isn’t a meaningfully distinct phoneme, we write it between [square brackets]. The first is a phonemic transcription and the second is a phonetic transcription.

Putting all this together, it means that longaevitās in Classical Latin would have been pronounced as [laŋˈɡaᵉ.wɪ.taːs] but perceived as /lanˈɡaᵉ.wɪ.taːs/, with /n/ pronounced as [ŋ] due to assimilation with the following /g/.

Confusingly but unsurprisingly, however, Latin pronunciation changed over the centuries. Even as Latin broke up into distinct languages like French, Italian, Portuguese, Romanian, Spanish and numerous others (what we today call Romance languages, or descendants of Latin), people still thought they were speaking Latin, and so they would read Latin texts with vernacular pronunciations. For example, the Latin word saeculum ‘generation’ was pronounced /ˈsaᵉ.ku.lum/ in Classical Latin, but in ninth-century Spain it was read as /sjeglo/ (Modern Spanish siglo ‘century’). To us, that might not seem like a very logical way to pronounce ⟨saeculum⟩, but the same kind of “illogical” transformation also happened in English: think about how English speakers today say knight as /naᶦt/ today rather than the original pronunciation /knɪxt/ (Wikipedia: Ecclesiastical Latin).

It took scholars centuries to recognize the Romance descendants of Latin as distinct languages. The first grammar of a Romance language wasn’t even published until 1492! It came in the form of Antonio de Nebrija’s grammar of Castilian Spanish (although we see hints of early Spanish as far back as the 900s in the Glosas Emilianenses, and early French in the Oaths of Strasbourg in 842).

As a result of the sound changes throughout Europe and in Italy specifically, the pronunciations used by the Catholic Church for Latin wound up being quite different from the original pronunciations of Classical Latin. Today we call this Ecclesiastical Latin, the variety of Latin used in Christian literature throughout the Middle Ages and in the Catholic Church today.

The genesis of Ecclesiastical Latin had an odd result for our history of longevity: because English scholars during the Renaissance only knew Ecclesiastical Latin pronunciation and not Classical pronunciation (linguists didn’t fully piece together how Classical Latin sounded until the florescence of historical linguistics in the 1800s), whenever they borrowed words from Latin, they did so with their Ecclesiastical pronunciations. This meant that when longevity first entered English in 1569, it was probably pronounced something like /lɑnˈd͡ʒɛ.vi.tas/ instead of the Classical [laŋˈɡaᵉ.wɪ.taːs]. The ⟨g⟩ was now pronounced as a “soft g” /d͡ʒ/, and the /n/ pronounced as [n] rather than the traditional eng [ŋ]. So when the word longevity first entered English, the ⟨g⟩ only represented the “soft g” sound /d͡ʒ/—it wasn’t yet doing double duty as it does for many speakers today. The eng pronunciation actually had to be reintroduced later, ironically making the pronunciation more like the original Classical Latin pronunciation.

What caused the eng pronunciation to come full circle like this? Well we know it’s not due to assimilation like it was in Latin. In Classical Latin, the /n/ in longevity was pronounced as [ŋ] because that made the sound more similar to the /g/ which followed it—both [ŋ] and [g] are pronounced with the tongue touching the velum (soft palate) at the back of the roof of the mouth. But this can’t have happened in English because the sound following the /n/ is /d͡ʒ/ rather than /g/, and /d͡ʒ/ is pronounced at the front of the mouth at the alveolar ridge. So pronouncing longevity with an eng [ŋ] arguably makes the word harder to pronounce because the tongue has to move rapidly between more distant articulatory targets.

Instead, it seems that the main reason for the shift in the pronunciation of longevity is because speakers made an analogy to the word long in English, which of course is pronounced with an eng, as /lɑŋ/. And this is a very reasonable thing to do! The word long is, after all, historically related to longus, which was part of that Classical Latin word longaevitās. The English word long wasn’t borrowed from Latin longus, though; instead they developed independently, in parallel, from the original Proto-Indo-European root *dlongʰos ‘long’. And even if you didn’t know that longevity actually comes from Latin, it’s entirely reasonable to make an association between long and longevity since the latter literally means ‘a long duration of individual life’. English speakers may not know the exact details of how long and longevity are related, but it seems to them that there must be some connection between them.

This makes the word longevity a subtle case of folk etymology—reinterpreting a word or phrase as having different elements than it originally did because the original elements are unfamiliar to the speaker. In this case, few English speakers know that longevity is actually from a Latin word longaevitās, so they reanalyze it as being from the English word long instead, and assume the -ev- part just has some other meaning they’re not aware of, maybe connected to words like brevity and levity. Of course, there is no actual connection between brevity/levity and -ev- or the original Latin aevum, but the mere existence of other words ending in -evity this makes it easier for speakers to analogically reinterpret longevity this way.

