Skip to main content
Social Sci LibreTexts

9.4: Understanding word combinations

  • Page ID
    • Catherine Anderson, Bronwyn Bjorkman, Derek Denis, Julianne Doner, Margaret Grant, Nathan Sanders, and Ai Taniguchi
    • eCampusOntario

    \( \newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \)

    \( \newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}} \)

    \( \newcommand{\id}{\mathrm{id}}\) \( \newcommand{\Span}{\mathrm{span}}\)

    ( \newcommand{\kernel}{\mathrm{null}\,}\) \( \newcommand{\range}{\mathrm{range}\,}\)

    \( \newcommand{\RealPart}{\mathrm{Re}}\) \( \newcommand{\ImaginaryPart}{\mathrm{Im}}\)

    \( \newcommand{\Argument}{\mathrm{Arg}}\) \( \newcommand{\norm}[1]{\| #1 \|}\)

    \( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\)

    \( \newcommand{\Span}{\mathrm{span}}\)

    \( \newcommand{\id}{\mathrm{id}}\)

    \( \newcommand{\Span}{\mathrm{span}}\)

    \( \newcommand{\kernel}{\mathrm{null}\,}\)

    \( \newcommand{\range}{\mathrm{range}\,}\)

    \( \newcommand{\RealPart}{\mathrm{Re}}\)

    \( \newcommand{\ImaginaryPart}{\mathrm{Im}}\)

    \( \newcommand{\Argument}{\mathrm{Arg}}\)

    \( \newcommand{\norm}[1]{\| #1 \|}\)

    \( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\)

    \( \newcommand{\Span}{\mathrm{span}}\) \( \newcommand{\AA}{\unicode[.8,0]{x212B}}\)

    \( \newcommand{\vectorA}[1]{\vec{#1}}      % arrow\)

    \( \newcommand{\vectorAt}[1]{\vec{\text{#1}}}      % arrow\)

    \( \newcommand{\vectorB}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \)

    \( \newcommand{\vectorC}[1]{\textbf{#1}} \)

    \( \newcommand{\vectorD}[1]{\overrightarrow{#1}} \)

    \( \newcommand{\vectorDt}[1]{\overrightarrow{\text{#1}}} \)

    \( \newcommand{\vectE}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash{\mathbf {#1}}}} \)

    \( \newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \)

    \( \newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}} \)

    One or more interactive elements has been excluded from this version of the text. You can view them online here:

    We’ve seen that babies start to learn the phonology of their first language very early — as soon as they’re born and maybe even earlier than that! And they’re starting to learn some lexical semantics before their first birthday. What about syntax? What do babies and young children know about the syntax components of their grammar? And how can we tell? We know that children start to speak or sign their first word around twelve months, and they start to combine two or more words around eighteen to twenty-four months. But if you’ve ever spent any time with young children, you know that they can understand a lot more than they can say! Their comprehension is often much more advanced than their ability to produce spoken or signed words. But comprehension is much harder to observe. How can we tell what babies and toddlers understand about language?

    How to be a linguist: Observing Preferential Looking

    One simple technique is called preferential looking. In this kind of experiment, researchers use a large screen or television. The baby or toddler sits strapped into a booster seat, facing the screen. The screen is split so that two different pictures appear, one on each side of the screen. While the pictures or are on the screen, a recorded voice speaks a sentence, maybe something like, “Look! Can you find the bear?” The idea is that if the baby understands the word bear, they’re going to look at the picture of the bear, not the picture of the bus. The researchers keep track of the direction of the baby’s head-turn, or they use eye-tracking to measure the baby’s eye movements. This kind of experiment has shown that babies pretty reliably look at the named object by about ten months, and even as young as six months, they’re looking at the named object more often than chance would predict. So at the age of six months, babies are already beginning to link up word forms with their meanings.

    Line drawing. Television screen with vertical line down the center. On the left is a pink teddy bear. On the right is an orange school bus. In front of the television is an infant strapped into a car seat.
    Figure 11.6. Preferential Looking.

    We can use this technique to figure out what kids know about syntax if we use the split screen to display two similar scenes instead of pictures of a single items. I’ve just put a simple picture here, but researchers often use short videos. The two scenes involve the same participants, but in different configurations. In the images here, the one on the left shows the adult chasing the kid, while on the right the kid is chasing the adult. The recorded sentence that plays, “Look! The kid is chasing the woman!” matches only one of the scenes. So if children look towards the matching scene, does that mean they know something about syntax? Or are they just paying attention to the word meanings?

