Social interactions and voice processing in PWS
Like individuals with ASD, participants with PWS display problems in social functioning, characterized by a reduced ability to interpret and respond to social information . Their empathy deficit, combined with social withdrawal, prevents them from engaging in harmonious peer-group relationships [6, 18, 19]. It is only natural to assume that their social interaction difficulties are related to deficits in processing the two major sources of information in human communication: the human face and voice [20, 21]. Concerning facial information processing, participants with PWS have a known facial recognition deficit related to an altered strategy of face exploration . Belin et al.  suggested that the human face and voice constitute a fused entity-in which case, face processing deficits should be accompanied by voice processing deficits. According to Salles et al. , participants with PWS present a specific deficit in distinguishing voices from nonvoices. However, it remains unclear whether their voice identification is also impaired and whether this is related to their performance for environmental sounds.
In the present study, we found that participants with PWS, especially UPD participants, were deficient in the recognition of voices and nonvoices. This deficit was slightly weaker for voices than for nonvoices in both PWS subgroups. It was accompanied by significantly longer RTs, with no difference between voices and nonvoices.
Given that no auditory deficit was reported for any of the PWS participants we tested, this voice recognition deficit is unlikely to be of sensory origin, except at an advanced level of sensory integration (e.g., multisensory integration). Salles et al.  demonstrated decreased multisensory benefits with an absence of violation of the race model indicating that multisensory information does not converge in participants with PWS. Though some of the participants were the same as in the present study, the analyses performed in the present study are different from those in Salle et al. , so that we cannot compare directly participants’ performances with Salles et al. .
Although this may depend on the particular task and the cognitive load, the participants with PWS seemed to have specific impairments that could not solely be explained by their ID. Even if top-down effects of impaired integrative functions cannot totally be excluded, the deficit in the perception of voice and nonvoice sounds was more likely to be attributable to impairment of the most integrative associative sensory areas (e.g. posterior portion of the superior temporal sulcus (STS) known to be involved in integrative and multisensory analysis, and temporal poles involved in voice processing). The temporal pole (Brodmann area, BA 38) was found to be hypoactive in a resting-state PET study of participants with PWS, as was the posterior temporal area (BA 22) . Individuals with ASD also show deficient activation during voice perception in the temporal voice areas, which are typically more sensitive to vocal stimuli . Moreover, there is a hypothesis that autism results from the disconnection of different brain areas owing to STS dysfunction . A variety of sensory disabilities have been reported in ASD , and similar ones may be present in participants with PWS.
Hit rates showed that UPD participants were more impaired on voice and nonvoice perception than DEL participants. These findings confirmed that the participants with PWS had a sensory integration deficit, but also indicated that their higher order integrative deficits needed to be considered, given that PWS is characterized by ID and impaired social adaptation. To unravel these effects, we looked for correlations between hit rates for voices and nonvoices and IQ and DBC scores, but no significant correlation was found.
To further check whether this difference could be due to ID, we ran a Mann-Whitney test to analyze the difference in IQ between UPD and DEL participants, but found that it was nonsignificant (p > 0.6). It is therefore unlikely that the differences between the UPD and DEL participants on voice and nonvoice identification were related to ID. This confirms the specific deficit of participants with PWS for voice/nonvoice discrimination, but also their heterogeneity , and explains the identification results we found.
Although participants with PWS had a voice perception deficit that could be predicted from their impaired social functioning, their deficit for nonvoices was even more pronounced. This finding may contradict the hypothesis of a centrally driven, highly integrative origin of the deficit, insofar as voices require a more integrative cognitive function related to the perception of identity and personality . Then again, PWS participants’ compensatory mechanisms for the recognition of such socially important stimuli as voices could be of central origin. However, given the social deficits of these PWS participants, it would be difficult to attribute this compensation for voices to social feedback or social adaptation, as opposed to the special role of the voice.
Decision Modelling with HDDM
The longer RTs for voice and nonvoice detection in PWS (Fig. 1b) raise the question of whether they were due to slower decision-making or to a general slowdown in perception and motor reactions. To address this question, we used the HDDM, which implies that before giving a response, individuals have to accumulate and integrate a certain amount of information. The precise amount of information they need to arrive at a decision is represented by a threshold, while the speed at which they reach this threshold is the drift rate. Importantly for our question, the model also deduces their nondecision time, reflecting the time it takes them to execute the motor control and detect the stimulus. We assumed that PWS participants’ nondecision time and drift rate would both be longer, owing to their general slowdown.
