AI Insight
This study investigated how prior exposure and image quality affect the brain's ability to predict upcoming visual objects. Using EEG recordings, researchers found that while participants were faster and more accurate at detecting expected objects compared to unexpected ones, the neural decoding patterns showed reduced accuracy for both expected and unexpected stimuli relative to random stimuli. The results suggest that prior exposure and image quality do not substantially interact with predictive processes during object recognition, and the brain may not represent expected versus unexpected objects as differently as some theories predict.
Why it matters
Understanding how the brain uses predictions to process visual information has implications for developing artificial intelligence systems and could inform approaches to treating perceptual disorders. The findings challenge some assumptions in predictive processing theories and suggest that the relationship between expectation and neural representation is more complex than previously thought.
Understand the Science
⚠️ Preprint – Noch nicht peer-reviewed
Dieser Artikel wurde noch nicht von unabhängigen Experten begutachtet. Die Ergebnisse sind vorläufig und sollten mit Vorsicht interpretiert werden.
According to predictive processing theories, the brain achieves efficiency in perception by comparing current sensory inputs with stored representations of the external world. Previous research has shown that manipulations of expectation can alter neural processing of low-level visual features as well as more complex, naturalistic objects. It remains unclear, however, precisely how the effects of expectancy are altered through learning and changes in stimulus fidelity. Here, we characterised behaviour and neural activity patterns while systematically varying individuals’ prior exposure to object sequences and the quality of the stimuli therein. Participants viewed rapid image sequences while we recorded their brain activity using electroencephalography (EEG). The stimulus sequences were probabilistically structured such that the appearance of each object was either expected, unexpected, or random. Participants were faster and more accurate in detecting cued target stimuli when these were expected relative to unexpected or random. Multivariate analysis of EEG activity patterns time-locked to the appearance of object images revealed reliably reduced decoding accuracy for both expected and unexpected stimuli relative to random stimuli. These patterns were consistent across exposure and image quality conditions, though exploratory analyses suggested that subtle changes in neural prediction effects may be linked to exposure. The findings do not provide strong support that the brain differentially represents expected versus unexpected object information and suggest that exposure and image quality do not reliably interact with predictive processes in object recognition.
Source: Exploring the role of prior exposure and image quality in neural and behavioural prediction effects