In the mid-1990’s, I was extremely fortunate to have the opportunity to explore how conceptual/semantic information was represented in the human brain.  One of these studies, published in Science in 1995, focused on the neural systems underpinning retrieval of information about the object-associated color and action, the other, published in Nature in 1996, focused on the neural systems underpinning identification of the object categories of animals and tools. As we stated at that time, the results of these studies suggested that object knowledge was stored as a distributed network of cortical regions and that the organization of these regions closely parallel the organization of sensory and motor systems in the human brain. This idea, in turn, was consistent with neuropsychological data on patients presenting with relatively focal deficits concerning a specific object attribute (e.g., patients with color agnosia), and from patients with so-called category-specific deficits (e.g., for animals and other living things). It was also consistent with data from nonhuman primates on the location of brain systems for processing specific object attributes like color and motion. Thus, these neuropsychological data provided both the motivation for our functional brain imaging studies and predictions for their outcome.

In stark contrast, cognitive psychology, dominated by amodal theories, had nothing to offer. Indeed, where information was located in the human brain was of little interest to cognitive psychologists and their theories were silent about how and where these ‘amodal’ symbols might be stored. It was against this backdrop that I became aware of Larry’s seminal publication, Perceptual Symbol Systems (Barsalou, Behavioral and Brain Sciences, 1999). Not only did Larry provide a strong argument against concept representations as amodal symbols but offered a compelling case for an alternative view of concepts as perceptual symbols, underpinned by what Larry termed, “records of the neural states that underlie perception” (Barsalou, BBS, 1999). So here was a fully fleshed out theory that offered predictions that were perfectly aligned with the results of our early studies on concept representation discussed above.

I’m not sure when I first met Larry, but I am sure it came about via a serendipitous meeting with one of his graduate students, Kyle Simmons, at a meeting of the Cognitive Neuroscience Society in San Francisco, probably in the early 2000’s. By this time, I was a big fan of Larry’s work, and I knew that he was aware of our brain imaging studies because he referenced both of the above noted publications in Perceptual Symbol Systems (PSS). Thus, I was thrilled when Larry invited me to collaborate on some functional brain imaging studies that he and Kyle were planning to further test predictions derived from the PSS model.

The first of these (Simmons et al., Cerebral Cortex, 2005) evaluated visual recognition of appetizing foods. The critical finding was that viewing foods selectively activated a specific region of the insula cortex. Importantly, based on previously published work, we noted that this region of the insula was also active when subjects experienced tastes. So, as would be predicted from PSS, simply viewing pictures of foods activated a seemingly modality-specific brain region associated with a critical food-related property, taste. This finding also underscored the multimodal nature of brain regions typically assumed to be unimodal – another finding consistent with PSS. [As an aside, our claim in that report that viewing food images activated taste sensitive cortex was made by appealing to previously published studies on taste perception. When we did our study, we lacked the MRI-compatible equipment needed to deliver tastants to our subjects so that we could directly localize taste-sensitive insula cortex. As a result, we really didn’t know if the region we reported that was active when viewing food was the same, or simply nearby, the region active when experiencing tastes. It was not until several years later, when Kyle

Simmons was a post-doc in my lab, that we developed an MRI-compatible taste delivery apparatus that allowed us to verify that, in fact, viewing food images activated the same region of insula cortex active when subjects experienced tastes (Simmons et al., Nature Neuroscience, 2013. Also see Avery et al., PNAS, 2021, for evidence of differential patterns of activity in this region of the insula depending on the quality of the tastants (sweet, salty, sour) as well as the dominant taste quality (sweet, salty, sour) associated with the food images (e.g., candy, pretzels, lemons).]

Our next collaboration was on the representation of color concepts. A written, property verification task was used to probe knowledge of object color, and a color perception task was used to localize color-sensitive brain regions (Simmons et al., Neuropsychologia, 2007). The choice of the color localizer was particularly important. The task was developed by Michael Beauchamp, another post-doc in my lab at that time (Ken McRae was also a member of the all-star team that Larry assembled for this project). The task was modeled after the well-known clinical test of color perception, the Farnsworth-Munsell 100 Hue Task. Michael had previously shown that when subjects simply stared at the color stimuli, activity was limited to the very posterior regions of occipital cortex, consistent with many previous studies of color perception. However, when the task required selective attention to subtle differences in hue, multiple color sensitive regions were revealed, extending from occipital cortex downstream into posterior, ventral temporal cortex (Beauchamp et al., Cerebral Cortex, 1999). Using this more demanding task, we found that the verbal, color property verification task activated a region in ventral temporal cortex that precisely overlapped with the more anterior regions activated by the color perception task. This was an extremely important finding. Previous studies found that conceptual knowledge about color was localized anterior to, but not within, color perception regions as defined by passive viewing of colored stimuli (e.g., Martin et al., Science, 1995; Chao et al., Journal of Cognitive Neuroscience, 1999). The current study showed that, in fact, information about an object property was stored within the same system active during perception – a key prediction of PSS (see Fig).

These two studies are among my all-time favorites, and I believe they have had considerable impact on the field (the food study cited over 700 times, the color study over 500 times, according to Google Scholar). Although during the years following these reports, we have had fewer opportunities to interact, Larry had, and continues to have, considerable impact on my thinking about the nature of representations of concrete, as well as abstract, concepts (see Wilson-Mendenhall et al., Journal of Cognitive Neuroscience, 2013).

So, Larry, thanks for that.