Citation:
Hoff, E. (2013). Language Development (5th ed.). Cengage Learning US.
CHAPTER 3Foundations of Language Development in Domain-General Skills and Communicative ExperienceSocial and Communicative Foundations of Language Development • The Communicative Function of Speech • Social Cognitive Skills of Infants• The Communicative Use of Gesture Sensory and Perceptual Foundations of Language Development • Methods of Studying Infant Perception • Infant Hearing and Prenatal Learning • Early Attention to Speech and to Speakers • Infant Speech PerceptionCognitive Foundations of Language Development• Conceptual Understandings of the Meanings Language Encodes• Domain-General Mechanisms of Learning and Development• Memory and Attentional Processes The Relation of Early Foundational Skills to Later Language Environmental Support for Language Development • Sources of Environmental Support • The Relation of the Availability of Environ-mental Support to Language AcquisitionSummary Key Terms Review QuestionsThe necessary and sufficient ingredients for language acquisition are a human child and an environment that provides the child with language experience. Explaining language acquisition requires specifying the properties of the child and the properties of language experience that produce this outcome. When the modern study of language acquisition began in the 1960s, the dominant view in the field was Noam Chomsky’s position that the relevant properties of the child are the child’s innate knowledge of Universal Grammar (UG) and the relevant property of experience is exposure to examples of grammatical sentences. The examples allow the child to test the possible grammars specified by UG against input in order to find which is the grammar of the language he hears. Chomsky argued that the communicative functions of language are irrelevant to the process of language acquisition and that children’s general learning abilities are too weak and children’s language experience too impoverished for language acquisition to be possible without recourse to innate knowledge (see also Chapter 1).In the decades since these arguments were first made, research on the communicative foundations of language, on children’s domain-general learning capacities, and on the role of language experience in language development has suggested an alternative account of language acquisition. According to this account, the interactions children experience and the speech children hear provide rich data and social support for the language acquisition process. Children, for their part, start life with social, perceptual, and cognitive skills that engage them in communicative interactions and allow them to find the meanings of words and the underlying structure of language in that experience.In this chapter we review the research that has been cited in support of the argument that children’s domain-general skills and communicative experiences are the foundation
of language development. We review evidence of the social, perceptual, and cognitive skills that infants bring to the language learning task and evidence that these early skills are related to subsequent language development. We also review evidence that the speech children hear provides information relevant to abstracting the underlying linguistic system and evidence that children make use of that information—including evidence that when children do not have adequate communicative experiences, their language development suffers as a result.Social and Communicative Foundationsof Language DevelopmentHumans are social creatures and language is a vehicle of social interaction. The undeniable fact that language is used for social purposes does not, however, necessarily mean that the social function of language is relevant to language development. In this section, we consider the arguments and evidence that the social function of language and the social capacities of children are part of the explanation of how children acquire language.The Communicative Function of SpeechIf you ask the proverbial man on the street why children learn to talk, you will probably hear that children do so in order to communicate their needs and desires. Relatedly, you sometimes hear a child’s delayed speech attributed to his or her mother or siblings who provide everything the child wants, thus removing the need to speak. This view is a bit simplistic. Babies who are very well fed and cared for prior to language reliably learn to talk nonetheless, and withholding food from children who do not talk will not spur their language growth.There are serious proposals, however, that a basic human desire to share one’s thoughts is part of what motivates the language acquisition process (L. Bloom, 1993; Snow, 1999). Consistent with this view, it has been noted that isolated children do not invent a language while groups of children such as the deaf children in Nicaragua do (Shatz, 1994a). It has been further argued that the desire to communicate provides not only the initial impetus but also ongoing motivation for language acquisition, so when children fail to be understood, the failure prompts them to find new and better ways of communicating (Golinkoff, 1983; Mannle & Tomasello, 1987). Consistent with this view, it has been found that children who have competition for access to their mother as a con- versational partner—that is, children who have older siblings—are particularly skillful at entering and sustaining participation in conversations (Bernicot & Roux, 1998; Dunn & Shatz, 1989; Hoff-Ginsberg, 1998).In addition to providing a motive, another reason that the communicative function of language might be relevant to language acquisition is that children are more likely to attend to and process language when they hear it in the context of communicative inter- action. Children appear not to learn language, for instance, if their sole means of expo- sure is through television. In the Netherlands, Dutch-speaking children regularly watch German television but do not acquire German as a result (Snow et al., 1976). More direct evidence comes from a study that compared learning from a live tutor to learning via video exposure (Kuhl, 2007). In this experiment, American babies were exposed to Man- darin either by being read to in person (in 12 sessions over the course of four to five weeks) or by watching a video recording of the exact same book-reading sessions. The infants who directly experienced Mandarin showed effects of that experience on their ability to hear sound contrasts that are unique to Mandarin. The infants who had the same language exposure but via video showed no effect of their experience. Kuhl has
described social interaction as serving a “gating” function, such that infants do not bother to analyze auditory signals unless they come from live humans (Kuhl, 2007). Such social gating of attention to sounds would clearly be adaptive, saving the infant from trying to figure out the pattern in door squeaks, traffic noises, and bird calls.Slightly older children may be able to learn from video—if what is on the video is communicative interaction between two people. A study of 2-year-olds exposed children to new words in four different conditions: they were taught the words in interaction with an adult, they saw a video of the adult teaching the new words, they observed an adult teaching the words to another adult, or they saw a video of an adult teaching new words to another adult. The children learned the new words in every condition except the direct video instruction (O’Doherty et al., 2011). That is, the children could learn from direct experience in interaction or from observing other interaction—live or on video. What was not sufficient was video-based instruction with no social interaction.The question of at what age and under what circumstances children can learn through media is an important and active area of research (Barr, 2008). Current evidence suggests that while older children may learn from video, infants do not—the claims of companies that market DVDs for infants notwithstanding. In fact, one study found that the number of hours that babies watched videos or DVDs per day was a negative predic- tor of their language development (Zimmerman, Christakis, & Meltzoff, 2007). Probably the videos themselves are not harmful, but they have a negative effect because they reduce the amount of time the infant experiences human interaction.It is important to distinguish the argument that the communicative function of language provides the motive for language acquisition or serves a gating function from the argument that social properties of humans explain how language is acquired. Even if communication provides the motive or labels the speech signal as worthy of analysis, communicative goals do not by themselves explain how children figure out the linguistic system (Maratsos, 1998). The link between wanting to communicate and being able to speak a language needs to be made by some learning mechanism. As Chomsky (1965) put it, some kinds of experience “may be required to set the language acquisition device into operation, although they may not affect the manner of its functioning in the least” (p. 33).Social Cognitive Skills of InfantsThere is evidence, however, that some social aspects of human nature are part of the explanation of how children acquire language. The research that supports this claim includes studies of children’s capacity for joint attention and studies of children’s ability to make use of cues in interpreting the meanings intended by other speakers.Joint AttentionBetween the age of 9 and 12 months, before children are doing much that would count as talking, their behavior changes in ways that suggest they are joining the communicative world. Prior to this age, babies will interact with objects, and they will interact with people, but they do not interact with people about objects. An example of relating to an object would be playing with a rattle or looking at a mobile over one’s crib. An example of relating to another person would be smiling or cooing. What emerges around 10 months of age is the ability to relate to another person about an object (Sugarman, 1984). Children demonstrate this ability when they hold their arms out to help in getting dressed or when they play exchange games in which they alternate between giving an object to someone and receiving it from that person (Tomasello, 2007). (11-month-olds will do this back-and-forth type of interaction for a long time for no other apparent purpose than the experience of doing it.) This developmental change has been described by Trevarthen and Hubley (1978) as a change from the capacity to share oneself with others (termed primary intersubjectivity) to the capacity
IG-3-1 Infants begin to point to communicate around the age of 12 months.to share one’s experiences with others (termed secondary intersubjectivity). It has also been described as the sharing of goals or the “cooperative dimension of human cognition” (Tomasello, 2007, p. 8). This new social cognitive capacity changes the infant from a solitary to a socially engaged creature.Other behaviors that emerge around this time are pointing, following the points of others, and following the eye gaze of others. In early infancy, babies do not point, and when someone else points, they are just as likely to look at the hand that is pointing as they are to look at the object being pointed to. Also, when infants first point, they don’t seem to have the expectation that anyone will follow their point. What emerges around 12 months of age is communicative pointing, in which the infant points and looks at a potential communicative partner at the same time (Bates et al., 1979; Desrochers, Morissette, & Ricard, 1995; Tomasello, Carpenter, & Liszkowski, 2007). Infants may produce communicative points to request something, but infants also point just to provide information. For example, infants will point to something if they see an adult searching for it (Tomasello, 2007). Also around the age of 10 and 11 months, infants first show gaze-following—they turn to look where another person is looking (Brooks & Meltzoff, 2005).As a result of these changes in their social cognitive skills, children between the ages of 9 and 15 months spend increasing portions of their time with others in what is referred to as joint attention (Carpenter, Nagell, & Tomasello, 1998) (see Figure 3.2). Joint attentional states are those in which the child and adult together attend to some third entity. Because joint attention is typically measured on the basis of where the infant and adult are looking, the term joint visual attention is often used (Meltzoff & Brooks, 2009). These episodes of joint attention are first nonverbal, but as children become older, nonverbal symbols such as gestures and verbal symbols (i.e., words) are increasingly used in these interactions (Adamson, Bakeman, & Deckner, 2004). Developmental changes in time spent in joint attention that were observed in two different studies are presented in Figure 3.2.