This is why I actually like calling folk etymology analogical reanalysis or etymological reinterpretation instead (Cienkowski 2008). The word folk has subtle pejorative implications of ignorance, when what’s actually happening here is some clever analogical reasoning on the part of speakers. And what’s pretty incredible about this is that almost no one is sitting around consciously thinking about longevity this way before they say it with an /ŋ/. All this analogizing is happening subliminally and automatically. Humans are incredible pattern-recognition machines, especially when it comes to language.

And so we now have the complete answer to our orthographic mystery: the ⟨g⟩ in longevity is pronounced twice because it was borrowed from Latin at a point when the pronunciation of Latin had undergone significant changes, and then it was reinterpreted as being based on the English word long rather than the original Latin word longaevitās. Along the way, we learned a little about how our brains process language, how spelling relates to pronunciation, and how language might change over time. So next time you notice someone pronounce longevity—or any word!—differently from how you do, it’s worth pausing to appreciate the amazing confluence of cognitive, historical, physiological, and analogical factors that brought those differences about, and yet still make us capable of understanding each other.

🙏 Credits

This issue of the Linguistic Discovery newsletter was edited by Amy Treber. If you’re looking for professional copyediting services, email Amy at amytreberedits@gmail.com.

The final responsibility for any mistakes or omissions is of course still wholly my own.

📑 References

• Cienkowski, Witold. 2008. The initial stimuli in the processes of etymological reinterpretation (so‐called folk etymology). Scando-Slavica 15(1). 237–245. https://doi.org/10.1080/00806766908600524.

• Gordon, Matthew K. 2016. Phonological typology (Oxford Surveys in Phonology & Phonetics 1). Oxford University Press.

• Jensen, John T. 2000. Against ambisyllabicity. Phonology 17(2). 187–235. https://doi.org/10.1017/S0952675700003912.

• Treiman, Rebecca & Catalina Danis. 1988. Syllabification of intervocalic consonants. Journal of Memory and Language 27(1). 87–104. https://doi.org/10.1016/0749-596X(88)90050-2.

The Amazon and Bookshop.org links on this page are affiliate links, which means that I earn a small commission from Amazon for purchases made through them (at no extra cost to you).

If you’d like to support Linguistic Discovery, purchasing through these links is a great way to do so! I greatly appreciate your support!

Check out my Amazon storefront here.

Check out my Bookshop storefront here.

🌟 Bonus Content: How do we know where syllable boundaries are?

Read more

For the not chronically online among you who might not have encountered the expression yet, 6-7 is meant to be nonsensical, playful, and absurd. Like most slang, it’s intentionally obscure and aims to disrupt common social expectations. A well-placed 6-7 is one that comes when it’s least expected.

6-7 is an example of a specific type of slang that has received its own category of late, called brainrot, referring to expressions that arise in “chronically online” communities and are said to be the result of consuming large amounts of low-value internet content.

It’s also one of a fascinating set of expressions where what you do with the word, socially, is more important than what is meant by the word. Ta-da! marks a reveal. Oops acknowledges a mistake. Ugh expresses disgust or frustration. The definitions of these words are the social deed you accomplish in the mere act of uttering them. In the technical jargon of philosophy of language, we’d say that the illocutionary or pragmatic meaning of these words is more prominent than their locutionary or semantic meaning. Although in some contexts 6-7 can mean ‘so-so’ or ‘maybe this, maybe that’ (especially when paired with an alternating hand gesture), its primary meaning is the social function of indicating playfulness, surprisal, and absurdity.

Yet even many of those who appreciate 6-7’s whimsy oppose Dictionary.com’s decision to recognize it, citing objections such as “It’s a passing fad” or “It’s meaningless.” Here are a few of the more common objections and what linguistics—the scientific study of how language works—has to say about them.

You’re reading Linguistic Discovery, a newsletter about the science and diversity of language—a field known as linguistics. I’m Danny Hieber, a Ph.D. in linguistics who works with indigenous communities to help them document and revitalize their languages.

Subscribe to the newsletter to receive weekly deep dives into how language works, language profiles, explainers of terms/concepts in linguistics, the latest news and research in linguistics, reviews, and more!

Paid subscribers also get bonus articles and early access to articles and chapters of my book!

Subscribe now

“It’s a phrase, not a word.”

Lots of dictionary entries are what are technically called multi-word expressions (MWEs), or compound words. Dictionaries record any MWE whose meaning is more than just the sum of its parts: give up and self-esteem are two examples.

6-7 isn’t structurally different from other hyphenated numeral expressions like to sixty-nine (the sexual act) or to eighty-six (to refuse to serve a customer; to eject or ban). Each of these are MWEs.

“Only kids use it.”

Many words are restricted in their usage or demographics. Only older generations say “That’s swell”, and only people in New England say “Wicked!” as an expression of approval.

Regional words or generational words are still words, even if not everybody uses them.

“It’s meaningless.”

No word in common use is truly meaningless. Why would people say something if not to express some kind of meaning? Just because the meaning isn’t obvious (to you) doesn’t imply it’s meaningless.

