More than just a memory: The nature and validity of working memory in educational settings
Darren S. Levin, S. Kenneth Thurman, and Marissa H. Kiepert
Since its origins in the 1960s cognitive revolution, “short-term” or “primary” memory has developed into the more sophisticated concept of working memory, in which information is not only retained for a brief period of time, but is also manipulated and closely involved in higher order processing activities such as comprehension, problem solving, and reasoning. While the dominant model of working memory has been Baddeley and Hitch’s multiple-component model (1974; Baddeley, 2000, 2007), there are several other theoretical views including Cowan’s embedded-processes model (1988, 1999, 2005) and Ericsson and Kintsch’s longterm working memory theory (1995). Assessments of working memory in children and adults are typically grounded or validated in light of a particular theoretical viewpoint. Research with children has focused on the relationship of working memory to academic achievement.This chapter reviews how working memory has been defined, assessed, and measured in experimental and applied (i.e., educational) settings with children. In light of this review, it raises serious questions regarding the ecological validity of much current research into the measurement of working memory in the classroom.
Models of working memory
How working memory is defined, assessed, and measured in laboratory or applied settings is guided by and closely related to a particular theoretical framework of working memory; moreover, how researchers define, assess, and measure working memory has implications for its applicability in real-world situations. Thus, it is
necessary to briefly present several theoretical views of working memory prior to discussing how this construct is used in educational settings.
Baddeley and Hitch’s multicomponent model
The dominant theory of working memory is the multicomponent Baddeley and Hitch model which was promulgated in 1974, and later revised by Baddeley (2000, 2007). This model defines working memory as “a limited capacity temporary storage system that underpins complex human thought” (Baddeley, 2007, p. 7). In contrast to Atkinson and Shitfrin’s (1968) “modal model” which depicted short-term memory as a temporary but unitary store of information, Baddeley and Hitch (1974) proposed a multicomponent system to account for several of the inconsistencies found between the modal model and existing data. Baddeley (2007) suggests that the modal model could not account for (1) evidence that short-term and long-term memory could not be neatly distinguished based on codes (i.e., semantic, phonetic, and visual), (2) findings from studies indicating that short-term memory could be severely disrupted, yet the transfer of information to long-term memory remained largely unaffected, and (3) data showing that concurrent tasks could selectively interfere with the uptake of information into long-term memory.
Baddeley and Hitch’s (1974) original model of working memory consisted of three components, each of which is responsible for different tasks. The “central executive” is the attentional control, decision-making system and is supported by two temporary storage subsystems (or “slave systems”) that are domain specific. The first slave system, the “phonological loop”, has received the most empirical investigation and support. The fimction of the phonological loop is to store and manipulate speech-based information through a subvocal rehearsal process known as articulatory control. The second slave system, the visuospatial sketchpad, stores and manipulates visual and spatial information. In 2000, Baddeley extended this model to include a third temporary storage subsystem known as the episodic buffer. This fourth component was proposed to account for data indicating that visual and phonological information are combined in some way (e.g., Logie, Della Sala, Val Wynn, & Baddeley, 2000; Saito, Logie, Morita, & Law, 2008), and helps explain data that could not be supported solely by these two existing slave systems. Thus, the episodic buffer was added as a subsystem that formed an interface between the other components and long-term memory (Baddeley, 2007).
Early research on the model focused heavily on the phonological loop. Initial laboratory evidence supporting the existence of a phonological loop includes: 
- • the phonological similarity effect: the effect that occurs when similar sounding words impair immediate serial recall (Conrad & Hull, 1964), and
- • articulatory suppression, which occurs when the prevention of the subvocal rehearsal processes severely disrupts performance on a memory list and also abolishes the word length effect (Baddeley, Lewis, & Vallar, 1984).
The concept of a phonological loop has not gone unchallenged, however. To date, the theoretical underpinnings of the phonological loop continue to be researched and have produced interesting developments in our understanding of language acquisition and processing (see Baddeley, 2007 for review).
The visuospatial sketchpad, researched to a lesser degree than the phonological loop, is also empirically supported. Evidence for the visuospatial sketchpad comes from studies on visual imagery (e.g., Baddeley, Grant, Wight, & Thompson, 1975) and more recently from neuropsychology research using brain-imaging techniques (e.g.,Jonides et al., 1993;Smith & Jonides, 1997;Della Sala & Logie,2002). Recent reviews of the advancements of the visuospatial sketchpad are given by Logie (1995, 2003), and Fletcher and Henson (2001).
Despite being at the core of the model, attempts to analyze the central executive came later than developments on these two subcomponents. However, great strides have been made in understanding the functions of the central executive since Baddeley’s first attempt in 1986, which comprised only a single book chapter. Review of recent developments in research on the central executive can be found in Baddeley (2007).
Other models of working memory have emerged in response to criticisms of Baddeley and Hitch’s multicomponent theory. Two of the most influential are Cowan’s embedded-processes model and Ericsson and Kintsch’s long-term working memory theory.
-  the irrelevant speech effect: the effect occurring when subject performance onremembering printed verbal items is disrupted by irrelevant spoken materialpresented at the same time (Colle & Welsh, 1976), • the word length effect: a phenomenon in which memory span coincides withthe spoken length of the word such that memory span for long words is smallerthan for short (Baddeley,Thomson, & Buchanan, 1975),