Restricted Receptive Fields Research Articles

Reduced Representation by Neural Networks with Restricted Receptive Fields

Model neural networks can perform dimensional reductions of input data sets using correlation-based learning rules to adjust their weights. Simple Hebbian learning rules lead to an optimal reduction at the single unit level but result in highly redundant network representations. More complex rules designed to reduce or remove this redundancy can develop optimal principal component representations, but they are not very compelling from a biological perspective. Neurons in biological networks have restricted receptive fields limiting their access to the input data space. We find that, within this restricted receptive field architecture, simple correlation-based learning rules can produce surprisingly efficient reduced representations. When noise is present, the size of the receptive fields can be optimally tuned to maximize the accuracy of reconstructions of input data from a reduced representation.

NMDA receptors and activity-dependent tuning of the receptive fields of spinal cord neurons.

After peripheral nerve section, sensory neurons regenerate but do not regain their original topographical position in the skin. Here we report that in the early stages of sciatic nerve regeneration, the cutaneous receptive fields (RFs) of dorsal horn neurons are larger than normal, reflecting the disorganized topography of the regenerated afferents. When nerve regeneration is complete, small contiguous RFs emerge, indicating a central compensation for the disrupted peripheral somatotopy. If the NMDA receptor antagonist MK801 is given during regeneration, RFs do not show this reorganization, but remain large and diffuse. We suggest that the coincident activity of afferents, newly innervating adjacent or overlapping cutaneous territory, acts through postsynaptic NMDA receptors to strengthen the central effectiveness of these inputs at the expense of other non-adjacent and non-coincidently activated inputs. In this way, dorsal horn neurons may attain and retain restricted RFs in the face of a spatially dispersed afferent input.

Evidence for a central component in the sensitization of spinal neurons with joint input during development of acute arthritis in cat's knee

1. In the spinalized cat, nociceptive spinal neurons with knee input show enhanced responses to mechanical stimulation of that joint once an inflammation has developed in the knee. Enhanced responses may result from increased afferent inflow as well as from modifications of the nociceptive processing within the spinal cord. To examine the significance of these components, we tested in 30 chloralose-anesthetized, spinalized cats whether, during development of arthritis, changes of responsiveness in spinal neurons are restricted to stimulation of the inflamed joint or whether responsiveness in these neurons is altered in general. While continuously recording from a neuron, we injected kaolin and carrageenan into one knee and tested the responses to mechanical stimuli applied to the joint and to regions adjacent to and remote from the knee during the developing arthritis. In addition, in six cats we monitored the neurons' responses to electrical stimulation of the sural nerves and the rostral lumbar spinal cord. 2. Of 32 neurons in laminae VI, VII, and VIII of the lumbar spinal cord, 15 ascending and eight nonascending cells were driven by mechanical stimulation of one or both knee joint(s). Nine of these were nociceptive specific (NS), responding exclusively or predominantly to noxious compression of the knee and other deep tissue, and 12 were wide-dynamic-range (WDR) cells with graded responses to gentle and noxious stimuli applied to the knee joint(s), deep tissue, and skin. Two neurons with high ongoing discharges had some excitatory joint input but showed marked inhibition by most stimuli used (INH neurons). The majority of the neurons had receptive fields on both legs. Nine of the 32 neurons had no input from the knee(s). 3. All 23 neurons with joint input became sensitive or more responsive to movements and gentle compression of the inflamed knee once the inflammation had developed. In general, these neurons also showed enhanced responses to compression of the adjacent muscles in thigh and lower leg. In 20 neurons, response properties were even altered for stimuli applied to regions remote from the inflamed joint, including the contralateral leg in 18 cases. We found expansion of initially restricted receptive fields (mainly in NS cells), enhancement of preexisting responses, and/or lowering of threshold to mechanical stimuli applied to these regions; few neurons developed inhibitory reactions.(ABSTRACT TRUNCATED AT 400 WORDS)

Comparative Physiology of Sound Localization in Four Species of Owls

Bilateral ear asymmetry is found in some, but not all, species of owls. We investigated the neural basis of sound localization in symmetrical and asymmetrical species, to deduce how ear asymmetry might have evolved from the ancestral condition, by comparing the response properties of neurons in the external nucleus of the inferior colliculus (ICx) of the symmetrical burrowing owl and asymmetrical long-eared owl with previous findings in the symmetrical great horned owl and asymmetrical barn owl. In the ICx of all of these owls, the neurons had spatially restricted receptive fields, and auditory space was topographically mapped. In the symmetrical owls, ICx units were not restricted in elevation, and only azimuth was mapped in ICx. In the barn owl, the space map is two-dimensional, with elevation forming the second dimension. Receptive fields in the long-eared owl were somewhat restricted in elevation, but their tuning was not sharp enough to determine if elevation is mapped. In every species, the primary cue for azimuth was interaural time difference, although ICx units were also tuned for interaural intensity difference (IID). In the barn owl, the IIDs of sounds with frequencies between about 5 and 8 kHz vary systematically with elevation, and the IID selectivity of ICx neurons primarily encodes elevation. In the symmetrical owls, whose ICx neurons do not respond to frequencies above about 5 kHz, IID appears to be a supplementary cue for azimuth. We hypothesize that ear asymmetry can be exploited by owls that have evolved the higher-frequency hearing necessary to generate elevation cues. Thus, the IID selectivity of ICx neurons in symmetrical owls may preadapt them for asymmetry; the neural circuitry that underlies IID selectivity is already present in symmetrical owls, but because IID is not absolutely required to encode azimuth it can come to encode elevation in asymmetrical owls.

