Mechanisms of Capacity Limitations in Working Memory

Gold and colleagues have provided an extremely elegant example of how a precisely controlled behavioral study can be used to directly test implications generated by neurobiological theories of cognitive impairment in schizophrenia.
Further, they have provided novel and important data in schizophrenia that should cause us to re-examine theories about the mechanisms underling working memory impairments in this illness.

As noted by Gold, it has been hypothesized that altered GABAergic, glutamatergic, and/or dopaminergic inputs into reverberating and oscillatory networks in prefrontal or parietal cortex among individuals with schizophrenia should render such networks unstable and lead to less precise working memory representations that are particularly prone to decay (; ; ; ).

However, Gold and colleagues have shown that working memory representations in schizophrenia (at least of color memory) are neither less precise nor show evidence of exceptionally rapid decay.
Instead, individuals with schizophrenia showed clearly reduced working memory capacity.

These data contribute to a systemic body of work generated by Gold and colleagues, who have investigated the many aspects of working memory that could be impaired in schizophrenia.

They have also shown
that iconic decay is not increased in schizophrenia (),
that feature binding is intact (), and
that certain aspects of attentional control over working memory are intact (),
though others are impaired ().

However, working memory capacity has consistently been shown to be reduced in schizophrenia across numerous studies (; ; ).
If we take these results seriously (and we should), they require us to look closely at the neural mechanisms postulated to modulate capacity limitations in working memory
in order to generate clues to the mechanisms that may be leading to reduced working memory capacity in schizophrenia.

The neural mechanisms leading to working memory capacity limitations are still very much an open source of debate.
However, one influential theory is that the number of “items” that can be maintained in working memory is limited by the number of gamma cycles (30-100 Hz) that can be embedded within a theta cycle ().

Related to the idea that originally drove the design of the Gold study,
Lisman and others have hypothesized that individual items within working memory are represented by oscillating neural populations with spike rates phase-locked in a gamma cycle.
The oscillatory activity representing different items must be kept isolated,
potentially by keeping gamma activity for different items out of phase with each other.
One way to accomplish this would be to couple such gamma cycles into a lower frequency theta oscillation
that can help regulate and separate activity associated with different items (as well as maintain information about order).

Lisman and others have argued that capacity constraints of approximately four items in working memory () thus reflect the number of gamma cycles that can be embedded in a theta cycle (approximately four) (; ).

Gold’s results suggest that it may not be
the maintenance of the individual gamma-oscillating neural populations representing individual items that is impaired in schizophrenia.
Instead, it may be either the ability to establish such synchronous neural activity associated with a specific item,
or the ability to couple a number of different gamma-oscillating sub-networks into a theta cycle.

Interestingly, a growing number of studies have shown altered gamma activity during working memory in schizophrenia
(; ; ; ), as well as some evidence for altered theta activity
().
However, additional work is needed to specifically examine gamma-theta coupling in schizophrenia and its role in determining capacity limitations in this disease.

The type of network models of working memory put forth by Wang and colleagues
suggest that the dynamics of excitatory and inhibitory inputs drive the number of independent “activity bumps” (i.e., items) that can be maintained in a network ().

A related idea about the mechanisms driving capacity limitations and variations in these limits across individuals has been
put forth by Klingberg and colleagues,
who have argued that the dynamics of such lateral inhibitory mechanisms in parietal cortex limit memory capacity to be between two and seven items ().

However, they have also argued that such capacity limits
can be overcome, at least temporarily,
by excitatory inputs into parietal cortex from prefrontal cortex ().

They have suggested that this provides a mechanistic account of top-down control over working memory capacity by prefrontal cortex.

As such, given the evidence for at least some types of abnormalities in top-down control of attention in schizophrenia (; ),
and evidence for altered connectivity between prefrontal and parietal regions (; ),
another possible source of reduced capacity in working memory in schizophrenia may be
a reduction in prefrontal-mediated excitatory input into parietal networks that maintain items in working memory.

One might argue that the same GABA, glutamate, or dopamine mechanisms thought to impair the maintenance of representations in working memory
could also impair the initial establishment of gamma oscillating networks representing items,
their coupling to a lower-frequency theta cycle,
or even the ability of prefrontal cortex to provide excitatory inputs into neural networks supporting the representation of items in working memory.
If so, such models will also need to explain how such impairments could lead to reduced working memory capacity in schizophrenia
without a change in precision or decay, a challenge for most current neural network models of working memory.
As such, the data provided by Gold and colleagues suggest an exciting new pathway for research on working memory in schizophrenia
that may allow us to develop more precise mechanistic hypotheses
as to the source of these cognitive impairments and their relationship to pathophysiology of this illness.

References
Lisman JE, Coyle JT, Green RW, Javitt DC, Benes FM, Heckers S, Grace AA. Circuit-based framework for understanding neurotransmitter and risk gene interactions in schizophrenia. Trends Neurosci. 2008;31(5):234-42.

Durstewitz D, Seamans JK. The dual-state theory of prefrontal cortex dopamine function with relevance to catechol-o-methyltransferase genotypes and schizophrenia. Biol Psychiatry. 2008;64(9):739-49.

Rolls ET, Loh M, Deco G, Winterer G. Computational models of schizophrenia and dopamine modulation in the prefrontal cortex. Nat Rev Neurosci. 2008;9(9):696-709.

Lewis DA, Cho RY, Carter CS, Eklund K, Forster S, Kelly MA, Montrose D. Subunit-selective modulation of GABA type A receptor neurotransmission and cognition in schizophrenia. Am J Psychiatry. 2008

t