Methods for Studying Language in Infants: Back to the Future
Golinkoff, R. M. & Hirsh-Pasek, K. (2012). Methods for studying language in infants: Back tothe future. In E. Hoff (Ed.), Guide to research methods in child language. NY: Wiley-Blackwell. 60-77.
Methods for studying language in infants: Back to the future
Roberta Michnick GolinkoffKathy Hirsh-Pasek
University of DelawareTemple University
Date: September 2, 2010
We thank Cora-Lee Picone and Sujeet Ranganathan for assistance with the bibliography of this chapter and Kelly Fisherand Erika Hoff for their helpful comments. This research was funded by joint grants to the authors: From NSF, SBR9615391 and from NIH, RO1HD050199.
Methods for studying language in infants: Our history and our future
These one-word speakers appear to be five- or six- word listeners…by 17 months of age …infants in the beginning stages of language learning can attend to syntactic cues [word order] and use these cues to distinguish between two relatively similar scenes in their environment
(Hirsh-Pasek & Golinkoff, 1996, p. 122).
…this study indicates that long before infants are producing wh-questions, they understand them in ways that few would have suspected…Infants responded appropriately to subject questions by 15 months of age and to both subject- and object-questions by 20 months
(Seidle, Hollich, & Jusczyk, 2003, p. 423, 434).
This prospective longitudinal study ... shows that individual differences in the efficiency of spoken language comprehension at the age of two years predicts children’s success in cognitive and language tasks in later childhood (Fernald, Zangl, Porillo, & Marchman, 2008. p. 131).
Researchers who study language acquisition wish to understand the very origins of the process, beginning from the earliest stages. To achieve this goal, organisms who cannot speak or even follow commands must be “tricked” into revealing what they know. This is the challenge our field has faced and, to some extent, conquered. Before the advent of methodologies to probe young children’s linguistic capabilities, the aforementioned findings would have been inconceivable.
In this chapter, we reflect on the recent history of methodological innovation and ponder how these methods and newer ones might shape the future of our field. The last 40 years took us well beyond counting and cataloguing children’s language productions (Carmichael & Mussen, 1954) as researchers began to probe children’s knowledgeof the elements and structures of language, and how this knowledge is used to bootstrap language competence – even before a single word is produced. We can only be selective in our brief review of the field’smethodological innovations, highlighting some of the key paradigmsand their significance.
In the Beginning
An intervention conducted in ancient Egypt may be the first study of language acquisition. Two infants were reared in silence to see which language would emerge first (Dale, 1972). While this methodology would now fail to pass human subjects scrutiny, reports of children raised without language input (Genie, “home-sign” deaf children) have contributed to our understanding of language development, e.g., Genie’s case (Curtiss, 1977; Rymer, 1993; Goldin-Meadow & Mylander, 1984).Baby biographies (e.g., Darwin, 1877; Stern Stern, 1907) gave way to more sophisticated diary studies(Dromi,1987; Halliday, 1975; Mervis, Mervis, Johnson, & Bertrand, 1992; Tomasello, 1992; Naigles et al., 2009) enabling us to paint a rich portrait of the growing language system. Studies by Brown(1973),Bloom (1970, 1973), and Nelson (1973) on children’s first words and word combinations were landmarks for the field and the CHILDES data exchange system (MacWhinney, 1991), available on the web, continues to illustrate the value of diary data.
Methods Moving Beyond Production
Two developments enabled researchers to move beyond observations of what children could produce. One development was work by Fantz on infants’ visual acuity; the other was work on the “orienting response” by Russian psychologists (Sokolov, 1963, 1969) that lead to the creation of the habituation method.
Fantz’s work. At a time when pediatricians believed that infants were incapable of much vision, Fantz (1961a, b) proved them wrong. Presented with two cards, one containing broad and one narrow stripes, Fantz showed that infants preferred to look at the cards with more stripes. As Fantz and Nevis (1967) wrote, “…it is now proven that even in the early weeks the young infant can resolve and discriminate patterns …(p. 78). These data suggested that infants’ acuity exceeded what was commonly believedand that infants were selective. Indeed, Fantz and Nevis argued that this selectivity “… is particularly revealing of early perceptual-cognitive development” (p. 78). Fantz and Nevis saw the potential of the paired comparisons method for predicting infants’ later capabilities, ending their article with “a bit of crystal-ball gazing” when they wrote, “Could it that the infant’s future prospects, as well as his past experiences and present interests, are reflected in his eyes?”