As discussed above, oftentimes the pragmatic meaning of a word (the contextual/situational meaning) is more important than the semantic (literal) meaning of a word anyway. That’s what’s happening with 6-7: it’s an intentionally absurdist use of otherwise dull words.

“It’s not even a word! It’s just numbers!”

A number is a mathematical object, a concept.

A numeral in linguistics is a word for a number.

6 and 6th are numbers. six and sixth are numerals (cardinal and ordinal numerals specifically).

So to be precise, Dictionary.com’s Word of the Year is the linguistic expression six-seven and not the numbers 6-7.

“It’ll be gone in a year. Dictionaries should only record words likely to stick around.”

It’s true that 6-7 has seen a rapid ascension to lexicographical greatness, from complete obscurity in January to one of the most widely-used and widely-discussed expressions of the year. And precisely because of its recency, Dictionary.com has for now marked the entry as slang and included it in its distinct slang dictionary.

And yes, the vast majority of slang—and words in general!—have short lifespans. But plenty of “real” words that once illuminated our lexicon have now dimmed into obscurity, such as wherefore, thither, and overmorrow. There are over 500,000 words in the OED, but the average native American English speaker only knows about 42,000, meaning that many words have fallen out of use. A century from now, readers will encounter 6-7 in texts from this period and want to know what it meant.

A dictionary’s job is to document all of the words, not just the ones you like. In all likelihood, we’ll have forgotten about 6-7 in a year. But who knows? Perhaps 6-7 will graduate from mere slang to the respected rank of Real Word, and in a hundred years we’ll be grateful to the dictionaries for documenting its whimsical rise to lexicographic legitimacy. Either way, we should appreciate the socially complex and linguistically creative nature of 6-7 and enjoy it while we still can.

📚 Recommended Reading

The Amazon and Bookshop.org links on this page are affiliate links, which means that I earn a small commission from Amazon for purchases made through them (at no extra cost to you).

If you’d like to support Linguistic Discovery, purchasing through these links is a great way to do so! I greatly appreciate your support!

Check out my entire Amazon storefront here.

Before I wrote The Language-Lover’s Lexipedia, and before I created my game League of the Lexicon—a quiz game about words and language—I ran a communications agency. For over a decade I told clients that it wasn’t what I knew about their business that mattered, but my curiosity about what I didn’t know. It might sound like an ingenious excuse for being ill-informed, but what I meant was that because I didn’t know what they knew, I asked questions; and because I asked questions, I saw their world differently.

I mention this because that’s how I think about words and language too. I’m neither a linguist nor a lexicographer—I didn’t even take English at university; but loving language doesn’t need an English degree, only curiosity. And it is this curiosity that has led me deep into the rabbit holes of language.

It often begins with a question—something that occurs to me mid-conversation or while staring at an ad on the London Underground. Questions like:

• Where does punctuation come from?

• How do Chinese keyboards work?

• How can I get a word in the dictionary?

• Why are pedants pedantic?

• Why are words with a K funny? (And why are words with a T not?)

• How are IKEA products named?

And on it goes.

For me, language is a loose thread that I can’t resist pulling on; and when I start, it’s hard to stop, because the thread is infinite. So I’m going to try an experiment for this article about linguistic curiosity; I’m going to write about the first thing my eyes land on, then keep pulling the thread to see where it leads…

keyboard (n.)

The first key-boards [sic] were found on musical instruments, specifically organs, pianos, harpsicords and the like, with the word first appearing in the Oxford English Dictionary (OED) in 1776. The first citation for a type-based keyboard is 1846, in relation to a telegraphy machine. Given the early design of these devices, it was an obvious choice, since they used actual piano keys, transposed from hammering strings to inputting letters.

Speaking of ancient keyboards…

…it seems strange to see the ampersand—the squiggly symbol that means and—featured on such an old keyboard; after all, it doesn’t get much use these days. But the truth is, the fortunes of this elegant glyph have plummeted over the last century.

Up to the 20th century, the ampersand was often taught as the 27th letter of the alphabet; today, it is relegated to corporate logos and space-squeezed signs. Indeed, as grammarian June Casagrande wrote in 2019: “Here’s a simple guideline for using ampersands: Don’t.” How the mighty have fallen.

The earliest ampersand was found on a wall in Pompeii (see below) where it was a ligature combining e and t of the Latin et, meaning ‘and’. Its name is a contraction of and per se and, a Latin-English hybrid that essentially means: ‘this symbol means “and”’. Children chanting their ABCs would conclude with: “X, Y, Z, and per se and”; spoken quickly, those last syllables combined to create its name, ampersand. When people wrote by hand—or carved on stone—the ampersand was no doubt useful; but modern keyboards have made its time-saving qualities redundant.

Incidentally, the ampersand wasn’t the first symbol to represent ‘and’. A century and a half before the destruction of Pompeii, Tiro, secretary and slave of the Roman statesman Cicero, invented the Tironian et, that looked like this: ⟨⁊⟩. It was used in the Gothic script known as blackletter, but is now only found in Irish Gaelic.

