Development of adaptive communication skills in infants of blind parents.

Interaction Sampling and Coding Procedure

For the PCI sample, parent-child dyads were seated on a picnic mat in the assessment room, and provided with a small set of age-appropriate toys. Mothers were asked to play with their children as they would usually do at home, making use of the toys if desired. The experimenter left the dyad to play alone for 10 min, capturing footage via a remote video recording system. The SCI sample was drawn from video footage of infants interacting with a sighted, unfamiliar female researcher (one of 6 members of our research center) within a semistructured play-based assessment; the Autism Observation Scale for Infants (AOSI; Bryson, Zwaigenbaum, McDermott, Rombough, & Brian, 2008 ). Developed as a standardized behavior sample from which to observe social-communication and other behaviors in 6- to 18-month-olds at risk of developing ASD, the AOSI includes presses to elicit specific infant behaviors (e.g., the ability to track moving objects, to imitate actions, to respond to name call, etc.) and two 3–5 min periods during which the examiner engages the child in free play with standard age-appropriate toys. The aim of these free-play periods was to observe infant’s referential behavior, spontaneous vocalizations, and spontaneous actions directed at the toys or at the adult. We therefore used the AOSI free-play periods as naturalistic samples from which to code infant communicative behavior with an unfamiliar, sighted adult. Experimenters were aware of the infants’ group membership, but naive to the current study hypotheses. When interacting with an infant, the experimenter did not use a script but she prompted the infant to explore the toys provided, and responded to the infant’s vocalizations and behaviors directed at her.

The toys used in the SCI were different from those used in the PCI, as was the setup with infants seated on the floor with their parents for PCI, and on their parents’ lap across the table from the experimenter for the SCI. For each of the PCI and SCI, the setup and available toys were identical for all participants.

We coded infants’ communicative acts during the first 6 min of each interaction sample—PCI free play with the blind or sighted parent, and SCI free play with the unfamiliar sighted examiner—using aspects of the social-communication coding protocol of Clifford et al. (2010) . Each infant communication act was assigned a specific function (i.e., initiation or response) and one or more forms (i.e., vocalization, action, and face gaze; see average scores in Table 1 ). An act was classified as an initiation if the infant’s communication behavior was not in direct response to a preceding adult behavior, and as a response when it followed on from something the adult had just said or done. The form of each act was classified as a vocalization when either a nonverbal vocalization, word approximation, or speech was used, as an action when there was some communicative movement of an object (e.g., holding something up to show it) or communicative use of the infant’s own body (e.g., reaching toward an object), and as face gaze when the infant looked toward the adult’s face or made a three-point gaze shift between the adult’s face and an object. Other more specific communicative forms were coded (e.g., pointing, giving/showing, head nodding/shaking, and following gaze), but these presented infrequently during the interaction samples for infants of this age and so were excluded from further analyses. Behavior combinations such as a vocalization accompanied by face gaze were coded as having only one communicative function but multiple communicative forms.

PCI coding from video footage commenced when the researcher left the parent and child to play alone and continued for 6 min. SCI coding from video footage commenced when the researcher placed the free-play toys on the table in front of the infant, and ended after 6 min (pausing when the researcher removed the toys at the end of the first AOSI free-play episode, and resuming when she returned these to the table for the second AOSI free-play episode).

To standardize the rates of communicative function codes across participants, we calculated an initiation-response index (IRI) by subtracting the number of responses from the number of initiations coded for each infant, and dividing this by the total number of communication acts. Hence, positive IRI values represent relatively more initiations and negative IRI values represent relatively more responses among an infant’s total communication acts. Similarly, the number of vocalizations, actions, and instances of face gaze were divided by the total number of infant communicative acts to obtain proportion measures of each communicative form (e.g., proportion vocalizations = number vocalizations/total communicative acts). As the communicative forms were not independent of one another, their sum could exceed 1. Total communication acts, IRI, and proportions of vocalizations, actions, and face gaze were then included in our key analyses.

