like playing a musical instrument shows that fl exibility can persist past the for-
mation of chunks, which would correspond to cognitive control capabilities
exerting their power on retrieved chunks.
As we can see, both memory systems are highly integrated with the MD
network through ventral or dorsal pathways, and they should support both
long-term memories and processing functions, blurring the dividing line between
the system to draw a picture that comes back to the global workspace model,
in which different brain areas contribute to processing only as part of a large-
scale network that handles the basic operations of integration of information
along a time dimension. In this sense, the declarative/procedural distinction is
useful in signaling that learning does not reduce to a clear-cut opposition
between conscious and unconscious ways of learning, or between content and
action, but that these distinctions can ultimately be traced back to a gradient
of cognitive control (Botvinick and Co hen 2014) that will determine the com-
plexity and degree of fl exibility during the process of extracting meaning from
the world. The implementation of syntax will necessarily have to navigate through
the altering fl ows of information that I have so far described. The remaining
of this paper will try to put forth some insights that could help to achieve this
quest someday.
2.1 Working memory and syntax
Poeppel and Embi ck (2005) provide an overview of the state of the art of the
interface between linguistic theory and cognitive neuroscience, arriving at
the conclusion that reasonable linking hypotheses between the two disciplines
are non-existent. They identify the cause for this lack of dialogue in the differ-
ent granularity and incommensurability of the concepts that each discipline deals
with. They also propose a solution: each fi eld of study should look at the com-
putational side, producing computational models of their respective phenomena
that can be associated with one another. In the case of linguistic theory, this
entails a decomposition of its concepts into sets of distinct, smaller processes
that could be performed by neuronal populations. This strategy should not only
ground linguistic concepts in biology, but also inform us about how the brain
functions to produce the hallmarks of human cognition, entailing a real inter-
disciplinary exchange.
Throughout this paper, I have listed a series of classic concepts in WM and
provided a description of how they are redefi ned in computational and neuro-
cognitive accounts, usually by being split into different stages or suboperations
that together produce what from the outside we perceive as a distinct category.
Such was the case of attention, which was reduced to localized, parallel encod-
ing during processing, distinguishing between processes that boost specifi c neural
assemblies while suppressing others, and a process of bottom-up encoding of
goals that may infl uence what gets boosted or suppressed across multiple cogni-
tive cycles. If WM and linguistic theory can initiate an interdisciplinary dialogue,
they will have to do it by agreeing on a common ground in which the concepts
114 Gonzalo Castillo