Edited by:
Reviewed by:
*Correspondence:
This article was submitted to Language Sciences, a section of the journal Frontiers in Psychology
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Infants are known to engage in conversation-like exchanges from the end of the second month after birth. These ‘protoconversations’ involve both turn-taking and overlapping vocalization. Previous research has shown that the temporal organization of adult–infant turn-taking sequences is similar to that of adult verbal conversation. It has also been shown that young infants adjust the quality of their vocalization in response to the quality and timing of adult vocalization. We present new evidence of turn-taking interaction in infants aged between 8 and 21 weeks based on the analysis of 176 samples of naturalistic face-to-face interactions from 51 dyads. We found high levels of latched turns as well as frequent initiation of turn-taking by infants at these ages. Our data do not support the hypothesis that turn-taking ability increases with age between 2 and 5 months but do suggest that infants are active participants in turn-taking from the earliest age and that mothers adjust turn-taking formats to infants.
Conversation is a complex communicative process in which visual and auditory signals are combined in a flow of turn taking. An important aspect of the temporal organization of conversation is the fluency between turns. The end of a speaker’s turn needs to be anticipated by the next speaker for the conversation to be fluent (
Yet a number of studies in psychology describe turn-taking interaction occurring between infants and adults as early as in the second month of life, long before any access to lexical information. A recent study conducted by
In the first weeks after birth, infants begin to produce vocalizations that are described in the literature as coos and murmurs (
The appearance of these kinds of vocalizations gives rise to vocal exchanges with adult partners that resemble conversations because they are characterized by alternating vocalizations separated by audible pauses (
Various approaches have been used to demonstrate the active role of both mothers and infants in the first half year of life in face-to-face interaction. A durational approach reveals stable timing structures in maternal behavior across vocal and kinesic modalities (
A related approach to studying protoconversational organization focuses on the match or cross-correlation between temporal patterns of maternal and infant vocalization in spontaneous face-to-face interaction. The fact that mothers and infants engaged in social interaction match each other for vocalization and pause duration suggests a mutual regulation of the turn-taking exchange (
If infants partake in a truly co-regulated turn-taking, they must have the ability to perceive the contingent relations between their own behavior and that of their partner. These contingent relations hinge on the perception of timing in social interaction. Research has shown that infants’ sensitivity to contingency changes right around the time when protoconversations appear, around the age of 2 months (
Starting around the age of 2 months, the cooing stage (
One longitudinal study has reported a quantitative increase in vocal turn-taking with less overlapping vocalization between 12 and 18 weeks of age (
The most detailed longitudinal study of the timing of turn-taking in infancy was conducted recently by Hilbrink et al. (submitted) on 12 infants aged between 3 and 18 months. These researchers report on the prevalence of turn-taking exchanges throughout the period they studied. They also find that turn-taking organization varies little between 3, 4, and 5 months of age but that turns slow down markedly around the age of 9 months. At this age infants respond to mothers’ utterances with longer gaps whereas maternal turn-timing remains stable over time. Furthermore, this study shows that the amount of overlap in turn-taking remains constant between 3 and 5 months of age but decreases thereafter. The slowing down of turn-taking can be associated with important qualitative changes in social interaction and communicative skills such as joint attention.
The present paper assesses developmental change in turn-taking organization between the 3rd and the 5th months of life. We first aimed to explore the hypothesis that turn-taking is, from the first weeks of life, a mutually coordinated effort and then to assess developmental change and stability in the amount of overlap, duration of switching pauses (gap) between turns and length of TTSs. We thus compared spontaneous vocal interactions between mothers and infants ages between 2 and 3 months with those of mothers with 4-to-5-months-old infants.
Fifty one MI dyads participated in this study. Infants were aged between 8 and 21 weeks [28 boys and 23 girls aged respectively 12.8 weeks ± 3.77 (range: 8–19) and 13.2 weeks ± 3.73 (range: 8–21)], were born full-term and in good health. Out of the 51 mothers, three mothers were bilingual speakers but spoke to their infants in French. The sample was divided into two age groups: a group of ‘younger’ infants (35 8–13 weeks, mean age 10.8 weeks ± 1.54, 15 girls) and a group of ‘older’ infants (15 17–21 weeks, mean age 18.3 ± 1.24, 8 girls). The present research was approved by the university ethics committee (CCP n°1450089).
