Social and linguistic factors influencing adaptation in children's speech

Journal of Psychotinguistic Research, Vot. 18, No. 5, 1989

Social and Linguistic Factors Influencing Adaptation in Children's Speech

Richard L. Street, Jr., 1'3 and Joseph N. Cappella z

Accepted August 29, 1989

The ability to appropriately reciprocate or compensate a partner's communicative response represents an essential element of communicative competence. Previous research indicates that as children grow older, their speech levels reflect greater adaptation relative to their partner's speech. In this study, we argue that patterns of adaptation are related to specific linguistic and pragmatic abilities, such as verbal responsiveness, involvement in the interaction, and the production of relatively complex syntactic structures. Thirty-seven children (3-6 years of age) individually interacted with an adult for 20 to 30 minutes. Adaptation between child and adult was examined among conversational floortime, response latency, and speech rate. Three conclusions were drawn from the results of this investigation. First, by applying time-series analysis to the interactants' speech behaviors within each dyad, individual measures of the child's adaptations to the adult's speech can be generated. Second, consistent with findings in the adult domain, these children generally reciprocated changes in the adult's speech rate and response latency. Third, there were differences in degree and type of adaptation within specific dyads. Chronological age was not useful in accounting for this individual variation, but specific linguistic and social abilities were. Implications of these findings for the development of communicative competence and for the study of normal versus language-delayed speech were discussed.

The authors wish to acknowledge two anonymous reviewers for their contributions to this manuscript.

~Department of Speech Communication, Texas A&M University. 2Department of Communication Arts, University of Wisconsin. 3Address all correspondence to Richard L. Street, Jr., Department of Speech Communication, Texas A&M University, College Station, Texas 77843-4234.

497

0090-6905/89/0900-0497506.00/0 �9 1989 Plenum Publishing Corporation

498 Street and CappeHa

One feature of children's social maturation is the ability to adapt communicative responses given the needs, characteristics, and responses of interloctuors (Bates, 1976; Flavell, Botkin, Fry, Wright, & Jarvis, 1968). By far the majority of research on children's communicative adaptations has examined linguistic variables such as grammatical complexity (Sachs & Devin, 1976), code-switching (Weeks, 1971), persua- sive message strategies (Clark & Delia, 1977), topic management (Foster, 1986), referential communication (Sonnenschein, 1988), utterance repairs and corrections (Evans, 1985), and requests (Ervin-Tripp & Gordon, 1986).

Another, much less studied, dimension of communication development is the ability to appropriately and effectively coordinate nonverbal communicative behaviors with a partner's behaviors, including gaze, touch, gestures, body position, and vocalics (Mayo & LaFrance, 1978). The purpose of this study was to investigate linguistic and social abilities related to the degree to which children adapt nonlinguistic aspects of speech relative to the speech of interlocutors. Speech behaviors examined included conversational floortime, response latency (pauses between speaker switches), and speech rate. We limited our investigation to these speech features because of their importance in social interaction. These behaviors represent part of a larger collection of nonverbal responses that regulate the intensity and tempo of conversation (Cappella, 1981). Also, the manner in which interactants adapt nonlinguistic aspects of speech influences the ease with which speaking turns are exchanged in dyadic interaction (Duncan & Fiske, 1977; Natale, 1975) and the impressions conversants form of their partners (Giles, Mulac, Bradac, & Johnson, 1987; Street, 1984).

SPEECH ADAPTATIONS IN SOCIAL INTERACTIONS

Adaptation covers two different aspects of the same process. Interactants may adapt their communicative behavior toward greater similarity relative to their partners' behavior, a phenomenon variously labeled reciprocity (Cappella, 1981), matching (Cappella & Planalp, 1981), congruence (Feldstein & Welkowitz, 1978), and convergence (Giles et al., 1987). In adult interactions, reciprocity has been observed among speech behaviors including accents (Giles & Powesland, 1975), speech rate (Street, 1984), pauses (Cappella & Planalp, 1981; Feldstein & Welkowitz, 1978), vocal intensity (Natale, 1975), and turn durations (Matarazzo & Wiens, 1972), as well as among other nonverbal behaviors

Children's Speech 499

such as posture and gestures (LaFrance, 1982; Maurer & Tindall, 1983), gaze (Kleinke, Staneski, & Berger, 1975), head nodding, and facial affect (Hale & Burgoon, 1984).

Though it is less common than reciprocity, interactants may also adapt communicative behaviors toward greater dissimilarity. Such adaptations have been labeled compensation (Cappella, 1981) and divergence (Giles et al., 1987). Compensatory responses have been reported among turn and vocalization durations (Cappella & Planalp, 1981; Matarazzo, Wiens, Matarazzo, & Saslow, 1968), body orientations (Hale & Bur- goon, 1984), gaze, changes in proximity (Cappella, 1981), and accents (Bourhis & Giles, 1977).

There is substantial evidence suggesting that speech adaptations represent a part of the normal communicative competence of adults (Wiemann, 1977). Speech matching promotes the meshing of communicative styles and has been associated with favorable impressions of interactants (Giles & Powesland, 1975; Street, 1982, 1984), with the intent to communicate effectively (Giles & Smith, 1979; Natale, 1975), and with preferred levels of nonverbal expressiveness (Cappella & Greene, 1982; Patterson, 1983). Compensatory responses, on the other hand, may function to change a partner's behavior to a more preferred level (e.g., increasing speech rate and loudness to stimulate a partner's involvement in a conversation; Cappella & Greene, 1982), to comple- ment a partner's communicative style (e.g., holding the floor for longer or shorter intervals in response to an interlocutor's decreases or increases in talk duration; Matarazzo et al., 1968; Street, 1983), or to show dissapproval of a partner (Giles et al., 1987).

The study of speech adaptation has for the most part focused on adult-adult interaction. Of particular interest is that competent adult communicators not only appropriately adapt speech but also make these adjustments with little expenditure of cognitive energy (Cappella, 1981). The automatic nature of adaptation suggests it is a funda- mental communicative skill that adults have routinized and habitu- ated. A growing body of evidence also indicates that infants exhibit various forms of mutual regulation in their interactions with adult caregivers (Beebe, Jaffe, Feldstein, Mays, & Alson, 1985; Cohn & Tronick, 1988). Surprisingly little attention has been directed to the study of adaptation in the intervening age spans, particularly among 3- to 6-year-old children. The limited research on nonverbal aspects of children's speech suggests that the ability to reciprocate or compensate a partner's response emerges as the child matures behaviorally and cognitively.