Speaking of Cicero…

…you may be more familiar with his work than you realise.

When graphic designers lay out work, they use placeholder text, and for as long as most people can remember, it’s been a chunk of Latin-sounding words starting “Lorem ipsum…” But where this text came from, and when it was first used, was a mystery for a long time.

It’s often said that the practice started among 16th-century printers, but this is untrue. The first evidence of Lorum ipsum is the 1960s, where it appeared in adhesive transfer sheets made by Letraset which were used for text layout. It then secured its place among graphic designers when it was adopted in 1985 by Aldus PageMaker, the desktop publishing software for Apple computers.

The text itself was assumed to be gibberish, until classicist Richard McClintock showed that it came from mangled fragments of Cicero’s De Finibus Bonorum et Malorum (‘About the Ends of Good and Evil’), written around 45 BC. It has thus become one of the most accidentally (mis-)quoted Latin texts in history.

Speaking of misquotations…

…did Groucho Marx really say, “I didn’t say that, and even if I did, I didn’t mean it”? Sadly not. Despite being widely shared online, it is, appropriately, a misquotation. Or rather, a misattribution. Likewise Yogi Berra’s “I never said half the things I said”, which he famously didn’t say, but people like to say he did.

Misquotations are nothing new, but thanks to social media and hallucinatory AI, they’re booming. There are three types of misquotation: the thing that was almost said, the thing that was never said and the thing that may have been said, but not by the person said to have said it. The problem is that determining the truth can be complicated, but sharing a misquotation is easy. Or as quotation compiler Robert Andrews is said to have said—but, again, didn’t—“Most quotations end up being misquotations.”

People love reading clever things said by famous people—Oscar Wilde, Coco Chanel, Mark Twain, Albert Einstein and the like. As a result, well-known figures are commonly attached to nuggets of wisdom they never uttered. But correcting these errors is difficult when people don’t want them corrected. “If you’re going through Hell, keep going” wasn’t Winston Churchill, but it’s what people want him to have said. Just as they want to believe Marilyn Monroe delivered the sassy line, “Well-behaved women seldom make history.” It’s a lot less exciting that the real author was Harvard professor, Laurel Thatcher Ulrich.

Mark Twain probably didn’t say “Never let the facts get in the way of a good story” either, making it an apt motto for those who share dubious quotations. However, if you belong to the minority of people who care about the accuracy of quotations, visit the Quote Investigator, who provides both facts and good stories.

Speaking of mistakes…

…errors in books are called corrigenda or errata. For a long time, such mistakes were common in printing. Shakespeare’s First Folio is full of them, and many first editions—from writers like Niccolò Machiavelli, Jane Austen, Oscar Wilde, Charles Dickens and more—feature errata slips. Even the Declaration of Independence has errata, with an erratum in the errata itself.

Today, however, mistakes can be embarrassing and costly. It certainly was for Penguin Australia who had to destroy 7,000 copies of The Pasta Bible in 2010 because the recipe for tagliatelle with sardines and prosciutto listed “freshly ground black people” in the ingredients.

Fortunately not all mistakes require a work to be pulped, often an errata slip will do—but these can be awkward too. When Faber and Faber published Moortown by Ted Hughes in 1979, there was a typo in “Night Arrival of Sea Trout”, that required the correction, “for ‘rape’, read ‘nape’.” But at least there was good news when the publisher of the Australian Dictionary of National Biography corrected one entry with the information: “For ‘died in infancy’ read ‘lived to a ripe old age at Orange’.”

Some mistakes are so bad they’re good. Susan Anderson says she’s still haunted by the typo on page 293 of her sizzling romance novel, Baby, I’m Yours: “He stiffened for a moment but then she felt his muscles loosen as he shitted on the ground.” And scientists at the Large Hadron Collider in Cern have stopped counting the number of references to their work at the “Large Hardon Collider”.

But not all mistakes are costly. One of the most valuable Harry Potter books sold in auction was a first edition hardback, where the title on the back cover is misspelled “Harry Potter and the Philospher’s Stone”. It’s not even the only mistake in the book. But thanks to these typos, it sold for £68,812 in 2019.

Speaking of typos…

…brings us back to keyboards and the origin of the QWERTY layout, a subject that has been much debated.

What’s known is that the QWERTY arrangement was created by Christopher Latham Sholes for the “Sholes and Glidden type-writer” [sic], which became the Remington No. 1, the first commercially viable typewriter.

It’s commonly said that Sholes’ layout, first documented in an 1878 patent, was designed to avoid the mechanical problem of type bars clashing within the machine. Far from being the best layout for speed, the story goes, QWERTY was designed to slow typists down so they wouldn’t jam the type basket. It’s also been claimed that the QWERTY layout was chosen so that salesmen could type TYPEWRITER with one finger from the first line of letters. It’s true that you can, but I wouldn’t read anything into it.