    Line drawing. Television screen with vertical line down the center. On the left is an adult chasing a small child. On the right is a small child chasing an adult.
    Figure 11.7. Preferential Looking.

    Do you remember the idea of compositionality? It says that syntax matters for sentence meanings — the meaning of a sentence comes not just from adding together the meanings of the words, but also from the way those words are combined, that is, from the syntax. As adults, we know that the sentence, “the kid is chasing the woman” can only mean this one, the one on the right. But if kids were understanding the sentence just by paying attention to the meanings of the words, well both scenes would be plausible: after all, both scenes show a kid, both scenes show a chasing event, and both scenes show a woman. It turns out, though, that children as young as 15 months, just a little over one year, look more often towards the correct image or video (Hirsh-Pasek & Golinkoff, 1996). That suggests that they’re not just adding together the meanings of the words in the sentence, but that they’re also sensitive to the way those words are combined, to the syntax, because that’s what distinguishes the kid chasing the woman from the woman chasing the kid. So kids who are just a little older than twelve months are already sensitive to syntactic constituent structure and its relation to meaning.

    What else do young children know about syntax? Some experiments with young children don’t involve screens, but actual toy items. In this experiment (Booth & Waxman, 2003), researchers introduced 14-month-old children to novel, made-up words, that the children would not already be familiar with. The experimenter presented small toys to the child and said, “These are blickets. This one is a blicket and this one is a blicket.” So the child has a couple examples of what a blicket is. Then the experimenter presents two new toys, one of which is from the same category as the earlier ones — in this case, the category of animals — but in a different colour, and the other of which is the same colour, but from a different category.

    Icon drawings of pink horse and pink pig.
    Figure 11.8a. Blickets.
    Icon drawings of green lamb and pink car.
    Figure 11.8b. More blickets.

    When the experimenter says, “Can you give me the blicket?“, if the child reaches for the new toy of the same category, that tells us they’ve figured that blicket means animal. But if they reach for the same colour, they’ve concluded that blicket means pink thing. The 14-month-olds reached for the new animal of a different colour more often than they reached for the toy that matched in colour. But the pattern was reversed for children who heard a different syntactic frame. If the toys had been introduced with the new word in an adjective position, “These are blickish. This one is blickish and this one is blickish,” then when the experimenter asked, “Can you give me the blickish one?”, the children were much more likely to choose the one that matched in colour. These results indicate that when one-year-olds hear a new word in a noun position, they conclude that it has a noun-like meaning, and refers to a thing or a category of things. But if it’s in adjective position, then its meaning is probably something more like an attribute or property. In short, one-year-olds seem to be sensitive to the differences between syntactic categories.

    There’s so much learning happening in that second year. Kids are learning new words very rapidly, and learning how words pattern in the morphological and syntactic behaviour. In fact, by the time they turn two, kids are sensitive to verb arguments and subcategories too! In another split-screen experiment (Arunachalam & Waxman, 2010), when experimenters presented the novel verb mooping in a transitive frame, like “The lady is mooping my brother,” then two-year-olds looked more often to the scene where one participant is doing something to the other, like pushing. But when the novel verb appeared in an intransitive frame, like “The lady and my brother are mooping,” then the children looked more often to the scene where the two participants are doing the same activity together, like waving.

    Line drawing. Television screen with vertical line down the center. On the left is a woman pushing a man in a wheelchair. On the right are a woman and a man, both waving and facing forward.
    Figure 11.9. Possible meanings for ‘mooping’

    This suggests that, by age two, children are sensitive not only to syntactic categories, but also to subcategories! So to go back to the question we started with, “What do kids know about syntax?” It turns out the answer is that, even before children start combining words to make phrases in their own speech or signing, they already know quite a lot about how words combine in the grammar.

    Check your understanding

    Query \(\PageIndex{1}\)


    Arunachalam, S., & Waxman, S. R. (2010). Meaning from syntax: Evidence from 2-year-olds. Cognition, 114(3), 442–446.

    Booth, A. E., & Waxman, S. R. (2003). Mapping Words to the World in Infancy: Infants’ Expectations for Count Nouns and Adjectives. Journal of Cognition and Development, 4(3), 357–381.