500 Street and Cappella

THE DEVELOPMENT OF SPEECH ADAPTATION

The propensity to adapt communicative responses appears early in life. When interacting with one another, infants and adults often modify their latency of response (Beebe et al., 1985), vocalizations (Rosenthal, 1982), body movements (Beebe, Stern, & Jaffe, 1979), and vocal pitch (Lieberman, Ryalls, & Rabson, 1982) with respect to the behavior of the other. Various theoretical perspectives have been formulated to explain adaptation. Because it appears so early in life, some have argued that adaptation is a genetically wired-in response in the human species (Hoffman, 1977; Meltzhoff, 1985). Others employ cognitive (Natale, 1975; Patterson, 1983) or learning theory (Matarazzo & Wiens, 1967) perspectives to explain adaptation because the degree and direction of communicative adjustments vary situationally with respect to a communicator's goals and perceptions of partners.

We have adopted a cognitive perspective for the purposes of this study. This perspective holds that certain information input, storage, and output (response) mechanisms are necessary to explain linkages between a partner's behavior and a communicator's response. For example, consider the resources necessary to adapt one's communicative response. Interactants must be interested and responsive in social encounters. They must have the perceptual ability to track ambient, external stimuli and be able to attend to these stimuli cognitively, store them for at least a brief period, and retrieve information about them. In order to respond with a well-timed reciprocal or compensatory reaction, an interactant must have control over, and flexibility in the use of, motor, vocal, and linguistic behaviors so that the rate and duration of utterance production can be manipulated. Thus, although various explanatory constructs seek to account for adaptation, we found the cognitive perspective to have the most heuristic value for our goals of explicating individual differences among children's production of speech adaptations.

Although the drive to adapt social behavior may be innate, the degree to which interactants are able to coordinate and adjust conversational speech behaviors in relation to those of partners is contingent upon certain social, linguistic, and cognitive skills that develop normally with increasing age. The presence of adaptation in the interactions of children and adults should be symptomatic of one or more of these underlying abilities, and lack of adaptation their deficit. The limited research on these issues indeed indicates that speech adaptation is related to developmental factors, such as age and social responsiveness, and is affected by developmental disorders, such as autism.


Age. The few studies on children's conversational speech adaptations have indicated that the degree to which children display reciprocal and compensatory response patterns increases with age. Welkowitz, Cariffe, and Feldstein (1976) report that, during interactions with peers, 6 1/2- to 7-year-olds matched the duration of pauses between speaker switches and the length of internal pauses (pauses within a speaking turn) more than did 5 1/2- to 6-year-olds. Garvey and Ben-Debba (1974) observed 4-year-olds displaying more equality among their numbers of utterances than did 3-year-old children. Finally, in interactions with adults, Street (1983) discovered that 5-year-olds reciprocated changes in the adult's turn durations, response latencies, and speech rates whereas 4-year-olds only matched the adult's speech rates and response latencies. Three-year-olds did not converge toward the adult's speech levels but did compensate their partner's turn durations.

Social Responsiveness. Yet another prerequisite for speech adaptation appears to be the desire for social involvement. Reticent children presumably experience social situations with negative affect; thus, it is not surprising that one study reported reciprocity among speech rate, utterance length, and response latency to be less evident in the responses of reticent children than of more talkative children (Street, Street, & Van Kleeck, 1983). Among adults, the degree to which one matches a partner's speech features has positively related to an interactant's interpersonal orientation (Street & Murphy, 1987), desire for social approval (Natale, 1975), and field dependence (Marcus, Welkowitz, Feldstein, & Jaffe, 1970). These measures purportedly index the degree to which individuals seek and desire social interaction and assimilation.

Autism. In addition to having linguistic deficits, autistic children are often less interested in engaging in conversation and display more self-centered behaviors in social interactions than do normally developing children (Feldstein, Konstantareas, Oxman, & Webster, 1982; Mundy, Sigman, Ungerer, & Sherman, 1986). Although autistic children may not differ from normal children regarding levels of certain expressive behaviors such as gaze, gestures, and vocalizations (Sigman, Mundy, Sherman, & Ungerer, 1986), they appear less communicatively adaptive (Crown, Feldstein, Jasnow, Beebe, & Jaffe, 1985). Consistent with this claim, Feldstein et al. (1982) reported that, when conversing with a parent or the experimenter, articulate and highly verbal autistic teenagers were less able to match the adult's internal pause and switching pause durations than were the two adults when interacting with one another. Crown et al. (1985) have argued that it is difficult for autistic interactants

502 Street and Cappeila

to temporarily integrate vocal and nonverbal behaviors with those of partners and that this may explain why their communicative responses reveal less adapation.

Purpose of Study

Current understanding of the development of children's speech adaptations is limited in several respects. First, although an integral feature of communicative competence, children's adaptation of nonlinguistic aspects of speech has received little empirical scrutiny, espe- cially among 3- to 7-year-olds in the formative years of verbal development. Second, many previous studies of adaptation give no information about the degree of the individual's adaptation. They proceed by obtaining mean behavioral scores for each person in a dyad and correlate the mean scores across dyads (Feldstein et al., 1982; Welkowitz et al., 1976). This procedure is recommended by its simplicity but carries no information about whether, how, or how much a given person changes in response to changes in a partner's behavior. In short, adaptation is assumed and not explicitly observed or described. Without individual measures, one simply cannot determine what factors are predictive of individual differences in adaptation. The procedures adopted here-- time-series analysis for each dyad--solve both problems. Time-series methods can be used to assess the degree of change over time relative to changes in the other and can do so for each individual in the sample.

Finally, although the research reviewed earlier indicated that speech adaptations are related to rather global dimensions of development, such as age, disorders, and sociability, little is known about specific linguistic, pragmatic, and cognitive resources facilitating speech adaptation. Three dimensions of linguistic and social abilities should be associated with children's speech adaptations: syntactic and lexical production, involvement in the interaction, and verbal responsiveness.