Since many of the early users of typewriters were morse code operators, for whom speed was a priority, it is highly improbable that Sholes would design a keyboard to deliberately inhibit speed. In the early days of the typewriter, inventors explored different layouts (see below), but all with the aim of improving typing speed. As John Harrison wrote in his 1888 Manual of the Type-Writer:

Why its alphabet should begin with Q, and end with M, is mysterious at first sight. But the reason of this disorder soon becomes clear. The keys are arranged so that greater facility is given to fingering letters than if the letters were placed in their usual order.

The question is not so much how QWERTY came to be devised as how it came to dominate, and the answer to this lies in Remington’s position as the leading manufacturer of typewriters. As their typing schools and teaching methods promoted QWERTY, other manufacturers fell into line and it became the de facto standard. QWERTY almost certainly isn’t the best keyboard layout—even Sholes proposed a new design before his death—but the combination of technology lock-in, the network effect and the efforts of the leading manufacturers cemented its position.

And thus we come full circle, almost as if I engineered it that way. But I had to. I could have gone from mistakes to malapropisms to humour to sound symbolism to language learning and beyond. Because the world of words never ends. And for those with a curious mind, that is the joy of it.

If you enjoyed this article, be sure to check out Joshua’s new book, The language lover’s lexipedia: An A–Z of linguistic curiosities (Amazon | Bookshop.org), and grab a copy of League of the Lexicon here!

The Amazon and Bookshop.org links on this page are affiliate links, which means that I earn a small commission from Amazon for purchases made through them (at no extra cost to you).

If you’d like to support Linguistic Discovery, purchasing through these links is a great way to do so! I greatly appreciate your support!

Check out my entire Amazon storefront here.

• Bookshop.org

• Amazon

• Barnes & Noble

What better way to leave our mark upon the world than to name something after ourselves?

If something is named after someone, or another proper noun like a place, that’s an eponym, from the Greek ‘given as a name’ or ‘named after’ (epi ‘upon, (called) after’ + onyma ‘name’).

For example, Vanderbilt University is eponymously named after the American industrialist Cornelius Vanderbilt. Shrapnel is named after Lieutenant General Henry Shrapnel (1761-1842), a British army officer and artillery specialist who invented a special kind of exploding shell—which produced, you guessed it, shrapnel.

One of our most common modern words is secretly an eponym: The word guy is ultimately derived from the name of the infamous Guy Fawkes, one of the conspirators in the 1605 Gunpowder Plot, which sought to blow up the Houses of Parliament. The plot is now remembered on Guy Fawkes Day, and his infamy (at least according to Protestant royalists) fully embedded Guy into the English language—but first, in the 1600s and 1700s as a pejorative, meaning a grotesque- or shabby-looking person. Its derogatory sense faded with popular usage, especially after it made its way into American English in the 1840s. Separation from the historical event ameliorated the term. Prior to guy, names such as Jack were often used as similar placeholders or generic terms—and we treat names in the same way today in phrases like average Joe.

Similarly, pants are named after a fictional Italian fellow. The word pants is a shortened version of the word pantaloons. The word pantaloons has referred to several types of legwear since the 16th century. Originally it was a word for men’s tights or hose. Later, pantaloons would refer to men’s knee breeches and women’s baggy under-trousers, both of which were gathered at the knee or the ankle. This eventually gave us the word panties for women’s undies, and the British English use of the word pants for underpants in general, regardless of gender. In the late 18th century, pantaloons was extended to our modern-day idea of long trousers and later shortened to pants as well. The etymological source of the word pantaloons, and therefore all pants, is Pantaloun (or Pantalone), a recurring comedic character in 16th-century Italian commedia dell’arte who famously wore red rights. Pantalone was a silly, avaricious old man who tended to get into humorous entanglements with women and servants who openly mocked him.

Another eponymous article of clothing is the fedora. Even though it’s largely associated with men’s fashion today, the fedora was first popularized by—and indirectly named after—a woman. If you’re a theater aficionado, you’ve probably heard of Sarah Bernhardt. She was a famous French stage actress in the late 1800s and early 1900s. She frequently played the female lead in popular French plays by the likes of Alexandre Dumas and Victor Hugo. And she was one of the first big-name actresses to act in motion pictures. But she was also well known for playing male roles, earning critical acclaim for her performance as Shakespeare’s Hamlet in 1899. She was a leading figure in the women’s rights movement, and was one of the first women to publicly wear a pantsuit. But back to the hat. In the 1880s, Bernhardt wore a hat in this style as she performed the role of the title character in the play Fédora by Victorien Sardou, which is about the Russian princess Fédora Romazov. Thanks to the popularity of Bernhardt’s performance, the hat was named after the character she played. Thereafter, around the turn of the century, it was adopted as a defining symbol of the women’s rights movement. It didn’t become a popular men’s fashion item until the fashion-forward Prince Edward started wearing one in 1924.

Sequoia trees are named after Sequoya, a Cherokee innovator who developed the Cherokee syllabary, the first Cherokee writing system. His work not only allowed Cherokee literacy to surge to nearly 100%, but also improved relations between the U.S. government and indigenous groups, and formed the basis for writing systems used in a total of 65 languages in North America, Africa and Asia. He also became a critical representative for the Cherokee people in a deeply contentious sociopolitical landscape.

