Too Little Strategy, Too Much Guessing: Problem-Solving in High-Functioning Adolescents with ASD

Background: Twenty Questions is a game that requires the player to select a target object from an array using information gathered by asking only twenty or fewer yes/no questions. It allows the exploration of problem-solving abilities in a highly-motivating context. Successful performance requires concept formation (grouping by abstract concepts, such as “living things”), verbal abilities (asking a question), strategy (asking questions in an efficient order), verbal working memory (remembering answers), and inhibition (not guessing until obtaining enough information). Previous research has shown that individuals with ASD generate less efficient questions (despite unimpaired recognition of efficient questions), make more premature guesses, and correctly identify fewer items than their typically-developing peers. Manipulations of the task have revealed that their poor performance is likely not due to attention or working memory problems.

Objectives: First, we sought to replicate the performance findings of previous research. Second, while previous research manipulated the task to explore factors theorized to underlie poor performance, we utilized parent-report and neuropsychological measures to explore such factors, namely concept formation ability, VIQ, inattention, impulsivity, and verbal working memory. Finally, we sought to investigate relationships between task performance and autism symptomatology as no relationships have previously been reported.

Methods: The sample consisted of 26 high-functioning (FIQ > 80) adolescents with ASD and 27 age (M 14.8 years)- and VIQ (M 102.7)-matched typically-developing controls. We used the Delis-Kaplan Executive Function System (D-KEFS) Twenty Questions task. Concept formation, VIQ, attention problems, impulsivity, and verbal working memory were measured using the D-KEFS Card Sorting task, WASI-II, Conners Parent Rating Scale, and WRAML-2 respectively. One-way ANOVAs explored group differences on task performance variables. Two-tailed Pearson correlations explored relationships between task performance, the factors listed, and autism symptomatology measured using the SRS-2.

Results: Consistent with previous findings, our ASD sample identified fewer items (F[1,51] = 6.34, p = .02, η² = .11) and were less efficient in their questioning (F[1,51] =4.51, p = .04, η² = .08) than their typically-developing peers. In contrast to previous findings, our sample was not less efficient in their questioning when questions that did not constitute a guess were excluded (F[1,51] =.24, p = .63). Successfully identifying items and question efficiency were associated with concept formation ability (r = .41, p = .04; r = .47, p=.02) and VIQ (r = .56, p = .003; r = .57, p = .002), while only identification success was associated with verbal working memory (r = .56, p=.003). There were no significant associations between task performance variables and inattention, impulsivity, or autism symptomatology.

Conclusions: We confirm past findings of impaired performance on Twenty Questions in ASD and significant associations between performance and concept formation ability, VIQ, and verbal working memory. Our new finding of impaired performance despite unimpaired generation of efficient questions (after removing guesses) reinforces a previous finding of impaired performance despite unimpaired recognition of efficient questions. Together, these findings point to a strategy deficit. The lack of association between performance and autism symptomatology weakens the theoretical significance of concept formation deficits in ASD.