Distinguishing Between Implicit and Explicit Measures of Metacognition in ASD

Background:

Metacognitive monitoring (awareness of one’s own mental states/cognition) is a key component of self-awareness and plays an important role in learning. The few existing studies of metacognitive monitoring in ASD have employed tests that require participants to make explicit judgements about the state of their own knowledge. The closer the correspondence between judgements of one’s knowledge and actual (objectively-measured) knowledge, the more accurate metacognitive monitoring is. Findings using such explicit tasks have been mixed. In the current investigation, we aimed to resolve a discrepancy in findings across previous studies by employing not only a standard test of explicit metacognitive monitoring, but also a paradigm adapted from one used in comparative psychology to assess metacognitive monitoring non-verbally/implicitly.

Objectives:

The aim of the present study was to investigate implicit and explicit metacognition in ASD. 

Methods:

Data collection is ~80% complete. 22 participants with ASD and 18 age- and IQ-matched comparison participants completed a “gambling” paradigm based on that used among monkeys by Son & Kornell (2005). This involved a visual discrimination task (e.g., judging most pixelated of two squares). After a decision had been made on each trial, participants were required to choose from one of two shapes (circle/triangle). Correct visual discrimination resulted in a payment gain, while incorrect visual discrimination resulted in financial loss, with subsequent shape selection dictating whether gain/loss was of high (triangle) or low (circle) value. Accurate metacognitive monitoring was indicated by a greater tendency to choose the high-value “Triangle” on successful visual discrimination trials than on unsuccessful trials, and to choose the low-value “Circle” more on unsuccessful than successful trials. In a second session, participants completed the same gambling task (but different visual discrimination task). This time, after each visual discrimination trial, participants were explicitly asked “Are you confident?”, with “Yes” (high-value gain/loss) or “No” (low-value gain/loss) chosen in response. Selecting “Yes” on successful visual discrimination trials more than on unsuccessful trials and, vice versa, by choosing “No” more on unsuccessful than successful trials indicated accurate metacognitive monitoring.

Results:

In both tasks, Gamma correlations were utilised to measure metacognitive performance. Across both tasks, between-group differences were apparent. Gamma correlations were lower among participants with ASD than comparison participants in both the implicit (ASD M = .32, SD = .59; Comparison M = .58, SD = .26; t = 1.78, p = .08, d = .58) and explicit tasks (ASD M = .56, SD = .39; Comparison M = .71, SD = .21; t = 1.49, p = .15, d = .49). Correlations between metacognitive monitoring performance and background cognitive measures of theory of mind and of ASD feature severity/traits will be analysed and discussed.

Conclusions:

These results complement the existing literature and provide evidence of an explicit metacognitive deficit in ASD. However, they extend the field significantly by providing the first evidence of a deficit in a non-verbal/implicit form of metacognitive monitoring. This suggests a pervasive difficulty with this ability at all levels among people with ASD, which may contribute to learning difficulties in this disorder.