1.4 Working memory and long-term memory
From the perspective of an organism, the physical world is ever-changing but
causal, as patterns and relationships can be extracted when gathering perceptual
data, and these patterns can be used in the future to obtain reliable outcomes.
WM and learning are two interconnected processes, as long-term memory can
be considered a repository in which the processing of present experiences can
be stored to help the processing of future ones. In the context of WM, a
description of the processes that operate as part of it cannot be complete unless
the act of chunking (producing complex, multimodal structures for processing
the present in connection with the past) is accounted for, including how pro-
cessing generates chunks, how it is possible to alter them to arrive at new
meaningful patterns, and how their formation maximizes WM capacities.
As mentioned before, the use of long-term memory activates the same sensory,
cognitive and motor areas that are present during the processing of external
stimuli, and a functional description of memory cannot be divorced from the
description of processing that we have established so far. Taking that into
account, the literature on long-term memory divides the object of its study into
two distinct processes that are specifically subserved by two different networks:
declarative and procedural memory. Declarative memories comprise episodes
(episodic memory) and facts (semantic memory) that are consciously recalled,
and their recollection mainly involves ventral pathways that connect prefrontal
cortex structures with medial temporal lobe structures, crucially including the
hippocampus (Ullman 2004), which seems to be especially relevant in keeping
multimodal information together and bound to a context during the first days
of memory formation, but that gradually loses importance as time progresses
and connections between the cortical regions that sustain that memory are
strengthened (Takashima et al. 2009). Rubin et al. ( 2014) go further, suggest-
ing that the hippocampus plays a crucial role in sustaining flexible behavior by
making multimodal representations available to control networks like MD, which
would connect to it through ventro-medial prefrontal cortex.
The procedural memory system, on the other hand, consists of the learning
of habits and procedures, defined as “sequential, repetitive, and motor behaviors
elicited by external or internal triggers that, once released, can go to completion
without constant conscious oversight” (Graybiel 2008). Examples of procedural
behaviors are skills like riding a bicycle or playing the piano, in which automa-
ticity develops with practice, but also the implicit acquisition of habits. In terms
of the brain, procedures are acquired by a prefrontal-thalamic-basal ganglia
network, which connects dorsolateral prefrontal cortex with striatum, and is
modulated by dopaminergic inputs that can produce two main states of activa-
tion in the network depending on where dopamine is expressed (cf. D’Esposito
and Postle 2015 for a review). In terms of WM, the striatum would favor task
switching, while the prefrontal cortex would be responsible for sustainment or
the deployment of the MD network, in a model analogous to the one by Jonides
et al. (2008) that was previously presented to explain short-term encoding.
112 Gonzalo Castillo