Syntactic and lexical production. For children to flexibly adapt speech, their linguistic repertoires must be sufficiently developed so that lexical choices can be made quickly, efficiently, and appropriately. If a child does not have sufficient verbal resources or control over these behaviors, he or she may have difficulty spontaneously adjusting the onset, duration, and tempo of an utterance. The impact of verbal ability on social interaction was highlighted in a recent review by Fey and Leonard (1983). These authors examined several studies indicating that a


subset of children with specific lanaguage impairments actively sought social contact and interaction, were nonverbally responsive as listeners, yet experienced difficulties in adapting communicative responses to the pragmatic demands of the interaction because of their limited verbal comprehension and syntactic production abilities.

For the present study, the index of syntactic ability chosen to predict children's speech adaptations was the Developmental Sentence Score (DSS; Lee, 1974). The scoring of DSS assigns variable numeric values to words representing various syntactic and semantic categories. A higher score is accrued for more appropriate and complex grammatical construc- tions. Thus, DSS should be highly correlated with another measure of language complexity, mean length of utterance (MLU). However, DSS appears to be a more sophisticated measure of grammatical ability than MLU because it takes into account both rule-appropriateness and complexity whereas MLU does not. Although controversy exists regarding the validity of such linguistic measures as indices of language development (Klee & Fitzgerald, 1985), there is little doubt that such measures reflect linguistic performance (Brown, 1973). For purposes of studying adaptation in a particular interaction, DSS is assumed to measure the child's linguistic performance (which is a product of capabilities and motivation) for that particular interaction.

Interaction Involvement. Interaction involvement is conceptualized as a conversant's interest and enthusiasm for a partner or topic. Typically, more involved individuals are more perceptive, attentive, and responsive during social interaction (Cegala, 1981). Involvement is often signaled by moderate to high levels of expressive behaviors, such as gestures, facial expressions, vocal variations, forward leans, and direct body orientations, and by the extent to which interactants reciprocate these responses (Cappella, 1983). A recent study by Coker and Burgoon (1987) indeed demonstrated that conversants highly involved in conversations were perceived to have speech and nonverbal behaviors that were more coordinated with those of partners than did interactants less involved in conversations.

Verbal Responsiveness. Verbal responsiveness represents the degree to which an interactant verbally responds to, elaborates on, and links topic changes to the content of an interlocutor's conversational contribution (Davis, 1982; Tracy, 1985). For present purposes, we categorized utterances into one of three responsiveness levels. Nonresponsive utterances are those acts in which a conversant refuses a verbal invitation to assume the floor (e.g., not answering a partner's question), produces a


topically irrelevant statement, or initiates a topic change unrelated to the partner's statement or topic (e.g., A utters "Did you have fun at your swimming lessons?" and B responds, "Here, you be this man"). Minimal responsiveness occurs when a conversant provides an essentially adequate response to a partner's utterance but fails to elaborate on it or provide additional content (e.g., A states, "And so that's what we did last night," and B responds only with "I see," "Uh huh," "That 's nice," "Good") . Finally, a highly responsive remark extends or elaborates on a partner's previous utterance or, if a topic change is made, connects the new topic with the previous one (e.g., A says, "That movie was really funny," and B replies, " I f you like movies with that kind of plot, you might enjoy this book I've been reading" or "Yeah, that part on the roller coaster was great").

Responsiveness reflects the extent to which interactants are cooper- atively integrating verbal responses to develop a topic. This integration may be reflected in the conversants' speech styles as well as they develop a common format for the tempo and structure of the interaction.

S U M M A R Y

The purpose of this study was to examine factors influencing adaptation in children's floortime, speech rates, and response latencies relative to an adult's performances of these behaviors. Through the use of time-series analysis, adaptation scores within individual adult-child dyads were ascertained by assessing the degree of change in the child's behavior from a baseline level and relative to the adult's behavior. Specifically, the following prediction was forwarded: In interactions with an adult, the degree of adaptation of children' s floortime, speech rates, and response latencies will be related to their DSS scores, involvement in the interaction, and verbal responsiveness.

M E T H O D

Subjects and Procedure

Thirty-seven volunteer subjects (20 girls, 17 boys) were solicited from day care centers to participate in this study. Their ages ranged from 3 years 2 months to 6 years 10 months. All children were screened by a


speech-language pathologist who verified that each child's speech and language development were normal.

A 24-year-old female graduate student in speech and hearing sciences conducted play sessions with the children. These were audio- recorded on separate channels using throat microphones to avoid problems of voice spillover when two microphones are used. A third microphone was used to obtain the content of the conversations. The graduate student was blind to the purposes of the study and was simply instructed to play and interact with the child. The interactions lasted about 25 minutes.

Speech Behavior Coding and Data Preparation

Human coders independently coded each channel of each audiotape for the presence or absence of speech using a hardware and software system developed in our laboratory and known as W5DATA. Human coders were used instead of a sound-activated system because the vocal tracks were often filled with a variety of sounds that were not conversational speech (e.g., giggles, coughs, grunts, objects banging together) and that would have triggered a sound-activated system. W5DATA uses input from digitized button boxes that act much like an event recorder and stores these data for later use. Subsequent computer programs interleave the codings for the adult and the child and produce a turn-by-turn analysis according to the categories developed by Jaffe and Feldstein (1970). Thus, for each speaker and each turn, the program produces the frequencies and durations of conversational turns, vocalizations, pauses within the turn, simultaneous speech, and the latency to respond. 4 The turn data were summarized into 1-minute units yielding the following measures for adult and child per minute unit: average floor duration, average latency, and average speech rate. Speech rate was operationalized by dividing the time spent speaking during a floorholding by the duration of the floorholding itself (which includes both speech and silence). This measure is highly correlated with another measure of speech rate, words per minute (Street, 1984). The speech data are then put into the form of a bivariate time series for each of the three variables.

Reliability for the coding of speech data was checked for both

4Our measure of "response latency" is essentially the same phenomenon as a "switching pause" as operationalized by Jaffe and Felstein (1970), with one important exception. Jaffe and Feldstein credit the silence between turns to the speaker relinquishing the floor. We attribute this silence to the speaker about to assume the floor.


unitizing and categorizing reliability (Krippendorff, 1980). No checks of categorizing reliability were done until units were agreed upon at very high rates (> 95%). Categorizing reliability for the three coders ranged from a correlation of .89 to .98 across time units.

