#whatWM? A Response to Lewandowsky, Oberauer, Morey, and Schweppe
Friday, December 2, 2016
Posted by: Stephan Lewandowsky
Nelson Cowan. Throughout most of my career, I think I was aware that there was some issue with the definition of working memory (WM), in that many psychological researchers often seemed to confuse the term WM with the multicomponent model of it that was proposed by Baddeley and Hitch in 1974. I started to go on a higher state of alert following a conference on WM, when many of the resulting chapters published in the volume edited by Miyake and Shah (1999) included definitions of WM that differed substantially from each other, largely because many of the supposed definitions were infused with descriptions based on the authors’ theories of WM.
Having written several review articles since about 2009 that tackled the issue of the definition of WM, finally the breadth of the problem seemed clear enough to me that I could report on the startling number of different definitions in the literature. I counted 9 of them, with a wide range of implied properties, and the resulting paper in the Psychonomic Bulletin & Review stimulated this digital event. WM seems analogous to the Egyptian Ennead, or 9 gods starting with the creator god Atum, who spawned 8 other gods. The purpose of the article was to improve the clarity of discourse by increasing the realization that when one refers to WM, others may not know just what is meant.
In brief, the 9 definitions I described, in order of their first appearance in the field, include computer WM (which is made temporary for convenience but has no strict capacity limit), life-planning WM (about using any mechanism available to bring to mind goals and sub-goals as needed), multicomponent WM (the traditional structure with at least separate phonological and visuospatial storage mechanisms and processes to coordinate them), recent-event WM (which tracks recent actions of the organism and their consequences), storage-and-processing WM (in which processes that act on items in temporary storage are considered part of WM), generic WM (which focuses on the limited quantity of information that is temporarily more available than other information), long-term WM (in which the long-term memory system is exploited to retain the structure of certain task-related events in a manner that allows their easy retrieval), attention-control WM (in which a very few active goals are kept in mind), and inclusive WM (in which WM is characterized as an ensemble of processes, short- and long-term, needed to carry out complex span tasks).
Following Stephan Lewandowsky’s very useful, general on-line summary of my paper, Klaus Oberauer, Candice Morey, and Judith Schweppe have now presented thoughtful and interesting commentaries on successive days, to which I respond here.
Oberauer suggests that although researchers do not agree on the definition of WM, they do generally agree on what procedures can measure WM. Although this is true, the procedures involved measure different aspects of WM and are not always in close accord. There may not even be a central, core common to all of the measures.
The idea that one knows it when one sees it, which appears to be at the core of Oberauer’s argument, is found in other areas as well, of course. Think of William James’ statement that everyone knows what attention is. In psycholinguistics, as a different kind of example, it is difficult to come up with a good definition of a spoken word. Even though one somehow feels confident in knowing what is or is not a word, words have variable numbers of morphemes (units of meaning) and are not consistently surrounded by silent gaps, so I know of no easy marker of words. In all of these areas, knowing it when we see it may be based primarily on learning specific instances rather than on principle. (Is memory of what you are going to get at the store in WM? Is something that you are trying to ignore part of what is attended? Is money-maker one word or two?)
Whereas I tended to think of the ideal definition as one that is neutral to contrasting theories, Oberauer pointed out that one mark of progress is when the definition can include more detail. The example he gave is that water can now be defined as composed of H2O molecules, in addition to the folk definition in which it is defined as an abundant, odorless, clear liquid. I agree that if, in the future, it turns out that there is a discrete brain system that is fully in charge of the temporary functions we imply with the term WM, then the definition could include some reference to that system and its properties. If, however, the term WM refers to a complex function that is handled by an ensemble of brain systems working in different combinations at different times, then it may always remain advisable to guard against the intrusion of more specific psychological or brain mechanisms into the term WM. As an analogy, we might never want the word “planning” to be defined with respect to the frontal lobe, even if this brain area is confirmed to be consistently involved in human planning.
Morey notes that “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.” That was what I was attempting to do with my favorite definition of WM, what I called the generic definition. I think that the function of WM is in principle distinct from long-term memory because of its temporary nature. The generic definition was that WM is “the ensemble of components of the mind that hold information temporarily in a heightened state of availability for use in ongoing information processing.” Oberauer thought that this definition implied a limited capacity or time limit. I did not mean for the definition to include those properties necessarily; the temporary nature of WM could instead come from interference by the stimuli and/or mental processes that inevitably follow an item’s presence in WM.
