#whatWM? Definitions of working memory do not need provocative claims
Wednesday, November 30, 2016
Posted by: Stephan Lewandowsky
Candice Morey. I was sitting in the audience, listening to a symposium speaker field questions, when someone asked about working memory. “Working memory? Does anyone actually believe in that anymore?” was the speaker's reply, and from the inside of my narrowly-focused scientific bubble, it astonished me.
I had only recently taken up my first tenure-track position as at Rijksuniversiteit Groningen and was still adjusting to being “the working memory person” on staff after spending the previous years working in laboratories devoted to working memory and cognitive control. In those settings, I took for granted that everyone knew what I meant by “working memory”, that I knew what everyone else meant by “working memory”, and of course that it was a vital cognitive phenomenon worthy of investigation. Outside those labs, I found that I needed to justify and defend my choices more vigorously. I was motivated to think about what working memory was and how I should explain it more carefully than I ever had before. But really, should I also have to defend its very existence? Beyond justifying that my research was theoretically interesting, practically useful, and worthy of attention and funding, did I now need to consider whether the topic of my research was actually a legitimate cognitive classification? How ridiculous, I thought at the time.
For a while after this conference, I occasionally thought about this speaker's remark, usually to mentally scoff again at how ridiculous it was. But as I carried on with my work, I grudgingly began to wonder whether he didn't have a point.
Working memory, to most people, seemed to be synonymous with the dominant multi-component model of working memory. My work had consistently turned up evidence contrary to this model, which proposes that verbal and visual memories are maintained in separate, distinct short-term memory buffers. At the time, my lab was producing some of the strongest evidence against this idea to date, which we were verifying and preparing to publish. As lecturer of introductory psychology I had become familiar with the most popular basic textbooks, and knew that the most salient “fact” students of psychology learn about working memory is that it consists of multiple short-term memory stores.
Furthermore, the language I encountered in so many published papers about working memory irked me. Titles like “Exploring the articulatory loop” or “Exploring the central executive” imprudently placed the disputed assumptions of a particular model, rather than the undisputed general phenomenon, at the forefront of the narrative. Authors regularly discussed theoretical metaphors as though they were literal structures one might actually isolate in a brain someday. Eventually, the speaker's remark no longer surprised me. If, to him, “working memory” was the multi-component model of working memory, and evidence against this framework was mounting, then why should he believe in it? I was even ready to affirm that I didn't believe in that particular characterization anymore either.
It is problematic that the phenomenon working memory is so synonymous with a particular model of working memory, and we would be wise to reconsider how to describe the phenomenon itself so that we avoid conflating it with any particular model. Cowan's paper that stimulated this digital event, in which he thoroughly delineates a taxonomy of descriptions of working memory, identifies the relevant disagreements and inconsistencies in this scientific space.
Most urgently in my opinion, if we want to avoid being accused of making models of nothing, we should take care to describe the phenomenon of working memory with as little theory-specific baggage as possible.
Let's consider two of the definitions surveyed by Cowan, 3) Multicomponent WM and 5) Storage-and-processing WM. According to the Multicomponent definition, WM must be a “multicomponent system that holds information temporarily and mediates its use in ongoing mental activities”. Similarly, according to the Storage-and-processing definition, WM is only measured when both the storage and processing components are in use. These definitions contain disputable points about how temporary maintenance and manipulation are accomplished which, if falsified, leave the concept of working memory vulnerable to irrelevance. However, these disputable points are unnecessary for conveying what working memory does: we do not need to say that separate components hold verbal and visual memories, or perform storage and processing functions, to say that working memory includes storage and processing of information in any sensory code.
A definition of working memory as a phenomenon should highlight points about working memory that are generally agreed to be true and minimize controversial assumptions. When discussing what working memory is, we can and should talk about what it does without making provocative claims about how it accomplishes those functions.
Working memory is a vital cognitive construct, and its validity doesn't depend on upholding the assumptions of any particular model. There is clearly something special about the information that is currently active in mind. The contents of working memory cannot be classified as “short-term” memory because they can include retrieved knowledge, but they also cannot be restricted to long-term knowledge, and thus cannot be subsumed within “short-term” or “long-term” memory classifications. In contrast to some realms of psychology, the evidence arising from paradigms designed to measure working memory is remarkably consistent. It is very clear that people can only remember a small amount of new, arbitrary information at once. It is very clear that we remember even less when we are required to try to remember new information while doing some other task. It is very clear that this loss is greater when the information maintained shares characteristics with the information processed. Working memory researchers do not dispute whether we expect these results, we dispute how best to explain these results within a model. If we let disputed theoretical assumptions be the most salient features associated with the idea of working memory, then the construct itself appears to be as controversial as the theory.