Linguistic and Social Predictors

The conversations were also transcribed and two language ability indices obtained from these transcripts. One hundred of the child's utterances were randomly selected from each transcript. Each utterance was then scored for developmental sentence score (DSS; Lee, 1974) and for verbal responsiveness (i.e., appropriateness of topic continuation). As a measure of syntatic complexity, DSS is scored taking into account both the complexity and sophistication of the utterance. For example, the sentence " I like eating cookies" has a DSS of 11--1 point for being a complete sentence, 1 point for having a primary verb, 1 point for using a personal pronoun, and 8 points for using the gerund, "eat ing." On the other hand, "he eats cookies" receives 1 sentence point, 2 points for the third-person pronoun, and 2 points for the appropriate formulation of the singular verb (eat + s), for a total DSS of 5. For this data set, the children's average DSS scores ranged from 3.64 to 9.82.

The verbal responsiveness measure was developed by the researchers as a 3-point behavioral judgment of the degree to which the child's speaking turn built upon by the previous remark by the adult: 3 = highly responsive utterance, 2 = minimally responsive remark, and 1 = nonresponsive utterance that was completely discontinuous with the adult's prior remark or question. Table I presents representative utterances for each of these categories. Reliability for DSS and verbal responsiveness measures was formulated by having another coder randomly select 100 utterances from the entire data set and score the utterances for DSS and responsiveness. For these measures, intraclass correlations comparing coded and recoded utterances were sufficient: .84 for DSS and .91 for verbal responsiveness. For each measure, the child's mean score was computed.

The final measure obtained was the adult interactant's subjective judgment of the child's involvement in the interaction. This was also a 3-point judgment of 1 = "not involved" (i.e., not interested in talking, not enjoying self, inattentive), 2 = "somewhat involved," and 3 = "highly involved" (i.e., enjoying the interaction, enthusiastic, very interested in the topic).

Children's Speech

Table I . Examples of Different Levels of Verbal Responsiveness

Responsiveness Level Utterance

507

Nonresponsive

Minimally responsive

Highly responsive

Ad: "What is your name?" Ch: " I wanna play this game." Ad: "Can I be the cow?" Ch: (Silence) Ad: "And then I ' ll drive to your house." Ch: "'Where are the blocks?" Ad: "What game would you like to play?" Ch: ' 7 don't know." Ad: "So you like to visit your grandmother's farm? Ch: "Uh-huh." Ad: "Here comes the momma cow!" Ch: "Cow.'" Ad: "This looks like a house" Ch: "'Yes, it is a very big house.'" Ad: "Do you want to play with the dolls?" Ch: "Yeah, but you have the Barbie, OK?'" Ad: ~ chive the truck over here." Ch: "OK, and F ll put the animals here."

Data Analysis

Time-series analyses (McCleary & Hay, 1980) were conducted for each of the three speech behaviors for each dyad separately using BMDP's time series software. Time-series analysis differs from the time-series regression in that in the latter case the researcher must specify a model to be tested both for the effects of one series on the other (including how far back in time the effects must be specified) and for the errors affecting the series. In well-developed disciplines and in cases of direct tests of theory, time-series regression procedures are common. However, our knowledge about the effects of one behavioral series on another is meager, and, hence, more empirical techniques like time-series analysis seem more appropriate. With time-series analysis both the error process and the model of the effects of one series on the other emerge from the data.

Each time-series analysis involves multiple steps: (a) determining trends within each series separately, (b) detrending the series, (c) obtaining cross-correlations for each pair at various lags, (d) testing cross-correlations, and (e) constructing a transfer function model. For our purposes only steps (a) through (d) are necessary since we are interested


only in the extent to which the behavior of the adult affects the child's behavior and at what lags. Thirty-seven separate time series on each of the three behaviors were run. The resulting cross-correlations for each pair of series represent the presence of adjustment by the child, the type of adjustment (compensatory when negative and reciprocal when posi- tive), and the magnitude of adjustment (through their absolute value). 5

The cross-correlations are the adaptation scores for individual children and therefore serve as the outcome measures in a stepwise regression analysis with DSS, verbal responsiveness, interaction involvement, age, and sex as predictors. Because of the exploratory nature of this research, the probability criterion for entry into the stepwise regression model was set at . 1. Of course, it is possible that some of these predictors may interact with one another (e.g., age and responsiveness) to influence adaptation. However, the relatively small number of observations (37) was insufficient to allow for an adequate test of interaction effects. 6

RESULTS

Description of the Data

Table II presents the means of the children's DSS, involvement, responsiveness, floortime, response latency, and speech rate levels as a function of the child's age and sex. Children did not differ among these variables as a function of age. However, girls displayed greater interaction involvement (F(1, 31) = 8.11, p < .01) and verbal responsiveness (F(1, 31) = 6.67, p < .05) than did boys. There were no other main or interactive effects on these measures as a function of age and sex.

Correlations

Table III displays relationships between children's speech behavior means, the adult's speech, and language ability measures. Children's

5We are not arguing that the child is the only interactant adapting in these interactions. To the contrary, the reverse is also occurring--the adult is adapting to the child. Previous research of adult interactions has revealed that patterns of communicative exchange are typically created mutually, not individually, by the participants; that is, each interactant tends to reciprocate or compensate the other's response (Cappella & Planalp, 1981; Street, 1984). Although the influence is bidirectional, we were interested in the child's ability to adapt and in the factors influencing this performance.

6Although it was not feasible to test for interaction effects, stepwise regressions with interaction terms as predictors were nonetheless conducted on the speech measures after the main effects were partialed out. None emerged as significant predictors of adaptation.


Table II. Means for the Speech, Language Ability, and Social Measures in Relation to the Child's Age and Sex

Speech Response Age/sex Floortime a rate b latency c DSS Responsiveness Involvement

Under 4 years 5 months Boys (n = 4) 2.05 0.66 0.81 5.65 2.35 2.50 Girls (n = 9) 2.40 0.73 0.65 6.14 2.61 3.00

4 years 5 months to 5 years 5 months Boys (n = 5) 1.95 0.73 0.81 4.61 2.42 2.00 Girls (n = 7) 2.41 0.74 0.81 6.60 2.65 2.75

Over 5 years 6 months Boys (n = 8) 2.50 0.76 0.57 7.00 2.68 2.67 Girls (n = 4) 2.50 0.75 0.64 6.40 2.73 3.00

aHoortime scores represent average seconds per speaking turn. bSpeech rate was coded as the ratio of the duration of vocalizing within a floorholding by the duration of the floorholding itself. ~Response latency scores represent average seconds per assumption of speaking turn.