Oberauer thought of another possible basis of the temporary nature of WM that did not exactly occur to me (though it seems similar to attention-control WM). He suggested that it is “the mechanisms and processes that hold the mental representations currently most needed for an ongoing cognitive task available for processing.” The implication here is that the limitation in WM occurs on the basis of a part of human information processing that determines what is most important for the ongoing task. This definition seems to match what a reviewer said regarding the Miyake and Shah volume, that what is being studied is actually something like “working attention” rather than WM (for use of the term working attention, see also Beaman, 2010).
If we assume Oberauer’s definition, though, I wonder what to call information that is temporarily available but not relevant to the current task. For example, suppose you have just been introduced to someone named Jan, four seconds ago, when a colleague enters the room. Now you suddenly have the task of introducing your colleague to Jan and you still recall her name, using just-acquired information that was not part of an ongoing task. Was Jan not in WM until the task of interpersonal introduction was formulated? As another example, suppose you have had two unrelated thoughts in close succession, when you suddenly realize that they fit together in an important way (e.g., my visit with my mother tomorrow; my shopping trip today…I just realized that I should ask my mother if she needs anything at the store). I see nothing wrong with defining WM as goal-dependent, provided that one can find another satisfactory term that includes the useful recent information that is not task-relevant (e.g., short-term memory?). I believe, however, that not everyone in the field would be on board with that definition. Certainly, for example, within the multicomponent approach, not all information currently in the phonological loop is needed for the ongoing task, yet researchers in that tradition might well nevertheless want this information to be considered part of WM. Given problems like this, I did not propose that everyone use a common definition, but proposed instead that each author should clarify his or her definition. Having a menu of at least 9 definitions to choose from should help.
Morey notes that “In contrast to some realms of psychology, the evidence arising from paradigms designed to measure working memory is remarkably consistent,” including limitations in how much new information can be remembered together and what kinds of conflicts tend to occur. To the extent that there is agreement about these phenomena, it might be suggested that we could discuss the phenomena directly, without even using the term WM. What may make such an approach difficult, though, is that as Schweppe pointed out, “the issue of different definitions of working memory with different theories in tow is not restricted to the field of working memory itself but goes a lot further and is even more problematic when research in other areas builds on these definitions.” Many (but not all) measures of what is considered WM appear to be excellent at accounting for individual differences in comprehension, reasoning, problem-solving, decision-making, concept-formation, and intelligence. One needs to settle on a definition of WM before one can make a theoretical statement about whether it is related to these various cognitive activities. In fact, I imagine that this consideration could have influenced the task-goal-based definition that Oberauer favored.
Schweppe points out that it is not always easy to know how to apply the definitions of WM that I summarized in my paper. She brings up the case of Jarvella (1979), who showed that memory for clauses and sentences heard in discourse remains relatively intact in verbatim form if the subsequent information is closely related to it, but that this memory for clauses and sentences tends to be retained only in gist form if the subsequent information is new and relatively unrelated. What WM definition, she asks, would apply? The type of information that interferes with previous contents of WM is, I reckon, an empirical phenomenon that does not influence the definition of WM. If one were not committed to a particular definition, I would want to use the most general one that is applicable and, therefore, probably something like the generic definition. Yet, in a literature review for such a study, it could be worth pointing out when there is relevant evidence from researchers who appear to have defined WM differently.
Schweppe concludes that “When I stumble upon distortions of the working memory construct in other research areas, I tend to blame these entirely on those who are borrowing, but upon reading Nelson’s paper, I couldn’t help but notice that I and others like me are just as much to blame.” Thinking further about this perspective, I wonder whether we need to have different goals for discourse (1) within the WM research community, versus (2) between the WM research community and other types of researchers and practitioners. Within the WM research community, specifying different definitions of WM for different purposes should be an advance over what we do now. In communication with the larger community of researchers and practitioners over various kinds, it possibly could be useful if we could come to agree on some standard definition that is simple, yet carefully-specified. I don’t know how close to that we could come.
Baddeley, A. D., & Hitch, G. (1974). Working memory. In G. H. Bower (Ed.), The psychology of learning and motivation, Vol. 8. New York: Academic Press. (pp. 47-89)
Beaman, C.P. (2010). Working memory and working attention: What could possibly evolve? Current Anthropology, 51, No. S1, Working memory: Beyond language and symbolism, S27-S38. University of Chicago Press on behalf of Wenner-Gren Foundation for Anthropological Research.
Cowan, N. (2016). The Many Faces of Working Memory and Short-term Storage. Psychonomic Bulletin & Review. DOI: 10.3758/s13423-016-1191-6
Jarvella, R. J. (1979). Immediate memory and discourse processing. Psychology of Learning and Motivation, 13, 379-421.
Miyake, A., & Shah, P. (eds.) (1999). Models of working Memory: Mechanisms of active maintenance and executive control. Cambridge, U.K.: Cambridge University Press.