The sketchpad model. A theory of consciousness, perception, and imagery.

Subjective consciousness suggests a unity of the sensing and perceiving self that is difficult to reconcile with the multiplicity of sensory analyzers and the absence of a convergence zone in the brain. This has led--on the one hand--to the dead-end assumption of a unifying sentient homunculus and--on the other--to a denial of conscious unity. The sketchpad model presented here avoids this dilemma by viewing conscious thought as a selfreferent loop of neural activity, rather than as the information content of a fictitious set of output neurons. Use is made of the numerous neural pathways that originate at various cortical and subcortical areas and terminate in the thalamus. The model assumes that primary sensory inputs are modified by such feedback in the thalamic relay nuclei through hill-climbing processes that tend to optimize global responses.

The inversion of sensory processing by feedback pathways: a model of visual cognitive functions.

The mammalian visual system has a hierarchic structure with extensive reciprocal connections. A model is proposed in which the feedback pathways serve to modify afferent sensory stimuli in ways that enhance and complete sensory input patterns, suppress irrelevant features, and generate quasi-sensory patterns when afferent stimulation is weak or absent. Such inversion of sensory coding and feature extraction can be achieved by optimization processes in which scalar responses derived from high-level neural analyzers are used as cost functions to modify the filter properties of more peripheral sensory relays. An optimization algorithm, Alopex, which is used in the model, is readily implemented with known neural circuitry. The functioning of the system is investigated by computer simulations.

Brainstem control of sensory information: a mechanism for perception.

We have proposed a theory in which pathways ascending from the brainstem reticular formation control sensory centers in the dorsal thalamus and neocortex. We assumed that the sensory messages received at a given level are transformed by a stochastic process, called Alopex, in a way which maximizes responses in central feature analyzers. Perception is seen as a process involving a close cyclic interaction between brainstem and sensory relays. We discuss the specific case of visual information flow and the proposed modification of visual images at the level of the dorsal lateral geniculate nucleus (dLGN). Computer simulations of a simple model, representing the dLGN and reafferent control emanating from the reticular formation, show that sensory features are effectively enhanced and--in the absence of sensory input--quasi-sensory features may be generated by feedback of a simple scalar variable that is formed by the non-linear superposition of the responses of any number of feature analyzers. The model proposes a specific mechanism for such processes as visual imagery, hallucinations, and dreaming, and provides a framework for further studies into the nature of cognitive brain functions.

Periodic and nonperiodic burst responses of frog (Rana pipiens) retinal ganglion cells.

Neural activity of class 3 retinal ganglion cells was recorded in frog optic tectum, using extracellular microelectrodes. The stimuli were rectangular patches of contrast (light-on-dark or dark-on-light), applied within the previously determined receptive fields, for periods ranging from a few milliseconds to several seconds. ON and OFF responses were recorded for as long as 1 s following stimulation. Poststimulus time histograms revealed two types of responses, labeled periodic and nonperiodic bursters. The periodic bursters were characterized by periods of high activity separated by silent or near-silent intervals. The bursts occurred rhythmically with frequencies roughly between 15 and 50 Hz. Nonperiodic bursters generally showed both broad and sharp peaks in activity, but no regular periodicities. Activity profiles were flat initially, with silent periods appearing after the first few stimulus presentations, suggesting an inhibitory nature of the bursting process. The records were shown to combine the activities of several neurons. Analysis of the waveforms in real time made possible isolation of some units. In these cases, neurons exhibited a high degree of selective synchrony, i.e., the sharing of a portion of the activity profile, and notable differences at other times. These data have implications for the processing of visual information.

The Alopex process: visual receptive fields by response feedback.

The determination of trigger features of single neurons in afferent pathways has been one of the central problems in sensory physiology. A novel method, called Alopex, has been developed, in which response feedback is used to construct visual patterns that optimize the responses. Data are presented which show the emergence of trigger features of cells monitored in frog visual tectum. The method id checked against results obtained by scanning the visual field with a small spot. Correlations between Alopex pattern and scann patterns are generally between 0.3 and 0.5 but may be as high as 0.9 when smoothing and/or averaging procedures are applied to the Alopex patterns. The dynamics of the Alopex process are discussed and details of the algorithms are presented. The series of experiments presented here has established the validity of the method and suggests that this approach should find wide application in receptive field studies. For that purpose data on the instrumentation and software are also presented.

Anistropic connectivity and cooperative phenomena as a basis for orientation sensitivity in the visual cortex.

A computer simulation model of the neural circuity underlying orientation sensitivity in cortical neurons is examined. The model consists of a network of 3000 neurons divided into two functionally distinct cell types: excitatory (E-cells) and inhibitory (I-cells). We demonstrate that both orientation sensitivity and shape selectivity can be accounted for by making the following assumptions: 1) thalamic afferents to a sheet of cortical neurons are retinotopically organized; 2) thalamic afferents come from a single neuron, or at most a few neurons, in the lateral geniculate nucleus; 3) cortical activity is cooperative, i.e. largely dependent on intracortical connections, some of which have anisotropies along directions parallel to the pial surface. Anisotropies are specified only by the distribution of cells which are postsynaptic to a particular neuron, without specifying the axonal or dendritic contributions. In this paper, orientation sensitivity arises through cooperative interactions among neurons having anisotropic excitatory, and isotropic inhibitory connections.

Some quantitative results on Golgi impregnated axons in rat visual cortex using a computer assisted video digitizer.

Axonal fiber distributions of pyramidal cells in the visual cortex of the albino rat have been investigated using the rapid Golgi method and modern data collecting techniques. Three dimensional coordinate information was extracted from Golgi-impregnated axonal networks using a computer-assisted video digitizer. Computer programs used this data to generate various statistical distributions. In particular, angular distributions of the initial collateral segments and their endpoints were examined and found to reveal anisotropies. Inspection of the spatial distributions of the endpoints indicated a clustering at two distinct levels with respect to the pyramidal cell from which they originate. Dynamic graphic displays of the three dimensional data have been obtained and presented in the form of computer tracings of various orthogonal projections.