Search

Machine learning (ML) approaches show increasing promise in their ability to identify vocal markers of autism. Nonetheless, it is unclear to what extent such markers generalize to new speech samples collected, for example, using a different speech task or in a different language. In this paper, we systematically assess the generalizability of ML findings across a variety of contexts. We train promising published ML models of vocal markers of autism on novel cross-linguistic datasets following a rigorous pipeline to minimize overfitting, including cross-validated training and ensemble models. We test the generalizability of the models by testing them on (i) different participants from the same study, performing the same task; (ii) the same participants, performing a different (but similar) task; (iii) a different study with participants speaking a different language, performing the same type of task. While model performance is similar to previously published findings when trained and tested on data from the same study (out-of-sample performance), there is considerable variance between studies. Crucially, the models do not generalize well to different, though similar, tasks and not at all to new languages. The ML pipeline is openly shared. Generalizability of ML models of vocal markers of autism is an issue. We outline three recommendations for strategies researchers could take to be more explicit about generalizability and improve it in future studies. LAY SUMMARY: Machine learning approaches promise to be able to identify autism from voice only. These models underestimate how diverse the contexts in which we speak are, how diverse the languages used are and how diverse autistic voices are. Machine learning approaches need to be more careful in defining their limits and generalizability., (© 2022 International Society for Autism Research and Wiley Periodicals LLC.)

Acoustic atypicalities in speech production are argued to be potential markers of clinical features in autism spectrum disorder (ASD). A recent meta-analysis highlighted shortcomings in the field, in particular small sample sizes and study heterogeneity. We showcase a cumulative (i.e., explicitly building on previous studies both conceptually and statistically) yet self-correcting (i.e., critically assessing the impact of cumulative statistical techniques) approach to prosody in ASD to overcome these issues. We relied on the recommendations contained in the meta-analysis to build and analyze a cross-linguistic corpus of multiple speech productions in 77 autistic and 72 neurotypical children and adolescents (>1000 recordings in Danish and US English). We used meta-analytically informed and skeptical priors, with informed priors leading to more generalizable inference. We replicated findings of a minimal cross-linguistically reliable distinctive acoustic profile for ASD (higher pitch and longer pauses) with moderate effect sizes. We identified novel reliable differences between the two groups for normalized amplitude quotient, maxima dispersion quotient, and creakiness. However, the differences were small, and there is likely no one acoustic profile characterizing all autistic individuals. We identified reliable relations of acoustic features with individual differences (age, gender), and clinical features (speech rate and ADOS sub-scores). Besides cumulatively building our understanding of acoustic atypicalities in ASD, the study shows how to use systematic reviews and meta-analyses to guide the design and analysis of follow-up studies. We indicate future directions: larger and more diverse cross-linguistic datasets, focus on heterogeneity, self-critical cumulative approaches, and open science. LAY SUMMARY: Autistic individuals are reported to speak in distinctive ways. Distinctive vocal production can affect social interactions and social development and could represent a noninvasive way to support the assessment of autism spectrum disorder (ASD). We systematically checked whether acoustic atypicalities highlighted in previous articles could be actually found across multiple recordings and two languages. We find a minimal acoustic profile of ASD: higher pitch, longer pauses, increased hoarseness and creakiness of the voice. However, there is much individual variability (by age, sex, language, and clinical characteristics). This suggests that the search for one common "autistic voice" might be naive and more fine-grained approaches are needed., (© 2021 International Society for Autism Research and Wiley Periodicals LLC.)

SEVERAL THEORIES HAVE ATTEMPTED TO CHARACTERISE AUTISM SPECTRUM DISORDERS (ASDS) AT THE COGNITIVE LEVEL, MOST NOTABLY: THEORY OF MIND (TOM), EXECUTIVE FUNCTION (EF), AND A LOCAL PROCESSING BIAS (LB). THE AIM OF THIS STUDY WAS TO INVESTIGATE HOW THESE COGNITIVE FUNCTIONS DEVELOP OVER TIME: The three cognitive domains (ToM, EF, and LB) were examined in a group of high-functioning children (age: 8-12, mean 10.85; IQ: 78-139, mean 105.48) with ASD and a matched group of children with neurotypical development (NTD) (IQ: 75-145, mean: 109.47), and several tasks were used within each domain to ensure the validity of the cognitive measures. Approximately 3 years later (mean age: 14.34), all children and their families were invited to participate in the follow-up (ASD, N = 21; NTD, N = 30). While the understanding of other's minds does improve from childhood to adolescence, ToM impairment persists in adolescents with ASD relative to their peers. Likewise, a development in EF was observed in the ASD group, while no significant improvement was seen in the NTD group, leading the ASD group to catch up in this domain. We did not detect any group differences at any time point regarding local bias processing (LB). Individual patterns of development were seen, but remarkably, ToM deficits were present in every child with ASD in whom we could detect any cognitive impairment at baseline, and a similar pattern was found at follow-up. These findings indicate that ToM is a persistent cognitive deficit in ASD. Autism Res 2018, 11: 1229-1238. © 2018 International Society for Autism Research, Wiley Periodicals, Inc. LAY SUMMARY: This was the first study to investigate the development of three well-known cognitive functions into adolescence: While the understanding of other's minds improves from childhood to adolescence, adolescents with ASD are still impaired relative to their peers. The EFs, however, seem to improve to a neurotypical level in ASD as children enter adolescence, while local processing bias seems to differentiate the groups only in early childhood., (© 2018 International Society for Autism Research, Wiley Periodicals, Inc.)

