Expression of and the Emergence of Language Author(s): Lois Bloom and Joanne Bitetti Capatides Source: Child Development, Vol. 58, No. 6 (Dec., 1987), pp. 1513-1522 Published by: Wiley on behalf of the Society for Research in Child Development Stable URL: http://www.jstor.org/stable/1130691 Accessed: 26-01-2017 19:00 UTC

REFERENCES Linked references are available on JSTOR for this article: http://www.jstor.org/stable/1130691?seq=1&cid=pdf-reference#references_tab_contents You may need to log in to JSTOR to access the linked references.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your of the Terms & Conditions of Use, available at http://about.jstor.org/terms

Society for Research in Child Development, Wiley are collaborating with JSTOR to digitize, preserve and extend access to Child Development

This content downloaded from 128.59.160.233 on Thu, 26 Jan 2017 19:00:22 UTC All use subject to http://about.jstor.org/terms Expression of Affect and the Emergence of Language

Lois Bloom and Joanne Bitetti Capatides Teachers College, Columbia University

BLOOM, Lois, and CAPATIDES, JOANNE BITETTI. Expression of Affect and the Emergence of Lan- guage. CHILD DEVELOPMENT, 1987, 58, 1513-1522. The relation between infant affect expression and the emergence of language was studied in 6 girls and 6 boys, from 9 months to 2 years of age. First words (FW) and vocabulary spurt (VS) were identified in the infants' transition from prespeech vocalizing to the emergence of language. Their expressions of affect were coded for the gradient properties of (positive, negative, neutral, mixed, equivocal) and intensity. Frequency of nonneutral affect expression at FW and VS was positively correlated with age at FW and VS (p < .02), meaning that the more frequently the children expressed , the older the age of lan- guage achievements. Time spent in neutral affect at FW and VS was negatively correlated with age at FW and VS (p < .02); the more time spent in neutral affect, the younger the age of language achievements. In addition, the measures of affect at VS predicted the eventual transition to mul- tiword speech, with more time spent in neutral affect expression at VS negatively correlated with earlier age in the use of sentences. We propose that neutral affect supports the early transition to language by allowing the reflective stance required for language learning, and has its antecedents in the quiet alert states which support the cognitive activity of early infancy.

The two systems of expression, affect and to the expression of and for speech, have different developmental histo- the origins of words. In the most influential of ries. Some expressions of affect are present these theories, Condillac, in the eighteenth virtually from birth (e.g., Darwin, 1913; Izard, century, proposed that the conventional signs Huebner, Risser, McGinnes, & Dougherty, of speech originated when the involuntary vo- 1980), and much of the form and content of cal gestures that expressed "the passions of communication between infants and their , of , or of ' were deliberately re- caregivers in the first year of life depends peated in the absence of their reflexive elicit- upon affective expression (e.g., Dore, 1983;ing conditions for the benefit of others (cited Emde, Gaensbauer, & Harmon, 1976; Lock, in Aarsleff, 1976, p. 10). These theories are 1978; Stechler & Carpenter, 1967; Stem, echoed in contemporary ontogenetic accounts 1977). In contrast, the forms of speech have to that trace the origins of language for com- be learned, and children usually begin to ac- munication to affective exchanges between quire words in the second year. The purpose infants and their caregivers. The first hy- of this study was to determine how affect ex- pothesis tested in the present study built pression relates to learning words in the pe- upon the developmental precedence of affect riod of transition from prespeech infancy to for communication. This hypothesis, the the emergence of language. hypothesis, was that the fre- quency with which infants display positive Given the temporal precedence of affect and negative emotion would predict (a) early for communication, the expression of emotion emergence of words and (b) frequency of ex- might be expected to facilitate learning pression with words. words. Indeed, among the classic phylogene- tic theories of language are those that appeal The second hypothesis built upon differ-

We are grateful for the support received from the Spencer Foundation for the conduct of this research, and we thank Richard Beckwith, Jeremie Hafitz, Karin Lifter, and Matthew Rispoli for their considerable involvement in all phases of the project. We thank Rick Booth, Abdul Raquib, and Donald Albury for their technical assistance; Margaret Honey, Tresmaine Roubaine, and Joy Vaughn for their assistance in data collection; Virginia Brennan, Suzanne Gottlieb, Geraldine McDonald, Marriete Newton-Danilo, Lisa Spiegel, Joy Vaughn, and Pia Wikstrom for their assis- tance in data transcription and coding; and Kathleen Bloom, Karin Lifter, and Carol Malatesta for their insightful comments and suggestions after reading an earlier draft of this paper. Our warmest appreciation goes to the children, for sharing their development with us, and their mothers, for their uninterrupted participation in the project. Requests for reprints should be addressed to Lois Bloom, Box 5, Teachers College, Columbia University, 525 West 120th Street, New York, NY 10027. [Child Development, 1987, 58, 1513-1522. ? 1987 by the Society for Research in Child Development, Inc. All rights reserved. 0009-3920/87/5806-0012$01.00]