-3-2 As infants’ social cognitive skills develop, they spend in- creasing portions of their time in joint engagement with another. The figure includes measures of the percentage of time in joint engagement during mother-child play from two independent studies. (dotted line [Bakeman & Adamson, 1984]; solid line [Carpenter et al., 1998]).12 10 8 6 4 208 9 10 11 12 13 14 15 16 Age in monthsSource: Carpenter, Nagell, & Tomasello, 1998.Several findings suggest that joint attention and the underlying social cognitive capac- ities it reflects support language acquisition. Infants who are more precocious at develop- ing joint attention skills are also more advanced in later measures of language development (Camaioni & Perucchini, 2003; Morales, Mundy, & Rojas, 1998; Mundy et al., 2007; Mundy & Gomes, 1998). Furthermore, among autistic children, who tend as a group to show impaired joint attention skills and impaired language, those who are less impaired in joint attention are less impaired in language (Parish-Morris, Hennon, Hirsh-Pasek, Golinkoff, & Tager-Flusberg, 2007). The amount of time infants spend in joint attention with an adult speaker is a positive predictor of their language development (Carpenter et al., 1998; Tomasello, Mannle, & Kruger, 1986; Tomasello & Todd, 1983). Individual differences in joint attention skills may be particularly important when these skills are developing. By the age of 18 months, all typically developing children can follow the speaker’s lead, and individual differences in joint attention and in maternal responsivity seem to be less important predictors of language development in studies of children older than 18 months (Hoff & Naigles, 2002; Morales et al., 2000; Pan, Rowe, Singer, & Snow, 2005).Intention ReadingAnother behavior that develops prior to language and may be a foundational skill for language acquisition is intention reading. This skill first involves understanding that other people have intentions—particularly communicative intentions—and, second, discerning what those intentions are. Early evidence that infants attribute intentions to others comes from the finding that they imitate others’ intended acts. While newborns have been shown to imitate behaviors like sticking out one’s ton- gue, only older children reproduce the aim of a behavior modeled for them. For example, if you take a string of beads and drop them in a cup, but miss, 18-month-olds will imi- tate that behavior by putting the beads in the cup (Meltzoff, 1995). Interestingly, they will not fix the observed behavior and produce an intended act if it is modeled by a mechanical device. Infants regard humans, not machines, as having intentions
Not only do prelinguistic infants attribute intentions to others, they appear to use attributed intentions as a basis for word learning. One clue to another’s intentions is eye gaze. Even when we don’t want to, we humans tend to give away what we are think- ing about by where we are looking. Infants use this information. When children hear someone produce a novel word, they will take it to be the label for what the speaker is looking at—even if that is different from what they themselves are looking at (Baldwin, 1993). And, reminiscent of Meltzoff’s finding that infants only imitate the intended behavior of humans, not machines, children only take new words as label for what the speaker is looking at if the speaker is a human, not if the voice comes from a robot (O’Connell, Poulin-Dubois, Demke, & Guay, 2009). As another example, Akhtar, Carpenter, and Tomasello (1996) found that 2-year-olds will infer that a novel label pro- duced with an expression of surprise refers to an object that is also novel to the speaker. And Tomasello and Barton (1994) found that 24-month-olds could distinguish acciden- tal actions from intentional actions. If an experimenter said, “Let’s go twang it” and then did something accidentally (i.e., clumsily) followed by doing something intentionally, children took twang to refer only to the intentional action.Although there is substantial evidence that joint attention and the intention reading it enables support the language acquisition process, concerns have been raised about attributing too much to these social supports. The findings that joint attention skill and language devel- opment are related could reflect the influence of early brain maturation on both processes. Support for this hypothesis comes from findings that brain maturation, measured by electro- encephalogram (EEG) coherence, is correlated with joint attention skill in 14-month-olds (Mundy, Fox, & Card, 2003). Other evidence that language acquisition also happens outside of joint attention is evidence that episodes of joint attention are not that frequent in chil- dren’s experience. In many cultures, adults do not talk to prelinguistic children, and thus children do not experience speech addressed to them in episodes of joint attention. Further- more, studies find that even Western middle-class mothers and children spend less than 25% of the time they are interacting in episodes of joint attention (Hoff & Naigles, 2002).There are a variety of ways to resolve the apparent conflict between the evidence that the mutual engagement of joint attentional episodes supports language acquisition and the evidence that children do not reliably experience a great deal of that sort of interac- tion. It may be that while joint attentional episodes are supportive, they are not essential, and children who experience little joint attention make use of other routes to figuring out language (Lieven, 1994). One such route may be learning from the speech among others that children overhear. There is evidence that at least by the age of 18 months children can, in some circumstances, learn from such speech and also that children who have more experience with multiple adults are more likely to be able to do so (Floor & Akhtar, 2006; Shneidman, Buresh, Shimpi, Knight-Schwarz, & Woodward, 2009). It may also be that children who experience little joint attention acquire language at a slower pace than children who experience a great deal of speech addressed to them in joint attentional episodes. Some evidence suggests this is the case (Hoff, 2006b).The Communicative Use of GesturePointing is often the first communicative gesture children use, but it is not the only one. In the typical course of development, children may use several different iconic gestures to communicate before they communicate with spoken words (Acredolo & Goodwyn, 1990; Cartmill, Demir, & Goldin-Meadow, 2012; Goldin-Meadow, 2007). Some iconic gestures are very common across children, such as reaching and opening and closing a hand to indicate “give.” Others are more idiosyncratic; examples are listed in Box 3.1. Even though pointing and iconic gestures are not arbitrary symbols the waywords are, these gestures are considered to be the beginning of symbolic communication because the gestures are conventionalized—they occur repeatedly in the same form not physically involving the object itself, and they occur outside the specific context in which they were first used (Acredolo & Goodwyn, 1990).Children’s use of communicative gesture predicts the language development that fol- lows. The objects a child points at are likely soon to become words in that child’s vocabu- lary (J. M. Iverson & Goldin-Meadow, 2005), and the number of different objects that the child points at early in communicative development predicts the size of the child’s com- prehension vocabulary later on (M. L. Rowe, Özçalişkan, & Goldin-Meadow, 2008). For children with pre- or perinatal brain lesions, early communicative gesture predicts which children will show later language delay (Sauer, Levine, & Goldin-Meadow, 2010). Gesture is predictive in the domain of grammatical development as well. At first, children’s com- municative gestures appear alone. Later, children produce gesture þ word combinations such as saying the word “eat” and pointing to a cookie. The age at which children produceBOX 3.1Examples of Early Communicative GesturesSource: Acredolo & Goodwyn, 1990.
