Refining the Pittsburgh Inference Test (PIT): A New Measure of Discourse Processing in Individuals with High-Functioning Autism

Background: Discourse processing is integral to appropriate communication and social interactions, which are known to be impaired in individuals with an autism spectrum disorder (ASD; APA, 2000).  Previous tests used to assess discourse processing in ASD are either too simple to assess older, high functioning individuals or are not appropriate for North American populations.  The Pittsburgh Inference Test (PIT) was developed as a measure of discourse processing that is sensitive enough to capture the challenges of verbal, older adolescents and adults with high functioning autism (HFA), especially in ascertaining emotion states.  However, further evaluation of the PIT is necessary to confirm and extend original findings.

Objectives: The objective of the current study was to further refine the PIT using a larger and more diverse sample.  

Methods: The PIT consists of 30 short stories about typical life situations that require inferences about events. Stories are followed by a verbal question that elicits a verbal response. The PIT is comprised of two types of questions that generate responses that: describe physical relationships only or describe physical, mental or emotional inferences (ToM). Responses for each story are scored as correct or incorrect and then categorized as a physical or ToM response.  ToM responses are further categorized by: emotion-ToM or other-ToM responses.  The PIT and Test of Language Competence–Expanded (TLC-E) were administered to a sample of 86 individuals with HFA (mean=20.6 years) and 65 typically developing (TD) individuals (mean=22.6 years) between the ages of 10-45 years, group-matched for age and IQ [FSIQ, VIQ, & PIQ, t(149) < 1.82, p > .07]. Autism diagnosis was determined with the ADOS, ADI, and clinical impression. All participants attained FSIQ’s > 85.

Results: After accounting for age and FSIQ, hierarchical regressions confirmed our original findings that individuals with HFA performed worse on the PIT in comparison to TD participants overall [t(147) = 4.5, pr² = .12, p < .01], particularly in ascertaining physical and emotional states [t(147) > 3.0, pr² > .06, p < .01]. Other-ToM inference making abilities remained intact [t(147) = .50, pr² = .002, p = .62]. Post hoc analyses revealed that only individuals with HFA tended to improve with age in the number of correct [t(144) = 1.9, pr² = .03, p =.05] and other-ToM responses [t(144) = 2.1, pr² = .03, p = .04].  However, no age-related improvements were noted in physical or emotion-ToM responses [t(144) < .85, pr² < .002, p> .39].  Confirming our previous findings, significantly moderate to strong correlations were evident between PIT (excluding physical responses) and TLC-E (subtest 1, 2, and 3) performance (r = .350 - .722, p < .001), indicating they assess similar underlying constructs of inference abilities. Significant negative correlations were evident for individuals with HFA between ADOS Total score and performance on the PIT, excluding physical responses (r range = -.562 to -.424, p < .01).

Conclusions: The current study confirmed and extended previous findings, which further highlight the usefulness of the PIT as a measure of discourse processing in individuals with HFA.