Computational Characterization of Language Use in Females and Males with Autism Spectrum Disorder

The female clinical presentation of Autism Spectrum Disorder (ASD) is not well understood, and is a contributing factor to underidentification and delays in the diagnosis of ASDs among females, particularly for those who are verbal and without cognitive impairments. Sex differences in clinical presentation may make it more difficult to recognize ASD symptoms in females, yet evidence for differences between females and males in the two core symptom domains—social communication/interaction and repetitive behaviors and restricted interests—has been inconsistent. Although some studies have found stronger social communication/interaction skills in affected females than in males, researchers and clinicians have cautioned that these strengths may be superficial and allow females to “mask” their ASD symptoms, further contributing to underidentification among females. Similarly, restricted and repetitive behaviors may be less common among females with ASD, or they may simply look different from clinical expectations based on male presentations, again allowing females to “fly under the radar.” Thus, there is a critical need to further understanding of sex differences in the fundamental patterns of behavioral and social functioning relevant to the clinical presentation of ASD. In the absence of such knowledge, females with ASD will continue to be underidentified; affected females will be identified at older ages and thus will fail to benefit from access to early intervention services.

The long-term goal of this project is to further understanding of the female clinical presentation of ASD, so that improved methods for identification, assessment, and treatment of females with communication difficulties such as ASD can be developed. Our objective is to identify features that characterize atypical language use—a core symptom of ASD—among females with ASD. In the parent R01, we have developed and validated new Natural Language Processing based methods that automatically measure features of atypical language use based on raw (i.e., not coded) transcripts of natural language samples. Our strong preliminary data indicate significant impairments in conversational reciprocity and repetitive speech behaviors among children with ASD. For this administrative supplement, we propose to increase sample size of females with ASD in order to increase statistical power to analyze sex differences among children with ASD in atypical language use. Our rationale for this project is that successful identification of observable and objective differences (and similarities) between females and males in atypical language use would allow clinicians to better recognize, assess, and potentially treat symptoms of ASD in females.

Avatar
Alison Presmanes Hill
Data Scientist & Professional Educator