gesture þ speech combinations of this sort predicts the age at which they first produce two- word utterances. Furthermore, the types of two-word constructions children produce are predicted by the type of gesture þ word combinations they produced earlier. Producing gesture þ speech combinations such as “Mommy” and pointing to the couch predicts the production of noun þ noun combinations in speech, whereas producing gesture þ speech combinations such as “drive” and pointing to a car predicts the production of verb þ noun combinations (Özçalişkan & Goldin-Meadow, 2005).There are several potential explanations of why children’s early use of communicative gesture predicts their subsequent language development. One possibility is that children’s use of gesture is an indicator of their developing communicative interest and the sorts of meanings they choose to convey, which are soon to be revealed in language. Another possibility is that their communicative gestures provide children with a way of entering into communicative interaction prior to language and that communicative experience supports language development. A third possibility, which perhaps best explains the close relations between the content of children’s gestures and the content of their subse- quent language, is that children’s gestures elicit speech from others. Goldin-Meadow and colleagues studied mothers’ responses to their children’s gestural communication and found that mothers often translate their children’s gestures into words. For example, when a child pointed to a cat, his mother said, “Yes, that’s a cat,” and when a child pointed to his baby sister and said “sleeping,” his mother said, “Yes, baby’s sleeping” (Goldin-Meadow, Goodrich, Sauer, & Iverson, 2007). These multiple possible bases for the relation between children’s communicative gestures and their subsequent language development are not mutually exclusive.Sensory and Perceptual Foundations of Language DevelopmentInfants come into the world with the ability to hear the language spoken around them, with the inclination to attend to it over other sounds in their environment, and with the capacity to match speech sounds with the faces that produce them. When they begin the language acquisition process, they are able to hear the difference between the sound contrasts that carry meaning, and this ability becomes tuned to the particu- lar language or languages they hear over the course of the first year of life. These abil- ities and inclinations in infants are the sensory and perceptual foundations of language acquisition.Methods of Studying Infant PerceptionScientific understanding of infants’ abilities has changed dramatically in the last 50 years, largely as a result of the development of ingenious behavioral methods for studying infants (Golinkoff & Hirsh-Pasek, 2012). The brain imaging methods discussed in Chapter 2 are one way researchers approach the task of studying infants. In addition, researchers have also figured out ways to use babies’ behavior as evidence of what they can discriminate and understand.Habituation procedures measure infants’ abilities to discriminate by taking advan- tage of the fact that babies get bored hearing the same sound over and over. Thus, a researcher can present one sound until the baby shows his or her boredom by responding less. Then the researcher changes the sound presented, and if the baby increases his or her level of responding, the researcher concludes that the baby noticed the change. The most frequently used response is sucking on a nonnutritive nipple, as shown in Figure 3.3. This nipple, essentially a pacifier, has a pressure transducer inside
FIG-3-3 In habitua- tion procedures, the nipple in the infant’s mouth is connected to a computer that records the frequency and strength of the infant’s sucking. Sucking de- clines as the infant ha- bituates to the stimulus (i.e., gets bored). The stimulus is then chan- ged, and an increase in sucking behavior indi- cates that the infant no- tices the change and thus is able to discriminate the first stimulus from the second stimulus.Source: Werker.which is connected to a computer, allowing the researcher to measure and monitor the strength and frequency of the baby’s sucking (Fennell, 2012). Because sucking is some- thing all babies do, researchers can use this method with infants as young as newborns.The conditioned head turn procedure requires somewhat older infants and involves training infants to turn their heads when the stimulus changes (see Figure 3.4). The pro- cedure begins by presenting an infant with sound changes that we know infants can detect. As soon as the sound changes, an interesting and rewarding visual stimulus is presented. Infants quickly learn to anticipate the interesting visual stimulus following a sound change, and they turn their heads in anticipation. Then researchers can use this conditioned response to ask the infants when they can discriminate a change in other stimuli.The intermodal preferential looking paradigm assesses language comprehension by showing infants side-by-side slides or videos, as the infant hears an audio presentation
FIG-3-4 In the conditioned head turn procedure, the infant, seated on her mother’s lap, is presented with repeated sounds. The mother wears earphones so she cannot hear the stimulus and thus cannot influence the infant’s behavior. When the sound stays the same, nothing happens. When the sound changes (e.g., from /pa/ to /ba/) an interesting visual and auditory stimulus is presented to the side of the infant, and the infant turns her head to look. Infants thus are conditioned to turn their heads when the stimulus changes, and then this head turning response can be used as a measure of whether the infant is able to discriminate between two sounds.Source: Werker.that matches only one of the videos (see Figure 3.5). So, for example, infants hear “Where is the dog?” as they are presented with a picture of a dog and a baby. If infants consistently look longer at the matching video, it is taken as evidence that they understand the lan- guage of the audio (Piotroski & Naigles, 2012). In the looking-while-listening procedure infants are presented with side-by-side videos and a paired audio, as in the preferential looking procedure, but in this procedure, the time course of the infants’ eye movements toward the pictures is measured. This method yields a measure not just of what infants understand but also a measure of how quickly they process the language presented (Swingley, 2012). (More comprehensive descriptions of these and other methods of study- ing infants’ language processing are available in Hoff, 2012.)Infant Hearing and Prenatal LearningAt one time, it was thought that babies were blind and deaf at birth and that basic sen- sory abilities matured only later. We now know that this is incorrect. Infants’ hearing is not quite as sensitive as adults’, but it is certainly adequate for hearing speech from the time infants are born (Fernald, 2001; J. R. Saffran, Werker, & Werner, 2006). In fact, the auditory system starts functioning in the fetus even before birth.Evidence that the fetus can hear comes from a study in which researchers played record- ings of their mothers’ and a stranger’s voice to 38-week-old fetuses (i.e., 2 weeks before they
FIG-3-5 Intheprefer- ential looking paradigm, the infant is presented with an audio stimulus (e.g., the word shoe) and presented with 2 pictures (e.g., one of a shoe and one of some- thing else). The infant’s looking behavior is re- corded through a peephole by a camera behind the screen. The amount of time the infant looks at the matching picture and the speech with which the in- fant first moves his eyes to look at the matching pic- ture are measures of word recognition and speed of word recognition.were due to be born), using a loudspeaker placed 10 cm away from the mothers’ abdomens. The fetuses’ heart rates went up in response to their mothers’ voices and down in response to a stranger’s voice, demonstrating that the fetuses were able to discriminate the two voices (Kisilevsky et al., 2003). (Each mother’s voice was the stranger’s voice to a different fetus, so the effect was not specific to one individual’s voice.) As newborns, infants show that they prefer their mothers’ voices to a stranger’s voice (DeCasper & Fifer, 1980).Newborns not only show memory for what their mother’s voice sounds like, as opposed to a stranger’s voice, they also show memory for features of what they heard their mothers saying while they were in utero. In one study, pregnant women read a particular passage aloud every day during the last six weeks of their pregnancy. When their babies were tested a few days after birth, these babies showed a preference for hearing that familiar passage over hearing a novel passage. A control group of newborns, whose mothers had not read either passage before their birth, responded equally to both passages (DeCasper & Spence, 1986). Other evidence that babies remember what they heard in utero comes from the finding that 4-day-old French infants could distinguish French from another lan- guage, but they could not distinguish two foreign languages (Mehler et al., 1988).Early Attention to Speech and to SpeakersNewborn infants would rather listen to the speech of humans than to other sounds. When presented with isolated syllables of human speech and with nonspeech stimuli that were carefully constructed to match on several acoustic parameters, babies from 1 to 4 days old worked harder (by sucking on a nipple) to hear the speech than the nonspeech stimuli (Vouloumanos & Werker, 2007). Infants also prefer listening to speech compared to other noises humans make such as coughs, hiccups, and sneezes, compared to the communica- tive calls made by rhesus monkeys, and compared to environmental noises such as run- ning water (Shultz & Vouloumanos, 2010). Those findings are presented in Figure 3.6. Vouloumanos and Werker (2007) describe the preference for human speech as a “sieve through which newborns could glean the acoustic signals important for communication” (p. 159). That is, in a world full of noises—some of which are relevant to the language acquisition task, but many of which are not—the baby has a mechanism that sorts the relevant from the irrelevant.
FIG-3-6 Three- month-old infants prefer to listen to speech than to sounds made by Rhesus monkeys and noncommunicative sounds made by humans (sniffs, coughs, sneezes, yawns, and throat clear- ings). The preference for speech over environ- mental sounds was not statistically significant.20 18 16 14 12 108**Rhesus macaqueSpeechHuman non- communicativeEnvironmentalSound typeSource: Schultz, S., and Vouloumanos, A. (2010). “Three Month Olds Prefer Speech to Other Naturally Occurring Signals.” Language Learning and Development, 6, 241–257.Babies are also particularly interested in visual stimuli that look like faces (Slater, 2001), and they seem to know from a very early age that speech comes from faces. Infants as young as 2 months of age become visibly distressed when they see their mother talking in front of them but hear her voice coming from elsewhere (Aronson & Rosenbloom, 1971). This expectation on infants’ part, that sounds and sights go together, is not limited to speech. Very young infants will look in the direction of a sound (Wertheimer, 1961), and 4-month-old infants prefer to watch a stuffed animal that bounces in rhythm to an accompanying sound rather than watch a stuffed animal whose bounces have a different rhythm (Spelke, 1979). Furthermore, infants as young as 18 weeks recognize that particular movements visible in faces correspond to particular speech sounds and look longer at faces whose movements match the speech sounds they are hearing (Hollich et al., 2005; Kuhl & Meltzoff, 1982, 1984). Measures of brain activity (ERPs) suggest that 4-month-old infants also expect speech to be accompanied by a face that is looking forward rather than a face with averted eye gaze (Parise, Handl, Palumbo, & Friederici, 2011).These early perceptual abilities mean that even though young infants do not under- stand language, the world they experience has some order to it. Sounds are discriminable as speech or nonspeech, and speech sounds come from faces. The particular faces pro- ducing speech sounds are likely to be looking at you and to be moving in ways that match the sound being produced. The inclination of babies to look to the faces that are source of the speech they are hearing puts them in a good position to notice eye gaze and other potential cues to the meaning of what is being said.Infant Speech PerceptionInfants’ Discrimination of Speech SoundsInfants’ perceptual systems do more for language acquisition than merely orient infants to speech sounds and the faces that produce them. Their perceptual systems allow infants to distinguish between different speech sounds. In order for languages to express different meanings with different sound sequences, the
users of those languages must be able to perceive those differences. The words pill and bill can mean different things only to hearers who perceive /p/ and /b/ as different sounds. Several decades of research on infant perception have established that infants come to the language learning task equipped with the ability to hear this and, in fact, most of the distinc- tions that any language requires. For example, infants as young as 4 weeks old can discrimi- nate vowel contrasts such as /u/ versus /I/ and /I/ versus /a/ (Trehub, 1973) and consonant contrasts such as /p/ versus /b/ and /d/ versus /g/. These abilities do not depend on experi- ence. Very young infants can discriminate contrasts that are not used in their ambient language. For example, English-learning babies can discriminate vowel contrasts that are present in French but not in English (Trehub, 1976), and they can discriminate consonant contrasts that Hindi uses but English does not (Werker, Gilbert, Humphrey, & Tees, 1981). Many studies have established the range of infant perceptual abilities using contrasts from many different languages (for summaries, see Aslin, Jusczyk, & Pisoni, 1998; J. C. Goodman & Nusbaum, 1994; Kuhl & Meltzoff, 1997; Werker & Polka, 1993).Categorical PerceptionOne feature of infants’ speech perception that was the focus of intensive research during the 1960s and 1970s is the phenomenon known as categorical perception. Categorical perception occurs when a range of stimuli that differ continuously are perceived as belonging to only a few categories. For example, the phonemes /b/ and /p/ differ along a single acoustic continuum, voice onset time (VOT)—which is the lag between when the air is released and when the vocal cords start to vibrate. However, when listeners are presented with stimuli that vary in VOT, they hear each stimulus as either a /b/ or a /p/; no acoustic signal is perceived as something in between a /p/ and a /b/. The data that indi- cate categorical perception in adults and infants are presented in Figures 3.7 and 3.8. Furthermore, listeners have no difficulty distinguishing two stimuli that are 20 msec apart in VOT if they are on different sides of the phoneme boundary, but they have difficulty hearing the difference within the phonemic category. For human perceivers, there is a place on the VOT continuum that is a phoneme boundary.FIG-3-7 Categorical Perception in Adults Adults switch from hearing [ba] to hearing [pa] at a voice onset time of þ20 milliseconds.100806040200 –50 –40 –30 –20 –10[pa][ba]304050607001020 Voice onset time (msec)Source: Reprinted with permission from “Discriminability, Response Bias, and Phoneme Categories in Discrimination of Voice Onset Time,” by C. C. Wood, 1976, Journal of the Acoustical Society of America, 1381–1389. Copyright © 1976 Acoustical Society of America.
