Associative memory in a multi-modular network

Cortical region interactions and the functional role of apical dendrites

The basal and distal apical dendrites of pyramidal cells occupy distinct cortical layers and are targeted by axons originating in different cortical regions. Hence, apical and basal dendrites receive information from distinct sources. Physiological evidence suggests that this anatomically observed segregation of input sources may have functional significance. This possibility has been explored in various connectionist models that employ neurons with functionally distinct apical and basal compartments. A neuron in which separate sets of inputs can be integrated independently has the potential to operate in a variety of ways not possible for the conventional neuron model, in which all inputs are treated equally. This article thus considers how functionally distinct apical and basal dendrites can contribute to the information-processing capacities of single neurons and, in particular, how information from different cortical regions could have disparate effects on neural activity and learning.

A balanced memory network

A fundamental problem in neuroscience is understanding how working memory—the ability to store information at intermediate timescales, like tens of seconds—is implemented in realistic neuronal networks. The most likely candidate mechanism is the attractor network, and a great deal of effort has gone toward investigating it theoretically. Yet, despite almost a quarter century of intense work, attractor networks are not fully understood. In particular, there are still two unanswered questions. First, how is it that attractor networks exhibit irregular firing, as is observed experimentally during working memory tasks? And second, how many memories can be stored under biologically realistic conditions? Here we answer both questions by studying an attractor neural network in which inhibition and excitation balance each other. Using mean-field analysis, we derive a three-variable description of attractor networks. From this description it follows that irregular firing can exist only if the number of neurons involved in a memory is large. The same mean-field analysis also shows that the number of memories that can be stored in a network scales with the number of excitatory connections, a result that has been suggested for simple models but never shown for realistic ones. Both of these predictions are verified using simulations with large networks of spiking neurons.

A critical component of cognition is memory—the ability to store information, and to readily retrieve it on cue. Existing models postulate that recalled items are represented by self-sustained activity; that is, they are represented by activity that can exist in the absence of input. These models, however, are incomplete, in the sense that they do not explain two salient experimentally observed features of persistent activity: low firing rates and high neuronal variability. Here we propose a model that can explain both. The model makes two predictions: changes in synaptic weights during learning should be much smaller than the background weights, and the fraction of neurons selective for a memory should be above some threshold. Experimental confirmation of these predictions would provide strong support for the model, and constitute an important step toward a complete theory of memory storage and retrieval.

Role of long-term synaptic modification in short-term memory

One way that some types of short-term or working memory may be implemented in the brain is by using autoassociation networks that recirculate information to maintain the firing of a subset of neurons in what is termed an attractor state. We describe how long-term synaptic modification is necessary to set up the appropriate stable attractors, each one of which corresponds to a memory of a particular item. Once the synapses have been modified, any of the short-term memory states may be triggered by an appropriate input which starts the neurons firing in one of the attractors, and then the firing is maintained in that attractor by the already modified synapses, with no further synaptic modification necessary. This analysis leads to the prediction that if this type of implementation is used for working memory, then long-term synaptic modification may be necessary only during an acquisition phase of a task, and once the task has been acquired, the performance of the working memory task should be unimpaired if no further synaptic modification is allowed. We show that a considerable body of research findings on the effects of agents that block synaptic modification on working memory tasks can be understood in this way. Many of the findings are consistent with the hypothesis that blocking synaptic modification in the hippocampus impairs the acquisition, but not the later performance, of hippocampal-dependent working memory tasks.

Dendritic mechanisms underlying the coupling of the dendritic with the axonal action potential initiation zone of adult rat layer 5 pyramidal neurons

1. Double, triple and quadruple whole-cell voltage recordings were made simultaneously from different parts of the apical dendritic arbor and the soma of adult layer 5 (L5) pyramidal neurons. We investigated the membrane mechanisms that support the conduction of dendritic action potentials (APs) between the dendritic and axonal AP initiation zones and their influence on the subsequent AP pattern. 2. The duration of the current injection to the distal dendritic initiation zone controlled the degree of coupling with the axonal initiation zone and the AP pattern. 3. Two components of the distally evoked regenerative potential were pharmacologically distinguished: a rapidly rising peak potential that was TTX sensitive and a slowly rising plateau-like potential that was Cd(2+) and Ni(2+) sensitive and present only with longer-duration current injection. 4. The amplitude of the faster forward-propagating Na(+)-dependent component and the amplitude of the back-propagating AP fell into two classes (more distinctly in the forward-propagating case). Current injection into the dendrite altered propagation in both directions. 5. Somatic current injections that elicited single Na(+) APs evoked bursts of Na(+) APs when current was injected simultaneously into the proximal apical dendrite. The mechanism did not depend on dendritic Na(+)-Ca(2+) APs. 6. A three-compartment model of a L5 pyramidal neuron is proposed. It comprises the distal dendritic and axonal AP initiation zones and the proximal apical dendrite. Each compartment contributes to the initiation and to the pattern of AP discharge in a distinct manner. Input to the three main dendritic arbors (tuft dendrites, apical oblique dendrites and basal dendrites) has a dominant influence on only one of these compartments. Thus, the AP pattern of L5 pyramids reflects the laminar distribution of synaptic activity in a cortical column.

Cortical instability and the mechanism of mania: a neural network simulation and perceptual test

BACKGROUND

A previous neural network simulation suggested that manic states arise from excessive levels of noise that destabilize neural representations. The Necker cube stick figure provides a simple perceptual task that assesses stability of gestalt-type representations.

METHODS

A neural network was developed that included a simulation of the Necker cube task. Noise was added to induce maniclike jumps from one representation to another. A parallel study of Necker cube perception was conducted with 16 patients diagnosed with manic-spectrum disorder, 18 patients with schizophrenia, and 19 normal control subjects. Cognitive speed and rate of indiscriminate responses were assessed using an auditory continuous performance task.

RESULTS

During processing of the "Necker cube" stimulus, the reversal rate of the noise-destabilized "manic" network was increased by 30%. In the human subject study, the median score of Necker cube reversal rates for manic-spectrum patients was roughly twice that of normal control subjects and patients with schizophrenia. Accelerated reversal rates in the manic-spectrum group were not attributable to excessive cognitive speed or higher rates of indiscriminate responses.

CONCLUSIONS

The two studies, considered together, support the hypothesis that excessive cortical noise destabilizes neural representations in manic-spectrum patients.