This content downloaded from 128.59.160.233 on Thu, 26 Jan 2017 19:00:22 UTC All use subject to http://about.jstor.org/terms 1514 Child Development ences in the cognitive requirements for learn- quence, and a snack after the first half hour. ing words and expressing emotion, and theEach session lasted 1 hour. The children difference between them in underlying were also visited at home every month until stance. Words are arbitrary units that have they to were 15 months old, and then every 3 be learned, and words require a reflective months thereafter. stance. The mental representations that un- derlie linguistic actions of expression and in- The playroom observations were video- terpretation are constructed from aspects recordedof in stereo, and a SMPTE time-code knowledge recalled from memory and thegenerator imposed a discrete, computer- data available in perception (Bloom & Beck- readable auditory signal on the second sound with, 1986). Language emerges in the child's track of each frame of the tape. This allowed endeavor to recall and recognize words that the data to be manipulated to within one- express these mental contents, and this thirtiethen- of a second (each second of vid- deavor entails cognitive effort (Bloom, 1981). eotape containing 30 frames) at the time of The expression of affect is automatic and data processing. The videotape deck was in- effortless (a) being virtually a symptom of terfacedthe at playback with a SMPTE time-code emotion it is an expression of, and (b) having reader and an Apple II plus computer for data been in place since early infancy. Underlying coding and transcription. an expression of affect is an evaluative stance-one in which the individual evalu- Three playroom observations from each ates the situation in relation to existing child con- provided the data for the study reported tents of a mental state (e.g., Sroufe, here.1984; These were the reference points used to Wozniak, 1986). The cognitive component define of the period of transition to language. an emotion is often the evaluation of the situa- The first was the same for all of the children, tion in relation to some goal (e.g., Campos, the prespeech observation at 9 months, before Barret, Lamb, Goldsmith, & Stenberg, 1983;words emerged. The second and third were Stein & Jewett, 1986). different for the individual children and de- termined by each child's progress in lanS- The evaluative stance underlying guage.emo- The second was the month when each tions and the expression of positive and childnega- reached the criterion for "first words" tive affect could preempt the resources (FW): the first use of at least one conventional needed for learning words. However, neutral word, said at least two times. The FW obser- affect would allow the reflective stance re- vation was the reference point for the begin- quired for the effort of accessing memory for ning of the transition into language. The third the contents of mental states that underlie reference point was a "vocabulary spurt" both expression and interpretation for learn- (VS), the first increase of at least 12 new ing words. The second hypothesis tested in words (utterance types) after the child had al- this study, the reflectivity hypothesis, was ready acquired at least 20 different words. that the time spent in neutral affect states This month was the reference point for would be associated with early achievements what we considered to be the end of the tran- in learning words. sition period. Any conventional words, in- Method cluding parts of routines and interjections, were counted for identifying the FW and VS Twelve infants were subjects in a longi- language achievements, but imitations of tudinal investigation from 9 months to 2 yearsmothers' speech and self-repetitions were of age.' The infants, six girls and six boys, not. The two reference points-beginning to came from homes in the New York metropoli- say words and a sharp increase in the number tan area, and were of mixed ethnicity and of so- words-have long been considered major cioeconomic background. Each infant developmentsand defining the transition to lan- mother visited our laboratory playroom guageonce in the single-word period (e.g., Bloom, each month from about 8 months to about 1973; 28 Dromi, 1982; McCarthy, 1954; Nelson, months of age and played with groups of 1973;toys Stern & Stern, 1907). Mean age at FW that were introduced on a schedule, at 8-min was 13.6 months (range = 305-510 days) and intervals. All of the children were presented mean age at VS was 19.6 months (range = with the same groups of toys in the same 392-755 se- days).

1 Fourteen subjects participated in the research project of which this study is a part; however, data collection was begun with two of the children later than 9 months, so that the prespeech comparisons reported here were not possible. For this reason, the present study includes only those children for whom data were available beginning at 9 months.

This content downloaded from 128.59.160.233 on Thu, 26 Jan 2017 19:00:22 UTC All use subject to http://about.jstor.org/terms Bloom and Capatides 1515