IG-3-8 Categorical Perception in Infants Infants recover from habituation and increase their rate of sucking when the stimulus changes from VOT 1⁄4 20 msec to VOT 1⁄4 40 msec (Panel 1). In contrast, they do not change their rate of sucking when the stimulus changes from VOT 1⁄4 60 msec to VOT 1⁄4 80 msecVOT = 20VOT = 40VOT = 60VOT = 80No change604530(Panel 2) or when the stimulus does not15 change (Panel 3).B543211234 B543211234 B543211234 Time (min)Source: “Speech Perception in Infants,” by P. D. Eimas et al., 1971, Science, 171, 303–306. Copyright © 1971 American Association for the Advancement of Science. Reprinted with permission from AAAS.This phenomenon was first discovered in adult perception and then demonstrated in infants (Eimas, Siqueland, Jusczyk, & Vigorito, 1971). The initial discovery of categorical perception for speech sounds led to the claim that speech perception is special because, for most physical continua, perception does not change abruptly at some point, and the ability to perceive a difference between two stimuli does not change abruptly in the mid- dle of the continuum. The finding that even babies perceived VOT categorically was taken as strong evidence that human babies come into the world specially prepared to acquire language (for a more detailed history of this argument, see Kuhl, 1987).Subsequent discoveries, however, cast doubt on the significance of categorical percep- tion for the argument that language is unique and even on the phenomenon of categori- cal perception itself. One study that burst the uniqueness bubble found that some nonspeech sounds are also perceived categorically (J. D. Miller, Wier, Pastore, Kelley, & Dooling, 1976). A second result with a similar effect was the finding that chinchillas (small rodents whose main claim to fame is their soft fur) also show the phoneme boundary effect for /b/ and /p/ (Kuhl & Miller, 1975), although chinchillas show no other signs of having evolved to acquire language. It seems that the phoneme boundary effect is a property of the mammalian aural system that language uses rather than a spe- cifically linguistic property of auditory perception (Kuhl, 1987; J. L. Miller & Eimas, 1994). In fact, language may have evolved to take advantage of this preexisting property of mammalian audition. It also seems to be the case that human listeners are not completely insensitive to acoustic differences within categories, so that while sounds are categorically perceived as one phoneme or another, within-category differences are also perceptible under some circumstances (McMurray & Aslin, 2005). Thus, categorical per- ception is more accurately described as a warping in perception of the physical continua on which stimuli differ such that some physical differences are compressed (i.e., the dif- ferences within categories) and others (i.e., the between-category differences) are expanded (Harnad, 2003). Some of the warping of acoustic continua in speech
perception seems to reflect the operation of the mammalian auditory system, but the warping can be acquired through experience as well.Early Tuning of Speech PerceptionThe warping of acoustic differences that occurs as a result of language experience tunes infants’ perception of speech to the sound contrasts in their language. Although infants start out able to make essentially all the discriminations used by the languages of the world (in Kuhl’s words, they are “universal listeners” [Kuhl, 2009]), with experience they become even better at making some of the discriminations their particular language uses and less able to discriminate some contrasts that are not used in their ambient language. The classic experiment that first demonstrated the experience- related decline in infants’ discrimination abilities was by Werker and Tees (1984). These researchers took advantage of the fact that Hindi and Inslekepmx (the language spoken by the native Salish of British Columbia) have consonant contrasts that English does not. They found that 6- to 8-month-old English-learning infants could make the Hindi and Inslekepmx discriminations, but very few could do so at 10 to 12 months of age. In contrast, 11- and 12- month-old Hindi-learning and Inslekepmx-learning infants were still able to make phonemic distinctions in the language they were acquiring.Since then, many studies have found effects of the target language on infants’ percep- tion of many contrasts. Effects on vowel discrimination have been found in infants as young as 6 months old (Kuhl, Williams, Lacerda, Stevens, & Lindblom, 1992). Vowel per- ception may be affected before consonant perception because there are fewer different vowels than consonants; thus, infants hear each vowel more frequently than they hear each consonant (Gerken, 2002). Effects on tone perception appear between 6 and 9 months of age (Mattock & Burnham, 2006). At 6 months of age, infants learning Chinese, in which tone contrasts can signal meaning differences, and infants learning English appear to be equivalent in their abilities to perceive tone contrasts in speech stimuli. By 9 months, the English-learning infants have declined in this ability while the Chinese-learning infants retain their abilities. At 6 months of age, Japanese and American infants are similar in their ability to distinguish the sounds /ra/ and /la/. At 12 months of age, American infants are better than 6-month-olds, and Japanese 12-month-olds are worse (Kuhl et al., 2006). These data are presented in Figure 3.9. In all these cases, it appears that exposure to the target language results in the tuning of speech perception to the ambient language such that the ability to hear a difference between sounds is diminished for contrasts that are notFIG-3-9 At 6 to 8 months, American and Japanese infants do not differ in their ability to hear the contrast be- tween /ra/ and /la/. By 10 to 12 months, the am- bient language has “tuned” infant speech perception such that American infants have improved in their ability to hear the contrast and Japanese infants have declined.90 80 70 60 500American infantsJapanese infants6–8 months Age of infantsSource: Kuhl et al., 2006.
used by the language and, in some cases, enhanced for contrasts that are used (Kuhl, 2007; Polka et al., 2007).This effect of the ambient language on speech perception appears to arise from infant’s detection of distribution of acoustic signals in the speech they hear and their for- mation of categories around frequently-occurring signals. If /d/ and /t/ are meaningfully different sounds in a language, then speech will contain many clear examples of /d/ and /t/ and not much that is in between. In contrast, if there were a language in which the difference between /d/ and /t/ were not phonemic (i.e., didn’t matter for meaning), then the speech signal would be more likely to include sounds that cover the whole acoustic continuum from /d/ to /t/ with equal frequency. If babies pay attention to acoustic prop- erties and the frequencies with which they occur, babies in the first case should learn that there are two distinct sounds on the /d/–/t/ continuum; babies in the latter case should not. Maye, Werker, and Gerken (2002) tested the hypothesis that this sort of learning, termed distributional learning, could account for ambient language effects on infant speech perception. They gave 6- and 8-month-old infants two different types of experi- ence hearing sounds along the /d/–/t/ continuum. Some infants heard sounds near the endpoints more frequently than they heard sounds in the middle of the range (a bimodal distribution of experience); other infants heard sounds in the middle more frequently than sounds at either end (a unimodal distribution of experience). Next, the infants’ abil- ities to make the /d/–/t/ discrimination were tested. The infants who heard the bimodal distribution were able to distinguish between /d/ and /t/; the babies who heard the unim- odal distribution could not. This experiment is described in Box 3.2: A similar process byBOX 3.2Learning Distributional Properties of InputSource: Maye, J., Werker, J. F., & Gerken, L. (2002). Infant sensitivity to distributional information can affect phonetic discrimination. Cognition, 82, 101–111.