The orienting response. Russian psychologists studied what Pavlov referred to as the “What-is-it-reflex,” having noticed that his lab dogs turned in the direction of a new stimulus. Called the “orienting reflex” (Sokolov, 1963, 1969), both humans and dogs would literally and figuratively ‘sit up and take notice’ at the first presentation of a novel stimulus. The orienting reflex, consisting of a variety of physiological and behavioral responses such as heart rate change, turning toward the new stimulus, eye movements, etc. (Cohen, 1976), is the basis of habituation and familiarization methods. Sololov grounded habituation in neurological functioning by claiming that it reflected not just moment-to-moment processing but comparison to existing memory traces and the build up of new ones (Bornstein, 1988).
Methods Multiply
Once researchers realized that they could capitalize on infants’ attentional responses additional methods appeared. There is no question that we would not be able to ask the sophisticated questions we do today were it not for the advent of these methods. Just as in chemistry and other sciences, new discoveries follow on the heels of the emergence of new methods. Nor is the list below complete; it favors the most frequentlyused methods.
The sucking response. The orienting response figured in work by Eimas, Siqueland, Jusczyk, and Vigorrito (1971). Eimas et al. shocked the field by showing that infants but a few weeks of age could discriminate between phonemes and demonstrate categorical perception for speech sounds. After infants’ sucking habituated to hearing the same sound in isolation repeatedly, a new sound (either in the same phonemic category or from a new category) was played. Infants then showed an orienting response to the out-of-category sound, recovering their sucking response, and revealing that theycould discriminate between phonemes. This important research told the field that infants begin language learning with some categories in place. However, as sucking wanes in its utility after about 4 months of age, other methods were needed by researchers to continue to explore infant language discrimination and categorization.
Heart rate deceleration. Kagan and Lewis (1965) published the first longitudinal study that used of a variety of responses (visual fixation, vocalizations, arm movements, and heart rate) to assess infants’ processing of both visual and auditory stimuli. Cardiac deceleration was touted as being a way to distinguish between “the empty stare” and “active assimilation” (p. 96) and a way to assess attention to auditory stimuli.
In one fascinating experiment, Kagan and Lewis (1965) crossed meaningfulness with ‘inflection’ to present 13-month-olds with 4 kinds of paragraphs: High meaning-high inflection (three complete sentences containing highly frequent words such as baby, mommy, daddy;read with normal intonation); high meaning-low inflection (one word per second; flat intonation); low meaning-high inflection(nonsense words; normal intonation): and low meaning-low inflection(nonsense wordsread one word at a time). Girls showed the greatest deceleration to high meaning-high inflection and to low meaning-high inflection paragraphs, suggesting that girls were motivated to find the meaning and perhaps attracted to infant-directed speech. Boys, on the other hand, showed the greatest deceleration to the high meaning-low inflection paragraph, as if needing to hear the word-by-word presentation to find meaning.
More provocative are the longitudinal links between attention to auditory stimuli (music and tones) at 6 months and attention to the paragraphs at 13 months. Boys showed low attention stability. For girls, high cardiac deceleration at 6 months predicted to greater attention to the novel paragraphs (both intonation types) with low meaning. These data suggest that sensitivity to nonlinguistic auditory stimuli at 6 months relates to infant language processing at 13 months.
Visual fixation. In 1974, Horowitz edited a monograph that asked whether visual fixation might be used as a window into language development. Visual fixation was so new as a dependent variable for this purpose that an appendix was devoted to procedures forestablishing an infant laboratory. Culp(1974), Self(1974), and Padden(1974) discovered that for 10- to-12- week-olds, introducing auditory stimuli after habituation increased visual fixation time and afforded assessment of infants’ discrimination between auditory stimuli. Horowitz suggested that, “…visual response decrement and recovery may be a very useful procedure for studying infant receptive language abilities” (p. 111).