Speech and affect expression were each neutral expression was defined by the face transcribed and coded independently by dif- being in a resting or baseline position, as de- ferent persons. Coders were naive both to the scribed by Ekman and Friesen (1975), and hypotheses to be tested and the units of analy- without body tension or affective vocalization. sis in the study. Mixed affect was an expression that included elements of both positive and negative va- Speech transcription.-Two transcrip- lence; equivocal affect was neither positive, tion passes were made through the data tonegative, nor neutral, as happened with ex- assure reliability. The first transcription of the pressions of or excitement. Nonneu- children's words (and all other nonword vo- tral affect expressions were also coded for calizations in the FW observation) was origi- intensity with three levels of intensity nally written out by hand, with nonword indicating the fullness of a display. Thus, the vocalizations and questionable words tran- coding scheme for describing the quality of scribed phonetically. A second transcriber expressed affect included three levels of in- then entered the words and vocalizations tensity: 1, 2, 3, and 5 qualities of valence: into the computer with times of onset neutral, and negative, positive, mixed, and equiv- offset, using the original transcription ocal.while (See Bloom, Beckwith, Capatides, & reviewing the video record. Differences Hafitz, 1987, for photographs of examples of between transcribers in identifying words these affect expressions.) were resolved by having both review the Ambiguous cases (the occurrence of a tape together; when agreement could momentary not vocalization or facial movement be reached, the vocalization was discounted that could not be assigned to one of the above as a word and not included in the analyses. categories) were also coded with time of on- Affect coding.-The emotional signal set. These were less frequent at VS (mean = carries both categorical and gradient informa- 10.7) and 9 months (mean = 15.4) than at FW tion (Stern, Barnett, & Spieker, 1983). (meanCate- = 19.8); they were not included in the gorical information is the particular emotion data analyses. In addition, the child was (such as , joy, or ) that issometimes ex- moving away from the camera, or pressed. We are unaware of any explicit thetheo- child's face was not visible for affect cod- retical or empirical claims for the relation ing withof no cues from body tension or affec- the discrete emotions to the emergence tive of vocalization. These intervals of "back- language. For this reason, a coding scheme turn" were coded for onset so that coding was devised to capture the gradient proper- affect expression in the stream of the child's ties of affective expression, valence (hedonic activity was not interrupted. (The coding tone), and intensity, and discrete categories scheme of used in this study is available from emotions were not identified or labeled. the(See authors.) Adamson & Bakeman, 1982; Ricciuti & Po- Three coders initially worked in rotating resky, 1972; and Stechler & Carpenter, 1967,pairs during training, in order to increase the for other studies that used gradient informa- accuracy of coding and to ensure initial confi- tion in the study of affect.) dence in their judgments. Training continued Every change in expressed affect in untilthe the paired coders (the three coders each stream of the child's activities was identified rotated with one another) achieved a prede- in the first half hour of the observations at 9 termined level of at least 85% agreement with months, FW, and VS, and entered into the a precoded segment of data, considered as a computer with the time of onset. The onset standard, that was coded by the investigators time of any change in affect expression was responsible for training. Segments were se- the offset time of the previous affect expres- lected randomly from different children for sion. This yielded a continuous record of (a) this posttraining test of reliability, and each affect expressions and (b) the duration of af- segment lasted from 2 to 5V2 min. Percent fect expressions from one shift in expression agreement was computed separately for the to another. An affect expression was any ob- categories of valence (positive, negative, neu- servable change in either valence or intensity tral, mixed, and equivocal) and for the inten- (using as cues the child's facial expression, sity (one, two, and three degrees) of all of the body tension or posture, and affective vocali- categories ofnonneutral valence. This level of zations such as whining, laughing, and the reliability for the coder pairs, after training like). and before the actual coding for the study was begun, was high: for valence, Pair 1 = 92%, These affect expressions were coded for Pair 2 = 90%, Pair 3 = 100%; and for inten- their valence, whether neutral, negative, posi- sity, Pair 1 = 89%, Pair 2 = 89%, and Pair 3 tive, mixed, or equivocal hedonic tone. A = 85%.

This content downloaded from 128.59.160.233 on Thu, 26 Jan 2017 19:00:22 UTC All use subject to http://about.jstor.org/terms 1516 Child Development

