Interpreting differential working memory test performance

Ah ha!! A small, yet bright (I proclaim) light bulb fired in my cortex this evening while skimming an article on the “bimodal format effects in working memory” (Note to self – I need to poll blogsters to see if they read similar exciting literature on Friday evenings). Why? Because I detected a link between the theory and research articulated in the article and clinical reports from assessment practitioners who wonder why some individuals perform differently on different working memory tasks.

In particular, I’ve heard anecdotal reports of individuals who do not experience problems with simple digit reversal tasks (e.g., Wechsler Digit Span Backwards; WJ III Numbers Reversed), but who turn around and perform either noticeably higher (or lower) on what appears (at face value) to be a more difficult task such as listening to numbers and words mixed together, and then being required to “list the numbers in order first, followed by the words in order” (e.g., WAIS-III Letter-Number Sequencing; WJ III Auditory Working Memory “How can this be?” The answer may lay in different working memory tasks placing differential demands on selective attention due to different format presentation modes.

As reported in a summary of two different experiments, Paula Goolkasian and Paul Foos’s (2005, American Journal of Psychology, 118[1], 61-77) research suggests that feature similarity effects can influence the ease of splitting divided attention. Given that working memory is hypothesized to include two different short-term working “slave” scratchpads (viz., articulatory or phonological loop for words and sounds; visual-spatial sketchpad for visual and figural stimuli), if all the material presented is from the same general stimulus family (e.g., words and sounds – both language), the scratchpad devoted to this form of processing (i.e., the phonological loop) may become overloaded and the chances increase that similar features will be “overwritten.” In other words, the phonological scratchpad may become too full if the two sets of stimuli (in a classic auditory working memory task) are more similar in general stimulus features.

Conversely, the more dissimilar features are in two sets (e.g., words and pictures), the greater the probability that the full resources of both scratchpads will be utilized. According to what the authors, the feature model predicts that “participants differentially allocate their attention resources across stimulus features, with distinctive features receiving more attention” (p. 63). Furthermore, the “probability of a feature being overwritten increases with increased similarity to subsequent events. Generalizing from this theory, picture and spoken word combinations should produce the most benefit because they are the most dissimilar.”

In the case of working memory tasks like the Wechsler Letter-Number Sequencing and WJ III Auditory Working Memory tests, the fact that both sets of stimuli are generally from the same class of stimulus features (viz., both spoken words), some examinees may overload their phonological loop scratchpad, resulting in a significant decrement in performance. In contrast, the same examinee might perform substantially different on a working memory task where the two stimulus sets are much different in format (e.g., spoken words and pictures of objects), given that each scratchpad resource can be maximized.

Long story short. Assessment personal, when comparing performance across different working memory tasks, should evaluate the extent to which different test paradigms increase or decrease the feature similarity of the classes of stimuli to which an examine must divide their selective attention. Based on their collective program of research, the authors offer the educationally relevant hypothesis (a bit broad and sweeping in my opinion), which also highlights the relative degree of difficulty of different working memory tasks derived from their program of research, when they conclude that:

• "When information is presented in educational and other settings, our results suggest that dual-modality presentation is very beneficial, particularly when pictures and words are combined. When only one presentation mode is possible, printed words should be your last choice. Finally, as one presents information, the last thing anyone wants is irrelevant speech, which interferes with both printed words and pictures" (p. 75).