Observing more languages

The typical study in 1974 described a small number of children, and those children were, for the most part, white English-learners from middle- to upper-class homes. One significant change that has taken place is that we now study language learning in children from a wide range of families representing the demographic diversity within the United States (e.g. Pan, Rowe, Singer & Snow, Reference Pan, Rowe, Singer and Snow2005), beyond the United States and, importantly, beyond English. For example, during the 1970s, 80% of data-oriented articles in the Journal of Child Language focused exclusively on English. By 1990, the percentage had dropped to 57% (Slobin, Reference Slobin and Slobin1992) and some aspect of language acquisition had been explored in thirty-six different languages.

What has studying children who are learning different languages taught us? Overall, we find that children graciously accept the differences in the languages to which they are exposed. They learn the particular properties that their language presents, and they do so from the earliest stages. But there are times when children veer from the input they receive, and these times are interesting because they provide insight into the preferences children bring to language learning, preferences with respect to both meanings and forms.

For example, regardless of the language they are learning, children begin by grammatically marking roles in highly transitive activities, that is, in events where an agent brings about a physical and perceptible change of state in a patient, either by means of direct body contact or an instrument (Slobin, Reference Slobin and Slobin1985). In Russian, accusative inflections must be placed on all words that fill the direct object slot, regardless of the type of event conveyed. But children first use the accusative inflection only in sentences describing manipulative physical actions (giving, putting, throwing; Gvozdev, 1949, as described in Slobin, Reference Slobin and Slobin1985). Similarly, in Kaluli, an ergative language in which an inflection must be placed on the agent in all transitive events, children first use the ergative inflection only in sentences describing concrete, manipulative actions (Schieffelin, Reference Schieffelin and Slobin1985). These highly transitive activities constitute a “central semantic organizing point for grammatical marking” (Slobin, Reference Slobin and Slobin1985, p. 7) – a starting point for the language learning child and a potential resilient property of language.

In addition to meanings that appear to be privileged in the early stages of child language, there are forms that children find easy to incorporate into their language. For example, children combine words into strings that follow particular ordering patterns, even when the language they are learning allows relatively free word order (Bates, Reference Bates1976; MacWhinney, Reference MacWhinney1977; Slobin, Reference Slobin, Smith and Miller1966). We can see the preferences children bring to language learning when we look at how they take in, and change, the languages of the world.

Observing more children from a wider range of families

Because transcribing and analyzing child talk is so labor-intensive, each individual language acquisition study in 1974 typically focused on a small number of children interacting with their primary caregiver at home. Advances in computer technology have made it possible for researchers to share their transcripts of child talk via the computerized Child Language Data Exchange System (CHILDES; MacWhinney, Reference MacWhinney1995). A single researcher can now call upon data collected from spontaneous interactions in naturally occurring situations across a wide range of languages, and thus test the robustness of descriptions based on a small sample.

In addition to increasing numbers by compiling data from independently conducted case studies, there now exist several studies containing large numbers of children followed longitudinally (e.g. Hart & Risely, Reference Hart and Risley1995; Pan et al., Reference Pan, Rowe, Singer and Snow2005). For example, our team at the University of Chicago has studied sixty-four typically developing children, chosen to reflect the demographic variability of the Chicago area (Goldin-Meadow, Levine, Hedges, Huttenlocher, Raudenbush & Small, Reference Goldin-Meadow, Levine, Hedges, Huttenlocher, Raudenbush and Small2014). We took videotapes of each child at home with a primary caregiver for 90 minutes every 4 months between the ages of 1;2 and 4;10, and continued to observe the children through 10 years, giving them tests of language, reading, math, higher-order thinking, etc. (future observations are planned through age fifteen). These data have thus far been used to explore variability in children's talk early in development (e.g. Vasilyeva, Waterfall & Huttenlocher, Reference Vasilyeva, Waterfall and Huttenlocher2008), and the relation between this early talk and later cognitive skills (Pruden, Levine & Huttenlocher, Reference Pruden, Levine and Huttenlocher2011; Levine, Suriyakham, Rowe, Huttenlocher & Gunderson, Reference Levine, Suriyakham, Rowe, Huttenlocher and Gunderson2010).

Widening the lens on our observations

In addition to widening the profile of who we study, we have also widened what we study. Recent studies of language learning often focus not only on what children say with their mouths, but also on what they do with their hands, that is, on their gestures. Children use gestures even before they begin to use words (Bates, Reference Bates1976), and gesture continues to be part of a child's communicative repertoire after the onset of speech, serving to extend the child's communicative range (e.g. a child says gimme while pointing at a cracker; gesture makes it clear what the object of gimme is).