After continuing to code in pairs for sev- quency of nonneutral expressions (i.e., the eral weeks, reliability was assessed for the in- shifts into + 1, - 2, etc. were often shifts from dividual coders working independently in the neutral and followed by shifts back into neu- following way. Each individually coded a 3?2- tral). More important, the reflectivity hypoth- min segment that included at least 25 coding esis was predicated on the assumption that entries. The percent agreement between each time in neutral was required for learning member of a pair coding independently was words. The frequency of affect expressions high: Pair 1 = 100%; Pair 2 = 94%; and Pair was the primary measure of nonneutral affect 3 = 94% for both valence and intensity. Atbecause the emotionality hypothesis was that point the coding was begun for the data predicated on the assumption that frequent used in this study, with the coders working expression of emotion would facilitate learn- independently. ing words. The percentage of time the infants spent in nonneutral affect was not a separate The margin of error for affect onset time measure since it was the inverse of the time was 16 video frames, or approximately '/2 sec, they spent in neutral affect. which was the mean discrepancy in coding onset time among the three coder pairs. Be- Secondary measures of affect were (a) the cause affect was coded continuously, the on- mean duration of neutral affect expressions, set time of an affect expression was the offset (b) the percentage of time spent in positive time of the previous expression. The margin and negative affect expressions, and (c) the of error for onset time of an affect expression frequency of positive and negative expres- was somewhat greater than the onset and off- sions. set times for speech, due to the fact that sev- eral kinds of continuous cues were used to The first analysis was a description code affect (i.e., facial expression, body ten- (means and standard deviations) of the two sion, affective vocalization). For transcription principal affect measures: percentage of total of speech (and all nonspeech vocalizations time at spent in neutral affect and frequency of FW) the margin of error for onset and offset nonneutral affect expressions. Both measures were then correlated between 9 months, FW, times was computed by having two indepen- dent transcribers judge the times of word andon- VS to determine their stability in this pe- riod. set and offset for a sample of 100 words, in blocks of 10, from different children selected In order to test the emotionality hypoth- at random, half at FW and half at VS. The esis-that frequency of margin of error for word onset time was would predict emergence and frequency of within two video frames (or /'A5 sec) in 92% words-the of frequency with which the infants the judgments of the two independent coders, expressed nonneutral affect at FW and VS and within five video frames (or /6s sec) inwas correlated (Pearson) with (a) age at FW 82% of their judgments for word offset time. and VS and (b) the frequency of words at FW The processed data were transferred and to VS. Secondary analyses of both frequency an IBM-XT computer for analyses. The coded of expressions and measures of time were affect was reduced using a program that (a)used to compare the separate effects of posi- counted all affect expressions (including in- tive and negative affect. tensity shifts, e.g., from + 1 to + 2, or from - 2 To test the reflectivity hypothesis-that to - 1); (b) counted the number of frames neutralfor affect would be associated with early the duration of each affect expression; and achievements (c) in learning words-the per- calculated the mean duration and total time centage of time spent in neutral affect at FW spent in 10 categories of affect expression: and VS was correlated with (a) age of FW and three degrees of positive valence; three de- VS, and (b) the frequency of words at FW and grees of negative valence; neutral, mixed, and VS. equivocal valence; and ambiguous cases. In addition, measures of affect at FW and Data analyses.-The two principal mea- VS were correlated with the age at which the sures of affect that were used to test the hy- children eventually reached the transition to potheses in this study were the percentage ofmultiword speech, to test whether affect in total time (in frames) that the infants spent inthe period of word learning predicted the be- neutral expression, and the frequency of their ginning of word combinations. The criterion nonneutral affect expressions. Time in neutral for the transition to multiwords was a mean states rather than frequency of neutral expres- length of utterance of 1.5 words, discounting sions was used in the analyses for two rea- imitations, self-repetitions, bound mor- sons. For one, the frequency of shifts into phemes, "yeah" and "no," and the articles "a" neutral states was conditioned by the fre- and "the" (after Rispoli, 1986).

This content downloaded from 128.59.160.233 on Thu, 26 Jan 2017 19:00:22 UTC All use subject to http://about.jstor.org/terms Bloom and Capatides 1517

TABLE 1

MEANS AND STANDARD DEVIATIONS FOR AGE AND AFFECT MEASURES (N = 12)

Vocabulary 9 Months First Words Spurt

Age (days): Mean ...... 274.75 408.58 588.08 SD ...... 6.0 56.7 108.26 Frequency of emotional expressions: Mean ...... 49.17 72.58 75.33 SD ...... 31.43 37.05 24.49 Percentage of time in neutral states: Mean ...... 869 .846 .823 SD ...... 096 .077 .079

Results The two principal affect measures, fre- quency of emotional expression and percent- The affect measures.-The means and age of time in neutral affect, were highly and standard deviations for age and the two prin- negatively correlated with one another, r(10) cipal affect measures, frequency of emotional = -.941, -.943, -.810 at 9 months, FW, and expression and percentage of time in neutral VS, respectively. Although the information affect expression, are presented in Table 1. provided by these two measures was some- The large variability in age and frequency of emotional expression reflected individual whatdif- redundant, the two measures together were more descriptive than either one alone. ferences among the children in the develop- ment of affect from 9 to 21 months (reported The emotionality hypothesis.-The re- in Bloom, Beckwith, & Capatides, 1987). sults of the analysis to test the hypothesis that frequency of emotional expression would be The differences among the infants were associated with age of language achievement stable in the period from FW to VS, but arenot presented in Table 3. These correlations between 9 months and either of the two lan- of frequency of emotional expression and age guage achievements, as can be seen in Table at FW and VS were all in the positive direc- 2; both measures were significantly correlated tion and statistically significant at the .05 level between FW and VS (p < .05). This means or less. Thus, the more frequently the chil- that frequency of emotional expression dren and expressed their emotions, the older they time spent in neutral affect were stable for werethe at the time of the language transitions. individual children in the period during which words were emerging, but not before. The infants' emotional expressions were positive affect 78% of the time, on average. The children were expressing neutral Frequencyaf- of positive expressions also cor- fect, on average, 84% of the time. When theyrelated with age at FW, r(10) = .608, p = were expressing nonneutral affect, that affect .036, and VS, r(10) = .613, p = .034, but the was positive most often. The mean proportion frequency of negative expressions did not. of time spent in positive affect varied from Using time (in frames) as a measure, the per- .099 (9 months), to .125 (FW), to .146 (VS). centage of time spent in positive affect also The mean proportion of time in negative correlatedaf- with age at FW, r(10) = .664, p = fect remained the same: .030, .026, and .028 .019, at and VS, r(10) = .765, p = .004, but the 9 months, FW, and VS, respectively. percentage of time spent in negative affect

TABLE 2

CORRELATIONS (Pearson r) OF AFFECT MEASURES BETWEEN LANGUAGE ACHIEVEMENTS (N = 12)

LANGUAGE ACHIEVEMENTS

AFFECT MEASURES 9, FW 9, VS FW, VSa

Frequency of emotional expression ...... 307 -.070 .636 (p = .026)b Percentage of time in neutral states ...... 074 -.136 .640 (p = .025)

a FW = first words; VS = vocabulary spurt. b 10 df, two-tailed test.