All dyads were recorded in naturalistic contexts, in their home, when infants were in a quiet alert state. Before each observation, consent forms were signed by the parents. Mothers and infants were placed in comfortable positions, facing each other. Mothers were asked to talk to their infants in their usual manner for approximately 10 min and to avoid using any toys. Video and audio recordings were made using two camcorders (Sony Handycam HDR-CX190) and a digital audio recorder (Korg Sound on Sound Unlimited Track Recorder) placed near the dyad. Only the audio recordings were used in this study.
A total of 176 audio samples were selected [3.5 audio samples on average per infant ± 2.2 (range: 1–9). Audio sample length was on average 106.7 s ± 57.9 (range: 18–252)]. In all 90 min and 41 s of interaction were analyzed. Samples lasted on average 31 s. The samples were taken from a large corpus of audio recordings, they were the first to meet following four selection criteria:
Each sample included at least one vocal contribution by the infant and no negative infant vocalizations (fuss, cry).
In each sample, mothers addressed their infant directly. Sequences including song to infants were discarded.
Samples were segmented based on pause duration: pauses between vocalizations that exceeded 3 s marked the end of the selected audio sample.
Recording quality was optimal for all samples.
We used Sound Analysis Pro (
Based on visualization of spectrograms and audio guidance, each sequence was manually segmented into 4 types of events: maternal vocalization, infant vocalization, overlapping vocalization and pause, according to the following criteria:
A vocalization (either by mother or infant) was defined as the production of vocal sound by one partner that was either continuous or included unvoiced segments of less than 300 ms. If the silent pause following an audible vocal sound was greater than 300 ms, two successive vocalizations were coded.
Overlapping vocalization was coded when either mother or infant vocalized over the vocalization of the partner. The entire vocalization was coded as overlapping even if it was only partially masked by the partner’s vocalization.
Pauses occurred either between two vocalizations by the same partner (within- speaker pause) or between alternating vocalizations (switching pause). Within-speaker pauses had a duration that was necessarily greater than 300 ms and lower than 3000 ms. Switching pauses could range from a few milliseconds to 3000 ms. Two alternating vocalizations were qualified as latched when the switching pause had a duration of less than 50 ms.
Vegetative sounds produced by infants such as burps, growls or hiccups, noise from the environment and vegetative sounds produced by mothers, such as coughs, were not coded.
A TTS was defined as sequence of vocalizations involving at least one alternation between speakers. Such a sequence could involve alternation between a vocalization of mother and of infant or between a vocalization of mother or infant and an overlapping vocalization. A TTS ended when the same speaker produced at least two vocalizations in succession or the pause following a vocalization was greater than 3000 ms.
Sound Analysis Pro (
Twenty percent of the data set were double-coded. Inter-coder reliability (Pearson product-moment correlations) was 0.87, 0.98, 0.78, and 0.95 respectively for number of infants’ vocalizations, mothers’ vocalizations, vocalizations involving overlap and pauses. Onset positions were considered identical if they occurred within 50 ms, thus measures of vocalization duration had an error tolerance of up to 100 ms. Both coders correctly identified 80.6% of all onset positions within the subset of double-coded sequences.
Analysis was done with Stata for Windows (version 12). Multiple regression was used to compute partial regression coefficients and logistic regression to estimate ORs and Wald 95% CIs (
The entire sample comprised a total of 2943 vocalizations of which 748 were produced by the infants alone, 1851 were produced by the mothers alone and 344 involved both mothers and infants vocalizing in overlap. Thus, the total sample of infant vocalizations comprised 1092 vocalizations and the total sample of maternal vocalizations comprises 2195 vocalizations. In total 2152 pauses were identified, of which 838 were switching pauses (38.9%).
Infant vocalizations occurring within less than 3 s of a maternal vocalization, i.e., involved in turn taking, represented 73.1% of all infant vocalizations (see
Proportions of infant vocalizations according to position in relation to other vocalizations and pauses.