FIG-3-10 Vocabulary Growth as a Function of Phonetic Discrimi- nation Abilities in InfancyNative predictorBetter discriminationPoorer discriminationNonnative predictorPoorer discriminationBetter discrimination0 14 18 24 30 14 18 24 30Age (months)Age (months) (a)(b)Source: Kuhl, 2009.which experience creates categories of sounds based solely on acoustic properties has been proposed to account for ambient language effects on vowel perception (Kuhl, 1999; Kuhl & Meltzoff, 1997). Werker and colleagues have analyzed the distributional properties of acoustic signals associated with different vowels in adults’ speech to infants and found that the speech signal does provide the evidence such a learning mechanism would require (Werker et al., 2007).This tuning of speech perception to the ambient language appears to be beneficial for language development. Among 11-month-olds, those who are better at discriminating con- trasts in the ambient (i.e., native) language have larger vocabularies than those who are less good at the native language discriminations (Conboy, Sommerville, & Kuhl, 2008). In fact, the benefit of this native language tuning appears to begin even earlier than 11 months: Infants who are better at discriminating contrasts in their native language between 6 and 7 months show more rapid language development between 11 and 30 months (Tsao, Liu, & Kuhl, 2004). We can tell that it is the language-specific tuning that is responsible for this relation—and not a general ability to hear contrasts—because infants who are better at distinguishing nonnative contrasts (i.e., the contrasts that are not relevant in their ambient language) at 6 to 7 months show slower subsequent language growth (Tsao et al., 2004). These data are presented in Figure 3.10. Put another way, it is not good for language development to keep on noticing distinctions that do not matter in the language you are acquiring (Kuhl, 2007, 2009; Kuhl et al., 2005). It is important to note, however, that the effect of early experience is not absolute. Infants appear not so much to lose the ability to hear nonnative contrasts as to learn to ignore the contrasts that don’t matter (Conboy et al., 2008). Also, some contrasts remain readily distinguishable even for nonnatives, and the perception of the difficult contrasts can be improved with training (Best, 1994).Cognitive Foundations of Language DevelopmentIn 1973, one of the pioneers of the field of child language described the cognitive prerequi- sites for the development of grammar as follows: the child must “be able to cognize the physical and social events which are encoded in language, and he must be able to process, organize, and store linguistic information” (Slobin, 1973, p. 176). Research since then has filled in some of the details regarding the conceptual understandings of the world, the pro- cessing abilities, and the memory capacities that children bring to the language learning task.
Conceptual Understandings of the Meanings Language EncodesBefore they start to talk, children already have some basic understandings of the mean- ings that words encode. Perhaps the most basic understanding that children bring to word learning is that things exist as separate, stable, individual things. This seems obvi- ous, but Xu and Carey (1995, 1996) found evidence that before about the age of 11 months—which is about when children start producing words—infants’ understand- ing of the world may be very different from ours. Imagine you saw a book emerge from behind an opaque screen and then return; then you saw a cup emerge from behind the screen and then return. What would you think was behind the screen? A book and a cup, most likely. If the screen were removed and only a cup was there, you would be surprised. However, you would not be surprised if the screen were removed and both objects were there; that would be just what you expected. When Xu and Carey per- formed this demonstration with babies, they found a difference between the responses of 10-month-olds and 12-month-olds. The 12-month-olds looked longer at the cup than at both objects, and looking longer is what babies do when they are surprised. Thus, the 12- month-olds responded the way adults would. In contrast, 10-month-olds, as a group, showed no surprise at seeing only a single object and thus no evidence that they expected two objects to be behind the screen (Xu & Carey, 1995, 1996). (There were actually several trials with different sets of objects.) It is as if 10-month-olds perceive things as “objects” but not as separate, different objects. Twelve-month-olds, on the other hand, share the adult view that a cup and book are two different things, not one thing that takes on different forms as it moves through space.Children bring other conceptual understandings of the world to the word learning task. They not only understand that objects are individual things, they also understand that objects can be organized into categories of like things. This is relevant to word learning because, apart from proper nouns, words are labels for categories of things and not for individual things. Children also know something about properties of objects and relations between objects, which are the conceptual underpinnings of adjectives and rela- tional terms. Children can recognize motion events, categorize motion events by their similarity, and are also sensitive at a very early age to the different components of motion events that verbs encode (i.e., manner of motion and path of motion) (see Poulin-Dubois & Graham, 2007; Pulverman et al., 2006, for reviews). To the extent that children’s understanding of the world matches the way language works, the task of lan- guage acquisition is only to find the correspondences between sound and meaning. For instance, when a child hears the word “cup,” he does not consider the possibility that cup means “curved edge on a table top” or “empty of breakfast beverage.” The child already organizes the world into concepts that words encode. (In Chapter 5 and Chapter 8, we will see that this is an oversimplification. Sometimes children do have to learn the concept as they learn the word, and sometimes learning words may even drive conceptual development.)Domain-General Mechanisms of Learning and DevelopmentIf one wants to make the case that children acquire language using only domain-general mechanisms of learning and development—without recourse to specifically linguistic innate knowledge, then the task is to identify domain-general mechanisms that could do the job. Chomsky’s arguments in the 1960s and 1970s convinced the world that the associative mechanisms that were the learning mechanisms of B. F. Skinner’s radical behaviorism could not (Chomsky, 1959). But the associative mechanisms of learning that Chomsky rejected as inadequate are not the only mechanisms of learning or devel- opment available to the child.
The Piagetian Account of Language AcquisitionOne theory that researchers first looked to for a nonlinguistic basis of language acquisition is the theory of cognitive development articulated by Jean Piaget. Piaget was more interested in how children come to understand their physical world than in how children learn to talk, but he did have a position on language development. Piaget and others working within the Piage- tian framework (Sinclair, 1969, 1971) argued that language is one manifestation of the child’s symbolic function. This symbolic function arises from the child’s sensorimotor interactions with the world but, when developed, allows the child to have mental repre- sentations in the absence of any physical referent. Language, like deferred imitation and pretend play, is the expression of this new form of intelligence in the child.Several studies did find relations between the achievement of cognitive milestones described by Piaget and milestones of language development. For example, measures of symbolic play, such as holding a banana and pretending it’s a telephone, are related to measures of children’s language (Bates, Benigni, Bretherton, Camaioni, & Volterra, 1979; McCune, 1995; Rescorla & Goossens, 1992; Shore, O’Connell, & Bates, 1984). Ultimately, however, the approach of trying to account for language development in Piagetian terms did not prove fruitful. The sorts of predictions that could be made on the basis of Piagetian theory were not very precise, and different studies obtained dif- ferent findings depending on details of how cognitive skills were measured (Corrigan, 1979). In sum, the cognitive underpinnings of language development seem not to be captured by a Piagetian description of children’s cognition. In contrast, other more recent developments in the study of children’s domain-general learning abilities have revolutionized the field.Statistical Learning as the Mechanism of Language AcquisitionBy far, the most influential argument that nonlinguistic cognitive processes could underlie language acquisition comes from research on a learning process called statistical learning. The notion was introduced to the field in 1996 in a short article in the prestigious jour- nal Science, which reported that 8-month-old babies could learn something about the patterns in sound sequences they heard (J. R. Saffran, Aslin, & Newport, 1996). The babies in this study listened for 2 minutes to a recording that presented four different “words” combined in random order in a single stream of sound. The words were, for example, tupiro, golabu, bidaku, and padoti; and thus the babies heard something like tupirogolabubidakupadotibidakugolabubidakutupiro. Next, the babies were presented with strings of the same “words” again on some trials, and on other trials, they were pre- sented with strings of “nonwords” made up of the same syllables combined in different orders. The result was that the babies listened longer to the nonwords than to the words. This finding is consistent with other research showing that babies of this age prefer novel stimuli to familiar stimuli. This experiment is described in Box 3.3.The question is, how did the babies know the difference? The answer has to be that the babies noticed that in the first sequence presented tu was always followed by pi and pi was always followed by ro, but the ro was followed by three different syllables, equally often. In the nonword sequence, these transitional probabilities were different, even though the particular syllables presented as words and nonwords were the same (see also Aslin, Saffran, & Newport, 1998). In other words, the babies were doing what is called statistical learning—counting the frequency with which one stimulus is followed by another. This example is of the statistical learning of conditional probabilities. The learning of distributional properties of the input illustrated in Box 3.2 is another form of statistical learning. They are both aspects of learning the statistical properties of the speech signal—the frequencies of sounds and the co-occurrence patterns among sounds