Table III. Correlations Among the Children's Characteristics, Children's Speech, and the Adult 's Speech Measures

Variable

Variable 2 3 4 5 6 7 8 9 10 11

Child's 1. DSS 2. Involvement 3, Responsiveness 4. Age 5. Sex 6. Floortime 7. Response latency 8. Speech rate

Adult's 9. Floortime

10. Response latency 11. Speech rate

< .05. ~'/~< .01.

.44 b .37 ~ .23 .08 .51 b - . 1 6 .23 - . 4 0 a - . 4 6 b .20 .30 .01 .38 a .36 a - . 3 2 ~ .31 - . 1 7 - . 38 ~ .35 ~

.30 .24 .34 ~ - . 2 9 .35 a - . 15 - . 35 a .34 ~ - .21 .11 .01 .14 - . 0 6 - . 1 2 - . 15

.12 - . 2 2 .17 - . 03 - .01 .27 - . 4 7 b .26 - .896 - . 4 0 a .72 b

.08 .37 - . 18 - . 7 0 b - . 05 - . 7 4 .05

.13 - . 7 0 b - .01

perceived involvement in the interaction was significantly correlated with the children's f loortime (r = .36) and response latency (r = .32). Responsiveness was related to the child's speech rate (r = .35) and


floortime (r = .34). DSS was correlated with floortime (r = .51). Interestingly, chronological age was not significantly related to any of the children's average speech levels and was only marginally correlated with responsiveness. These findings affirm our assumption that children's speech levels are related to their language and social abilities more than to age per se, and that children within a similar age range may variously exhibit these social and linguistic skills.

Speech Adaptation

Error Processes. The first step of any time-series analysis requires estimation of the degree and type of autocorrelation present in the input series. On the basis of the form of this autocorrelation, if any, "prewhitening" of the input and output series is done. All 37 dyads were investigated separately for each variable to determine the nature of autocorrelation function necessary for purifying the series. Eight lags back in time (8 minutes) were used as the basis of estimating these patterns. No clear trends emerged across the series of variables as to the type of autocorrelation function needed. Many series did not exhibit strong patterns of autocorrelation and so no prewhitening was necessary. No series required a transformation greater than three lags back, and most transformations were simple first differences or first-order autoregressive modifications. Appropriate prewhitening was carried out for each variable in the series.

Cross-Correlations. Cross-correlations for each pair of the series, one for the child and one for the adult, are next carried out at various lags. At lag-0 the scores for the child and the adult at the same time period are correlated; at lag-l, the scores for the adult at the prior time period are correlated with those of the child at the current period, and so on for different lags. Cross-correlations were generated for 6 lags back in time for each variable in each dyad. The child's series always served as the output series or outcome score because we were interested in the degree to which the child adjusted to the adult's behavior. The overwhelming majority of significant (> 2 SEs) or near-significant (> 1 SE but < 2 SEs) cross-correlations occurred for lag-0 and lag-1. Thus, for this sample the effects of adaptation can be restricted to being within the same 1-minute period or at most lagged back 1 minute. Effects for 2 to 6 minutes back in time were ineffective in general and hence are disre- garded.


Table IV . Resu l t s o f T ime-Ser ie s Ana lys i s for Three Speech Behav iors

Variable Weighted Significance Significance (lag) correlation (df = 37) a (df = 1,094) b

Floortime (0) - . 0 5 7 n.s. p < .07 + n = 16 C

- n = 2 1 range: .44/- .44 d

Floortime (1) .039 n.s. n.s. + n = 19 - n = 18 range: .36/- .45

Speech rate (0) .307 p < .05 p < .01 + n = 33 - n = 4 range: . 85/-. 34

Speech rate (1) .057 n.s . p < .07 + n = 17 - n = 2 1 range: .50/- .38

Response latency (0) .097 n.s . p < .01 + n = 28 - n = 9 range: .51/- . 36

Response latency (1) .059 n.s. p < .08 + n = 26 - n = 11 range: .60/-. 19

~l 'hese degrees of freedom represent the number of dyads. bThese degrees o f freedom represent the number of t ime units across dyads. c,, + , , refers to the number of dyads in which the chi ld 's adaptation was in the direction of reciprocity. ' . . . . refers to the number of dyads in which the child 's adaptation was in the direction of compensation. dThese figures represent the range of correlations for this behavior across the 37 dyads.

General Trends of Speech Adaptation. In Table IV, the column for "weighted correlation" represents the mean cross-correlation weighted by sample size (the number of time-series points within the dyad) across 37 dyads and were derived from meta-analytic procedures (Hunter, Schmidt, & Jackson, 1982). Table IV also displays a summary of the number of children adapting in either a reciprocal or a compensatory manner for each behavior at lag-0 and lag- 1. Two tests of significance are offered in the last two columns, one based upon the dyad as the unit of analysis (n = 37) and the other based upon the observation (minute of interaction) as the unit of analysis (n = 1094). Both significance

512 Street and Cappeila

estimates are presented because it is not clear whether one should use the number of dyads or the number of observations as the basis for judging overall significance. Hence, both are presented to the reader for more informed judgments.

Using the dyad as the unit of analysis, the only significant effect was observed for speech rate at lag-0 as children generally reciprocated the adult's adaptations in speech rate. Using the larger degrees of freedom, two of the six coefficients achieved standard levels of significance, speech rate and response latency, both at lag-0. A fair conclusion would be that, for both rate and latency, there is a trend across all dyads for the child to reciprocate the adult's speech rates and response latencies even though individual children may vary in their degree of reciprocation. Not quite reaching significance (p < .09) were compensation for floortime at lag-0 and matching of rate and latency at lag-1.

A couple of comments are in order regarding the size of these correlations, which represent general trends of adaptation. Compared with adult conversations, the degree of adaptation in these adult-child interactions seems rather low. However, the magnitude and sign of the weighted correlations themselves may not be informative. For example, the correlation may not be significant either because there is little evidence of adaptation in any dyad or because there is variation within each dyad in terms of the direction and degree of adaptation. The latter would be the case if one had two sets of correlation coefficients such that one group had a cross-correlation of .40 and the other - . 40 . Assuming the two groups had the same number of observations, the overall weighted correlation would be zero. Yet this value would be unrepresen- tative of the kind and degree of adaptation at the individual dyad level. Explaining individual differences in adaptation within each dyad is the goal of the next set of analyses.