Infant vocalizations ( |
||||||
---|---|---|---|---|---|---|
All infants ( |
8–13 weeks ( |
17–21 weeks ( |
||||
% | % | % | ||||
Vocalizations preceded by another infant vocalization | 140 | 12.8% | 66 | 9.2% | 74 | 19.8% |
Preceded by mother vocalization with pause | 252 | 23.1% | 195 | 27.1% | 57 | 15.3% |
Preceded by mother vocalization without pause | 202 | 18.5% | 145 | 20.2% | 57 | 15.3% |
Involving overlap | 344 | 31.5% | 226 | 31.4% | 118 | 31.6% |
Neither preceded nor followed by any other vocalization | 154 | 14.1% | 87 | 12.1% | 67 | 18.0% |
Among the vocalizations occurring within TTSs, some were overlapping vocalizations (
Infant vocalizations lasted on average 869.7 ms ± 662.3 (range: 50.2–3640.0). Multiple regression conducted on the durations of vocalizations revealed no effect of age (
Maternal vocalizations occurring within less than 3 s of an infant vocalization – i.e., involved in turn taking – represented 35.8% of all maternal vocalizations (see
Proportions of maternal vocalizations according to position in relation to other vocalizations and pauses.
Mother vocalizations ( |
||||||
All mothers ( |
8–13 weeks ( |
17–21 weeks ( |
||||
---|---|---|---|---|---|---|
% | % | % | ||||
Vocalizations preceded by another mother vocalization | 1113 | 50.7% | 928 | 54.5% | 185 | 37.7% |
Preceded by infant vocalization with pause | 257 | 11.7% | 202 | 11.9% | 55 | 11.2% |
Preceded by infant vocalization without pause | 184 | 8.4% | 136 | 8.0% | 48 | 9.8% |
Involving overlap | 344 | 15.7% | 226 | 13.3% | 118 | 24.0% |
Neither preceded nor followed by any other vocalization | 297 | 13.5% | 212 | 12.4% | 85 | 17.3% |
Contingency analysis conducted on type of vocalization revealed a significant age effect (Chi square = 51.6;
Mothers’ vocalizations lasted on average 1144.7 ms ± 904.9 (range: 50–9803). Multiple regression conducted on the durations of the vocalizations showed a significant effect of age: the vocalizations of mothers of younger infants were on average 399 ms longer than those of mothers of older infants (
Vocalizations involving overlap lasted on average 1550.7 ms ± 1090.1 (range: 110–4928). Multiple regression conducted on the durations of the 344 overlapping vocalizations revealed a significant effect of age: overlapping vocalizations involving younger infants were 812 ms longer than those involving older infants (
Among the 454 turns produced by infants 44.5% (
Pauses between vocalizations by the same speaker, which were by definition restricted to the range of 300 to 3000 ms, lasted on average 745.5 ms ± 557.0 (range: 310–2643.6) for infants and 967.3 ms ± 606.4 (range: 300–2994.8) for mothers. Multiple regression analysis revealed that within speaker pauses were not significantly different according to age, neither for infant (
The samples comprised 838 switching pauses lasting on average 730 ms ± 543.6 (range: 50–2974). Multiple regression analysis showed that switching pauses were on average 174.0 ms longer in samples involving older infants (
Among the switching pauses, 60.5% concerned pairs of vocalizations that were either between an IM pair of vocalizations (30.1%) or between a MI pair of vocalization (30.4%), i.e., switching pauses which did not involve an overlapping vocalization. Both types of switching pause were analyzed separately. Multiple regression analysis confirmed the overall age effect for all switching pauses (
In total, 489 TTSs were identified, lasting on average 5.68 s ± 4.39 (range: 0.3–28.2) and ranging from 2 to 18 turns (mode of three turns, i.e., three alternating vocalizations). Because duration and number of turns were highly correlated [
Multiple regression conducted on the number of turns per TTS did not reveal any significant effect, neither of age (
TTSs were divided into two subgroups depending on which partner initiated the sequence. Sequence initiation was coded when a vocalization occurring after a pause lasting at least 3 s was followed by at least one vocalization by the partner within less than 3 s. Among the 489 TTSs collected, 44.8% were initiated by infants. Logistic regression showed that TTS initiation by infant was most likely to occur in older infants (
Turn-taking sequences were also divided in two subgroups according to which partner terminated the sequence. A sequence termination was coded when a vocalization following at least one vocalization by the partner within less than 3 s was followed by a pause of at least 3 s. Among the 489 TTSs observed, 37 (7.6%) ended with a co-vocalization and were then excluded from analysis. Twenty four percent of the remaining 447 TTSs were ended by infants. Logistic regression showed that TTSs were more likely to be ended by infants in older infants (
In the present study, durational features of vocal interactions between mothers and infants from two age groups (2–3 and 4–5 months) were collected using acoustic analysis software. The aim of the study was, first, to ascertain whether vocal exchanges involving young infants has a turn-taking format and, second, to investigate developmental change in vocal turn-taking.