For both voice and nonvoice identification, participants with PWS exhibited a similar pattern of parameters in the HDDM. They had a higher threshold for both types of auditory stimuli, meaning that they needed to accumulate more information to identify them than controls did. This need for more information can be explained by a lack of integrative brain capacity, linked to their general ID. This alone would have been enough to slow down their responses, but they were also slower at accumulating the necessary information (lower drift rate).
As a results, both factors (higher threshold and lower drift rate) contributed to the long RTs of participants with PWS, which were nearly twice as long as those of controls (Fig. 1b). Moreover, PWS participants had longer nondecision times than controls (Fig. 2b), which also contributed to their longer RTs.
Thus, the HDDM demonstrated that a number of different processes contribute to the behavioral slowdown in participants with PWS. Furthermore, the initial bias parameter indicated that participants with PWS were predisposed to the perception of voices, but to a lesser extent than controls were. It is curious that, despite the significant difference in performance between the UPD and DEL participants, the HDDM did not indicate any difference between the two subgroups on any of the parameters. This may mean that the UPD and DEL participants used similar cognitive strategies, but were more or less efficient in doing so, leading to the significantly different performances. According to the HDDM, participants with PWS needed more time to accumulate information for decision-making and were predisposed to voice perception. The sound categorization tests highlighted categorization strategies similar to those of controls, although PWS participants had more problems describing the categories they had created.
Auditory free sorting task
Our exploration of PWS participants’ performances on the identification of predefined categories of voices and nonvoices led us to postulate that the deficit we observed was not related to higher-order cognitive functions, but instead to a deficit in integrative sensory processing in the temporal lobes. To further verify this hypothesis, we examined the results of an FST task that required greater involvement of cognitive and intellectual abilities such as similarity judgment, working memory, and executive functions . The additional load on high-order integrative functions was generated by requiring participants to establish the categorization criteria/principles for themselves. In an FST, participants may group items according to a variety of subjective criteria, but sounds are usually grouped according to their common semantic or acoustic properties [16, 28, 29]. We found that controls divided the sounds they heard into voice, instruments and environmental categories (i.e., categorization predominantly based on semantic information as a consequence of identifying the sources of the sounds). Both the tree diagrams (Fig. 3A) and MCA maps (Fig. 3b) showed that participants with PWS created the same voice, instruments and environmental categories. This means that participants with PWS used the same semantically based cognitive strategy as controls. Furthermore, no correlations were found between categorization and IQ. PWS participants’ IQ therefore only weakly influenced their ability to establish sound categories. This weak influence could be detected at the within-category level, where the within-group distances between stimuli were significantly greater than they were for controls. The within-category dispersion can be explained by subgroups of PWS participants with low IQ who did not categorize the sounds as the other PWS participants did. The outlier PWS participants in the participant maps (Fig. 4a) had a lower IQ than the other PWS participants (there were no outliers in the control group). However, the correlation with IQ disappeared when we considered it from the opposite direction, in that not all PWS participants with low IQ were outliers with poor categorization performances.
As demonstrated by the word clouds (Fig. 4b), participants with PWS were less accurate in the description of the stimuli because of their poorer vocabulary, which may have been related to their ID. We also noticed that they tended to tell stories involving the stimuli, instead of providing an exact description of each category they formed.
As previously discussed, the HDDM indicated that participants with PWS needed more time to accumulate information to make a decision. Higher information accumulation demands may explain their relatively good results on categorization, where no time limits were imposed. This observation evokes the theory that ASD is the phenotypic expression of spatiotemporal processing disorders, which may result from multisystem brain disconnectivity-dissynchrony, defined as an increase or decrease in functional connectivity and neuronal synchronization within/between multiple neurofunctional territories and pathways . Consequently, the world changes too fast for these participants, but given enough time, their brain can find compensatory pathways and circuits.
Differences between UPD and DEL participants
Hit rates indicated that UPD participants had poorer voice and nonvoice perception than DEL participants (Fig. 1a). Similarly, UPD participants had longer RTs for both voices and nonvoices (Fig. 1b). This is in line with the finding of Salles et al.  that UPD participants have a greater deficit for the discrimination of voices and environmental sounds than DEL participants. However, the HDDM did not reveal any differences between the UPD and DEL participants on the decision making parameters. For sound categorization in the FST (Fig. 4a), three of the five PWS participants with the poorest performances were DEL participants, and the remaining two were UPD participants, so no conclusion can be reached as to possible differences between these subgroups. This may mean that UPD participants had more problems with the explicit task and fewer problems with the more implicit FST. Considering the absence of differences on the FST and the HDDM for decision-making, our overall results suggest that the differences in voice and nonvoice perception between the UPD and DEL participants concerned integrative sensory processing rather than the higher cognitive functions related to decision-making and ID.