BOX 3.3Learning Conditional Probabilities in InputSource: Saffran, Aslin, & Newport, 1996.The finding that infants could learn the conditional probabilities among syllables in the speech stream demonstrated that babies were more powerful learners than had pre- viously been thought. Related findings support the notion that although these mechan- isms are recruited for language acquisition, they are not language specific, because they can operate on nonspeech stimuli as well (Aslin & Newport, 2009). Subsequent research has pursued the question of how much of language acquisition statistical learning might explain. Studies with artificial stimuli have demonstrated that statistical learning could contribute to learning aspects of the phonology, the lexicon, and the syntax of language (see McMurray & Hollich, 2009; Saffran, 2009; for reviews). There is also evidence that babies can track the transitional probabilities in real speech, not just in artificial stimuli, which supports the idea that statistical learning contributes to real-world language acqui- sition (Pelucchi, Hay, & J. R. Saffran, 2009).Rule Learning and Language AcquisitionBabies may be able to do more than mere statistical learning. In 1999, another article in Science claimed that 7-month-old infants were capable of learning rules (Marcus, Vijayan, Bandi Rao, & Vishton, 1999). The babies in this study also heard sequences of syllables for 2 minutes. For half the babies, the sequences followed an ABA pattern (e.g., ga ti ga, li na li), and for the other half, the sequences followed an ABB pattern (e.g., ga ti ti, li na na). Then the babies heard sequences of entirely new syllables that either matched the pattern they had heard or matched the other pattern. So, for example, the ABA sequence was wo fe wo and the ABB sequence was wo fe fe. When the babies were presented with the new sylla- ble sequences, they were able to tell the difference between the pattern they had heard and the new pattern, even though they had never heard any of those sounds before. This experiment is described in Box 3.4. Marcus et al. (1999) claimed on this basis that babies can do more than learn the co-occurrence patterns among sounds they actually experience. Instead, babies can learn a pattern that must be described in terms of sym- bols (or variables) that stand for any sound. As Marcus and colleagues put it, the babies were learning algebraic rules, not just statistical regularities, although not everyone agrees
BOX 3.4Rule Learning: The Learning of Abstract Rules from InputSource: Marcus et al., 1999.that rule learning is the only way to explain the infants’ performance in this study (Christiansen & Curtin, 1999; Seidenberg & Elman, 1999).Other research has demonstrated that children between 1 and 2 years can learn pat- terns that are more abstract than just regularities in the particular stimuli they have expe- rienced. A series of studies by LouAnn Gerken and Rebecca Gómez have demonstrated that children can learn something like a grammar to capture the patterns in stimuli they hear (Gómez, 2002; Gómez & Gerken, 1999). For example, Gómez (2002) presented 18-month-olds with sequences of the form AXB and CXD, meaning only B-type items could follow A-type items and only D-type items could follow C-type items, but X could be anything. She then presented the infants with sequences that either conformed to this miniature grammar or that violated it. She found that the infants learned the grammar when lots of different items were presented in the X slot, but not when there were only a few X-type items. She interpreted this to mean that the infants tried to learn the pattern among the actual stimulus items and only learned the more abstract pattern when remembering all the sequences became impossible. This result and others suggest that babies are able to generalize beyond the particular stimuli they experience if the sti- muli they are presented require and support making generalizations (Gerken, 2007, 2009). The questions for research then become what sort of data are necessary to support making generalizations and are such data available in the speech children hear? We will consider those questions more fully in the second half of this chapter. For now, we add only the evidence that children’s inclination and ability to extract abstract patterns from the speech they hear does not depend on their data being perfect.Children can learn a system from imperfect data. The phenomenon of creolization, dis- cussed in Chapter 2, is one example of this. Another demonstration comes from a case study of sign language development in a child whose only exposure to sign was the signing
of his parents, who were not native signers (Singleton & Newport, 2004). The parents were deaf, but they had both learned American Sign Language (ASL) after the age of 15. Exper- imental tests of their ASL competence revealed they had a very imperfect mastery of verb morphology, and thus they could provide their child with only an errorful sample of verb morphology. Nonetheless, this child surpassed his parents in terms of his ability to use verb morphology and scored, on experimental tests, on par with children exposed to the signing of native users. Singleton and Newport suggest that the child was able to do this by finding and storing the regularities that did exist in his parents’ signing and essentially ignoring the data that didn’t fit the pattern. It appears from this result and others (Newport & Aslin, 2000) that children have available to them learning mechanisms that make them very good at extracting patterns from input, thus lessening the dependence of the language acquisi- tion process on either data that perfectly reveal the systematicity of language or innate knowl- edge that obviates the need for finding the systematicity in input.Memory and Attentional ProcessesImagine the young language-learning child sitting in his high chair, in the kitchen with his mother. His mother says, “Do you want marmalade on your toast?,” while she is reaching into the refrigerator. Meanwhile, the child is busy studying a bit of dried oat- meal stuck to the high chair tray, and the dog is barking in the background. The word marmalade is new to this child, and this interaction provides him the opportunity to learn that new word. But, in order to make use of this opportunity, he must be able to remember the new sound sequence and he must be able to direct his attention away from the dried oatmeal and the dog and to the cues to the meaning of this new word (it goes on toast, and it’s in the refrigerator). The ability to remember newly encountered sound sequences and the ability to coordinate attention are two components of memory that have established relations to language development.Phonological Memory Phonological memory is the capacity to remember newly encountered sound sequences. Some children are better at this than others, and individual differences in the ability to remember sound sequences predict individual differences in the rate of language development. The standard task for measuring phonological memory is the nonword repetition task. Children (or adults) are given nonwords such as grall, ballop, and brasterer and asked to repeat them. This can be done with children as young as 22 months by embedding the task in a game, and this procedure can also be used to measure individual differences in adults by making the stimuli longer (Coady & Evans, 2008; Dollaghan, Biber, & Campbell, 1995; Gathercole, 2006; Hoff, Core, & Bridges, 2008). Individual differences in the accuracy with which children repeat the sounds in these nonwords are related to their concurrent vocabulary, predict future vocabulary growth, and, among adolescents, predict vocabulary learning in foreign language instruction as well (Gathercole, 2006; Parra, Hoff, & Core, 2011; Service & Kohonen, 1995). Nonword repetition accuracy also predicts gram- matical development (Adams & Gathercole, 1995, 1996; Parra et al., 2011), and children with language disorders reliably score below typically developing children on nonword repetition tasks (Gathercole, 2006; Schwartz, 2009).In sum, the component of working memory that provides temporary storage of auditory stimuli, known as phonological memory or the “phonological loop,” is one of the capacities that the child brings to the language learning task. It is also a capacity that develops as a result of experience. Although phonological memory was first conceptualized as an unlearned, unchanging cognitive capacity (Baddeley, Gathercole, & Papagano, 1998), more recent treatments incorporate the idea that phonological memory relies on the quality of children’s phonological representations, which are built on the basis of language experience. Children who are better at identifying phonemes and producing rhymes show betterphonological memory skills (Bowey, 2001); and adolescents show better memory for sound sequences that conform to the language they know than for sound sequences in a foreign language (Service & Kohonen, 1995). In bilingual children, accuracy of nonword repetition is related to children’s exposure to that particular language (Parra et al., 2011; Summers, Bohman, Gillam, Peña, & Bedore, 2010). This relation of language experience to phonologi- cal memory should not be surprising as memory capacity in other domains has been shown to depend on a representation system for encoding the to-be-remembered stimuli (Chase & Simon, 1973; Schneider, Gruber, Gold, & Opwis, 1993).Central Executive Function in Working MemoryThe central executive is the component of working memory that allocates mental resources among currently com- peting demands. Several accounts of children’s language abilities make reference to the notion that children (like adults) are limited-capacity processors and that performance is not just a matter of having or not having the required knowledge, it is also a matter of being able to handle the multiple demands of the task. For example, infants who can discriminate two sounds are not necessarily able to distinguish new words on the basis of that sound distinction (Werker, Cohen, Lloyd, Casasola, & Stager, 1998; Werker & Curtin, 2005). As another example, very young children’s contributions to conversation tend to be less related to the ongoing topic than older children’s—not because of dif- ferences in linguistic skill, per se, but rather, according to Bloom and colleagues, because of differences in the ability to allocate cognitive resources. The argument is that it is cognitively more demanding to pay attention to the ongoing context and also formulate a message than it is just to formulate a message and that younger children have trouble managing those multiple demands of conversation (Bloom, Rocissano, & Hood, 1976).The studies that have demonstrated a relation between central executive function and language have compared children with language disorders to typically developing children on tasks that measure the ability to handle competing demands. For example, in the Com- peting Language Processing Task (CLPT) children are presented with sentences such as “Pumpkins are purple,” “Fish can swim” and they have to do two things with these sen- tences: They have to say whether the sentences are true or false and they have to remem- ber the last word in each sentence for later recall. Several studies have found that children with language disorders have more difficulty with this sort of task than do typically devel- oping children (Montgomery, Magimairaj, & Finney, 2010; Schwartz, 2009). One study using the CLPT procedure also found that among the language disordered children (but not the typically developing children), those who had the most difficulty with the CLPT task also had the most difficulty with the sentence comprehension task (Montgomery & Evans, 2009). The argument that individual differences in some attentional processes con- tribute to individual differences in language development among typically developing chil- dren is supported by findings that measures of infants’ orientation to new stimuli and measures of their sustained attention to those stimuli at 3 to 9 months of age predicted measures of those children’s vocabularies at 2 years and 4 years of age (Colombo, Shaddy, Richman, Maikranz, & Blaga, 2004; Colombo et al., 2009).Memory, Sleep, and Language LearningMemories are consolidated during sleep, and in a variety of domains, adults and infants have been found to show improved mem- ory or improved performance when there is a period of sleep in between training and test (Fagan & Rovee-Collier, 1983; Gómez, 2011). Sleep seems to benefit language learn- ing as well. Babies who nap after being exposed to sound sequences remember the abstract pattern underlying those sound sequences better than babies without a postex- posure nap (Gómez, Bootzin, & Nadel, 2006), and this benefit of napping is observable even after a 24-hour delay (Hupbach, Gómez, Bootsin, & Nadel, 2009).
