Asking the Right Questions: Planning Differences During Verbal Problem-Solving In Children with Autism Spectrum Disorders

Background: Proficient verbal problem-solving requires a range of executive and linguistic skills; on the Twenty Questions Task (TQT), categories must be used to systematically eliminate a set of possibilities.  Children and adults with autism spectrum disorders (ASDs) show difficulty in using category-based questions on TQT despite having good category recognition skills. Specifically, ASD participants ask questions that eliminate fewer items than typically-developing (TD) controls, leading to less efficient overall search.

Effective TQT searches narrow down possibilities hierarchically (living; animal; etc). Groups with frontal-executive impairments tend to show difficulty using narrowing and hierarchical planning on TQT while ASD participants show executive impairments on non-verbal planning measures. These findings provide a potential explanation of the ASD TQT profile; inefficient performance as the result of specific difficulties in constructing effective question plans.

Objectives: First, we aimed to examine in detail the planning processes of  ASD participants prior to attempting TQT, including assessment of question discrimination, plan construction, “narrowing” (the extent to which general questions precede specific questions) and question content. Second, we assessed how this affected TQT performance, to evaluate whether the encouragement to plan improved problem-solving skills.

Methods: 22 children with an ASD and 21 age- and IQ-matched controls attempted a forced-choice question discrimination task, a 5-question plan construction task, and six TQT trials (Age_M=13:7; FSIQ_M=96.42). For question discrimination participants identified the most effective question from two options across 10 trials. In plan construction, participants selected five category questions from a 40-question array and arranged them into a TQT search plan. The TQT consisted of a 24-picture array of everyday objects. The experimenter selected a target and participants asked yes/no questions to establish its identity. Changes in TQT performance were assessed in a repeated measures design comparing “baseline” performance (2 trials) with scores on TQT trials following the different planning tasks (2 trials each).

Results: No group differences were observed in question discrimination. On plan construction, ASD participants selected more specific questions at the start of their plans than controls. However, ASD performance also differed markedly with VIQ. When groups were split into participants above/below the mean VIQ (95.91), problems with narrowing were only seen in the Lower-ASD group (n=12), who demonstrated a flat profile of questions and poor question efficiency. Higher-ASD participants (n=10) constructed narrowing plans, but their questions were consistently inefficient compared to controls. Participants in both ASD subgroups showed moderate improvements on TQT following planning activity. Further results on question content will be discussed.

Conclusions: Inefficiencies on TQT appear to relate to poor plan construction for individuals with ASD. However, the nature of planning problems depends on VIQ: lower-ASD participants failed to use narrowing and selected inefficient questions, indicating a failure to generate categorical search beyond restricted semantic groupings. Higher-ASD participants utilised narrowing but still selected inefficient questions, suggesting intact hierarchical planning abilities in the context of irregular or weakened semantic relations. This variation of planning difficulties with VIQ is likely to impact on everyday functioning and has implications for the education and training of problem-solving skills in ASD.