Predictors of Speech Adaptation. Floortime compensation at lag-0 was predicted by involvement (B = - .124 , SE = .067, F(1, 35) = 4.68, p < .038, R 2 = .12). There were no significant predictors of floortime adaptation at lag-1. Speech rate matching at lag-0 was positively related to verbal responsiveness (B = .447, SE = .204, F(1, 35) = 4.80, p < .036, R 2 = .12). At lag-l, rate matching was again positively related to responsiveness (B = .276, SE = . 145, F(1, 34) = 3.65, p < .065, R 2 = .09) but negatively associated with involvement 03 = - . 15 , SE = 0.56, F(1, 34) = 7.20, p < .02, R 2 = .11). Response latency matching at lag-0 was marginally, but positively, related to DSS (B = .055, SE = .022, F(1, 35) = 2.95, p < .10, R e = .08). Overall,


Table V. Correlations Among Children's Speech Adaptation Scores, Linguistic and Social Predictors, Age, and Sex

Speech adaptation variable

Floortime Speech rate Response latency

Linguistic, social, age, and sex variables Lag-0 Lag-1 Lag-0 Lag-1 Lag-0 Lag-1

DSS - .15 -.01 - .03 - .12 .28 a -.11 Involvement - .34 b -.01 .08 -.340 - .07 .07 Responsiveness - .28 a - .09 .350 .19 - .19 .06 Age .03 .03 - .10 .23 - .10 .14 Sex -.23 - .06 - .05 - .10 - .19 .07

~P< .05.

age and sex exhibited no explanatory power in accounting for variation in the children's adaptation scores. However, other indicators did, with verbal responsiveness and judged involvement showing the most impact in accounting for variation in the degree of adaptation. Table V presents the correlations between adaptation scores (at lag-0 and lag-l) for each of the three speech behaviors and the language, sex, and age measures.

DISCUSSION

The purpose of this study was to identify specific linguistic and social abilities predictive of children's speech adaptations in social interactions. In order to achieve this goal it was necessary to employ a measure of temporal adaptation unique to each child observed in conversation. This measure was provided by the cross-correlations generated from time-series analyses of each behavior for each child. The results were generally supportive of our original hypothesis. Several conclusions were drawn from the results of this research.

First, with the use of time-series analysis , researchers can generate meaningful measures of adaptation for the individual interactant. How- ever, one must ensure an adequate number of observations in the data series. By coding each behavior for 1-minute intervals, we generated 20 to 30 data points per dyad. This is at best the minimum number necessary to allow sufficient power for discovering autoregressive processes. Future research may need to utilize longer interactions or a more frequent rate of


sampling (assuming stable estimates of the behavior can be computed). Second, specific findings and trends across the group of dyads paralleled previous findings for these behaviors in the adult domain. Specifically, the children showed tendencies to reciprocate the adult's speech rates and response latencies. For floortime, there was a nonsignificant trend toward compensation. Matching on speech rate and response latency, and compensation on floortime are typical of findings from temporal analyses of adult social interactions (Cappella & Planalp, 1981; Feldstein & Welkowitz, 1978; Matarazzo et al., 1968; Street, 1984).

Third, the presence of adaptation effects cannot automatically be treated as uniform across a group of subjects. For example, children may collectively reciprocate adults' changes in speech rate. However, this strong effect is moderated by the observation that there exists noticeable variability in the degree to which children individually adapted speech behaviors. This variability was frequently related to indices of linguistic and social ability.

Children who were more verbally responsive (that is, whose comments were more appropriate to and extended the adult's preceding remark) converged more to the adult's rate of speech during both the current and the previous minute of the interaction. In addition to possessing the pragmatic skills necessary for topic development, the verbally responsive child also may be holistically attentive to other characteristics of a partner's response. That is, in addition to integrating utterances to the content of the partner's speech, the responsive child is able to accommodate to nonverbal features, such as tempo, of a partner's utterances as well.

Children's involvement in the interaction was related to floortime compensation at lag-0 and speech rate compensation at lag-1. As a child becomes more interested and enthusiastic in the interaction, he or she may speak for relatively long periods of time while the adult listens. The adult may remain quiet, encouraging or allowing the child to talk, and may later assume the floor for a relatively long period of time. This explanation seems likely since the child's involvement was positively related to the child's floortime (r = .36). In effect, the adult and more involved children tended to exchange roles as the predominant speaker during the course of the interaction, a pattern much like what typifies adult conversations (Cappella & Planalp, 1981; Street & Murphy, 1987).

Regarding the relationship between involvement and speech rate compensation at lag-l, longer floorholdings usually are produced at slower rates. Thus, the interactant who is the predominant speaker for 1 minute of interaction may talk more slowly than a partner who talks less


during that minute. To the extent that the role of the predominant speaker is exchanged during the course of the interaction (e.g., every couple of minutes or so), a compensatory relationship may emerge between an interactant's rate for 1 minute of the interaction and the partner's rate for the previous minute. 7

The syntactic complexity of the children's utterances (as measured by DSs) was marginally predictive (p < .1) of the extent to which children reciprocated the adult's response latencies. Although this finding did not meet strict standards of statistical significance, children producing more complex syntax may have greater flexibility in the use of their verbal repertoires and thus may be more capable of controlling the timing of their contributions than are children exhibiting less complex syntactic structures.

Every linguistic and conversational skill measure showed stronger relationships to speech adaptation than did chronological age. These findings lend support to our claim that specific communicative skills account for the age effects reported in previous research regarding speech adaptations. This seems particularly true since these measures normally are only moderately correlated with age, and were not significantly correlated in this study (see Table III).

