Dissociating accuracy and metacognition in nutritional judgements

The gap between nutritional knowledge and dietary behaviour remains poorly understood, yet may partly reflect systematic biases in how individuals evaluate food and assess their own nutritional judgements. Here, we introduce the MetaBites task, a novel two-alternative forced-choice paradigm assessing nutritional judgements under adaptive difficulty control. Thirty-two healthy volunteers compared foods on either caloric density or nutritional quality, the latter quantified using the Nutrient-Rich Foods (NRF) 9.3 index. Participants provided trial-level, block-level, and task-level confidence ratings, enabling assessment of metacognitive monitoring across multiple levels. The task showed adequate validity, supported by staircase convergence, a clear speed-accuracy trade-off, and higher confidence on correct relative to incorrect trials. Results revealed a dissociation across conditions: participants showed higher objective performance when judging nutritional quality, yet reported significantly higher confidence for calorie judgements, and task-level global confidence decreased significantly in the NRF condition but remained stable for calorie judgements. Metacognitive efficiency did not differ between conditions. Familiarity with food items was associated with choice behaviour in opposing directions across conditions, consistent with a tendency to perceive familiar foods as more nutritious and less calorie-dense. Liking rather than familiarity was associated with higher trial-level confidence. Together, these findings show that objective performance and subjective confidence can dissociate in nutritional judgements, with distinct heuristic cues shaping choice and confidence. The MetaBites task offers a promising tool for investigating metacognitive contributions to nutritional decision-making in both healthy and clinical populations.