Evaluation of Interrater Agreement

Footage was coded by one of two raters, neither of whom was aware of the infants’ group status or age, or the study hypotheses. Interrater reliability was established by having both raters code a subset of clips, selected unsystematically, representing both the SIBP ( n = 13 clips) and control groups ( n = 30 clips) across both PCI ( n = 27) and SCI ( n = 16) contexts. Two-way mixed intraclass correlation coefficients (ICC 2,2 with absolute agreement; see Trevethan, 2016 ) were used to evaluate interrater agreement across the key measures (see the Results section for a description of the measures). ICCs were adequate to excellent ( Fleiss, 1986 ) for all the measures except for the IRI: total communication = .82 (ICC 2,1 with absolute agreement); IRI = .62; proportion vocalizations = .91; proportion actions = .72; proportion face gaze = .87. The lower reliability score for the IRI may have been due to the fact that with very young infants it was more difficult to judge when they initiated communication than when they responded to the parent (ICC 2,1 scores for Initiations = .45, and Responses = .77). ICC 2,1 scores for the raw number of communicative forms are reported in the online supplemental materials . Note that the form of the ICC model changes for ICC 2,2, to ICC 2,1 because the total number of communication acts and the raw number of communication forms were single measures, that were not averaged prior to the analysis.

Results

We conducted a series of three-way analyses of variance (ANOVAs)—with group varying between participants and communication context and timepoint varying within participants.

The three-way ANOVA on total communication showed main effects of communication context ( F (1, 40) = 76.81, p < .001, η p 2 = .66) and timepoint ( F (1, 40) = 36.36, p < .001, η p 2 = .48), as infants communicated more often during SCI ( M = 33.35, SD = 8.14) than PCI ( M = 18.08, SD = 6.97), and more often at Time 2 ( M = 30.56, SD = 6.68) than at Time 1 ( M = 20.87, SD = 7.55). The latter main effect was qualified by a significant Timepoint × Group interaction term ( F (1, 40) = 4.81, p = .034, η p 2 = .11) such that controls used significantly more total communication acts at Time 2 ( M = 31.84, SD = 7.07) than Time 1 ( M = 20.05, SD = 6.65), t (27) = 7.96, p < .001, d z = 1.50, whereas the differences in total communication acts between timepoints did not reach significance in SIBP (Time 2: M = 28.00, SD = 5.13; Time 1: M = 22.5, SD = 9.13), t (13) = 1.98, p = .07. The significance level for these post hoc tests and the ones reported hereafter was lowered to p = .025 after applying Bonferroni correction for multiple comparisons. Only those comparisons where p < .025 were reported as significant. Crucially, neither the main effect of group, F (1, 40) = .15, p = .70, nor the Communication Context × Group ( F (1, 40) < .001, p = .98), nor the three-way interaction term, F (1, 40) = .65, p = .43 reached significance ( Figure 1 ).

The mean IRI composite score was negative, overall, suggesting that the majority of infant communication functions were responses rather than initiations to the adult partners ( Figure 2 ). However, results of the three-way ANOVA showed that IRI was modulated significantly by group membership and communication context. That is, there were significant main effects of group ( F (1, 40) = 11.03, p = .002, η p 2 = .22) and communication context ( F (1, 40) = 131.01, p < .001, η p 2 = .77). These effects were qualified, however, by a significant Group × Communication Context interaction term ( F (1, 40) = 36.37, p < .001, η p 2 = .48). Observed power was 90% for the significant main effect of group, 99% for the significant main effect of communication context, and 99% for the significant interaction. Follow-up analyses revealed that controls ( M = −.07, SD = .31) initiated relatively more than SIBP ( M = −.52, SD = .18) during PCI, t (40) = 5.07, p < .001, d s = 1.77. Indeed, IRI of controls during PCI was very close to zero, implying a more balanced initiation and responses in this condition. By contrast, SIBP ( M = −.78, SD = .15) initiated relatively more than controls ( M = −.90, SD = .10) during SCI, t (19.28) = 2.86, p = .01, d s = .94. No other main effects or interactions reached significance (Time-point effect, F (1, 40) = .108, p = .74; Group × Timepoint, F (1, 40) = .001, p = .98; Communication Context × Timepoint, F (1, 40) = .78, p = .38; three-way interaction, F (1, 40) = .39, p = .54).