This one sounds far too fanciful to be true, but I assure you it is an eponym: Sideburns are named after the resplendent and voluminous whiskers of Union Civil War General Ambrose E. Burnside. Starting around the 1870s, they were called burnsides, but shortly thereafter, the words were transposed. But the transposition of burnsides was—er, burned by etymologists, lexicographers and grammarians of the time. Literary legend Ambrose Bierce included sideburns on a “blacklist” of words to never use in his 1909 book Write it Right, instead insisting that burnsides is correct:

Sideburns for Burnsides. A form of whiskers named from a noted general of the civil war, Ambrose E. Burnside. It seems to be thought that the word side has something to do with it, and that as an adjective it should come first, according to our idiom.

In addition to being famously pedantic and a fellow Ambrose, Bierce was also a Union Civil War veteran so it makes sense that this malapropism would have rankled.

Embedded brands: Proprietary eponyms

Eponyms usually refer to people giving human names to things—like the Fahrenheit scale for temperature, which is named after Gabriel Daniel Fahrenheit—but eponyms can also come from brands.

You probably know that Kleenex and Band-Aid are genericized trademarks, but you might not know that they’re also proprietary eponyms, meaning ‘something named after its ownership’. (Proprietary means ‘owned’ or ‘pertaining to ownership’, from the Latin proprietarius ‘owner of property’.)

Some proprietary eponyms, like Kleenex and Band-Aid, have been genericized for so long that they are almost fully divorced from their original brand name usage. Here are a few others:

Bubble Wrap is owned by Sealed Air Corporation. The stuff was invented by two guys who were attempting to make three-dimensional plastic wallpaper. They failed, but they did make a useful (and entertaining) packing material.

Moxie, used generally from 1930, comes from the brand name of a bitter syrup first marketed as the medicine Moxie Nerve Food in 1876, then sold as a soft drink starting in 1884. The brand may be from a Native American Abenaki word for ‘dark water’ that appears in the names of several lakes and rivers in Maine.

Ping-Pong is currently owned by Parker Brothers, though it was first trademarked by the British manufacturer J. Jaques & Son Ltd in 1901. The game ping-pong, or table tennis, did exist before that, though, and the term may have been used before the trademark. Same with Breathalyzer, Jacuzzi, Rollerblade, Super Glue, Windbreaker, Zipper and even TV Dinner.

The word soccer is a twisted eponym—one named after an association rather than a commercial brand. Soccer is short for association because, in the 1800s, in British universities, the sport was differentiated from Rugby football by calling it Association football, after the International Football Association Board that governed the sport. It was originally called Assoc, but in British university slang you’d often add -er onto a syllable of a word, so rugby became rugger or footer, and Assoc football became soccer. So despite the British insistence that Americans call “football” by the wrong name, it was the Brits themselves who first gave it that name.

Gasoline is oddly somewhat eponymous and somewhat proprietary. It all has to do with John Cassell, an Englishman who sold refined petroleum for powering lamps. Based on his name, he called it Cazeline. You might think he used the Latin-derived chemical or elemental suffix -ine, to make it sound sciency, but nope: he used a Greek derived suffix -elene, from Greek elaia, ‘olive’.

He even had a fancy ad placed for it: “The Patent Cazeline Oil: safe, economical, and brilliant … possesses all the requisites which have so long been desired as a means of powerful artificial light.”

Shortly thereafter, an Irishman named John Boyd began selling counterfeit cazeline in his shop. Cassell found out and sent a nastygram insisting he desist. Boyd changed the spelling of his product to begin with a G: Gazeline. American refineries took the name gazeline and turned it into gasoline. Here’s where a little bit of chemistry logic does start to come in: , gas- because it combusts, -oli- from the Latin oleum meaning ‘oil’ (just like in petroleum), and -ine from that Latin elemental ending -ine/-ene.

The use of gasoline in the States and petrol or petroleum in most other places pretty much boils down to marketing. Petroleum is first recorded in the 1400s, first in reference to crude oil—or literally ‘rock oil’ (Latin petra ‘rock’ + oleum ‘oil’). The shorter petrol was also used as a word for crude oil as early as the 1580s. Petroleum was applied to the refined liquid used in internal combustion engines starting in the late 1850s, and petrol was first used to market refined petroleum as a solvent in the 1860s by British wholesalers, and then as a fuel shortly thereafter. Entirely coincidentally, the first British company to use the word petrol to describe its products was co-founded by a fellow inaptly named Eugene Carless. British refineries also used motor spirit as a generic name for automotive fuel and aviation spirit for aviation fuel.

If this kind of etymological spelunking delights you, you’ll find more from Jess Zafarris @jesszafarris on TikTok, @uselessetymology on Instagram, in the podcast Words Unravelled, and in the book Useless Etymology: Offbeat Word Origins for Curious Minds. Get it at the links below!