This content downloaded from 128.59.160.233 on Thu, 26 Jan 2017 19:00:22 UTC All use subject to http://about.jstor.org/terms 1518 Child Development

TABLE 3

CORRELATIONS (Pearson r) OF FREQUENCY OF EMOTIONAL EXPRESSION WITH AGE AT FIRST WORDS AND VOCABULARY SPURT (N = 12)

AGE AT LANGUAGE FREQUENCY OF ACHIEVEMENTS EMOTIONAL EXPRESSION Age FW Age VSa First words ...... 658 (p = .020)b .593 (p = .042) Vocabulary spurt ...... 662 (p = .019) .684 (p = .014)

a FW = first words; VS = vocabulary spurt. b 10 df, two-tailed test. did not. These results mean that the relation Frequency of words.-In the second test between frequency of emotional expressions of the emotionality and reflectivity hypoth- and later age of language achievement was eses, the numbers of words (both utterance dependent on positive affect primarily. tokens and types) were correlated at FW and VS, and with (a) frequency of emotional ex- The reflectivity hypothesis.-The results pression and (b) the percentage of time spent of the analysis to test the hypothesis that time in neutral affect at FW and VS. None of the in neutral affect would be associated with correlations was significant. A regression early age of language achievements are pre- analysis to test the interaction between fre- sented in Table 4. The correlations of age at quency of emotional expression and age at FW and VS with percentage of time spent in FW, with the number of word types at FW as neutral expression, at FW and VS, respec- the dependent variable, was not significant, tively, were in the negative direction and F(2, 9) = 2.679, p = .122. significant, r(10), p < .02. The children who spent more time in neutral affect showed ear- The transition to multiword speech.-At lier achievements in word learning. the children's eventual transition to multi- An additional measure was the mean du- word speech (mean length of utterance M = ration of neutral expressions, which was how 1.50, SD = .111), their mean age was 24.23 long a neutral expression lasted on average. months (SD = 4.25, range = 514-960 days), Early word learning was significantly cor- and the mean span between VS and the tran- related with longer neutral expressions. A sition to multiword speech was 4.63 months problem in reporting this result, however, (SD = 1.28, range = 76-210 days). The affect was that the mean duration of neutral expres- measures at VS predicted age of multiword sions covaried with the amount of time the speech: frequency of emotional expression, infants had their backs to the camera or were r(10) = .668, p = .018, and percentage of otherwise not available for affect coding. time in neutral affect, r(10) = -.711, p = Thus, this measure was not appropriate em- .009. While not significant, the correlations pirically, although it is, at least potentially, an with the affect measures at FW and age of appropriate measure conceptually. multiword transition were high, for example,

TABLE 4

CORRELATIONS (Pearson r) OF TIME IN NEUTRAL AFFECT WITH AGE AT FIRST WORDS AND VOCABULARY SPURT (N = 12)

PERCENTAGE AGE AT LANGUAGE OF TIME IN ACHIEVEMENTS NEUTRAL EXPRESSION Age FW Age VSa First words ..... -.697 (p = .012)b -.578 (p = .049) Vocabulary spurt ...... -.522 (p = .082) -.828 (p = .001)

a FW = first words; VS = vocabulary spurt. b 10 df, two-tailed test.