The first finding was that three out of four infant vocalizations are either followed or preceded by a maternal vocalization. Infants in both age groups vocalize in turn as frequently, and the most common turn-taking format at both ages is one involving three turns (two alternations between mother and infant). Indeed, infants respond to mothers’ vocalizations as often as mothers respond to infants’ vocalizations at both ages. Although the turn-taking format is observed in vocal interactions of infants aged between 2 and 5 months, the question that needs addressing is whether mothers are responsible for creating these formats or whether infants actively partake in shaping TTSs.
One way to investigate infants’ active role in turn-taking is to ask whether they demonstrate anticipation of the mothers’ turns and whether they show initiative in generating a vocal response from the mother. The present study provides partial answers to these questions because it is based on durational measures and does not include other acoustic features such as intonation contour, pitch matching or amplitude modulation. Some of our findings support the hypothesis that, already from the 2nd month of life, infants anticipate turns and initiate TTSs.
Infants at both ages used a high proportion of latched turns, that is they began to vocalize within a few milliseconds (less than 50 ms) of the end of the mother’s vocalization, without any overlap. To our knowledge, no study of early MI vocal interaction has reported on the frequency or significance of latched turns. This finding is significant in teasing apart the roles of adults and infants in turn-taking exchanges. Indeed, although latched turns are less frequent than overlapping vocalization, it seems implausible that they would occur randomly at the frequencies we report. The temporal window for initiating vocalization is markedly smaller for latched turns than for turns involving a pause and for overlapping vocalizations. Indeed, the average time frame within which an infant vocalization is described as an ‘overlapping vocalization’ in our study is 1145 ms (the average duration of a maternal vocalization) and the average time frame within which an infant vocalization is involved in a turn (excluding latched turns) is 730 ms (the average duration of a switching pause). The high frequency of latched infant vocalizations suggests that infants actively shape the turn-taking organization of an exchange and anticipate events within it. Based on a comparison of the probabilities of infants producing different prosodic contours, a previous study showed that 3-months-old selectively match their vocalizations to preceding maternal vocalization (
The finding that infants in the 3rd months already frequently perform latched turns poses a major challenge. How do young infants project the end of a turn as precisely as adults do in verbal conversation, with ‘no-gap-no-overlap’ (
Furthermore, it is interesting to explore the possible functions of latched turns in protoconversational exchange. The experience of a seamless transition between self-expression and other-expression may be highly relevant for a young infant, reinforcing an emerging sense of agency in the 1st months of life. Latched turns may be conceived as joint action, where each individual’s actions are coordinated so as to achieve a joint outcome and where each individual’s action cannot be understood in isolation from the others’ (
Overall, comparison of the two age groups revealed both continuity and change in turn-taking organization. We do not find any differences in terms of mean frequency of vocalizations involving overlap between the two ages. Nor do we find that the duration of switching pauses decreases with age. On the contrary in our data they are longer among the older infants. Older infants do not perform proportionately more latched turns than younger infants. We did find that switching pauses were more frequent in younger infants’ interactions. This finding should entail that younger infants partake in a greater number of alternating vocalizations but we did not find that the number of turns within TTSs increases with age. This inconsistency may be explained by the quantification method we used for turn-taking. Overlapping vocalizations and latched vocalizations were taken into account in the quantification of number of turns whereas the frequency of switching pauses was quantified on the basis of non-overlapping vocalizations alone. It is therefore not possible to describe the contribution of each type of vocalization within the TTSs. Future studies should be aimed at more precise description of overlapping vocalization in MI interaction.