For vocalization, there was a significant main effect of communication context ( F (1, 40) = 96.51, p < .001, η p 2 = .71), with relatively more vocalization during PCI ( M = .56, SD = .19) than SCI ( M = .26, SD = .12; Figure 3 ). This was qualified by a significant Timepoint × Communication Context interaction term ( F (1, 40) = 7.95, p = .007, η p 2 = .17). Observed power was 99% for the significant main effect of communication context and 80% for the significant interaction. Follow-up analyses revealed that infants’ vocalizations increased between Time 1 ( M = .20, SD = .16) and Time 2 ( M = .32, SD = .19) during SCI, t (41) = 3.02, p = .004, d z = .48, but not during PCI, t (41) = .61, p = .55 ( M Time1 = .58, SD Time1 = .25; M Time2 = .55, SD Time2 = .25). No other main effects or interactions reached significance (group effect, F (1, 40) < .001, p = .99; timepoint effect, F (1, 40) = 2.57, p = .12; Group × Communication Context, F (1, 40) = 1.74, p = .19; Group × Timepoint, F (1, 40) = 1.69, p = .20; three-way interaction, F (1, 40) = .45, p = .51).

A significant main effect of communication context for proportion of actions ( F (1, 40) = 87.74, p < .001, η p 2 = .69) reflected infants’ greater use of communicative actions during PCI ( M = .48, SD = .17) compared with SCI ( M = .21, SD = .08; Figure 4 ). This effect was qualified, however, by a significant Group × Communication Context interaction term ( F (1, 40) = 10.04, p = .003, η p 2 = .20). Observed power was 99% for the significant main effect of communication context and 87% for the significant interaction. Follow-up analyses revealed that, during PCI, SIBP ( M = .38, SD = .13) used relatively fewer actions than controls ( M = .52, SD = .17), t (40) = 2.72, p = .01, d s = .93, whereas there was no such between-groups difference during SCI (SIBP: M = .22, SD = .08; control: M = .20, SD = .08), t (40) = .93, p = .36. No other main effects or interactions reached significance (group effect, F (1, 40) = 3.28, p = .08; timepoint effect, F (1, 40) = .009, p = .93; Group × Timepoint, F (1, 40) = .80, p = .38; Communication Context × Timepoint, F (1, 40) = .03, p = .86; three-way interaction, F (1, 40) = 1.84, p = .18).

Finally, for proportion of face gaze, there were significant main effects of group ( F (1, 40) = 4.60, p = .038, η p 2 = .10), communication context ( F (1, 40) = 235.11, p < .001, η p 2 = .86), and timepoint ( F (1, 40) = 12.73, p < .001, η p 2 = .24). Observed power was 54% for the significant main effect of group, 99% for the significant main effect of communication context, and 93% for the significant main effect of time. These were such that SIBP used more face gaze ( M = .60, SD = .09) than controls ( M = .52, SD = .11), and all infants used more face gaze during SCI ( M = .77, SD = .08) than PCI ( M = .33, SD = .18), and at Time 1 ( M = .59, SD = .14) compared with Time 2 ( M = .51, SD = .13; Figure 5 ). The Commun

