Asymmetric Compression of Representational Space for Object Animacy Categorization under Degraded Viewing Conditions

Abstract

Animacy is a robust organizing principle among object category representations in the human brain. Using multivariate pattern analysis methods, it has been shown that distance to the decision boundary of a classifier trained to discriminate neural activation patterns for animate and inanimate objects correlates with observer RTs for the same animacy categorization task [Ritchie, J. B., Tovar, D. A., & Carlson, T. A. Emerging object representations in the visual system predict reaction times for categorization. PLoS Computational Biology, 11, e1004316, 2015; Carlson, T. A., Ritchie, J. B., Kriegeskorte, N., Durvasula, S., & Ma, J. Reaction time for object categorization is predicted by representational distance. Journal of Cognitive Neuroscience, 26, 132-142, 2014]. Using MEG decoding, we tested if the same relationship holds when a stimulus manipulation (degradation) increases task difficulty, which we predicted would systematically decrease the distance of activation patterns from the decision boundary and increase RTs. In addition, we tested whether distance to the classifier boundary correlates with drift rates in the linear ballistic accumulator [Brown, S. D., & Heathcote, A. The simplest complete model of choice response time: Linear ballistic accumulation. Cognitive Psychology, 57, 153-178, 2008]. We found that distance to the classifier boundary correlated with RT, accuracy, and drift rates in an animacy categorization task. Split by animacy, the correlations between brain and behavior were sustained longer over the time course for animate than for inanimate stimuli. Interestingly, when examining the distance to the classifier boundary during the peak correlation between brain and behavior, we found that only degraded versions of animate, but not inanimate, objects had systematically shifted toward the classifier decision boundary as predicted. Our results support an asymmetry in the representation of animate and inanimate object categories in the human brain.

Finding decodable information that is read out in behaviour

“…In both experiments, we found correlations between distance and reaction time for animate 403 stimuli, but none for the inanimate stimuli. This is consistent with previous work (Carlson et al,404 2014; Grootswagers, Ritchie, et al, 2017;Ritchie et al, 2015), which argued that this discrepancy 405 might be caused by inanimate being a negatively defined category (i.e., "not animate"). Under this 406 hypothesis the animacy categorisation task can be performed by collecting evidence for animate 407 stimuli and responding inanimate only when not enough evidence was accumulated after a certain 408 amount of time.…”

“…This goes against the notion of the negative definition of inanimate as the main reason for a 411 lack of correlation. However, it still is possible that observers still treated these tasks as 'A' or 'NOT 412 A', with 'A' being the category that is easiest to detect (Grootswagers, Ritchie, et al, 2017). For 413 example, perceptual evidence for a face would be easier to obtain that evidence for a body-part, Gomez-Marin, MacIver, & Poeppel, 2017).…”

Using neural distance to predict reaction time for categorizing the animacy, shape, and abstract properties of objects

“…In particular, Carlson et al 15 found a robust RT-distance effect in human VTC, which was driven entirely by the animate exemplars, and taking a searchlight approach to NDBA Grootswagers et al 17 found that neural distance for animate exemplars correlated with RTs in LOC and VTC, and largely tracked the areas of peak decoding. The same asymmetry has also been observed when applying NDBA to human MEG time-series data 16,18 . The fact that a weak RT-distance effect was also observed for animate images in ECV is also consistent with previous findings of Carlson et al 15 , and suggests that the visual properties represented in the region influence how easily a stimulus can be categorized even in situations in which these low-level properties by themselves are not sufficient for categorization.…”

“…Second, previously we have found that observer choice accuracies also correlate with neural distance. When both accuracy and RT distributions are jointly modeled using evidence accumulation models, this provides a more complete description of the observer performance, and the drift rate parameters of these models can also be correlated with neural distance 18 . Efficiency scores provide a simpler method for combining both measures, and in the present study we used the recently proposed LISAS 23 .…”

“…Based on the predictions of classic psychophysics 12,13 , if the neural activation space reflects the representation that determines observe categorization response, then RTs should negatively correlate with distance from the decision boundary 14 . Recent studies have used NDBA to predict observer RTs for animacy categorization tasks based on neural distances measured with human fMRI and MEG 15–18 .…”