It goes without saying that studies like the present one need to be replicated, generalized, and extended to a broader and more precise set of individual difference variables. The long-range goal is to discover the underlying cognitive, linguistic, social, and perceptual abilities that make adaptation to one's partner possible. It will be important to know what behaviors for example, vocal or kinesic ones--most successfully dis- criminate adaptive abilities. Recent research has found that very young infants exhibit adaptation in the temporal patterning of responses in both vocal and protogestural actions (Beebe et al., 1985; Cohn & Tronick, 1988; Penman, Meares, Baker, & Milgrom-Friedman, 1983). It will be important to show how early in the infant's development these adaptations occur, and, if they occur in one context with a particular partner, do they also occur in another context with the same partner? Perhaps the crucial step will be the association between individual measures of

7As one reviewer aptly observed, these results do not "prove" that more involved children compensate a partner's floorholding and speech rate more than do less involved children. It could be that children making these adaptations were perceived by the adult to be more involved. However, the results are at best suggestive and consistent with our claims that one's involvement in an interaction influences the nature of speech adaptation.


adaptation and individual differences in empathy (Bryant, 1982), in interpersonal awareness (Borke, 1971) and expressiveness (Buck, 1977), and between adaptation and differences in dimensions of neonatal development and expression among infants (Brazelton, 1984; Knobloch, Stevens, & Malone, 1980). Perhaps nonverbal adaptations and their failure can serve as useful diagnostic tools for underlying developmental problems (cf. Crown et al., 1985).

The present study has advanced us a very small way toward this goal by demonstrating that individual measures of adaptation derived from time-series techniques can be usefully employed in establishing and accounting for variation between persons in degree of communicative adaptation.

REFERENCES

Bates, E. (1976). Language and context: The acquisition of pragmatics. New York: Academic Press.

Beebe, B., Jaffe, J., Feldstein, S., Mays, K., & Alson, D. (1985). Interpersonal timing: The application of an adult dialogue model to mother-infant vocal and kinesic interactions. In T.M. Field & N.A. Fox (Eds.), Socialperception in infants (pp. 217-247). Norwood, NJ: Ablex.

Beebe, B., Stern, D., Jaffe, J. (1979). The kinesic rhythm of mother-infant exchanges. In A.W. Siegman & S. Feldstein (Eds.), Of speech and time (pp. 23-34). Hillsdale, NJ: Erlbaum.

Borke, H. (1971). Interpersonal perception of young children: Egocentrism or empathy? Developmental Psychology, 5, 262-269.

Bourhis, R.Y., & Giles, H. (1977). The language of intergroup distinctiveness. In H. Giles (Ed.), Language, ethnicity, and intergroup relations (pp. 119-135). London: Academic Press.

Brazelton, T.B. (1984). Neonatal behavioral assessment scale (2nd ed.). London: Blackwell Scientific Publications.

Brown, R. (1973). A first language: The early stages. Cambridge, MA: Harvard University Press.

Bryant, B.K. (1982): An index of empathy for children and adolescents. Child Development, 53, 413-425.

Buck, R.W. (1977). Nonverbal communication accuracy in preschool children: Relationships with personality and skin conductance. Human Communication Research, 3, 225-236.

Cappella, J.N. (1981). Mutual influence in expressive behavior: Adult-adult and infant-adult dyadic interaction. Psychological Bulletin, 89, 101-132.

Cappella, J.N. (1983). Approaching and avoiding others: Involvement in dyadic interaction. In J.M. Wiemann & R.P. Harrison (Eds.), Nonverbal Interaction (pp. 113-148). Beverly Hills, CA: Sage.

Cappella, J.N., & Greene, J.O. (1982). A discrepancy-arousal explanation of mutual influence in expressive behaviors for adult-adult and infant-adult interactions. Communication Monographs, 49, 89-114.


Cappella, J.N., & Planalp, S. (1981). Talk and silence sequences in informal conversations III: Interspeaker influence. Human Communication Research, 7, 117-132.

Cegala, D. (1981). Interaction involvement: A cognitive dimension of communicative competence. Communication Education, 30, 109-121.

Clark, R.A., & Delia, J. (1977). Cognitive complexity, social perspective-taking, and functional persuasion skills in second- to ninth-grade children. Human Communication Research, 3, 128-134.

Cohn, J.E., & Tronick, E.Z. (1988). Mother-infant face-to-face interaction: Influence is bidirectional and unrelated to periodic cycles in either partner's behavior. Child Development, 24, 386-392.

Coker, D., & Burgoon, J.K. (1987). The nature of conversational involvement and nonverbal encoding patterns. Human Communication Research, 13, 463-494.

Crown, C.L., Feldstein, S., Jasnow, M., Beebe, B., & Jaffe, J. (1985). A strategy for investigating autism as a prelinguistic disorder of social development. Australian Journal of Human Communication Disorders, 13, 61-76.

Davis, D. (1982). Determinants of responsiveness in dyadic interaction. In W.I. Ickes & E.S. Knowles (Eds.), Personality, roles, and social behaviors (pp. 85-139). New York: Springer-Verlag.

Duncan, S., & Fiske, D. (1977). Face-to-face interaction. Hillsdale, NJ: Erlbaum. Ervin-Tripp, S., & Gordon, D. (1986). The development of request. In R. Schiefelbusch (Ed.),

Language competence: Assessment and intervention (pp. 61-95). San Diego, CA: College-Hill Press.

Evans, M.A. (1985). Self initiated speech repairs: A reflection of communicatlve competence. Child Development, 21, 365-371.

Feldstein, S., Konstantareas, M., Oxman, J., & Webster, C.D. (1982). The chronography of interaction with autistic speakers: An initial report. Journal of Communicative Disor- ders, 15, 451-460.

Feldstein, S., & Welkowitz, J. (1978). A chronography of conversation: In defense of an objective approach. In A.W. Siegman & S. Feldstein (Eds.), Nonverbal behavior and communication (pp. 435-499). Hillsdale, N J: Erlbaum.

Fey, M., & Leonard, L.B. (1983). Pragmatic skills of children with specific language impairment. In T.M. Gallagher & C.A. Prutting (Eds.), Assessment and intervention issues in language (pp. 65-82). San Diego, CA: College-Hill Press.

FlaveU, J., Botkin, P.T., Fry, C.L., Wright, J., & Jarvis, P. (1968). The development of role-taking communicative skills in children. New York: Wiley.

Foster, S. (1986). Learning discourse topic management in the preschool years. Journal of Child Language, 13, 231-250.

Garvey, C., & Ben-Debba, M. (1974). Effects of age, sex, and partner on children's dyadic speech. Child Development, 45, 1159-1161.

Giles, H., Mulac, A., Bradac, J., & Johnson, P. (1987). Speech accommodation theory: The first decade and beyond. Communication yearbook, 10, 13-48.

Giles, H., & Powesland, P.F. (1975). Speech style and social evaluation. London: Academic Press.

