Until a few weeks ago, I didn’t know what it meant to go berserk. Or, more accurately, I didn’t know what it really meant to go berserk.

I was used to examples like this (thanks, COCA!):

• “My God. It’s like the Richmond Botanical Gardens gone berserk. I’ve never seen plants this big.”
• “Annie hadn’t given him a goodbye today, but she would have, wouldn’t she, if that phone hadn’t gone berserk?”
• “If Fido goes berserk at the sight of the mail carrier, close the curtains to block the view.”

I understood the action to be something vaguely like ‘going wild’. Fine. This level of understanding has gotten me through every use of the phrase until now, and I have never had to ask for clarification.

It turns out the phrase refers to the trance that Berserkers, a type of Old Norse warriors, entered during battle. This sort of thing involved the guys donning bear pelts, drinking heavily and/or doing drugs, and then the following series unfolding:

“Shivering, chattering of the teeth, and chill in the body, and then the face swelled and changed its color. With this was connected a great hot-headedness, which at last gave over into a great rage, under which they howled as wild animals, bit the edge of their shields, and cut down everything they met without discriminating between friend or foe.”

See below for an illustration:

When reading this description, to what extent did you simulate past sensorimotor experiences to understand the scene? Did you draw on your experiences of feeling a chill, shivering, biting on metal (a fork, maybe?); your experiences of seeing someone’s face swelling and turning ruddy or pale? Did you draw on your experiences of fighting or witnessing fighting, and of the feelings you had for friends and enemies?

Now that you’re armed with this elaborated understanding of the action, which is rich with details about visual, acoustic, and tactile properties, to what extent will you draw on this knowledge the next time you encounter or use the phrase? Is your conceptualization of the action different from other people’s, and different from that of your past self?

More broadly, the #symbodiment special issue and this week’s posts have grappled with question of how we conceptualize the words on this screen, actions, events, and objects, and to what degree our conceptualizations draw on our sensorimotor experiences.

One theme that has emerged is that “it depends”. Our conceptual representations are flexible (see contributions by Binder; Dove; Kemmerer; Yee & Thompson-Schill; Zwaan) and depend, among other factors, on our past experience, on relevance to the task, and on the level of abstraction that’s available to us.

It depends on our past experience

The degree to which we can draw on relevant sensorimotor experiences depends, of course, on the kinds of experiences we’ve had. In a post earlier this week, Lupyan stressed the importance of examining how individual differences in knowledge can stem from differences in experience (e.g., an expert musician’s vs. a novice’s representation of an instrument).

Yee and Thompson-Schill reviewed evidence that experience can change our conceptual representations. To give one example, experience playing a sport can change the way we process language related to that sport. Hockey players show greater activation in premotor regions when reading sentences about hockey actions (“The hockey player finished the shot”) than do hockey novices. In contrast, there’s no difference between novices and experts when reading sentences about everyday actions (“The individual pushed the doorbell”), which both groups have experienced.

On a much shorter time-scale, even a few weeks’ experience can change our conceptual representations of objects.

The experiences we can draw on change over a lifetime, and individuals’ conceptualizations may differ as a result of current and past experiences. Going back to going berserk, if you live in an area where there are a lot of dogs or coyotes, you might have more experience than others hearing animals howling. When you read the description of warriors howling like wild animals, you might have drawn on this experience to understand the scene.

It depends on relevance to the task

Kaschak and Lupyan both stressed the importance of examining behavior in context. The context of conceptual processing influences how it occurs, and sensorimotor properties will not be relevant to all tasks.

As brought out in Mahon and Hickok’s introductory article, the idea is that:

“concepts (i.e., the “representations themselves”) are dynamic with dissociable components, such that one aspect of a “concept” may be used in one context or task, but another aspect of the concept may be used in another… concepts do not have cores that are retrieved each time a concept is tokened: Overlapping subsets of conceptual information, which collectively form the “full” concept, can be solicited in a flexible manner according to the task constraints.”

