Do we read our own minds with the same system we use for others?
When the ability to read other people's minds is weakened, either in autism or by a competing mindreading task, people also become worse at judging their own confidence, suggesting one shared system for knowing self and others.
Source
Linking metacognition and mindreading: Evidence from autism and dual-task investigations
Study at a glance
- Design
- Human experiment — Experiment 1 compared autistic and IQ-matched neurotypical youths on implicit (gambling) and explicit (confidence) metacognition tasks; Experiments 2 and 3 randomly assigned neurotypical adults to single-task or dual-task conditions with a mindreading or non-mindreading secondary task.
- N
- No single N. Experiment 1: 25 autistic and 25 neurotypical children/adolescents. Experiment 2: 160 university students, 40 per task-by-condition cell. Experiment 3: two new groups of 31 adults doing the confidence task with a gender-judgment secondary task.
- Population
- Autistic and matched neurotypical children and adolescents (Experiment 1); neurotypical university students and community adults in the UK (Experiments 2-3)
- Outcome
- Metacognitive efficiency (M-ratio, meta-d' divided by d') on an explicit judgment-of-confidence task and an implicit gambling task; mindreading task scores
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
In Experiment 1, 25 autistic and 25 matched neurotypical children and adolescents made perceptual judgments (which patch is lighter or denser) and then either said whether they were confident (explicit metacognition) or chose a high- or low-risk bet (an implicit gambling task like those used with monkeys). In Experiment 2, 160 students did one of these tasks either alone or while simultaneously judging speakers' mental states from their voices. In Experiment 3, two new groups of 31 adults did the confidence task while judging only the speaker's gender, using either normal sentences or neutral sentences played in reverse word order.
What they found
Autistic participants had lower metacognitive efficiency than neurotypical participants on the explicit confidence task but not on the gambling task. In neurotypical adults, doing a mindreading task at the same time lowered confidence-judgment accuracy but left gambling accuracy unaffected, with a Bayes factor supporting no effect on gambling. Gender-judgment secondary tasks did not reduce confidence accuracy, so the drop seemed specific to mindreading rather than to any extra load. Implicit and explicit task performance were uncorrelated, and explicit metacognition correlated modestly with one mindreading measure (the Animations task) but not with the Reading the Mind in the Eyes test.
The limits
What it doesn't show
The group-by-task interaction in Experiment 1 was only just significant, and the samples were too small to reliably test correlations between metacognition and mindreading, as the authors acknowledge. The mindreading secondary task in Experiment 2 may simply have been harder or more demanding of working memory than the gender tasks in Experiment 3, so a general-load explanation is not fully ruled out. The results cannot exclude a version of the metacognition-is-prior view in which a metacognitive deficit causes the mindreading problems in autism. Finally, showing that humans can gamble accurately without metacognition does not prove that monkeys do the same, only that such tasks do not require it.
Key terms
- Metacognition
- Thinking about your own mental states, such as knowing how confident you should be that an answer is right.
- Mindreading (theory of mind)
- Attributing mental states such as beliefs, intentions and emotions to other people.
- One-system theory
- The view that a single metarepresentational system handles both our own and other people's mental states, so weakening one should weaken the other.
- Metacognitive efficiency (M-ratio)
- How well confidence tracks accuracy (meta-d') relative to how well the person performed the underlying task (d'), giving a measure not inflated by simply being better at the task.
- Dual-task paradigm
- A method where people do two tasks at once; if performance on one drops, the tasks are assumed to compete for a shared mental resource.
- Bayes factor (BF10)
- A ratio of how much the data favour the alternative hypothesis over the null; values below 1 favour no effect and values above 3 give firm evidence for an effect.
Flashcards
0 of 10 answers reviewed
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
Which theory do the authors conclude their results best support?
Common questions
Why include a gambling task at all?
Gambling tasks are used to argue that monkeys have metacognition. If autistic people and dual-tasking adults perform them normally while their explicit metacognition drops, the tasks can be solved without metacognition, perhaps by the good or bad feeling of an easy or hard judgment.
How does the dual-task design support the one-system view?
If self-knowledge and other-knowledge use separate systems, keeping the mindreading system busy should not hurt judgments about your own confidence. It did, and non-mindreading secondary tasks did not, which fits a shared resource.
Could autistic participants just have misunderstood the confidence task?
Both groups remembered the points rules similarly well, and their perceptual (object-level) performance was measured separately so the M-ratio controls for it. That makes simple misunderstanding or poorer perception unlikely explanations.
More on Metacognition