Giles, H,, & Smith, P. (1979). Accommodation theory: Optimal levels of convergence. In H. Giles & R. St. Clair (Eds.), Language and social psychology (pp. 45-65). Baltimore: University Park Press.

Hale, J.L., & Burgoon J.K. (1984). Models of reactions to changes in nonverbal immediacy. Journal of Nonverbal Behavior, 8, 287-314.

Hoffman, M.L. (1977). Empathy, its development, and prosocial implications. In C. Keasey

518 Street and CappeHa

(Ed.), Nebraska symposium on motivation 1977 (pp. 169-218). Lincoln, NE: University of Nebraska Press.

Hunter, J.E., Schmidt, F.L., & Jackson, G.B. (1982). Meta-analysis: Cumulating research findings across studies. Beverly Hills, CA: Sage.

Jaffe, J., & Feldstein, S. (t970). Rhythms of dialogue. New York: Academic Press. Klee, T., & Fitzgerald, M.D. (1985). The relation between grammatical development and

mean length of utterance in morphemes. Journal of Child language, 12, 251-269. Kleinke, C.L., Staneski, R.A., & Berger, D.E. (1975). Evaluation of an interviewer as a

function of interviewer gaze, reinforcement of subject gaze, and interviewer attractive- ness. Journal of Personality and Social Psychology, 31, 115-122.

Knobloch, H., Stevens, F., & Malone, A.F. (1980). Manual of developmental diagnosis. Hagerstown, MD: Harper & Row.

Krippendorff, K. (1980). Content analysis. Beverly Hills, CA: Sage. LaFrance, M. (1982). Posture mirroring and rapport. In M. Davis (Ed.), Interactions rhythms:

Periodicity in communicative behavior (pp. 279-298). New York: Human Sciences Press.

Lee, L. (1974). Developmental sentence analysis. Evanston, IL: Northwestern University Press.

Lieberman, P., Ryalls, J., & Rabson, S. (1982). Some acoustic aspects of early imitation in speech. Journal of the Acoustical Society of America (Supplement 1), 72, SI01.

Marcus, E., Welkowitz, J., Feldstein, S., & Jaffe, J. (1970, April). Psychological differen- tiation and the congruence of temporal speech patterns. Paper presented at the Eastern Psychological Association, Atlantic City, NJ.

Matarazzo, J.N., & Wiens, A.N. (1967). Interviewer influence on durations of interviewee silence. Journal of Experimental Research in Personality, 2, 56-69.

Matarazzo, J.D., & Wiens, A.N. (1972). The interview: Research on its anatomy and structure. Chicago: Aldine-Atherton.

Matarazzo, J.D., Wiens, A.N., Matarazzo, R.G., & Saslow, G. (1968). Speech and silence behavior in clinical psychotherapy and its laboratory correlates. In M.J. Schlien, H. Hunt, J.D. Matarazzo, & C. Savag e (Eds.), Research in psychotherapy (Vol. 3, pp. 347-394). Washington, DC: American Psychological Association.

Maurer, R.E., & Tindall, J.H. (1983). Effects of postural congruence on clients' perception of counselor empathy. Journal of Counseling Psychology, 30, 158-163.

Mayo, C., & LaFrance, M. (1978). On the acquisition of nonverbal communication: A review. Merrill-Palmer Quarterly, 24, 213-228.

McCleary, R., & Hay, R.A. (1980). Applied time series analysis for the social sciences. Beverly Hills, CA: Sage.

Meltzhoff, A.N. (1985). The roots of social and cognitive development: Models of man's original nature. In T.M. Field and N.A. Fox (Eds.), Social perception in infants (pp. 1-30). Norwood, NJ: Ablex.

Mundy, P., Sigman, M., Ungerer, J., & Sherman, T. (1986). Defining the social deficits of autism: The contribution of nonverbal communication measures. Journal of Child Psychology and Psychiatry, 27, 647-656.

Natale, M. (1975). Convergence of mean vocal intensity in dyadic communication as a function of social desirability. Journal of Personality and Social Psychology, 32, 790-804.

Patterson, M. (1983). Nonverbal behavior: A functional perspective. New York: Springer- Verlag.


Penman, R., Meares, R., Baker, K., & Milgrom-Friedrnan, J. (1983). Synchrony in mother-infant interaction. British Journal of Medical Psychology, 56, 1-7.

Rosenthal, M.K. (1982). Vocal dialogues in the neonatal period. Developmental Psychology, 18, 17-21.

Sachs, J., & Devin, J. (1976). Young children's use of age-appropriate speech styles in social interaction and role-playing. Journal of Child Language, 3, 81-98.

Sigman, M., Mundy, P., Sherman, T., & Ungerer, J. (1986). Social interactions of autistic, mentally retarded, and normal children and their caregivers. Journal of Child Psychol- ogy and Psychiatry, 27, 647-656.

Sonnenschein, S. (1988). The development of referential communication: Speaking to different listeners. Child Development, 59, 694-702.

Street, R.L., Jr. (1982). Evaluation of noncontent speech accommodation. Language and Communication, 2, 13-31.

Street, R.L., Jr. (1983). Noncontent speech convergence in adult-child interactions. Commu- nication Yearbook, 7, 369-395.

Street, R.L., Jr. (1984). Speech convergence and speech evaluation in fact-finding interviews. Human Communication Research, 11, 139-169.

Street, R.L., Jr., & Murphy, T. (1987). Interpersonal orientation and speech behavior. Communication Monographs, 54, 42-62.

Street, R.L., Jr., Street, N.J., & Van Kleeck, A. (1983). Speech convergence among talkative and reticent three year-olds. Language Sciences, 5, 79-96.

Tracy, K. (1985). Conversational coherence: A cognitively grounded rules approach. In R.L. Street, Jr., & J.N. Cappella (Eds.), Sequence and pattern in communicative behavior (pp. 30-49). London: Edward Arnold.

Weeks, T.E. (1971). Speech registers in young children. Child Development, 42, 1119-1131. Welkowitz, J., Cariffe, G., & Feldstein, S. (1976). Conversational congruence as a criterion

of socialization in children. Child Development, 47, 269-272. Wiemann, J. (1977). Explication and a test of a model communicative competence. Human

Communication Research, 3, 195-213.

Social and linguistic factors influencing adaptation in children's speech

Documents