Turn-taking sequences were found to be longer for older infants despite the fact that infant vocalizations have similar durations in both age groups and that mothers’ vocalizations are shorter in interactions with older infants. Longer TTSs at 4–5 months than at 2–3 months are most likely explained by the increase in switching pause duration between the two ages. Furthermore, a greater number of infant vocalizations were either responded to or were responses to the mother among the younger infants, and older infants vocalized more often twice in succession. These two age-related changes may reflect a transition between the highly social cooing stage and the more solitary exploratory sound play that follows in vocal development (
Our findings confirm the active role of mothers in adjusting their vocal behavior to infants. Mothers’ vocalizations were found to be longer with younger infants, and they produce more successive vocal utterances when interacting with younger infants. Mothers perform more latched turns with older infants and switching pauses between IM pairs of vocalizations are shorter on average than between MI pairs, suggesting that mothers respond to infants faster than infants respond to mothers. In a previous study, the duration of switching pauses in vocal turn-taking between 2 and 3 months-old infants and mothers was found to vary cross-culturally in relation with parenting styles and cultural representations (
The significance of our findings on initiation and termination of TTSs is ambiguous. We found that older infants were more likely to initiate a TTS. However, initiating a sequence does not imply that older infants show greater initiative in turn-taking, as this finding could be explained by mothers responding more often to the vocalizations of older than of younger infants. In fact, a recent study has shown that, without knowledge of infant age, naïve listeners implicitly attribute greater communicative intentionality to vocalizations of 5 months-old infants than to those of 1 month-old (Gratier et al., submitted). Our data also show that older infants more frequently terminate TTSs than younger infants. This may also be seen to reflect the mother’s stance toward the infant rather than a change in the way infants organize turn-taking because mothers may not respond as often to terminal vocalizations with older infants. It would be interesting to take into account the acoustic characteristics of these initial and terminal infant vocalizations in order to gain insight into the question of infant initiative. Prosodic cues such as intonation contour may signal initiative to mothers, for both initiating and terminating turn-taking. It is also possible that mothers do not respond to terminal vocalizations because the TTS has reached a durational threshold, her unresponsiveness would then serve a regulatory function. It is worth investigating the transformations of the two partners’ roles in turn-taking across ages. Indeed, mothers may progressively give the infant more and more prominence in the interaction, considering her more and more as an active partner with “something to say” while making the rules of turn-taking more salient. Infants may become less interested in a vocal turn-taking and more interested in multimodal turn-taking around object play or shared activities.
Overall our findings do not clearly support the hypothesis that infants’ become more competent at turn-taking between the 3rd and 5th months of life. Indeed, we find neither shorter switching pause durations nor less overlap between these ages. Rather, our findings suggest that turn-taking organization is sensitive to changes in infants’ social motives, reflecting a growing involvement in object play and joint activity. Although mothers are clearly highly adaptive and active in vocal exchanges with infants, we have found evidence that already very young infants play an active role in shaping the unfolding of TTSs. This exploratory study on the early development of turn-taking points to the possibility that turn-taking in preverbal interaction adapts to infants’ changing motives for communicating and learning, and paves the road to the crucial socio-cognitive skills that precede and enable language use.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
We would like to thank all the families who participated in this study. We especially thank Anne Bobin-Bègue and Giulietta Von Salis for help with data collection, doctors Bourrelier and Guillemot for providing access to MI health care services and Rose-Marie Guérinel for help contacting families. We also gratefully acknowledge the help of Ofer Tchernichovski on the use of Sound Analysis Pro. This research was supported by the Fondation de France (reference # 00015165) and by a CAPES grant (BEX-10506/12-8) from Brazil.
confidence interval
infant–mother
mother–infant
odds ratio
turn-taking sequence