The Pittsburgh Inference Test (PIT): a Pilot Study Evaluating a Measure of Discourse Processing in Individuals with High-Functioning Autism

Background: Making inferences requires the integration of linguistic and social information, a process that occurs automatically in individuals with typical development.  In autism, inference making has been found to be impaired even in relatively able individuals (Dennis, Lazenby, & Lockyer, JADD, 2001; Happé, JADD, 1994; Mason et al., Neuropsychologia, 2008). There are few measures that evaluate this aspect of discourse processing in individuals who speak Standard American English.

Objectives: The Pittsburgh Inference Test (PIT) is being developed to assess the ability to make inferences during discourse processing in individuals with ASD. This pilot study evaluated the sensitivity of the PIT at the identification of difficulty with inference making in HFA.

Methods: The PIT consists of 34 short stories of common situations that require inferences about physical events, emotional states, or mental states. A question that requires the participant to make an inference about the situation follows each story. The participant generates unique verbal responses that are recorded verbatim by the examiner. The participant’s inferences are classified as correct or incorrect, and further coded as physical, emotional, motivational, or nonsensical. The PIT and the Test of Language Competence – Expanded (TLC-E) were administered to 22 adults with HFA (16-40 years; mean = 25.6 years) and 19 adults with TD (16–42 years; mean = 28.1 years), group-matched for age [t(39)=-1.57, p=.12] and IQ [FSIQ t(39)=-1.54, p=.13; VIQ t(39)=-1.69, p=.10; PIQ t(39)=-.40, p=.69]. Autism diagnosis was determined with the ADOS, ADI, and clinical impression. All participants attained Full Scale IQ’s ³ 97. The TLC-E is a formal measure consisting of four subtests that samples metalinguistic abilities. If the PIT and the TLC-E measure related underlying cognitive and linguistic constructs, correlations between these measures would be expected. However, the PIT may be more sensitive to difficulties in discourse processing because of the nature of the tasks.

Results: A repeated measures ANOVA indicated that the HFA participants performed worse than TD participants on the PIT, with significantly fewer correct responses on the three types of items [F(1, 39)=4.70, p=.04]. Post hoc analysis indicated that the autism group had the most difficulty with items requiring inferences about emotional states [t(39)= -2.04, p=.05]. The correct responses for the HFA group for the emotional states items were moderately to strong correlated with performance on 3 of the 4 subtests of the TLC-E; correct responses for the HFA group on the motivational states were correlated with performance on the “Making Inferences” subtest of the TLC-E. HFA participants were more likely to provide irrational and nonsensical responses to items on the PIT in comparison to TD participants [t(39)=2.62, p=.01]. 

Conclusions: The PIT identified impaired performance in HFA in making inferences, particularly in ascertaining emotional states. Analysis of the data is being used to make further refinements in this tool that is intended to be used for evaluating discourse processing in high-functioning individuals with autism.