head in one direction, and the adult-directed speech tape was played if the babies turned their head in the other direction. The infants chose to hear infant-directed speech more frequently than they chose to hear the adult-directed speech. This preference has also been demonstrated in newborns and 1-month-olds (Cooper & Aslin, 1990), and it holds true across languages—English-learning babies prefer infant-directed Cantonese to adult-directed Cantonese and Cantonese-learning babies prefer infant-directed English to adult-directed English (Werker, Pegg, & McLeod, 1994).What is it about infant-directed speech that makes it so interesting to babies? One possibility is simply that the exaggerated intonation contours produce a stimulus with high contrast. According to this view, babies like infant-directed speech for the same rea- son they like bold colors and black on white patterns (Vihman, 1996). Fernald (1992), on the other hand, has suggested that infant-directed speech and infants’ responses to it have a deeper basis. She argues that mothers’ speech to infants functions like the calls of other species, directing infants’ attention and calming or arousing them, depending on the particular call. Fernald points out that when mothers say things like “No!” or “Don’t touch that!” the intonation is very different from when they say things like “Good!” or “Clever girl!” These different intonations for both prohibitions and praise tend to be much the same across different languages, and adults can tell the difference between prohibitions and approval statements addressed to babies based on intonation alone (Fernald, 1989).Speech to infants and children is not only more interesting than adult-directed speech, it is also more informative. When mothers talk to infants, their vowels are particularly clean examples of the vowel being produced (Kuhl, 1999; Kuhl, Andruski, Chistovich, & Chistovich, 1997). Also, the slow rate of infant-directed speech causes vowels to be prolonged, which makes them more discriminable. Infants appear to make use of this, because mothers who produce more discriminable vowels in their infant-directed speech have infants who demonstrate better speech perception skills in laboratory tests (Liu, Kuhl, & Tsao, 2003). In talking to children, adults often produce single word utterances, and the presence of isolated words helps children segment the words from the speech stream (Lew-Williams, Pelucchi, & Saffran, 2011). The exaggeration of prosody in infant-directed speech may make it a particu- larly good database for language acquisition because prosodic features of speech coin- cide with structural properties of the language. There is experimental evidence that infants use prosodic cues in segmenting speech (Seidl, 2007; Seidl & Cristiá, 2008), and that at least very young language learners (aged 21 months) learn new words pre- sented in infant-directed speech but not when the same words are presented in speech with the intonation pattern of adult-directed speech (Ma, Golinkoff, Houston, & Hirsh-Pasek, 2011).The content of child-directed speech is also more accessible than the content of speech among adults. We tend to talk to young children about what is currently going on rather than about past or future events. In addition, adults use gestures to secure children’s atten- tion, so the speech children hear is likely to be about the things the child is focusing on (Zukow, 1990). These features of child-directed speech make the child’s task of mapping words to their referents easier than if the child were watching speeches on the Senate floor, for example. Furthermore, in talking to children, adults also provide explicit instruc- tion about word meanings. Adults offer new labels (e.g., “That’s a chair”) and corrections of children’s imprecise lexical choices (e.g., Ch: “Mommy, where my plate?” M: “You mean your saucer?” Ch: “Yeah”). Adults also provide information about word meanings (e.g., “It’s called an eel. It’s like a snake, only it lives in the water”) (Chouinard & Clark, 2003; S. A. Gelman, Coley, Rosengren, Hartman, & Pappas, 1998).
Speech to children is very repetitious. Caregivers say things like, “Put the doll in her crib. Yes, the doll. That’s right, in her crib.” They also repeat and expand children’s utterances:Child: Milk. Adult: You want some milk?These repetitions and expansions of children’s incomplete utterances might serve as little language lessons, revealing the component structures that make up sentences (Onnis, Waterfall, & Edleman, 2008).The special characteristics of child-directed speech are less pronounced in speech to older children than in speech to infants; in particular the exaggerated prosody of infant-directed speech diminishes (Liu, Tsao, & Kuhl, 2009). Other differences remain, however, and as children acquire more experience with language, they appear to be able to discriminate speech likely to be addressed to them on bases other than prosody. For example, infant- and child-directed speech has very few grammatical errors (Newport, Gleitman, & Gleitman, 1977) and few disfluencies, such as pauses filled with uh or um). By the age of 22 months, although not at 10 months, children notice the difference between speech with and without such disfluencies, and they prefer fluent to disfluent speech (Soderstrom & Morgan, 2007).The Role of Feedback In 1970, Roger Brown and Camille Hanlon studied the mothers’ speech in the transcripts of Adam, Eve, and Sarah, and they found that the mothers did not correct their children’s ungrammatical utterances. The mothers did cor- rect factual errors, mispronunciations, “naughty” words, and some overregularizations, such as goed. However, syntactic errors such as “Why the dog won’t eat” passed without comment. On the basis of this finding, the general consensus for a long time was that children do not receive negative feedback. More recently, others have looked to see whether parents respond differentially to children’s grammatical versus ungrammatical utterances in ways that are more subtle than overt correction but are nonetheless poten- tially useful to the child. Researchers have found that adults are more likely to repeat verbatim children’s well-formed sentences than their ill-formed sentences, repeat with corrections children’s sentences that contained errors, and ask for clarification of ill- formed sentences (Bohannon & Stanowicz, 1988; Chouinard & Clark, 2003). They have also found that adults reformulate their children’s errorful utterances more frequently than they repeat or revise their error-free utterances.Just how much children use such feedback is a matter of debate. The fact that researchers can find a statistical difference between responses to grammatical and ungrammatical utterances by looking at hundreds of children’s utterances doesn’t mean that children could ever discern a pattern in their parents’ behavior. Marcus (1993) has estimated that a child would have to say the same ungrammatical sentence 85 times to have enough data to determine that the sentence was ungrammatical. Chouinard and Clark (2003) argue, in contrast, that children interpret it as a correction when the adult expresses the same meaning as the child just has, but uses a different form because chil- dren frequently directly acknowledge such corrections or use the new forms in their sub- sequent utterances. Saxton and colleagues found evidence that, in some cases, children tend to show improvement that is related to how much implicit feedback they receive (Saxton, Backley, & Gallaway, 2005).The issue of whether or not children get feedback is important if language acquisition is a process of testing hypothesized grammars against evidence from the input. No feed- back means no negative evidence—that is evidence that the hypothesized grammar is
wrong. This was the model of language acquisition in the 1970s, and the putative absence of negative evidence was crucial in theoretical debates (see Marcus, 1993). How- ever, the issue may be moot if the process of language acquisition is one of finding pat- terns in the input. Statistical learning does not rely on feedback, only on exposure to stimuli that contain regularities (Aslin & Newport, 2009). Furthermore, infants seem able to distinguish stimuli that contain learnable patterns from those that do not. Gerken and colleagues found that 17-month-olds attended longer to stimuli with learnable pat- terns than to stimuli with unlearnable patterns (Gerken, Balcomb, & Minton, 2011). Thus, children might be able to monitor their own progress in finding the pattern, anal- ogous to the way adults can tell if they are making progress on a crossword or jigsaw puzzle—and continuing only if things are fitting.The Role of Maternal ResponsivityIn Western cultures, mothers typically treat babies as conversational partners from birth. Young infants do not do much in terms of holding up their end of the conversation, but mothers build conversational sequences around the smiles, burps, and other noises their infants do produce. The following exam- ple from Snow (1977) illustrates this kind of interaction.3-Month-Old: (smiles) Mother: Oh what a nice little smile! Yes, isn’t that nice? There. There’s a nice little smile. 3-Month-Old: (burps) Mother: What a nice wind as well! Yes, that’s better, isn’t it? Yes. 3-Month-Old: (vocalizes) Mother: Yes. Yes! There’s a nice noise.Researchers have suggested that through such interactions, adults draw infants into communicative exchanges that become the basis for the later emergence of true inten- tional communication (Camaioni, 1993; J. L. Locke, 1993). Snow (1984) has described this phenomenon as mothers “pulling intentionality out of the pre-intentional child.” Experimental findings support the view that babies notice the effects of their vocaliza- tions. When researchers instructed mothers to keep their faces completely still while looking at their infants (something that mothers do not normally do and that infants clearly do not care for), the infants increased the rate of their vocalizations (Goldstein, Schwade, & Bornstein, 2009).Not only are mothers’ responses to vocalizations noticed by their preverbal infants, these responses may lay the groundwork for subsequent verbal interaction. Bell and Ainsworth (1972) studied the responsiveness of 26 mothers to their infants’ crying during the first year of the infants’ lives. Observers visited each home at approximately three-week intervals, recording the frequency of infant crying episodes, the length of time the infant cried without obtaining a maternal response, and the effectiveness of maternal interven- tions at calming the baby. Bell and Ainsworth found that infants who had the most responsive mothers when they were 6 to 12 months old cried less at the age of 12 months than did the infants with less responsive mothers. Also, the infants who cried less at 12 months were more communicative in terms of both their vocalizations and their nonvocal behavior. In the Goldstein et al. (2009) experiment in which babies increased vocalizations to their mothers’ still faces, the infants who showed a bigger increase (i.e., found no maternal response more disturbing) in the experimental setting at 5 months showed more advanced language comprehension at 13 months. Even for somewhat older