Infant language capabilities are more than about the auditory signal itself. Learning language entails forming the concepts languages encode in their semantic structures, such as who does what to whom (Fillmore, 1968). Many theorists believed semantic roles were the first way that infants understood sentences (Golinkoff, 1981a; Pinker, 1984). Golinkoff (1975) presented infants with silent, filmed, dynamic events using familiarization and recovery of the visual fixation response or a combination of visual fixation and heart rate deceleration (Golinkoff & Kerr, 1978). Since these studies preceded videotape, human actors were filmed in super 8mm film. Both studies pitted a change in the roles of agent and patient against a change in the event’s directionacross the screen. Fourteen-month-olds watched action role changes more than direction changes indicating that they could tell agents from patients (Schoppner, Sodian, & Pauen, 2006).
The link between dynamic events and auditory stimuli appeared in Spelke (1976). Sheshowed infants two simultaneous visual displays (a donkey jumping up and down versus a person clapping her hands) accompanied by a single auditory stimulus matching only one of the displays in the paired comparisons method (Fantz, 1961a,b). When 4-month-olds watched the target – the matching event -- more than the non-target, it became clear that this method had great promise for testing language comprehension -- promise that Golinkoff, Hirsh-Pasek, Cauley, and Gordon (1987) capitalized on bydeveloping the “Intermodal Preferential Looking Paradigm” (IPLP) for studying language comprehension (see Hirsh-Pasek & Golinkoff, 1996; Piotroski & Naigles). That first paper validated the IPLP by testing 17-month-oldson the comprehension of nouns and verbs and 28-month-olds -- already using multi-word utterances -- on the use of word order in sentence comprehension (e.g., “Big Bird is tickling Cookie Monster.”
Hirsh-Pasek and Golinkoff (1996) then showed that infants as young as 13 months of age recognized that words come in ‘packages,’ specifying unique events in the world. When shown a video of a woman kissing a set of keys and holding a ball in the foreground vs. the same woman kissing the ball and dangling the keys in the foreground, babies looked more at the matching event when they heard, “She’s kissing the keys!” This result could only have emerged if infants were doing more than processing individual words as both videos contained a “she,” the action of “kissing,” and the target item (keys). By 17 months, infants, with as few as 2 words in their productive vocabularies, could use word order as a guide to watch specific events that matched the linguistic stimuli (Hirsh-Pasek & Golinkoff, 1996) – if they knew the names of the Sesame Street characters. Thus, infants not yet speaking were capable of comprehending not only action role relations in language but probably the grammatical categories of subject and object of the sentence.
Hollich, Hirsh-Pasek, and Golinkoff (2000) introduced a three-dimensional version of the IPLP -- the Interactive Intermodal Preferential Looking paradigm. It presented real objects to infants (10 to 24 months) and allowed for the manipulation of social cuesin an online word learning task. Pruden, Hirsh-Pasek, Golinkoff, and Hennon (2006) and Hollich et al. (2000) reported that the cues infants used for word learning changed over time, moving from perceptual salience, or the attractiveness of the object, to the use of social cues such as eye gaze. Nurmsoo and Bloom (2008), using another paradigm, showed that by preschool, children do not blindly follow eye gaze but switch to using linguistic cues when these are put into conflict, as the emergent coalition model (Hollich et al., 2000) predicts.
Yet another use of the IPLP is a nonlinguistic version (Pruden, 2006) that presents single and paired dynamic events. It is designed to assess which relationsinfants can discern that they will later express linguistically (Goksun, Hirsh-Pasek, & Golinkoff, 2010; Golinkoff & Hirsh-Pasek, 2008). Infants appear to form a range of nonlinguistic categoriesthat sometimes change as they learn the particular perceptual distinctions their language encodes. Concepts such as containment and support, path and manner, figure and ground, source and goal have been explored with the nonlinguistic IPLP or single stimulus habituation studies (e.g., McDonough, Choi, & Mandler, 2003;Pulverman, Sootsman, Golinkoff, & Hirsh-Pasek, 2008; Pruden, Goksun, Roseberry, Hirsh-Pasek, & Golinkoff; 2010a; Pruden, Goksun, Roseberry, Hirsh-Pasek,& Golinkoff, 2010b;Goksun, Hirsh-Pasek, and Golinkoff, under review; Lakusta, Wagner, O’Hearn, & Landau, 2007; Hespos & Spelke, 2004; Casasola & Cohen, 2002; Casasola, Cohen, & Chiarello, 2003). Thus, methods relying on visual fixation are probing infants’ perceptionof the event components that languages encode and, in some cases, showing language heightening or dampening these distinctions (e.g., Goksun et al., under review).