But there is variability in gesture use across children, and it turns out that this variability can predict differences in the onset of linguistic milestones. For example, the range of different meanings children convey in gesture at 1;2 predicts the size of their spoken vocabulary at 4;6 (Rowe & Goldin-Meadow, Reference Rowe and Goldin-Meadow2009a). As another example, children vary in the age at which they produce combinations in which speech conveys one idea and gesture another (e.g. nap + point at bird) and this variability predicts the age at which children produce their first two-word utterance (bird nap; Goldin-Meadow & Butcher, Reference Goldin-Meadow, Butcher and Kita2003; Iverson & Goldin-Meadow, Reference Iverson and Goldin-Meadow2005). Once two-word speech has begun, the types of gesture+speech combination children produce continue to predict the onset of different types of multiword combinations (e.g. predicate+predicate: Özçaliskan & Goldin-Meadow, Reference Özçalişkan and Goldin-Meadow2005; determiner+noun: Cartmill, Hunsicker & Goldin-Meadow, Reference Cartmill, Hunsicker and Goldin-Meadow2014).

Why does early gesture forecast later language? Early gesture might be an index of global communicative skill. For example, children who convey a large number of different meanings in their early gestures might be generally good language learners and/or have high levels of intelligence. If so, not only should these children have large vocabularies later in development, but their sentences ought to be relatively complex as well. The data do not support this hypothesis. Instead, we find that particular types of early gestures are specifically related to particular aspects of later spoken language use (Rowe & Goldin-Meadow, Reference Rowe and Goldin-Meadow2009b). More specifically, controlling for early vocabulary, number of different meanings conveyed in gesture early in development predicts the size of the comprehension vocabulary several years later, but number of gesture+speech combinations produced early in development does not; in contrast, controlling for early syntax, number of gesture+speech combinations produced early in development predicts syntactic complexity several years later, but number of different meanings conveyed in gesture early in development does not. The selectivity with which gesture predicts different linguistic skills suggests that the gestures are reflecting not just general intelligence or overall language learning ability, but rather skills specific to learning vocabulary or to learning syntax.

The fact that children express sentence-like ideas in a gesture+speech combination several months before they express these ideas in speech alone makes it clear that they have an understanding of these notions before they are able to express them in speech, thus eliminating one frequently held explanation for the slow acquisition of certain structures – the cognitive explanation, that is, that children do not express a given structure because they lack an understanding of the notion underlying the structure. Widening our lens to include gesture thus allows insight into when cognition does, and does not, shape the course of language learning.

In addition to serving as a window into the earliest steps a language learner takes, gesture has the potential to play a role in getting the learner to take those steps. For example, LeBarton, Raudenbush, and Goldin-Meadow (Reference LeBarton, Raudenbush and Goldin-Meadow2014) have recently found that encouraging children aged 1;5 to gesture during a book-reading task not only increases the child's gesture rate, but also increases the size of the child's spoken vocabulary. Gesture can be part of the mechanism of change, another area in which advances have been made since 1974.

Variations in language learners

Children who are deaf and born to deaf signing parents are exposed to a language system, but one that differs dramatically from the typical. It is language by hand and eye, rather than by mouth and ear. Does language learning take a different course when it is conducted in a different modality? The answer is that, for the most part, it does not – children learn language in the manual modality as easily and on the same timetable as children learning language in the oral modality (Newport & Meier, Reference Newport, Meier and Slobin1985). Similarly, children who are born blind and cannot take advantage of the visual input the world and their parents provide nevertheless acquire spoken language following the same trajectory as sighted children (Landau & Gleitman, Reference Landau and Gleitman1985). Language learning is surprisingly resilient to variations in how linguistic input is taken in.

Another way to extend variation in language learners is to examine children who early in development suffer an insult to the brain. In general, children with unilateral brain injury to either hemisphere tend to acquire the early-appearing aspects of language on time or with minimal delays (e.g. Marchman, Miller & Bates, Reference Marchman, Miller and Bates1991). However, these children experience iterative difficulty with each aspect of language development as these skills come on line (Stiles, Reilly, Levine, Trauner & Nass, Reference Stiles, Reilly, Levine, Trauner and Nass2012), including complex syntactic (Levine, Huttenlocher, Banich & Duda, Reference Levine, Huttenlocher, Banich and Duda1987) and narrative (e.g. Demir, Levine & Goldin-Meadow, Reference Demir, Levine and Goldin-Meadow2010) skills. But there is considerable variation in the language learning skills of children with unilateral brain injury.

