The inferior temporal cortex is taken to be the termination of what is termed the ventral visual stream. The ventral visual stream describes the concept that visual information is built up from representing simple visual features such as bars and edges in primary visual cortex to representing whole images or objects at the termination of a sequence of visual processing steps in inferior temporal cortex. What is labeled as inferior temporal cortex covers differently labeled brain regions, TEO, TE and often the more medial structure perirhinal cortex. Thus, inferior temporal cortex is an imprecise description of a set of architectonically defined brain regions. When inferior temporal cortex is damaged, old-world monkeys and humans are impaired in performing image recognition tasks normally. Our data show that after selective damage to each of these brain regions, the deficits in behavioral performance are easily distinguished from one another. After bilateral rhinal cortex removal, Rhesus monkeys no longer distinguish among the values of visual cues, something that normal monkeys learn to do quickly and accurately. However, the monkeys are unimpaired at distinguishing among members of visually distinguishable categories such as cats vs dogs. They also assign blended stimuli (so-called morphs) with the same degree of accuracy as control monkeys. Taken together these results seem to show that rhinal cortex has an important role in reward assessment, but not in visual recognition. When area TE is removed bilaterally, monkeys seem to show a mild to moderate deficit in categorizing cats vs dogs, and they are able to learn to categorize cars vs trucks seemingly as easily as control monkeys. As in earlier studies, we used stimuli made up from morphed (blended and warped) cats and dogs ranging between 0 and 100% dog, with a distribution biased around the category boundary (11 levels, 0, 25, 35, 40, 45, 50, 55, 60, 65, 75, 100% of dog). Recently we showed that removing TE caused moderate impairments in visual categorization using a cat vs. dog category discrimination in a visually-cued two-interval forced choice paradigm. The partial sparing of categorization after the TE removals surprised us. There are two new important findings. First, the monkeys with TE removals, those with TEO removals, and those with combined TEO-TE removals all improve progressively with practice with the stimulus set. However, in all three groups when a new set of stimuli is introduced the deficit returns to its most severe impairment (most severe is different for all three groups). Thus, the monkeys are learning to discriminate by learning features of the specific stimulus set. This does not shed light on what brain regions are allowing this to happen. Second, when the monkeys are switched to a task we call running recognition, TE and TEO lesions have different effects. In this task, never-before-seen stimuli are presented sequentially and continuously. For every stimulus the monkey has to report whether it has seen the stimulus before or not, that is, the monkey is trained to distinguish truly new stimuli, from those seen previously. In this experiment each visual stimulus is presented exactly twice. The second presentation is after some number of intervening stimuli between 1 and 128. Each of the intervening stimuli is either new or being seen for a second time. Normal monkeys are able to distinguish between new and familiar stimuli when up to 32 stimuli appear from the first (unique) presentation to the repeat, and they perform at above chance levels to at least an interval of 128 stimuli. We expected the monkeys with TE, TEO and combined TEO-TE removals to be impaired in this task. The monkeys with TE removals and monkeys with combined TEO-TE removals were equally severely impaired even at the short intervals with two intervening stimuli. By an interval of 8 intervening stimuli, both groups were performing at just above chance levels. To our great surprise the monkeys with TEO only removals were unimpaired on this task. Thus, area TE is critical for this type of familiarity judgement, whereas area TEO seems to play no role. These results raise a severe challenge for the idea that processing in the ventral stream is feed-forward in a strictly sequential manner. It is currently problematic to understand how visual information gets to area TE after TEO removals since it has been thought that visual information is carried in a strictly sequential feed-forward manner. Because the combined lesion leads to a more severe deficit in categorization, the results from the categorization experiments are compatible with the ideas that areas TEO and TE might be part of a monolithic processing area of the brain with equipotency of function. The results from the running recognition make it clear that there is an as yet undefined organization for image processing and recognition that must include other brain regions.
Showing the most recent 10 out of 14 publications