• Bookshop.org

• Amazon

• Barnes & Noble

It is a sunny day, and a man takes shelter from the heat in a street not far from the east bank of the river Nile. Between the houses, he can catch glimpses of that mighty river, full of ships travelling up and down. He, a humble merchant, had himself arrived by one such boat, travelling upriver to visit this city of Thebes, which is prosperous and bustling.

The crowds increase as he walks towards the temple of the god Amun, where he is to meet with a potential client. The temple is a grand site with royal patronage, having been constructed on the orders of pharaoh Senusret I himself. The current ruler is Amenemhat III, and Egypt is enjoying a state of unity and stability that historians of the future will call its Middle Kingdom, 1,800 years before Christ—not that our merchant thinks of his day and age as being ‘before’ anyone or anything.

Looking up at the great temples, statues and obelisks, it is clear to the visitor that these Egyptians are a highly literate people; the buildings are covered with hieroglyphs, many painted and carved with painstaking detail.

These are official inscriptions intended to impress and to endure, existing somewhere between writing and art. However, when he had stopped at a tavern for a drink, the visitor had also seen people writing in a more casual style, composing messages and lists with a brush, ink and papyrus.

The local language is Egyptian, which our visitor to Thebes understands and speaks pretty well, but the written language remains tricky. Having been around for over a thousand years, this system of pictures and symbols is sophisticated and complex. He knows that some hieroglyphs simply stand for a particular object or idea, the thing that they appear to represent. Others stand for a sound, or two sounds together, or even three. Others are used instead to tell you something about the type of the thing that precedes them—whether it is a man or a woman, a god, a building, a drink, a boat. To complicate matters, in principle, a given hieroglyph can be put to all three of these uses in different contexts and by different scribes. Moreover, the number of hieroglyphs that the Egyptians use seems beyond count; future experts will reckon around 700. Try as he might, Egyptian writing still makes the visitor’s head spin.

He himself is not Egyptian. His origins lie far to the north-east, beyond the kingdom of the Nile, and his mother tongue is very different. What he and his people speak is one dialect of an expanding family of languages that will one day include internationally famous members, such as Arabic and Hebrew. His people can be found throughout Egypt, though especially concentrated in Lower Egypt, where the Nile spills out into the Mediterranean, and where he has set up a business trading papyrus. Many of his people have been employed by the Egyptians as soldiers and servants. Though he lives and works among the Egyptians, the visitor is a linguistic outsider to this complex world of hieroglyphs and the Egyptian language that they write down. Their symbols and speech share an ancient association, while his own language is a newcomer on the scene. If he wants to write in his language, not theirs, he and his people must either invent their own writing system from scratch, or adapt the Egyptians’ characters. They are opting for the latter. Their adaptation has already begun, and with it, the first shoot of a new tree of writing is sprouting.

With his business in Thebes concluded, our visitor begins his return journey northwards. However, this time, rather than sailing along the Nile around the Qena Bend, he chooses to cut a corner and take the Farshût road over land. It’s a sensible decision; it’s quicker, and the waters of the Nile can be difficult and unkind. The route takes him away though from the lush, green landscape of the Nile valley and off into wilder land. Halfway through this shortcut, the man stops in a valley that will be known in Arabic as Wadi el-Hol. There he chats to his travelling companions, eats and drinks with them and takes a break from the journey. Wadi el-Hol is a popular spot; soldiers, pilgrims and merchants are all passing through, and many people have left behind messages in the surrounding limestone rock.

To pass the time, the man decides to add to the lithic literature. Rather than in the Egyptians’ formal hieroglyphs, or their more casual ‘hieratic’ script, he carves from right to left 16 hieroglyph-like symbols that work according to a much newer system. It is something his people have been developing recently, and it is not the first example at Wadi el-Hol; he spots a previous message by a fellow compatriot, written instead from top to bottom.

Compared with the complexities of hieroglyphs, the principle behind the man’s system is simple: every image looks like a thing, and it stands for the first sound of the word for that thing in his language. For example, he carves an ox’s head, with eyes and horns, facing to the left:

For him, an ox is a ʾalp. The word ʾalp begins with a glottal stop. This is a sound produced when the glottis, an opening between the vocal folds in our throats, quickly closes then opens. In doing so, it briefly stops any exiting airflow. It’s a very common sound in speech, made when English speakers for example will say uh-oh. The traveller’s revolutionary principle says that, because the word for an ox begins with a glottal stop, the symbol of an ox now stands for that sound. No more ambiguities about whether the symbol refers to a thing, an idea, a category, a sound, two sounds, three sounds—just one symbol, for one sound.

His inscription completed and his companions well rested, the man continues on his way north, and so vanishes from history and our imagination. In truth, we know nothing about the individuals who carved the two inscriptions in the new script at Wadi el-Hol, nothing certain about their origins, languages, profession, status or gender. All we have is what they left behind there, but it’s enough to know that they were part of something big. Over the course of almost 4,000 years, that head of cattle would develop beyond recognition and travel far, all the way to this very text. Its horns would become a letter’s legs, and its mouth would come to point upwards.

