Single-unit data from the cortex of monkeys performing working-memory tasks support the main point of the target article. Those data, however, also indicate that the activation of long-term memory is essential to the processing of all cognitive functions. The activation of cortical long-term memory networks is a key neural mechanism in attention (working memory is a form thereof), perception, memory acquisition and retrieval, intelligence, and language.

Neural models have proposed how short-term memory (STM) storage in working memory and long-term memory (LTM) storage and recall are linked and interact, but are realized by different mechanisms that obey different laws. The authors' data can be understood in the light of these models, which suggest that the authors may have gone too far in obscuring the differences between these processes.

Modern investigators of cognition ask about the conditions under which faculties occur rather than about their existence. This tendency, combined with the axiom of parsimony, emphasizes a paradigm shift in the fundamental principles of economic thought in science, mimicking evolutionary conceptualizations. The Ruchkin model of memory-related brain activity replaces less economic models. From interdisciplinary approaches, proceduralist models for other memory-related processes analogously support this model.

Evidence for the contribution of the neocortex to memory is overwhelming. However, the theory proposed by Ruchkin et al. does not only ignore subcortical contributions, but also introduces an unnecessary and empirically unsupported division between the posterior cortex, assumed to represent information, and the prefrontal cortex, assumed to control activation. We argue instead that the representational power of the neocortex is not restricted to its posterior part.

In the verbal domain, there is only very weak evidence favoring the view that working memory is an active state of long-term memory. We strengthen existing evidence by reviewing two recent fMRI studies of verbal working memory, which clearly demonstrate activation in the superior temporal lobe, a region known to be involved in processing speech during comprehension tasks.

Attentional processes that operate on the contents of memory to produce the activation that is described as working memory by Ruchkin et al. and others, involve a network of brain regions that include both prefrontal and parietal sites. This network appears to mimic the one that is activated by attentional processes that operate on information entering via the senses.

Ruchkin et al.'s view of working memory as activated long-term memory is more compatible with language processing than models such as Baddeley's, but it raises questions about individual differences in working memory and the validity of domain-general capacity estimates. Does it make sense to refer to someone as having low working memory capacity if capacity depends on particular knowledge structures tapped by the task?

We focus on the functional specificity of theta and alpha oscillations and show that theta is related to working memory, whereas alpha is related to semantic long-term memory. Recent studies, however, indicate that alpha oscillations also play an important role during short-term memory retention and retrieval. This latter finding provides support for the basic hypothesis suggested by Ruchkin et al.

Studies of neuropsychological patients are relevant to models of how long-term memories are stored. If amnesia is considered a binding deficit and not a difficulty in transferring information from short-term to long-term memory, it is unclear why context-free semantic learning is impaired. Also the model should account for the reverse temporal gradient seen in patients with semantic dementia.

Ruchkin et al. make a strong claim about the neural substrates of active information. Some qualifications on that conclusion are: (1) Long-term memories and neural substrates activated for perception of information are not the same thing; (2) humans are capable of retaining novel information in working memory, which is not long-term memory; (3) the content of working memory, a dynamically bound representation, is a quantity above and beyond the long-term memories activated, or the activity in perceptual substrates.

Working-memory retention as activated long-term memory fails to capture orchestrated processing and storage, the hallmark of the concept of working memory. The event-related potential (ERP) data are compatible with working memory as a mental workspace that holds and manipulates information on line, which is distinct from long-term memory, and deals with the products of activated traces from stored knowledge.

We challenge Ruchkin et al.'s claim in reducing short-term memory (STM) to the active part of long-term memory (LTM), by showing that their data cannot rule out the possibility that activation of posterior brain regions could also reflect the contribution of a verbal STM buffer.

Ruchkin et al. offer a compelling case for a model of short-term storage without a separate buffer. Here, I discuss some cognitive neuropsychological data that have been offered in support of and against their model. Additionally, I discuss briefly some new directions in cognitive neuropsychological research that bear on the role of attention in Ruchkin et al.'s model.

Consistent with Ruchkin and colleagues' proceduralist account, recent research on grouping and verbal-spatial binding in immediate memory shows continuity across short- and long-term retention, and activation of classes of information extending beyond those typically allowed in modular models. However, Ruchkin et al.'s account lacks well-specified mechanisms for the retention of serial order, binding, and the control of activation through attention.

The view that posterior brain systems engaged in lower-order perceptual functions are activated during sustained retention is challenged by fMRI data, which show consistent retention-related activation of higher-order memory representations for a variety of working-memory materials. Sustained retention entails the dynamic link of these higher-order memories with schemata for goal-oriented action housed by the frontal lobes.

There is remarkable agreement between Ruchkin et al.'s psychophysiological views and my own model, based on developmental-experimental evidence, of working memory as activated long-term memory (LTM). I construe subvocal rehearsal as an operative scheme that maintains order information and demands attentional resources. Encoding and retrieving operations also demand attention. Another share of resources is used for keeping activated specific LTM representations.

The current model, based on event-related potential (ERP) studies, posits that the working-memory system is a state of activated long-term memory; this appears comprehensive, but it needs further detailed analysis of functional neural connectivity analysis within the prefrontal cortex (PFC) and between the posterior and prefrontal cortex. Specifically, the role of dorsolateral PFC and anterior cingulate cortex (ACC) is probably critical for PFC's attentional controller. Neural implementation of the executive function in working memory appears critical to build a firm model.

In this commentary, we focus on four points. First, we discuss the assertion that the unitary model explains dissociations that implicate multiple systems. Second, the distinct nature of information utilized in immediate- and delayed-recall supports the distinct memory systems view. Third, the variable nature of capacity limits corroborates this view. Finally, we review event-related fMRI results that suggest support for multiple systems.

Although visual long-term memory (VLTM) and visual short-term memory (VSTM) can be distinguished from each other (and from visual sensory storage [SS]), they are embodied within the same modality-specific brain regions, but in very different ways: VLTM as patterns of connectivity and VSTM as patterns of activity. Perception and VSTM do not “activate” VLTM. They use VLTM to create novel patterns of activity relevant to novel circumstances.

We argue that an approach that treats short-term memory as activated long-term memory is not inherently in conflict with information recycling in a limited-capacity or working-memory store, or with long-term storage based on the processing in such a store. Language differences aside, real model differences can only be assessed when the contrasting models are formulated precisely.