Jonides et al. (2008) provide a detailed model of short-term memory encod-
ing and maintenance in which three different stimuli get serially encoded and sustained
until the first one reappears and is matched to its memory (cf. Figure 7.3). Distrib-
uted patterns of sustained activity represent each stimulus during the encoding
phase, which simultaneously and within the same location trigger different
plasticity mechanisms (Zucker and R egehr 2002) that are responsible for short-
term retention. After the stimulus is gone, and before presenting a new one,
sustained activity would start to decrease while plasticity mechanisms decay
much more slowly. This decrease of sustained activity should take a few milli-
seconds after the stimulus is gone, so it could come back more quickly if the
same stimulus was presented again instead of a new one. This would explain
why there is a difference in reaction time between retrieving the most recently
presented item as compared to those that are supposed to be stored in a magical
number four format (e.g. McElree and Dosher’s 1989). When the second and
third stimuli are presented, synaptic weight changes are affected by the same
nature of some of their visual features, which may interfere with the retrieval
of the first item (the objective of the task). The last presented item, which
should have the same perceptual identity as the first and be recognized as such
(a process known as updating), would trigger the same pattern of sustained
activity by means of the pattern-completion property of attractor networks
(Hopfield 1982 ).
The model could be almost directly translated into a dynamic neuronal
dimension by proposals such as Dipoppa and G utkin (2013) or Roux and
Uhlh aas (2014), in which different kinds of oscillatory activity would synchronize
to perform the encoding, sustainment and updating operations. The difference
is that here the sustainment operation is not performed by plasticity, but by
synchronization that is strong enough to deflect potentially irrelevant encodings
(interferences). This idea is in line with McElree (2006 ), who reviews a series
of experiments in which the distinction between short-term and long-term
memory is blurred by the fact that no qualitatively different reaction times are
observed during the retrieval process from each purported component. If short-
term memory and long-term memory form a continuum, capacity limitations
such as Cowan’s (2001 ) magical number four may be better described as pro-
cessing constraints that can reduce to rapid periods of refreshing. Instead of
interference phenomena overriding encoded material in short-term memory
(e.g. Oberauer 2013) , what we would find is that performance is reduced by
interference as a result of new encodings that are not related to the task that
needs to be performed. Similarly blurring the dividing line between processing
and storage, Barrouillet et al. (2011) propose the notion of cognitive load, a
function of the amount of time it takes to process a specific stimulus, which
has been shown to be negatively correlated with the amount of information
that can be sustained in WM. The debate is still open regarding the distinctive-
ness of short-term storage and processing functions, and the modeling of
interference phenomena.
110 Gonzalo Castillo