This content downloaded from 128.59.160.233 on Thu, 26 Jan 2017 19:00:22 UTC All use subject to http://about.jstor.org/terms Bloom and Capatides 1519 percentage of time in neutral affect, r(10) = cation). Several convincing arguments against -.538, p = .071. Early age of transition to the notion that is a discrete emotion multiword speech was predicted by more were put forth by Wozniak (1986), and we time in neutral affect at VS, and more fre- suggest that interest is part but not all of what quent emotional expression at VS predicted was coded here as neutral affect. The category later age of transition to multiword speech. of neutral affect needs to be explored further in future research, accounting as it did for Discussion such a large proportion of the children's affect One of the most frequently cited facts expression. of One thing was clear, however; child language is that children differ neutralwidely affect was not "flat affect." The infants in onset and rate of language development. in neutral In states were best described as quiet, the present study, such differences were alert, as- and attentive. sociated with differences in the quality Although of the neutral affect described expressed affect. Language emerges herein the has its antecedents in the quiet alert child's endeavor to recall and recognize states of early infancy, the affect measures at 9 words that express the contents of their men- months were not correlated with the mea- tal states. Such cognitive activity requires a sures at FW and VS. This long-term discon- reflective stance, one which, we suggest, tinuity was probably the result of the consid- would be more likely to occur with neutral erable variability among the infants in the affect than with expression of the positive and length of time that elapsed between 9 months negative emotions. For the children in this and VS (range = 4 to 16 months). One possi- study, more time spent in neutral affect was ble explanation of the stability from FW to VS associated with early language achievements. is that learning to say words may have a The more time the children spent in neutral stabilizing effect on emotional expression. affect, the younger they were at the time of Given that affective communication is in FW and VS, and more time spent in neutral place from early infancy, learning language affect at VS predicted earlier transition to entails a shift in attentional resources. Change multiword speech. in one domain (i.e., language) may require stability in another (i.e., emotionality) (Bloom We propose that the neutral affect we & Wickstrom, 1987). Other evidence of stabil- have described here has its antecedents in the ity in the children's affect expression between "quiet alert states" during which neonates FW and VS was reported in Bloom et al. and young infants attend to visual and audi- (1987). tory stimuli (e.g., Brazelton, 1979). These quiet alert states are initially brief and occur The children in this study differed in rarely, but they increase in frequency and du- their expression of affect, and these differ- ration with development. They are generally ences were reflected in the emergence of lan- considered to be the moments during which guage. Because the subject population in- infants are perceiving and learning from ob- cluded only 12 subjects, formal tests of group jects and events in their environments (e.g., differences were not possible. However, we Lamb & Campos, 1982; Olson & Sherman, can report that children of both sexes, and 1983). Just as quiet alert states support the different ethnic and economic backgrounds, cognitive activity of the younger infant, neu- were among the children who expressed tral affect supports the cognition required for emotion frequently and talked later, and also early word learning in the transition to lan- among the children who spent relatively more time in neutral affect and talked earlier. guage in the second year. The emotionality hypothesis was not Both neutral affect at the age range supported. Frequency of emotional expres- studied here and the quiet alert states of early sion was not significantly correlated with infancy are relevant to the emotions category number of words at FW or VS. Further, emo- of "interest." Interest was considered to be a tional lability was associated with later age of source of energy in the creation of new cogni- language achievements, suggesting that emo- tive structures by Piaget (1954/1981), and in- tional expression does not facilitate early lan- terest was described as an "organizer and guage learning. motivator of attention" by Izard (1986, p. 27). These views are compatible with the claim Two points deserve emphasis in inter- we are making for the importance of neutral preting these results. The first is that we are affect for language learning. And indeed, the not talking here about language delay. All of neutral affect expression coded in this study the infants reached the FW and VS achieve- included expression of the emotions category ments, and none were at risk for language dis- of interest (C. Malatesta, personal communi- orders in this period. Several factors could

This content downloaded from 128.59.160.233 on Thu, 26 Jan 2017 19:00:22 UTC All use subject to http://about.jstor.org/terms 1520 Child Development have contributed to the differences among for the infants we studied. However, we ex- them. Caregivers in Western societies regu- pect that both factors, cognition and tempera- larly attribute meaning to their infants' ex- ment, covaried in their influence on the chil- pressions of affect (e.g., Charlesworth, 1982; dren's language learning. Johnson, Emde, Pannabecker, Stenberg, & The results reported here indicate that Davis, 1982; Lewis & Michalson, 1983; neutral affect supports early language learn- Newson, 1979; and Trevarthan, 1979). Infants ing, in particular. The amount of neutral af- may be less inclined toward expression fect may not be the relevant variable for chil- through another system when a caregiver is dren who talk later, having had the benefit particularly successful in attributing meaning of further maturation and more exposure to to their emotional expressions. Also, master- words for their language learning. Neither the ing the phonology of language depends on frequency of emotional expression nor time maturation. Children who are not ready to an- spent in neutral affect predicted the fre- alyze the sounds of speech may, instead, in- quency of words at the language achieve- crease in . ments. Thus, the differences among the in- fants were associated with age of language The second point is that the early word achievement, and not with talking more or learners who spent more time in neutral saying more different words. We would ex- states certainly expressed positive and nega- pect, however, that when equated for age, the tive affect. In fact, the early word learners and children who made the transition earlier were later word learners did not differ in their af- ahead of those who made the transition fect at 9 months (Bloom et al., 1986). How- later-on these quantitative measures as well ever, after 9 months, some infants increased as on other more qualitative measures as, for in frequency of emotional expression while example, they were with the transition to other infants began to use words. In infant- multiword speech. caregiver interaction, only some minimal level of emotional expression may be neces- To conclude, the meaning of language is sary to sustain intersubjectivity, and may be not transparent in its forms, as is the case with under the infant's control from an early age affect, but depends, instead, on acts of expres- (Sroufe, 1979; Tronick, Als, & Adamson, sion and interpretation. Such actions entail 1979). Also, mothers have been shown to dis- certain cognitive capacities. These include, at the frequency of their infants' emo- the least, accessing knowledge from memory tional expressions in the context of their so- in relation to the data of perception for repre- cialization practices (Malatesta & Haviland, sentation in the contents of states of mind, 1982). Thus, less frequent emotional expres- and recalling and recognizing words that ex- sion can be associated with learning words, press those contents (Bloom & Beckwith, without sacrifice to intersubjectivity. 1986). The results of the present study sug- gest that neutral affect supports the reflective Because the data reported here were cor- stance that such processes require for early relational, causal direction cannot be as- word learning. sumed. The relation between affect and lan- guage learning probably results from several factors and we have argued for the role played References by cognitive process in these developments. A second possible influence on the relation Aarsleff, H. (1976). An outline of language-origins between affect and language learning is theory since the Renaissance. In S. Harnad, H. underlying temperament. Because the study Steklis, & J. Lancaster (Eds.), Origins and evo- reported here concerned individual differ- lution of language and speech (pp. 4-13). New ences between the children in their affect ex- York: Annals of the New York Academy of Sci- pression, which were associated with differ- ences (Vol. 280). ences in their language learning, the results Adamson, L., & Bakeman, R. (1982). Affectivity and are relevant to the study of temperament. Be- reference: Concepts, methods, and techniques ginning with Allport's (1937, p. 54) definition in the study of communication development of of temperament as "an individual's emotional 6- to 18-month-old infants. In T. Field & A. nature, including... quality... fluctuation Fogel (Eds.), Emotion and early interaction and intensity in ," the idea that emotion- (pp. 213-236). Hillsdale, NJ: Erlbaum. ality is a function of temperament has been Allport, G. (1937). Personality: A psychosocial in- fundamental in temperament theory and re- terpretation. New York: Holt. search (Buss & Plomin, 1986; Goldsmith & Bloom, L. (1973). One word at a time: The use of Campos, 1982; Rothbart, 1986). We do not single word utterances before syntax. The have independent measures of temperament Hague: Mouton.