Discussion

This study represents a unique investigation of the communication behavior of SIBP, adopting a prospective follow-up design to examine interaction with both a blind parent and a sighted unfamiliar adult. We examined various aspects of infant communicative behavior—including both the function of communication acts and various forms of signaling these to the partner (i.e., via vocalization, action, and face gaze)—and found significant interactions between child group and social partner for some of these. Specifically, when they interacted with their blind parents, compared with control infants interacting with their own sighted parents, SIBP showed marked differences in both the function and the form of communication including using relatively more responses than initiations, and fewer communicative actions. By contrast, during interaction with a sighted unfamiliar adult, SIBP initiated relatively more than controls, with both groups using similar levels of communicative actions. A similar trend was observed for face gaze, where SIBP showed more face gaze than controls during interaction with their parents, but with no between-groups differences during interaction with a sighted stranger. Interestingly, both groups used similar levels of vocalizations, and vocalized more during the interaction with the parent than with a sighted stranger, and more at Time 2 than at Time 1. The results suggest that SIBP are flexibly and adaptively switching the style of their communication when with blind parents versus a sighted experimenter. This is consistent with the prediction derived from the interactive specialization model ( Johnson, 2011 ), which hypothesizes that infants develop optimized communication behavior adaptive to the given communicative context. By contrast, it is inconsistent with the prediction derived from the affective learning viewpoint, which hypothesizes that infants learn the reward value of communication behavior through interaction with parents/caregivers and generalize this to other communicative contexts.

The directions of group differences in both the function and the form of communication are also informative, and somewhat counterintuitive. As for communicative function, SIBP responded more toward their parents than did controls, but initiated relatively more (or rather, responded relatively less) toward the sighted experimenter than did controls. This might suggest that SIBP have acquired skills to more effectively (or frequently) initiate communication to compensate for their parents’ difficulty to notice a visual form of communication. It may also be that this between-groups difference during PCI simply reflects a stronger tendency for initiated communication by blind (compared with sighted) parents—hence eliciting relatively more responses by their infants. However, this latter interpretation cannot account for the group differences also observed in communicative functions during the SCI condition (i.e., SIBP initiated relatively more than controls), in which both groups of infants were communicating with unfamiliar sighted adults.

As for the form of communication, SIBP used fewer communicative actions than controls, only when interacting with their parents, suggesting that SIBP also flexibly change the channels of communication depending on their communicative partner. It seems rational not to use actions—such as showing or reaching for an object—when these cues are less likely to be picked up by their blind parents. However, these results also showed that SIBP used a similar amount of these actions when they interacted with sighted communicative partner, suggesting that they can still use this channel of communication when it is efficient. In addition, overall higher use of face gaze by SIBP—particularly during interaction with their blind parents—may seem inconsistent with a previous study ( Chiesa et al., 2015 ) which found shorter overall face gaze in SIBP. Possibly, this discrepancy is due to the adoption of different coding schemes. We coded the frequency of each form used in successful communication events, whereas Chiesa et al. (2015) coded the total frequency of each behavior during an observation period regardless of whether or not behaviors lead to successful communicative exchanges. Thus, it is possible that SIBP overall spend less time attending to parents’ faces, but efficiently respond to parental communicative bids with face gaze.

Methodological differences between studies may also explain the apparent contradiction between the results of the current study and those of our recently reported eye-tracking studies ( Senju et al., 2015 ). Senju et al. (2015) found that SIBP and controls differ in terms of their gaze following behavior and face scanning pattern. Specifically, when presented with video clips of a female actress which looks directly toward the infant and then gazes at one of two objects in front of her, SIBP and controls follow equally frequently the gaze of actress to the object, but SIBP look for a shorter period of time at the gazed-at object that controls do. On the contrary, when watching a silent video of a dynamic female face, SIBP look more at the mouth than at the eyes area, whereas controls show the opposite face-scanning pattern, looking more at the eyes than at the mouth. The findings reported in the current paper, in contrast, are based on successful communication bids between infants and adults, and quantify different forms of communication among which is the proportion of looks to the adult’s face, irrespective of the part of the face attended to. In fact, given the interaction setup in the current study, it would be very difficult for us to report which part of the adults’ face infants gazed at when communicating. We therefore cannot rule out that the face-scanning pattern observed in the SIBP group by Senju et al. (2015) is specific to certain communication partners. Interestingly, Senju et al. (2015) found that SIBP and co