Children with autism spectrum disorders (ASD) often show changes in (automatic) auditory processing. Electrophysiology provides a method to study auditory processing, by investigating event-related potentials such as mismatch negativity (MMN) and P3a-amplitude. However, findings on MMN in autism are highly inconsistent, partly due to small sample sizes in the studies and differences in MMN paradigms. Therefore, in the current study, MMN and P3a amplitude were assessed in a relatively large sample of children with ASD, using a more extensive MMN paradigm and compared with that of typically developing children (TDC). Thirty-five children (aged 8-12 years) with ASD and 38 age and gender matched TDC were assessed with a MMN paradigm with three types of deviants, i.e., frequency, duration and a combination of these two. MMN elicited by duration and frequency-duration deviants was significantly reduced in the ASD group. P3a-amplitude elicited by duration deviants was significantly increased in the ASD group. Reduced MMN in children with ASD suggests that children with ASD may be less responsive to environmentally deviant stimuli at an early (sensory) level. P3a-amplitude was increased in ASD, implying a hyper-responsivity at the attentional level. In addition, as similar MMN deficits are found in schizophrenia, these MMN results may explain some of the frequently reported increased risk of children with ASD to develop schizophrenia later in life. Autism Res 2017, 10: 1857-1865. © 2017 International Society for Autism Research, Wiley Periodicals, Inc., Lay Summary: Automatic detection of deviant sounds in the environment, such as upcoming traffic, is often affected in children with autism spectrum disorders (ASD). Mismatch negativity (MMN) is a way to quantify automatic deviancy detection, using electroencephalography. In this study, auditory MMN was assessed in 35 children with ASD and 38 matched control children, revealing significantly reduced MMN in the ASD group. This may indicate that children with ASD are less able to automatically detect environmentally deviant stimuli., (© 2017 International Society for Autism Research, Wiley Periodicals, Inc.)

The autism spectrum is characterized by genetic and behavioral heterogeneity. However, it is still unknown whether there is a universal pattern of cognitive impairment in autism spectrum disorder (ASD) and whether multiple cognitive impairments are needed to explain the full range of behavioral symptoms. This study aimed to determine whether three widely acknowledged cognitive abnormalities (Theory of Mind (ToM) impairment, Executive Function (EF) impairment, and the presence of a Local Processing Bias (LB)) are universal and fractionable in autism, and whether the relationship between cognition and behavior is dependent on the method of behavioral assessment. Thirty-one high-functioning children with ASD and thirty-seven children with neurotypical development (NTD), comparable in age, gender and Intelligence Quotient (IQ), completed several tasks tapping into ToM, EF, and LB, and autistic symptomatology was assessed through parental and teacher questionnaires, parental interview and direct observation. We found that ToM and EF deficits differentiated the groups and some ToM and EF tasks were related to each other. ToM and EF were together able to correctly classify more than three-quarters of the children into cases and controls, despite relating to none of the specific behavioral measures. Only a small subgroup of individuals displayed a LB, which was unrelated to ToM and EF, and did not aid diagnostic classification, most likely contributing to non-diagnostic symptoms in a subgroup. Despite the characteristic heterogeneity of the autism spectrum, it remains a possibility therefore that a single cognitive cause may underlie the range of diagnostic symptoms in all individuals with autism. Autism Res 2016, 9: 1328-1339. © 2016 International Society for Autism Research, Wiley Periodicals, Inc., (© 2016 International Society for Autism Research, Wiley Periodicals, Inc.)

Autism spectrum disorders (ASD) and schizophrenia are separate disorders, but there is evidence of conversion or comorbid overlap. The objective of this paper was to explore whether deficits in sensory gating, as seen in some schizophrenia patients, can also be found in a group of ASD children compared to neurotypically developed children. An additional aim was to investigate the possibility of subdividing our ASD sample based on these gating deficits. In a case-control design, we assessed gating of the P50 and N100 amplitude in 31 ASD children and 39 healthy matched controls (8-12 years) and screened for differences between groups and within the ASD group. We did not find disturbances in auditory P50 and N100 filtering in the group of ASD children as a whole, nor did we find abnormal P50 and N100 amplitudes. However, the P50 amplitude to the conditioning stimulus was significantly reduced in the Asperger subgroup compared to healthy controls. In contrast to what is usually reported for patients with schizophrenia, we found no evidence for sensory gating deficits in our group of ASD children taken as a whole. However, reduced P50 amplitude to conditioning stimuli was found in the Asperger group, which is similar to what has been described in some studies in schizophrenia patients. There was a positive correlation between the P50 amplitude of the conditioning stimuli and anxiety score in the pervasive developmental disorder not otherwise specified group, which indicates a relation between anxiety and sensory registration in this group., (© 2015 International Society for Autism Research, Wiley Periodicals, Inc.)

The relation between autism spectrum disorders (ASD) and schizophrenia is a subject of intense debate and research due to evidence of common neurobiological pathways in the two disorders. The objective of this study was to explore whether deficits in prepulse inhibition (PPI) of the startle reflex, as usually seen in schizophrenic patients, can be replicated in a group of children with ASD in comparison with a group of matched neuro-typically developed (NTD) controls. An additional aim was to explore possible psychophysiological subgroups within our ASD sample. In a case-control design, 35 ASD patients and 40 matched NTD controls were tested in a psychophysiological test battery. The PPI of the acoustic startle reflex was analyzed in 18 ASD subjects and 34 NTD controls. Habituation and sensitization were analyzed in 23 ASD subjects and 39 NTD controls. In trials with less intense prestimuli (76 dB), patients with ASD did not display the drop in percentage PPI normally found in healthy controls. In addition, ASD patients showed significantly increased sensitization compared with NTD controls. Combined, our results may reflect the hypersensitivity to sensory information in children with ASD. The relation to PPI deficits observed in schizophrenia is not apparent. Future research should study the developmental course of PPI deficits in ASD patients in a longitudinal design., (© 2013 International Society for Autism Research, Wiley Periodicals, Inc.)