This content downloaded from 128.59.160.233 on Thu, 26 Jan 2017 19:00:22 UTC All use subject to http://about.jstor.org/terms Bloom and Capatides 1521

Bloom, L. (1981). Language development in rela- Goldsmith, H., & Campos, J. (1982). Toward a the- tion to cognition. In R. Stark (Ed.), Language ory of infant temperament. In R. Emde & R. behavior in infancy and early childhood (pp. Harmon (Eds.), The development of attach- 395-398). New York: Elsevier/North Holland. ment and afjiliative systems (pp. 161-193). Bloom, L., & Beckwith, R. (1986). Intentional- New York: Plenum. ity and language development. Unpublished Izard, C. (1986). Approaches to developmental re- manuscript. search on emotion-cognition relationships. In Bloom, L., Beckwith, R., & Capatides, J. (1987). De- D. Bearison & H. Zimiles (Eds.), Thought and velopments in the expression of affect. Manu- emotion: Developmental perspectives (pp. 21- script submitted for publication. 37). Hillsdale, NJ: Erlbaum. Bloom, L., Beckwith, R., Capatides, J., & Hafitz, J. Izard, C., Huebner, R., Risser, D., McGinnes, G., & (1987). Expression through affect and words in Dougherty, L. (1980). The young infant's ability the transition from infancy to language. In P. to produce distinct emotion expressions. De- Baltes, D. Featherman, & R. Lerner (Eds.), velopmental Psychology, 16, 132-140. Life-span development and behavior (Vol. 8). Johnson, W., Emde, R., Pannabecker, B., Stenberg, Hillsdale, NJ: Erlbaum. C., & Davis, M. (1982). Maternal perception of Bloom, L., & Wikstrom, P. (1987, July). The role of infant emotion from birth through 18 months. temperament in language development. Paper Infant Behavior and Development, 5, 313-322. presented at the meeting of the International Lamb, M., & Campos, J. (1982). Development in Congress for the Study of Child Language, infancy. New York: Random House. Lund, Sweden. Lewis, M., & Michalson, L. (1983). Children's emo- Brazelton, T. (1979). Evidence of communication in tions and moods: Developmental theory and neonatal behavioral assessment. In M. Bullowa measurement. New York: Plenum. (Ed.), Before speech: The beginning of inter- Lifter, K., & Bloom, L. (1986). Object play and the personal communication (pp. 79-88). Cam- emergence of language. Manuscript submitted bridge: Cambridge University Press. for publication. Buss, A., & Plomin, R. (1986). The EAS approach Lock, A. (1978). The emergence of language. In A. to temperament. In R. Plomin & J. Dunn Lock (Ed.), Action, gesture and symbol: The (Eds.), The study of temperament: Changes, emergence of language (pp. 3-18). London: continuities, and challenges (pp. 67-79). Hills- Academic Press. dale, NJ: Erlbaum. Malatesta, C., & Haviland, J. (1982). Learning dis- Campos, J., Barret, K., Lamb, M., Goldsmith, H., & play rules: The socialization of emotion expres- Stenberg, C. (1983). Socioemotional develop- sions in infancy. Child Development, 53, 991- ment. In M. M. Haith & J. J. Campos (Eds.), 1003. P. H. Mussen (Series Ed.), Handbook of child McCarthy, D. (1954). Language development in psychology: Vol. 2. Infancy and developmental children. In L. Carmichael (Ed.), Manual of psychobiology (pp. 783-915). New York: child psychology (pp. 492-630). New York: Wiley. Wiley. Charlesworth, W. (1982). An ethological approach Nelson, K. (1973). Structure and strategy in learning to research on facial expressions. In C. Izard to talk. Monographs of the Society for Re- (Ed.), Measuring emotions in infants and chil- search in Child Development, 38(1-2, Serial dren (pp. 317-334). Cambridge: Cambridge No. 149). University Press. Newson, J. (1979). The growth of shared Darwin, C. (1913). The expression of the emotions understandings between infant and caregiver. in man and animals. New York: Appleton. In M. Bullowa (Ed.), Before speech: The begin- Dore, J. (1983). , form, and intention in the nings of interpersonal communication (pp. baby's transition to language. In R. Golinkoff 207-222). Cambridge: Cambridge University (Ed.), The transition from prelinguistic to lin- Press. guistic communication (pp. 167-190). Hills- Olson, G., & Sherman, T. (1983). Attention, learn- dale, NJ: Erlbaum. ing, and memory in infants. In M. M. Haith & Dromi, E. (1982). In pursuit of meaningful words: J. J. Campos (Eds.), P. H. Mussen (Series Ed.), A case study analysis of early lexical develop- Handbook of child psychology: Vol. 2. Infancy ment. Unpublished doctoral dissertation, Uni- and developmental psychobiology (pp. 1001- versity of Kansas, Lawrence. 1080). New York: Wiley. Ekman, P., & Friesen, W. (1975). Unmasking the Piaget, J. (1981). Intelligence and affectivity: Their face: A guide to recognizing emotions from fa- relationship during child development (T. cial cues. Palo Alto, CA: Consulting Psycholo- Brown & C. Kaegi, Trans.). Palo Alto, CA: An- gists Press. nual Reviews. (Original work published 1954) Emde, R., Gaensbauer, T., & Harmon, R. (1976). Ricciuti, H., & Poresky, R. (1972). Emotional be- Emotional expression in infancy. New York: havior and development in the first year of life: International Universities Press. An analysis of , approach-withdrawal,

