A harsh review can sometimes feel like a wrecking ball to one’s work, and, as academic lore has it, it’s often ‘Reviewer 2’ who wields the ball. However, a critical review I received in the late 1990s became a turning point in my research, thanks to Larry Barsalou’s (1999) groundbreaking article on perceptual symbol systems.
Let me start at the beginning.
In the late 1990s, I had submitted a paper to Memory & Cognition. In it, I described tests of a model of text comprehension called the event-indexing model. The model had been introduced in earlier papers and was intended to capture the comprehender’s unfolding mental representation of the state of affairs denoted in a text.
The units of the model were events, and they were thought to be linked on several dimensions, including time, location, and protagonists. It is a model fully in the tradition of semantic and propositional networks, with labelled nodes and links between them.
Reviewer 2 was not impressed, saying, “Those nodes in that network of yours don’t refer to anything. The model has no meaning. It is not grounded in perception and action.”
Actually, the review was much more constructive in tone, but this is what I took from it. However, since the criticism in it could have been levelled at pretty much any paper on language comprehension, I felt especially unlucky at being singled out.
However, the reviewer was right. What the reviewer was in fact pointing to, is that the event-indexing model was suffering from the grounding problem.
The grounding problem was first described by Stevan Harnad in 1990. It inflicts all cognitive models relying on semantic and propositional networks. As Harnad notes, the labelled nodes in these models do not refer to anything outside the model. Referencing occurs in the researcher’s head when reading the labels but these connections are not part of the model itself.
I found myself in a difficult spot. Simply acknowledging the reviewer’s point wasn’t going to be sufficient. Addressing the grounding problem required a fundamental conceptual shift, so much was clear.
As luck would have it, an article was about to be published that turned out to offer a way out of these doldrums, as I was to learn a little later. I am referring, of course, to Larry Barsalou’s famous article on perceptual symbol systems.
In this article, Larry proposes a theory of knowledge that is grounded in perception. His central idea was that perceptual experiences are stored in sensorimotor areas of the brain and can be partially reactivated during later processing.
I already knew Larry’s ingenious work on ad hoc categories, but this new approach intrigued me. I added the article to our lab meeting’s reading list.
As we discussed the article, it became clear how Larry’s arguments could address the grounding problem that had been vexing me for weeks. It was rich with insightful analyses and useful references to research in cognitive linguistics and neuroscience.
Larry’s proposal suggested to us that language could be viewed as a set of cues that reactivate aspects of perceptual experiences in the comprehender’s mind. Grounding problem solved.
Solved, that is, if empirical evidence for grounded cognition could be found. And this was the tenor of the commentary on Larry’s article that my students and I wrote. We expressed support for the ideas in the article but noted that empirical tests of perceptual symbol systems were currently lacking.
As we put it at the time: “Because we find Barsalou’s view congenial and inspiring, we attempt to identify some existing and possible lines of research that might have diagnostic value with respect to the comparison between perceptual and amodal symbol systems and as such could bolster the case for perceptual symbol systems empirically.”
We subsequently set out to develop such lines of research. Not just my students and me, but for many other researchers, who were impacted by Larry’s article on perceptual symbol systems. No wonder it has become a citation classic!
In the conversations I have had with Larry over the years since the publication of his 1999 article— from the desert of Palm Springs, with its spray-painted front yards, to my living room in leafy Zeist in the center of the Netherlands—he has always shown himself to be a kind-hearted, erudite, and stimulating conversation partner. He seemed in possession of an endless supply of constructive ideas.
Over the years, through both our conversations and his published work, Larry Barsalou has remained a significant influence on my research. Today, as I integrate my work on text comprehension with grounded cognition, his ideas continue to influence my thought processes. What began as a deflating experience became a catalyst for innovation, and for that, I have Larry to thank.
1 Comment
Prior to becoming interested in grounding, I’d been closely following your work, Rolf, on situations and events. At the time, I was trying to understand the mechanisms that contextualize concepts dynamically, producing different forms of the same concept on different occasions. I was thinking that dynamic frames played a central role, where it was content in a frame that contextualized a concept and varied across occasions. After discovering your work, I became convinced that it was content for situations and events that populated these frames. Later, when I became interested in grounding, the importance of situations receded to the background for a while. Eventually, though, I wanted to understand how the multimodal simulations of a concepts vary across contexts, and again your work and work on situated action led me back to the conclusion that situational variability was responsible. Since then, I’ve always assumed that multimodal simulations are situated, grounded in all the modalities that process situations. I became further convinced that the brain evolved to be a situation-processing architecture.
When your first articles on grounding came out in Psychological Science, I was again relieved, analogous to what I mentioned in my response to Alex Martin’s post. As you said in your post, at the time, we needed solid evidence for multimodal processing in semantic processing. Whereas Alex provided compelling neuroscience evidence, you provided compelling behavioral evidence. Immediately after these initial articles of yours on grounding came out, I could feel things starting to change. People started taking the idea seriously that multimodal simulations could be involved in conceptual processing and language semantics. It was no longer a crackpot idea. Many other people in multiple disciplines followed suit, developing paradigms to potentially demonstrate further multimodal effects. To my knowledge, though, no other group came close to producing the large number of ingenious paradigms and results that you reported. What a great body of work. And it has led to so much later work across disciplines, such as the paradigms that Gary Lupyan and Markus Ostarek developed.
I remember well the visit to Zeist. It absolutely poured on returning by bike, such that when my trainers eventually dried out, they started squeaking and have never stopped. I still wear these trainers when I work out, and when they inevitably start squeaking, I’m reminded of that visit.