Conditioned head turn and the head turn preference procedure. Colombo and Bundy (1981) offered a procedure to assess infant auditory selectivity, which was a forerunner of the Head Turn Preference procedure (Fernald, 1985; Hirsh-Pasek, Kemler-Nelson, Jusczyk, Cassidy, Druss, & Kennedy, 1987; Kemler-Nelson, Jusczyk, Mandel, Myers, Turk, & Gerken (1995). Two - and 4.5-month-olds, in an auditory analogue to Fantz’s (1956) paired comparisons method, were offered the opportunity to look at one of two red light rings composed of light diodes placed approximately 1.3 feet apart. A look at one of the rings played a female voice; alook at the other ring, yielded silence or white noise. Thus, this was an “infant-control” method, as children could make the stimulus begin. Two-month-olds did not look selectively but 4.5-months-olds preferred the voice over either silence or white noise. Columbo and Bundy wrote, “The finding that, by 4 months, infants show preference for the female voice over both silence and white noise is not an intuitively surprising one, but the demonstration of a valid auditory selectivity paradigm is of more interest” (p. 222).
Werker, Polka, and Peg (1997) suggested that the conditioned head turn procedure was adapted from work testing auditory perception in preschoolers(Dix & Hallpike, 1947;Suzuki & Ogiba, 1960). Eilers (1977) and Kuhl (1985) adapted the procedure to test infants’ perception of sounds by conditioning infants to turn their heads to see a reinforcer (e.g., a bear playing cymbals) when they detected a change in the auditory stimulus(Kuhl, Williams, Lacerda, Stevens, & Lindblom, 1992). For example, babies were able to override changes in voices, gender of the speaker, and age of the speaker to turn to the reinforcer only when a vowel sound changed from /i/ to /a/ (Kuhl, 1983; and see Trehub, Bull, & Thorpe 1984; Trehub, Schneider, Thorpe, & Judge, 1991).
Fernald (1985)used a head turn procedure with loudspeakers placed to the left and right behind infants, training them to activate an auditory stimulus by turning theirheads. Four-month-olds infantsturned their heads more to hear passages spoken in infant-directed versus adult-directed speech. Prior to this time, studies had examined infants’ sensitivity to phonological aspects of language, taking a bottom-up approach. Fernald’s (1985) study suggested that infants might be sensitive to larger stretches of text than just individual phonemes or syllables.
The Headturn Preference Procedure further increasedin utility and sensitivity in a study designed to evaluate whether 7- to 10-month-old infants might be sensitive to the temporal and intonational properties of passages (Hirsh-Pasek, Jusczyk, Kemler Nelson et al., 1987). Hirsh-Pasek et al. used a passage of infant-directed speech in which one-second pauses were either inserted at random points or at clausal boundaries where they are usually found. Their methodological innovation was to code not only the side to which infants turned to hear a language sample, but also how long infants attended to a side after turning to it. They found that even when the number of head turns to a side did not differ significantly, the duration of infants’ looks revealed a clear preference for the natural passage. Using this method a developmental sequence emerged: Infants preferred to hear appropriately segmented clauses before they preferred to hear appropriately segmented phrases (Jusczyk, Hirsh-Pasek, Kemler Nelson,Kennedy, Woodward, & Piwoz, 1992) and before they noticed the difference between compound versus non-compound words (e.g., “night rate” vs. “nitrate” (Myers, Jusczyk, Kemler-Nelson, Charles-Luce, Woodward & Hirsh-Pasek, 1996).as a way to test for infants conduct top-down analyses as well as bottom-up segmental analyses.
The Headturn Preference Procedure has been used in studies addressing a wide range of questions. For example, Mandel, Jusczyk, and Pisoni (1995) discovered that babies recognize the sound patterns of their own name at 4.5 months. Bortfeld, Morgan, Golinkoff, and Rathbun (2005) found that if a word in a passage occurred after the baby’s own name versus someone else’s name, 6-month-old babies recognizedthe word when played in isolation. The procedure has also been used to test for grammatical sensitivity. Santleman and Jusczyk (1998) found that 18-month-olds preferred to hear an auxiliary and a verb that pattern together (… is baking) over an auxiliary and a verb that are ungrammatical (…can baking).