This content downloaded from 128.59.160.233 on Thu, 26 Jan 2017 19:00:22 UTC All use subject to http://about.jstor.org/terms 1522 Child Development

and affective responses. In A. Pick (Ed.), Min- infants and children (Vol. 2, pp. 238-267). nesota symposium on child psychology (Vol. 6, Cambridge: Cambridge University Press. pp. 69-96). Minneapolis: University of Min- Stem, D. (1977). The first relationship. Cambridge, nesota Press. MA: Harvard-University Press. Rispoli, M. (1986). The development of the transi- Stem, D., Barnett, R., & Spieker, S. (1983). Early tive/intransitive distinction: The conceptual transmission of affect: Some research issues. In and semantic origins. Unpublished doctoral J. Call, F. Galenson, & R. Tyson (Eds.), Fron- dissertation, Teachers College, Columbia Uni- tiers of infant psychiatry (pp. 74-84). New versity. York: Basic. Rothbart, M. (1986). Longitudinal observation of in- Stem, W., & Stem, C. (1907). Die Kindersprache. fant temperament. Developmental Psychology, Leipzig: Barth. 22, 356-365. Trevarthan, C. (1979). Communication and cooper- Sroufe, A. (1979). Socioemotional development. In ation in early infancy: A description of primary J. Osofsky (Ed.), The handbook of infant devel- intersubjectivity. In M. Bullowa (Ed.), Before opment (pp. 462-516). New York: Wiley. speech: The beginning of interpersonal com- Sroufe, A. (1984). The organization of emotional de- munication (pp. 321-347). Cambridge: Cam- velopment. In K. Scherer & P. Ekman (Eds.), bridge University Press. Approaches to emotion (pp. 109-128). Hills- Tronick, E., Als, H., & Adamson, L. (1979). Struc- dale, NJ: Erlbaum. ture of early face-to-face communicative inter- Stechler, G., & Carpenter, G. (1967). A viewpiont actions. In M. Bullowa (Ed.), Before speech: on early affective development. In J. Hellmuth The beginning of interpersonal communica- (Ed.), Exceptional infant: The normal infant tion (pp. 321-347). Cambridge: Cambridge (Vol. 1, pp. 164-189). Seattle: Special Child University Press. Publications. Wozniak, R. (1986). Notes toward a co-constructive Stein, N., & Jewett, J. (1986). A conceptual analysis theory of emotion-cognition relationships. In of the meaning of basic negative emotions: Im- D. Bearison & H. Zimiles (Eds.), Thought and plications for a theory of development. In C. emotion: Developmental perspectives (pp. 39- Izard & P. Read (Eds.), Measuring emotions in 64). Hillsdale, NJ: Erlbaum.

This content downloaded from 128.59.160.233 on Thu, 26 Jan 2017 19:00:22 UTC All use subject to http://about.jstor.org/terms