    Hirsh-Pasek, K., & Golinkoff, R. M. (1996). The origins of grammar: Evidence from early language comprehension. MIT Press.

    11.7: Syntax in early utterances

    One or more interactive elements has been excluded from this version of the text. You can view them online here:

    Beginning at about age 1;6 to 2;0, most children start to combine words into phrases of two or even three words. At this early stage, their utterances are usually telegraphic[1], containing mostly content words like nouns and verbs, with few function words and few inflectional morphemes. It can be hard to figure out whether these short utterances have any syntactic structure to them. Let’s look at some of the evidence.

    Syntactic Categories

    Even children’s first short sentences give us evidence that their mental grammar has already grouped words into syntactic categories. In a couple of large corpus studies of children’s utterances (Cazden, 1968; Maratsos, 1982), researchers looked for examples where the children had used a word in a way that violated its syntactic category, for example, if they had used a noun as if it were a verb. In over 200 hours of recorded speech, the researchers found almost no examples of children producing syntactic category errors. Here are a few of the examples they found:

    • Mummy trousers me.
    • I’m crackering my soup.
    • I want to comfortable you.
    Four saltine crackers.
    Figure 11.10. Crackers.

    I suppose these sentences are technically “errors”. The word cracker is a noun in English but the child in this example has used it as a verb. But the grammar of English often allows verbs to be derived from nouns. For example, salt is a noun but there’s also a verb form to salt, which means to sprinkle salt onto something. So the child who uses the same process to derive the verb crackering from the noun cracker is maybe not really making a mistake, but is actually using their grammar generatively.

    So the evidence from these and other large corpus studies shows that syntactic categories are quite robust in English-acquiring children’s grammars, even in their early utterances. They rarely use words in positions that would be ungrammatical in adult English, and their few productions that aren’t adult-like give evidence of generativity.

    Asking Questions

    Corpus data (StromswoId, 1995) show that English-acquiring children start to ask questions not long after they begin combining words in their utterances. Simple questions like, “Where kitty?” and “Who crying?” are common in children’s speech, with the wh-word moved to the beginning of the sentence just like in adult English. Why questions show up a little later, but as any parent of a preschooler will tell you, the why stage feels like it lasts forever!

    This kind of question is a little harder to observe in sign-acquiring children’s productions because there are several different grammatical ways to form them, but a careful analysis found that children produced wh-questions as young as age 1;7 in ASL and in Brazilian Sign Language (LSB) (Lillo-Martin & de Quadros, 2006).

    For simple structures that are frequent in children’s language, corpora are valuable for providing a large volume of data. But what about more complex structures like embedded clauses and questions? You could record an awful lot of hours of kid speech without ever capturing an example of a question with an embedded clause, but that doesn’t necessarily mean that these structures don’t exist in kids’ mental grammar.

    How to be a linguist: Elicitation

    Researchers sometimes use puppets and toys to elicit sentences of interest from children, that is, to encourage them to produce a particular kind of sentence. Often this involves asking the child to ask the puppet something. For example, in an experiment about embedded wh-questions, Rosalind Thornton had a bear puppet wearing a blindfold. While the bear was blindfolded, Thornton asked the child to choose small toys to hide inside a set of small boxes. Once each box contained one toy, she removed the bear’s blindfold then asked the child to ask the bear to guess what was in each box. There are many different ways that children might form this question, but this elicitation context at least makes it likelier that they’ll produce the wh-question of interest, whereas in a free play situation they’re relatively unlikely to produce these complex questions.

    Crain and Thornton (1991) were interested in whether children’s grammars included questions where the wh-word originates was from an embedded clause, like these ones:

    • Who do you think ___ will win the election?
    • What did Lexi say she wanted ___ for lunch?

    In these sentences, notice that the wh-word at the beginning of the sentence logically originates in a position inside the embedded clause: this becomes obvious if we compare the declarative versions of the sentences:

    • You think someone will win the election.
    • Lexi said she wanted something for lunch.

    (If you need a refresher on how wh-questions are formed, look back at the Syntax chapter!)

    Crain & Thornton used elicitation to try to get preschoolers aged three to five years to produce complex questions like these. The kids rarely produced adult-like examples questions, but their questions still revealed something interesting about their mental grammar. Some of the children in this study asked the bear puppet questions like these:

    • Who do you think who wants to hug Grover?
    • What do you think what’s in that box?

