The representation of object concepts across the brain
Loggia, S. R.; Feibes, H. E.; Duffield, S. J.; Kasi, K.; Conway, B. R.
Show abstract
The mechanisms supporting concrete object concepts remain elusive. Here, macaque monkeys spent years learning object-concept primitives defined by color and shape. During fMRI, the monkeys held in mind one concept feature cued by the other. Using a whole-brain searchlight, we observed cross-feature decoding not in retinotopic cortex, but bilaterally in anterior inferior temporal cortex (aIT), temporal pole (TP), and frontal cortex (FC). Only TP and aIT had response patterns to concept shapes matching color-space geometry, and only PF had responses that decode behavioral errors. Two additional imaging experiments showed that color-associated versus non-color-associated shapes, and color-shape combinations congruent versus incongruent with the concepts, were distinguished only by extrastriate visual cortex, suggesting extrastriate cortex underpins color perception and perceptual feature binding. Together, the results provide a blueprint for how the brain generates stable object concepts and provides hypotheses for how conceptual representations engage perceptual representations through selective feedback during vision.
Matching journals
The top 4 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Confirmation Bias through Selective Readout of Information Encoded in Human Parietal Cortex 98%
- Inferring Latent Behavioral Strategy from the Representational Geometry of Prefrontal Cortex Activity 97%
- Object representations in the human brain reflect the co-occurrence statistics of vision and language 97%
Similar papers in this journal
- Joint representation of working memory and uncertainty in human cortex 98%
- Entorhinal and ventromedial prefrontal cortices abstract and generalise the structure of reinforcement learning problems 97%
- Human hippocampus and dorsomedial prefrontal cortex infer and update latent causes during social interaction 97%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.