7 Some things to learn from the
intersection between language
and working memory
Gonzalo Castillo
Working memory is a set of cognitive models that attempt to capture informa-
tion processing and the short-term storage function that supports it. This rapidly
expanding fi eld of research started by proposing an encapsulated system with
no commitments regarding the nature of language, but it has now reached a
stage in which fruitful theoretical exchanges with linguistics are possible when
an effort is made by both disciplines to approach a biologically grounded dis-
course. Here, I intend to confl ate a specifi c view of WM and syntax that I
believe already matches that requirement: state-based models of WM and
Boeckx’s (2013) Unbounded Merge. This exercise will hopefully reveal a com-
mon theme about how to look for the neurocognitive underpinnings of language,
which I take to be one of the most challenging and important problems of
modern linguistic science. More specifi cally, this chapter is organized in two
parts: fi rst, the main concepts of WM will be introduced; second, it will be
discussed to what extent Unbounded Merge can accommodate these concepts
by providing a computational description for them. I will fi nally argue that the
results of this exchange open new avenues of research for biologically minded
linguists.
1.1 The original multicomponent model
Working memory (WM) models were born to explain the necessary connection
that exists between paying attention, understanding and keeping things in short-
term memory. The most infl uential WM account is the original multicomponent
model proposed by B addeley and Hitch (1974), which I will introduce here by
taking into account its most recent incarnations (B addeley and Logie 1999,
B addeley 2000, B addeley 2012, cf. Figure 7.1) to defi ne a set of concepts that
will be useful to keep in mind throughout this paper.
An important fact about the multicomponent model is that the function of
storing content for short-term recall is performed in separate buffers. To handle
the visual and auditory modalities, a visuospatial sketchpad and a phonological
loop are respectively proposed. The phonological loop can maintain auditory
content for approximately two seconds unless verbal rehearsal (repeating what
one has just heard) is used to prevent decay (Bad deley and Ecob 1970).