    While these preschoolers hadn’t really mastered this complex structure, their utterances suggest that their mental grammar contains a structure that parallels that of adult grammar. They’ve moved the wh-word to the beginning of the question the way adults do, and they’re also pronouncing that wh-word in its original position in the embedded clause.

    Check your understanding

    Query \(\PageIndex{1}\)


    Cazden, C. B. (1968). The acquisition of noun and verb inflections. Child Development, 39(2), 433–448.

    Crain, S., & Thornton, R. (1991). Recharting the Course of Language Acquisition: Studies in Elicited Production. In N. A. Krasnegor, D. M. Rumbaugh, R. L. Schiefelbusch, & M. Studdert-Kennedy (Eds.), Biological and behavioral determinants of language development (pp. 321–337). Psychology Press.

    Lillo-Martin, D., & de Quadros, R. M. (2006). The Position of Early WH-Elements in American Sign Language and Brazilian Sign Language. In K. U. Deen, J. Nomura, B. Schulz, & B. Schwartz (Eds.), The Proceedings of the Inaugural Conference on Generative Approaches to Language Acquisition (pp. 195–203).

    Maratsos, M. (1982). The child’s construction of grammatical categories. In E. Wanner & L. R. Gleitman (Eds.), Language Acquisition: The state of the art (pp. 240–266). Cambridge University Press.

    StromswoId, K. (1995). The Acquisition of Subject and Object Wh-Questions. Language Acquisition, 4(1 & 2), 5–48.

    1. Before the days of email, text messages, long-distance phone calls, if you needed to send an urgent message over a long distance you could send a telegram. You had to pay per word in the message, so you'd use as few words as possible.

    9.4: Child Language Acquisition Linguistics

    Child Language Acquisition Linguistics, from Sarah Harmon

    Video Script

    Let's break down those stages of child language acquisition a little bit more, and specifically the patterns that we observed of how children learn their first language or languages. I keep phrasing it this way, because not all children grow up in a monolingual society; many children grow up in a at least a bilingual society, if not a multilingual society. What we can say is that, regardless of how many linguistic inputs they are receiving with respect to individual languages, we see the following patterns.


    With phonology, there is a progression on the sounds that children tend to learn first. Whatever the language or languages that they're learning, the progression tends to be the same. When we're talking about manner of articulation, nasal and glides are first, because there's the least amount of control needed to produce those sounds. You don't have to move your articulators so much. Then come stops, then liquids, then fricatives, then affricates.

    Let's think about the very young children that you are around. If you think about how they talk at their youngest stages, the first words that they come up with all have lots of nasals, some glides and some stops. If it's a language that's heavy on the fricatives and affricates, those sounds aren’t produced very early on. Think of the French-German example that we saw in very first section of this chapter; the child having the worst time with not one, but two different consonant clusters. One of the consonant clusters was an actual affricate, the [pf]; the [kn] is just a consonant cluster. In either case, the child couldn't do it, and simplified everything; he used a basic nasal and a basic stop in their places.

    The same is true with place of articulation: labials and velars are the early sounds, and then the sounds in the middle come later, with more precision of the articulators. Palatals are universally the last sounds that the child acquires and articulate well.

    I really want to focus on this articulation aspect, that it's the performance that comes late, not the competence. They understand and recognize those sounds, even as early as six months of age. However, producing the sounds that is a different story. The distinction of voiced versus voiceless is an articulation that comes pretty early. Children produce more voiced sounds to start off, and then voiceless sounds come a little bit later.

    Think about early words that most children acquire, frequently having to do with mom or dad, because those are the folks that they're mostly around. Those sounds tend to have a lot of nasals and a lot of labials. There early errors in pronunciation, but they're always rule governed. Remember what we said about language change: it's always rule governed. Acquisition is no different.

    There is a very strong tendency for children to simplify consonant clusters; we saw this with the French-German example, and you know plenty of other examples. The voicing of final consonants, not just with English, but in general, is also a strong tendency, and it makes sense. When we are trying to articulate the end of a lexicon, getting that pronunciation at the end is very crucial. The more consonants you have in the coda of the last syllable, the harder it gets to articulate them, and young children tend to chop off the end of the lexicon, and especially the last coda. There's always this consistent voicing of initial consonants and consonant harmony; again, creating patterns and using them in the production of the language.