It would become our letter A.

This short narrative is the opening section of ‘Chapter A’ in Why Q Needs U, a new book about the history of the alphabet and how we today spell the English language. It sets out to answer, in an accessible and entertaining style, two broad questions. First, where did we get the alphabet from? Second, why does English use it in such a strange way?

To tackle the first topic, Why Q Needs U starts off in Ancient Egypt. It assembles the story of the alphabet’s birth from ancient fragments of text and modern academic scholarship. The new principle mentioned in that story is a milestone in human history; it was a deliberate departure from the Egyptian tradition of writing, and so launched a new tradition that has endured all the way up to the very letters that your eyes are reading right now. We humans as a species come out well from this story, because it shines a light on our shared linguistic ingenuity. The evidence suggests that the alphabet’s ancient innovators were humble people, whose status as ethno-linguistic outsiders in the Kingdom of Egypt motivated their repurposing of old symbols. They understood the value of writing, what it offered their speech. They also understood what in the new medium their language needed to have represented, if it was to be written down and read successfully.

Clearly, though, from those first sparks of alphabetic fire, our letters have a long way to go—our capital letter A definitely doesn’t look like an ox’s head anymore! To tell the full tale, Why Q Needs U spends time with each character in the cast of this long story. For instance, Chapter A goes on to track the alphabet’s anti-clockwise journey around the Mediterranean Sea, until it lands in the hands of the Greeks. It was in Greece that A was reassigned to represent a vowel sound, no longer a glottal stop. It was also in Greece that a left-to-right direction of writing first won out, having fought off the older right-to-left direction that Arabic and Hebrew still go by today. This shift, and its possible psychological origins, are discussed in Chapter B.

Indeed, every letter is a chance to explore some aspect of the international collaborative project that is the alphabet. Some of the chapters of Why Q Needs U spend time with certain peoples who have played a part in how we now spell English; Chapter C exalts the Etruscans, while Chapter K acknowledges the linguistic imports of the Vikings. Other chapters, like Chapters G, J and P, look at how specific individuals can leave a lasting legacy, if the societal conditions are right. Others, like Chapters O and V, look at the multi-faceted shape of the English language, with all its different dialects and disagreements in sound. Letters like E, H and N invite us in their chapters to step back and admire ingenious systems quietly at work in how we spell, such as the matter of doubled consonant letters that assist in the spelling of the vowel that they follow.

Why Q Needs U is a linguistic love letter to language, speech and spelling, and an attempt to capture in one book the world of human experience behind our letters, from A to Z.

Why Q Needs U is written by Dr Danny Bate, and is published by Blink, an imprint of Bonnier Books UK. It is available to buy from October 2nd and can be ordered here.

If you enjoy this newsletter and want to support Linguistic Discovery’s mission to educate the world about the science and diversity of language, consider becoming a supporter! You’ll get bonus articles and early access to articles and chapters of my book!

Subscribe now

So imagine my surprise earlier this week when, as I was reading the latest issue of Babel: The Language Magazine, I learned that the most frequent consonant in spoken English isn’t /t/ at all—it’s /n/!

ℹ️ How to read linguistics

In linguistics, ⟨angle brackets⟩ refer to the written (orthographic) representation of a word or letter, while /slashes/ refer to the spoken (phonemic) representation of a word or sound. For example, the word house is written as ⟨house⟩ but pronounced as /haʊs/ in most dialects of American English. You can learn more about how to read linguistics on this page.

What’s going on here? Why the discrepancy?

The fact that the relative frequency of ⟨t⟩ and /t/ differ immediately tells us something interesting: either a) many of the written ⟨t⟩s in English are not pronounced as /t/ (or are silent), or b) in speech we use more words with /n/ and fewer words with /t/ than we do in writing. So which is it?

This is a special bonus issue for paying subscribers. In the rest of this article, we’ll look at the frequencies of sounds in English and discover a fascinating pattern about how they compare to sounds in the rest of the world’s languages. If you’d like to read the rest of the article, become a supporter here:

Subscribe now

Read more

These two little mouth-watering words—pumpkin spice—may not seem like anything special as far as the lexicon of English goes, but if we dig a little deeper, we’ll see that these two words can actually teach us a great amount about language change, language ideology, and Indigenous vs. Western perspectives on history.

In this 20-minute talk, I take you on an etymological adventure through the winding history of these two words—and the people who shaped them. Happy Fall!

This talk was originally presented at Nerd Nite Edmonton in 2022, and a modified version of it at Nerd Nite Chicago in 2023.

If you’re enjoying this newsletter, consider becoming a subscriber! Issues include deep dives into how language works, language profiles, explainers of terms/concepts in linguistics, the latest news and research in linguistics, reviews, and more!

Paid subscribers also get bonus articles and early access to articles and chapters of my book!

Subscribe now