Although many studies have explored the relation between differences in language skill and biological characteristics (lesion laterality, location, and size; seizure history) in children with brain injury (e.g. Feldman, Reference Feldman2005), few studies have examined these differences in relation to linguistic input. Doing so has resulted in interesting findings. For example, Rowe, Levine, Fisher, and Goldin-Meadow (Reference Rowe, Levine, Fisher and Goldin-Meadow2009) analyzed vocabulary growth (word types) and syntactic growth (MLU) in children with brain injury, comparing these children to the typically developing children in the Chicago sample. They found, not surprisingly, that children with less input at 1;2 had lower vocabularies at 2;6 than children with more input, and that acceleration in growth was more profound for children with high than low input. The relation between parent speech and vocabulary did not differ based on brain injury status. However, MLU showed a different pattern – there was a bigger difference between rates of growth in MLU for high- and low-input children with brain injury than for high- and low-input typically developing children. Thus the effect that linguistic input has on children with brain injury, compared to typically developing children, can differ as a function of linguistic property.

In addition to looking at natural variation in language learners, we can also vary the learner experimentally by substituting adults as learners in a paradigm called ‘human simulation’ (Gleitman, Reference Gleitman1990). If language learning is affected by the cognitive and social maturity of the learner, adults should go about learning in a very different way from children. To the extent that the trajectory is the same across child and adult learners, we have evidence that the to-be-learned information may be constraining the steps the learner takes, not the child's cognitive skills. Gillette, Gleitman, Gleitman, and Lederer (Reference Gillette, Gleitman, Gleitman and Lederer1999) explored this question with respect to noun and verb learning. Children have been found to learn nouns earlier than verbs (e.g., Gentner, 1982). Gillette et al. (Reference Gillette, Gleitman, Gleitman and Lederer1999) found that adults had more difficulty identifying verbs than nouns in mother-to-child talk, particularly when the words were presented without linguistic context. The findings suggest that the information requirements, rather than the conceptual requirements, of verb learning determine its late acquisition relative to noun learning.

Taking advantage of an unusual naturalistic situation to complement the human simulation approach, Snedeker, Geren, and Shafto (Reference Snedeker, Geren and Shafto2007) examined the acquisition of nouns and verbs in internationally adopted preschoolers who were older, and thus more cognitively mature, than children are when they typically acquire their first nouns and verbs. They found that the adoptees initially learned a disproportionate number of nouns relative to verbs, thus supporting Gillette et al.'s (Reference Gillette, Gleitman, Gleitman and Lederer1999) view that verbs are not acquired later than nouns because of the child's conceptual limitations (see also Gleitman, Cassidy, Nappa, Papafragou & Trueswell, Reference Gleitman, Cassidy, Nappa, Papafragou and Trueswell2005).

Taken together, these findings suggest that the age of the learner has little impact on when verbs are learned in relation to nouns. However, age does have an impact on the acquisition of some properties of language. For example, Newport (Reference Newport, Carey and Gelman1991) observed deaf individuals exposed to American Sign Language (ASL) for the first time at varying ages, and found that some properties of language (word order) were not affected by the learner's age, whereas other properties (morphological marking) varied systematically as a function of age – individuals exposed to ASL later in life had more difficulty mastering the morphological system of ASL than individuals exposed at birth or before age six. Some properties of language thus appear to be relatively resilient with respect to variations in the learner's state; others are relatively fragile.

Variations in language learning environments

Another way to explore the biases children bring to language learning is to observe learning in circumstances that vary in how much linguistic input children experience. The assumption is that children bring the same processing biases to whatever circumstances they encounter. To the extent that child outcomes remain the same across these various input situations, we have evidence that the child's processing biases are themselves important in determining the language the child develops – that the child's developmental trajectory is buffered from vagaries in the input. To the extent that each varying input situation results in a different child outcome, we not only have evidence that input matters, but we can begin to explore the patterns between input and child outcome to make inferences about the child's biases and processing strategies. For obvious ethical reasons, we cannot tamper with the circumstances under which children learn language. But we can take advantage of the varied circumstances that children find themselves in when attempting to learn language.