Two primary processes that are critical for normal higher level visual performance are the ability to distinguish one from another while recognizing that that the items are similar, e.g. distinguishing 2 cats is different than distinguishing cats and dogs irrespective of which cat or dog and, second, memories for items we have seen recently. Visual perceptual categorization is the process through which we assign objects into categories based on some similarity in their appearance. Previously we have shown that rhesus monkeys quickly learn (in less than one testing session) a visual category task. Inferior temporal (IT) cortex is considered a late stage of visual processing in the ventral visual pathway. Monkeys with lesions of IT cortex are impaired in discriminating between pairs of patterns. Single neurons in IT cortex, i.e. area TE, show selectivity for visual stimuli, most notably faces or complex objects. Physiological recordings suggest that there are groups/patches of cells that selectively respond to like stimuli, e.g. face and non-face patches in area TE. These findings lead to the inference that area TE plays a critical role in generating perceptual categories for faces or other objects. Previously we reported that, despite the importance of TE in visual perception and learning, removal of this cortex did not disrupt the acquisition and retention of perceptual categories. This striking result led us to look elsewhere for candidate areas important for acquisition of perceptual categories, such as the rhinal cortex. Rhinal cortex (Rh) has extensive input from TE and other visual association areas. It is important for stimulus-stimulus and stimulus-reward associations, and is required for generalization across new views and discrimination between ambiguous stimuli. Therefore we tested 3 Rh-lesioned monkeys on a task of visual categorization. Rh lesion had no effect on learning and generalization of two visual categories (e.g., dogs vs cats;cars vs trucks). Therefore we modified the task by making the category task more challenging by using computer-generated images that are morphs between many dog-cat pairs. Three control monkeys (Ctl), 3 with bilateral TE ablations and 3 with bilateral Rh ablations, reported whether the morphed image was more dog-like or more cat-like. All groups performed well on the original category task, ie. cat vs. dog discrimination. In the modified version, the 3 controls performed at a high level, having difficulty only when the stimuli approached the category boundary. The TE-lesioned monkeys were impaired whenever the original images were morphed to comprise 10% or more of the opposing category. Monkeys with Rh lesions performed as well as controls. Thus, Rh cortex appears to play no role in the perceptual component of categorization. Area TE, however, seems to contribute to the fine discrimination required to distinguish between categorical images. To study how short-term memory is done, we have measured the performance of normal rhesus monkeys on two visual short-term memory tasks: Match Any and Match First (Wittig and Richmond 2011). In both tasks, monkeys viewed sequences of 2 to 8 images and reported whether the last (test) image had appeared previously in the current sequence. In Match Any, the task was to report whether the test image matched any of the preceding images in the current sequence. In Match First, the task was to report whether the test image matched the first image and not any other distracting image from the sequence. We reported that Match Any performance was best for sample images that were presented near the end of a sequence. We reported that monkeys relied heavily on a similar strategy when solving Match First, except in this case the monkeys reported whether an image had occurred near the beginning of the sequence. Monkeys augmented this proximity strategy by reporting whether the test image specifically matched the first image, but only when task conditions were carefully tuned (e.g., by lengthening the sample presentation time and increasing the probability of a distractor in the second position). Now we report data from three monkeys trained on the tasks after bilateral lesions of lateral prefrontal cortex (dorsal and ventral). Initially, the lesion group had a learning deficit when the test image extinguished 1 second before the monkey was cued to respond to the trial. When this delay was eliminated, monkeys from the lesion group learned Match Any at the same rate as those from the control group (6, 10, 15 versus 10, 16, 18 sessions), but were impaired in learning Match First, taking more than twice as long (34, 36, 41 versus 65, 76, 91). Once learned, overall performance on both tasks was indistinguishable between control and lesion groups for 8-image sequences. However, the lesion group was impaired for short sequences (2 &4 images) when intermixed with long sequences, and for short sequences when only a few images (2 or 5) were used repeatedly throughout a session.
Showing the most recent 10 out of 14 publications
