There is a debate over the adequate conceptualization of verbal short-term or working memory, that is, the retention of a limited number of items and their position within a respective sequence. Some researchers view short-term memory as involving separate stores or buffers that hold information independently of long-term memory. Others conceive of working memory as activated long-term memory (henceforth aLTM), supplemented by a mechanism to specifically represent order information (e.g. Majerus, 2019) or maintain items in a highly accessible state (e.g. focus of attention, Cowan, 2019). Colleagues and I reported a double dissociation among two groups of individuals with neuropsychological impairments that we took as support for cognitive models with separate buffers. Majerus, Cowan and Oberauer put forward a critique of our conclusions to which I shall respond. We would first like to state that we really see the value of their position (e.g. Bormann et al., 2025). We also agree that dissociations in individuals with neuropsychological impairments require great methodological scrutiny. Majerus et al. put forward three arguments against our conclusions, and we shall respond to each of their points below. We believe that each of their arguments has weaknesses. In general, we believe that they focus too much on one aspect of the reported double dissociation and pay too little attention to the opposite pattern of impairments, namely HS's better short-term retention of verbal information despite her severe impairments at multiple levels of word processing. Majerus et al.'s first argument is that those accounts which conceive of working memory as aLTM do not predict that each or any language impairment should affect repetition. They note that language processing involved several different levels and representations and not all are involved in repetition. We would like to respond that proponents of working memory as aLTM often refer to correlations between impaired word processing and impaired verbal short-term memory and view these correlations as evidence for their theoretical position. For example, they refer to studies by Patterson et al. (1994), Majerus et al. (2007) as well as Martin et al. (1996) reporting correlations between single word or semantic processing and retention of verbal information. Majerus et al. do not propose criteria on which neuropsychological studies to consider and which studies to ignore. Crucially, our participant HS exhibited severe impairments at multiple levels of word processing. This included auditory lexical decisions but also spoken word-to-picture matching reflecting both a considerable deficit of lexical processing as well as comprehension of a spoken word's meaning. Thus, we demonstrated a severe deficit in HS in mapping auditory input onto semantic representations. If semantic processing is mandatorily involved in short-term retention tasks as argued by Majerus et al. (see Majerus et al., 2007; Patterson et al., 1994), HS with her severe deficit in accessing semantic information from auditory input should be affected. However, HS also had poor naming, that means, impaired access to output phonological information from semantics. It is, thus, fair to say that HS had multiple deficits in processing spoken words both on the input as well as the output side and it is also fair to say that her deficits were severe. Yet she was better at repeating back strings of words than individuals with better preserved word processing and naming. The second argument of Majerus et al. is that in their favoured family of cognitive models, impaired verbal short-term memory does not necessarily predict impaired language processing. They refer to a study by Martin et al. (1996) who ‘lesioned’ a computational model of word repetition. With artificial lesions, the model produced errors in word repetition tasks that resembled the errors of an individual, NC, with deep dysphasia. Crucially, the error patterns changed when model parameters were altered: By severely increasing activation decay, the model produced semantic errors in word repetition. A milder increase of decay altered the error patterns to formal and non-word errors. A very similar pattern was observed for their participant NC who apparently exhibited some recovery of functions. This went along with a mild increase in short-term memory capacity. Majerus et al. argue that within the same computational model, qualitatively different patterns of impairments could be simulated. This reflects an argument often found in the literature, namely, that previous STM patients have not been investigated thoroughly enough to rule out subtle deficits at the single word level (e.g. Allport, 1984; Buchsbaum Foygel writing – original draft; writing – review and editing. Data sharing is not applicable to this article as no datasets were generated or analysed during the current study.
Tobias Bormann (Tue,) studied this question.