children, experiencing a responsive mother (or other caregiver) appears to promote lan- guage development. For example, children whose mothers are responsive to their vocaliza- tions at 13 months produce their first word earlier and reach a 50-word vocabulary earlier than children who have less responsive mothers (Tamis-LeMonda, Bornstein, Kahana- Kalman, Baumwell, & Cyphers, 1998).The Relation of the Availability of Environmental Support to Language AcquisitionInput as a Source of Individual Differences in Language DevelopmentIf lan- guage acquisition depends on environmental support, then children who receive different levels of environmental support should show differences in the rate or outcome of their language development. There is abundant evidence that this is the case. Differences among children in the communicative experiences they have with parents, teachers, and peers all have been shown to produce differences in the rate of children’s language devel- opment and thus in the level of language skill they display.Many different properties of children’s experience influence their language develop- ment. There are effects of responsivity, as just described. In addition, there are also effects of what is called contingency. Mothers who not only respond, but respond in a manner that is related to the child’s previous verbal or nonverbal move appear to benefit their children’s language development. For example, the frequency with which mothers follow their 13-month-old children’s focus of attention in their talk is related to those children’s vocabularies at 22 months (Akhtar, Dunham, & Dunham, 1991). Also, chil- dren whose mothers more frequently followed their attentional focus in interaction at 18 months had more advanced language at 30 months (Adamson et al., 2004).The amount of speech children hear is related to the rate at which their language skills grow. Children whose mothers address a great deal of speech to them develop vocabulary more rapidly and are faster at processing the language they hear than children whose mothers talk to them less (Hart & Risley, 1995; Hurtado, Marchman, & Fernald, 2008; Huttenlocher, Haight, Bryk, Seltzer, & Lyons, 1991). The number of different words mothers use, the number of different grammatical structures they use, and the grammatical complexity of their speech are all positive predictors of their children’s vocabulary, grammatical development, or both (Hoff, 2003a, b, and c; Huttenlocher, Waterfall, Vasilyeva, Vevea, & Hedges, 2010). Language development is also fostered by conversational experience. The number of conversational exchanges children experience can be more predictive than the amount of speech they hear (Zimmerman et al., 2009).Properties of the speech children hear from their teachers also have effects on chil- dren’s language development. The vocabulary preschool teachers use predicts both chil- dren’s currently developing oral language skills and also their reading and language skills in the fourth grade (Dickinson & Porsche, 2011). The syntactic complexity of teachers’ speech in kindergarten classrooms is a positive predictor of the children’s syntactic development over the course of the school year (Huttenlocher et al., 2002). As is the case for home interactions, it is not just properties of teachers’ speech that matter; it is also the conversational exchanges that occur. Children benefit from classrooms with more teacher-child exchanges and less teacher-only talk (Dickinson, 2012).The language children hear from their peers also influences their language develop- ment. Children who attended pre-kindergarten with other children who had high expres- sive language skills developed more over the course of the pre-K year than children who attended Pre-K with other children who had low expressive language skills (Mashburn, Justice, Downer & Pianta, 2009). The less-advanced children were more affected by their
FIG-3-12 Speed of word recognition is influenced by the amount of speech children hear. The two bars in the graph show that children whose mothers talk to them less during a recorded play session (the “fewer words” group) take longer to recognize words than children whose mothers talk to them more (the “more words” group). These monolingual Spanish-learning children were presented with the sentence, Donde está el perro? (Where is the dog?), and the time it took them to look at the picture of the dog was measured in the looking-while-listening task.¿Dóndeestáel PERRO?Fewer wordsMore words0 200 400 600 800 1,000 1,200 Time in ms from noun onsetSource: Hurtado, N., Marchman, V. A., and Fernald, A. (2007). “Spoken Word Recognition by Latino Children Learning Spanish as their First Language.” Journal of Child Language, 34, 227–249. Copyright © 2007 Cambridge University Press. Reprinted with the permission of Cambridge University Press.classmates’ level of language skill than the more advanced children (Justice, Petscher, Schatschneider, & Mashburn, 2011).Input as a Source of Differences in Language Development Related to Socioeconomic Status If individual differences in environmental support create individual differences in language development, then group differences in environmen- tal support should create group differences in language development. There is good evidence that this is the case for children who differ in socioeconomic status (SES). (By socioeconomic status, we mean, essentially, parents’ education, income, and occu- pational prestige.) Children in higher SES homes are talked to more, and the speech they hear is more supportive of language development than is the case in lower SES homes (Fernald & Marchman, 2011; Fernald, Marchman, & Weisleder, 2012; Hoff, 2006; Hoff, in press; Hoff, Laursen, & Tardif, 2002). Higher SES parents also use ges- ture more in communicating with their children (Rowe & Goldin-Meadow, 2009a). These SES-related differences in children’s communicative experience create SES- related differences in children’s language development (Fernald & Marchman, in press; Hoff, in press).In one of the most famous studies to demonstrate SES-related differences in talk to children and in children’s language development, researchers Hart and Risley (1995) studied 42 children from the time they were 9 months to 3 years of age (and beyond, in subsequent studies). Every month for 21⁄2 years, researchers went into the children’s homes and recorded an hour of the children’s daily lives. Everything that was said to the child and everything that the child said was then transcribed for analysis
The children came from three different socioeconomic strata: 13 came from high SES homes in which the parents were professionals (doctors, lawyers, professors); 23 came from middle to lower SES homes in which the parents worked in offices, in construction, in factories, and so on; and 6 children came from homes in which the parents were on public assistance. Group differences in the children’s vocabulary size were noticeable from almost the beginning of speech, and they increased with development. By 3 years of age, the mean cumulative recorded vocabulary for the higher SES children was over 1,000 words and for the lower SES children it was close to 500.There were also SES-related differences in the amount the parents talked to their chil- dren and in the quality of that talk. Estimates of the children’s language experience based on the recorded samples suggested that over the course of 1 week, the children of high SES parents heard 215,000 words, the children of the middle SES parents heard 125,000 words, and the children of the lower SES parents heard 62,000 words. By the time the children were 3 years old, the children in the higher SES families heard some 30 million more words than did the children in the lower SES homes (Hart & Risley, 1995). Not only did the higher SES parents talk much more to their children, they also displayed to their children a richer vocabulary and more complex structures. Also, when the high SES parents talked to their children they were more likely to give their child affirmative feedback than were the lower SES parents. The speech of lower SES parents to their chil- dren contained more negative imperatives (e.g., “Don’t,” “Stop,” “Quit”). Findings are shown in Figures 3.13 and 3.14.Hart and Risley’s findings are striking, but they are not unique. SES-related differ- ences in how mothers talk to children have been found by many different researchers in many different countries (Hoff, Laursen, & Tardif, 2002). SES-related differences in children’s language development are also reliably found, and several studies have done the statistical analyses required to show that differences in language experience are anFIG-3-13 The rate of vocabulary growth in the first three years of life is different for children from different socioeco- nomic backgrounds. The graph shows the growth in the number of differ- ent words children were heard to produce for children from profes-600 sional, working class,1,200 1,00013 higher SES children (professional)80023 middle/lower- SES childrenand families on public assistance.(working class)6 welfare children400 200010 12 14 16 18 20 22 24 26 28 30 32 34 36 Age of child in monthsSource: Hart & Risley, 1995.
FIG-3-14 The amount of time adults spend interacting with children and the amount of speech addressed to children also differs as a function of socioeco- nomic status, and these differences in children’s20 experience are a large part of why the chil-10 dren’s language skills differ.0 10 14 18 22 26 30 34Age of child in months912131617202124252829323336 Age of child in monthsSource: Hart & Risley, 1995.50 40 3013 professional parents23 working- class parents6 welfare parents2500 2000 1500 1000500 013 professional parents23 working- class parents6 welfare parentsimportant part of why children from different socioeconomic strata have different levels of language skill (Hoff, 2006; Hoff, in press; Huttenlocher, Waterfall, Vasilyeva, Vevea, & Hedges, 2010; Vasilyeva, Waterfall, & Huttenlocher, 2008).It is important, to point out, however, that there are also differences within socio- economic strata in how much adults talk to their children and in the support they provide for language development. Huttenlocher et al. (1991) found differences in the amount mothers talked to their children that predicted children’s vocabulary growth within a middle-class sample. Other studies have found differences in mothers’ speech and consequent differences in children’s language within low SES samples (Pan et al., 2005). In both Hart and Risley’s study and in a study of high and mid SES families by Hoff (2003), there were differences in maternal speech and resulting differences in the children’s language skills within each level of SES, as well as between groups.The idea that children find significant support for language acquisition in their com- municative interactions with others is not new. In making the argument that language acquisition was not accomplished by a Language Acquisition Device (LAD) alone,