    We see these patterns and tendencies in language after language, in child after child; we see this both in monolingual and multilingual situations. For example, if you have a child who grows up in a bilingual house with bilingual parents, they're going to go through this process with both languages. Remember, competence before performance; we know these children, even at six months, are starting to differentiate the sounds of the language or languages around them. By one year of age, they understand the sounds of whatever languages are around them. They may not be able to produce the sounds correctly, but they understand and recognize them.


    With respect to morphology, there are some pretty basic changes that we observe. Overgeneralization—we've heard that one before. An example is the foot, foots or feet, feets, along those lines. We see this in pluralization and in all aspects of language; anytime there's any kind of irregularity—and every language has irregularity—then there's going to be some overgeneralization always.

    We're still trying to understand why it is the case that children pick up and produce standard pluralization very quickly, whatever language. This is also true of other basic inflection and morpho-syntactic structures. Think back to syntax with respect to subject-verb agreement and the Linear Agreement Rule; that's something that's learned very quickly. The same is true with noun-adjective agreement; if the noun has a certain gender and/or number classification, and it gets inflected as well on the adjective, then that is learned very quickly. The basic derivational rules are also learned very early on; that's why children are able to overgeneralize, because they have learned the basic derivational rules.

    One of my favorite activities to do with young children, even as young as two or three, once they start putting those two- to five-word sentences together, I have them tell me stories. I love hearing the combinations that they put together with respect to lexicon; they come up with the best ones.

    Syntax, Semantics and Pragmatics

    When we talk about syntax, semantics and pragmatics, let's be clear: syntax is kind of on one end of the learning spectrum, and semantics and pragmatics is the other. Basic syntax is learned pretty early on; as the child acquires more is able to produce more. But semantic elements are harder to measure. It's difficult to measure, outside of asking the child, “What does this?”, “What do you call this thing?” We can do that kind of lexical analysis as to how many words the child can produce, as well as how many they can understand. Studies where you have a child in a lab, and the researcher holds this thing (🍼) up using the incorrect name for the item; the babies and toddlers know it's not the correct name. They just may not be able to say ‘bottle’ or [baba]. For example, if I say, “it's zork,” they're going to say, “No, that's not a zork. It’s a bottle.”

    Pragmatic elements are even more difficult to measure. That's true just because of the development of a child's brain; these concepts of politeness, of reading between the lines, of implicature and presupposition, the Maxims of the Cooperation Principles, these are things that a child does not acquire until well into childhood. Most commonly, children don’t understand these concepts well until between the years of seven or six, perhaps not until 9-12, and not until teenage years in some cases. Much of this is culturally simulated, as you would expect, and that's why young children don't understand this. The phrase, “Honesty out of the mouths of babes,” or something along those lines, is seemingly ubiquitous because children don't know when not to say something or how to rephrase it.

    That all being said, syntax-semantic relationships, like auxiliaries (like be, have, will or would) and modals (like could, should, must, or do) are learned not too early, but not too late; usually somewhere in that three- to four-year range. The more stimulus the babies and toddlers are around, the quicker they pick things up, but also the more complex they become earlier on.

    That being said, that doesn't mean all of their input has to be super complex for a number of years. It honestly seems to be cyclical; there have been many times, where the prevailing theory is that you have to surround your children with high level language from the earliest of times. That's not necessarily going to ensure that your child understands, let alone produces, those terms. It is important to understand, though, that just because they hear Cookie Monster all the time, that doesn't mean they're going to talk like Cookie Monster.

    Frank Oz and Cookie Monster's lack of influence over child language acquisition

    This is something that Frank Oz, who created the character of Cookie Monster, as well as voiced him for many decades. I love this; I guess this has happened throughout his career, that the parents were concerned about Cookie Monster's vocabulary and syntax. Somebody asked him if he thought that Cookie’s way of speaking could corrupt the children that were watching Sesame Street. He said that he didn't foresee a child growing up to become a lawyer and saying, “Me want to represent you.”

    I think Frank is on to something.

    This page titled 9.4: Understanding word combinations is shared under a CC BY-NC-SA 4.0 license and was authored, remixed, and/or curated by Catherine Anderson, Bronwyn Bjorkman, Derek Denis, Julianne Doner, Margaret Grant, Nathan Sanders, and Ai Taniguchi (eCampusOntario) via source content that was edited to the style and standards of the LibreTexts platform; a detailed edit history is available upon request.