Dove’s, Yee and Thompson’s, and Kemmerer’s contributions explore the idea that conceptual representations are context-dependent. To bring out one example, if we attend to a certain modality (e.g., tactile sensation), we are better able to access conceptual information that involves that modality (does an ant tickle?). We are also better able to access concepts that involve the modality we’re currently using.

When reading words (i.e., visual modality), we are faster at identifying words that have a strong visual component (e.g., small) than words that do not (e.g., husky). This pattern reverses when naming words (i.e., auditory modality)—we are faster at naming words that have a strong auditory component (husky) than words than do not (small).

Even the most salient sensorimotor features of concepts are not activated automatically and can be influenced by context—including color features in the Stroop task, number magnitude in the SNARC paradigm, and valence in affective priming.

It depends on the level of abstraction available to you

Constantly simulating sensorimotor experiences would presumably require effort. If there’s one thing we know about people, it’s that we rarely want to expend mental energy to do things “the right way”. Instead, we use heuristics and try to reduce computational effort.

Abstracting away from sensorimotor experiences offers us a way to process information faster and with less effort. It’s resourceful. What would happen if every time you encountered the word sister, you were to simulate past experiences of interacting with your sister (if you have one) or of witnessing sisters interacting, or of the feelings you’d experienced when you heard the word before?

In nearly all cases, these details would not help you understand the use any better than if you had not accessed this information. For most purposes, you can use a more abstract meaning of sister (a female sibling) and move on quickly. If you represent a concept at multiple levels of abstraction (see here, here, here, and here), then you can draw on the level that allows you to act with the least effort—which is, often times, the most abstract level.

Consider metaphorical uses of words, which abound in language. Metaphorical uses of verbs (Clinton and Trump are running for office) and prepositions (in love, under stress) are especially common. What if every time you encountered one of these uses (in love, in a moment, in time, in a war, in exchange), you were to model all of the sensorimotor properties associated with the word? You would have to find the few relevant properties (in the case of prepositions, control) among the many that are irrelevant every time you encountered a use. In this case, drawing on sensorimotor properties instead of a more abstract representation could actually impair comprehension.

Even though you now know the origin of the phrase going berserk, and may have more visual, acoustic, and tactile properties associated with the action, your abstract representation (‘going wild’) hasn’t disappeared.

Now, let’s say you encounter a new use: “Strong electromagnetic fields, as are generated by loudspeakers and the electric doors of a commuter train, cause the watches to go berserk, reported owners, and heat and cold can be disastrous to accuracy.” Did you draw on all those properties? Probably not—operating at an abstract level is still easier when you want to understand the word in context.

Abstract representations are useful, but how do we get them? Barsalou in his contribution to the #symbodiment special issue laid out three general routes to abstraction: multimodal compression, distilled abstraction, and distributed linguistic representation. Multimodal compression refers to the integration of information from different modalities into a single representation. Distilled abstraction refers to abstracting only the relevant aspects of experience to form a sparser concept. Finally, distributed linguistic representation refers to the use of words as symbolic placeholders that may or may not be enriched with sensorimotor simulations.

All of these routes allow us to build up and use stand-alone abstract representations that don’t necessarily require sensorimotor simulation. These representations may only be heuristics, as Barsalou points out, but heuristics are good enough for most situations. For example, understanding what it means to go berserk.

A broader view

Taking a broader view of embodied cognition, Killeen characterized it as a “movement that recognizes that the whole body and parts of its environment often play crucial roles that interact with other cognitive processes to guide perception and action”. Indeed, the theme that emerges is that we can adapt the way we conceptualize the world based on our experiences, our current environment, the task at hand, and so on.

In the future, I hope we further our understanding of how exactly we adapt to our (internal and external) environments. And given the divisions of the past, I hope that work in this domain is nourished by a spirit of generosity and camaraderie. The approaches we take are often on “different, often complementary, levels of discourse. When they do overlap, important value is added”, as Killeen put it. There are many ways to understand mind and behavior, and there is room for multiple levels of analysis.