Our brain is highly effective at predicting future sensory input based on learned associations between stimuli. For instance, ample previous research has shown that if we can predict an upcoming perceptual stimulus because it always reliably follows another perceptual stimulus, we respond faster and more accurately to the expected stimulus, and our brain response to the expected stimulus is suppressed (i.e., expectation suppression). However, in our everyday life we often build associations between concepts rather than percepts. It is unclear whether conceptual priors also modulate sensory processing of an expected stimulus, or whether its processing is only affected at the post-perceptual level.

In the current study, the authors examined whether and how priors derived from learned conceptual associations modulate visual processing of the expected stimulus. To test this, subjects first learned arbitrary object associations in sequences of words (e.g., the word car always followed the word dog). The next day subjects performed a categorization task on the same object associations in the MRI scanner, but while the first object was still always presented as a word, the second (expected) object was now presented as an image instead.

First, while subjects responded with equal accuracy to expected vs. unexpected images, possibly due to a ceiling effect (i.e., >97% accuracy), they responded significantly faster to expected images, indicating a generalization of (conceptual) associations from word-word to word-image pairs. Second, overall BOLD responses were suppressed for expected vs. unexpected images along the ventral visual stream from early visual cortex (EVC) to the object-selective lateral occipital cortex (LOC) and the ventral temporal cortex (VTC). Importantly, this expectation suppression was specific to the voxels preferentially responding to the category of the expected image, but not for voxels responding to another category, indicating that sensory predictions were category-specific.

In sum, these results indicate that learned conceptual associations result in category-specific predictions that modulate perceptual processing and behaviour. Our brain is thus able to not only use prior knowledge derived from concrete perceptual associations to make predictions, but also prior knowledge derived from abstract conceptual associations.

Follow this link to read more about this interesting study: https://doi.org/10.1523/JNEUROSCI.1874-22.2023