Comparative Physiology of Sound Localization in Four Species of Owls (Part 2 of 2)

Bilateral ear asymmetry is found in some, but not all, species of owls. We investigated the neural basis of sound localization in symmetrical and asymmetrical species, to deduce how ear asymmetry might have evolved from the ancestral condition, by comparing the response properties of neurons in the external nucleus of the inferior colliculus (ICx) of the symmetrical burrowing owl and asymmetrical long-eared owl with previous findings in the symmetrical great horned owl and asymmetrical barn owl. In the ICx of all of these owls, the neurons had spatially restricted receptive fields, and auditory space was topographically mapped. In the symmetrical owls, ICx units were not restricted in elevation, and only azimuth was mapped in ICx. In the barn owl, the space map is two-dimensional, with elevation forming the second dimension. Receptive fields in the long-eared owl were somewhat restricted in elevation, but their tuning was not sharp enough to determine if elevation is mapped. In every species, the primary cue for azimuth was interaural time difference, although ICx units were also tuned for interaural intensity difference (IID). In the barn owl, the IIDs of sounds with frequencies between about 5 and 8 kHz vary systematically with elevation, and the IID selectivity of ICx neurons primarily encodes elevation. In the symmetrical owls, whose ICx neurons do not respond to frequencies above about 5 kHz, IID appears to be a supplementary cue for azimuth. We hypothesize that ear asymmetry can be exploited by owls that have evolved the higher-frequency hearing necessary to generate elevation cues. Thus, the IID selectivity of ICx neurons in symmetrical owls may preadapt them for asymmetry; the neural circuitry that underlies IID selectivity is already present in symmetrical owls, but because IID is not absolutely required to encode azimuth it can come to encode elevation in asymmetrical owls.

Overlapping taste and tactile maps of the oropharynx in the vagal lobe of the channel catfish, Ictalurus punctatus.

Microelectrode mapping experiments indicate an ipsilateral representation of the oropharynx and a well-defined, bilateral input from the proximal portion of the maxillary barbels and snout region within the vagal lobe of channel catfish. The map of the oropharyngeal epithelium is distorted so that the gill arches are rotated through an angle of 90 degrees along the transverse plane, and the dorsally mapped region of the gill rakers is tilted posteriorly in the sagittal plane of the vagal lobe. Multiunit recording studies fail to provide definitive boundaries of adjacently mapped domains of oropharyngeal structures. Gustatory receptive fields of neurons in the vagal lobe correspond to their location on the topological map obtained by tactile stimulation of the oropharyngeal epithelium. A few single unit recordings indicate restricted receptive fields and different response patterns of taste, tactile, and proprioceptive neurons in the vagal lobe of catfish.

Consistent features of the representation of the hand in area 3b of macaque monkeys.

Multiunit microelectrode recordings were used to explore the responsiveness and somatotopic organization of the representation of the hand in area 3b of anesthetized macaque monkeys. Major findings were as follows: Recording sites throughout the hand representation were activated by low-threshold cutaneous stimulation. Simple, punctate mechanical stimuli were highly effective in activating neurons. Neurons had small, restricted receptive fields. Representations of nearly all skin surfaces of the hand were demonstrated in individual monkeys. The basic topographic pattern found in all monkeys included the following: a large sequential representation of the glabrous digits from thumb to little finger from lateral to medial in cortex, and from proximal to distal hand parts in cortex extending down the caudal bank of the central sulcus; moderately large representations of radial and ulnar pads of the palm in respective lateral and medial cortical locations in the hand representation; and a relatively small, fragmented representation of the dorsal hand and dorsal digits, with the fragments interspersed within the representation of the glabrous hand. The proportions of the proximal, middle, and distal glabrous digits varied, so that the representation of the distal phalanx sometimes approached the dorsal border of area 3b with area 1. A comparison of the present findings with previous results from macaque monkeys indicates that the above-described features have been revealed under a variety of recording and anesthetic conditions. Consistencies in previous and present results strongly support the conclusions that the hand representation in area 3b of macaque monkeys is activated by cutaneous receptors throughout; is composed of neurons with relatively simple, small, cutaneous receptive fields; includes all skin surfaces of the hand; and is somatotopic for the glabrous skin with small, discontinuous, intercalated representations of fragments of the dorsal skin.

Cutaneous facilitation of transmission in Ib reflex pathways in the human upper limb.

Group I effects from wrist extensors to flexors, i.e. an early inhibition followed by a relative facilitation which are thought to be Ia and Ib in origin respectively, were compared in control conditions and when preceded by a weak cutaneous stimulation. Stimulating the skin of the dorsal side of fingers II-III, which did not modify the test reflex size when applied alone, increased Ib facilitation. By contrast this Ib effect was not changed by stimulation of the skin of the palmar side. It has previously been shown that heteronymous Ib inhibition in the lower limb during voluntary contraction is facilitated by cutaneous stimulation from restricted receptive fields. Thus, present results lend support to the idea that cutaneous facilitation of transmission in Ib reflex pathways might be functional in curtailing exploratory movements.

Auditory receptive fields in primate superior colliculus shift with changes in eye position.

The process by which sensory signals are transformed into commands for the control of movement is poorly understood. A potential site for such a transformation is the superior colliculus (SC), which receives auditory, visual and somatosensory inputs and contains neurones that discharge before saccadic eye movements. Along the primary sensory pathways, signals coding the spatial location of auditory, visual and somatosensory targets are based on distinctly different coordinate systems, and it is not known whether each type of sensory input uses a separate motor pathway or if they are converted into a common coordinate system in order to share a single pre-motor circuit. Sensory neurones in the SC have spatially restricted receptive fields (RFs) and are organized into maps across the collicular surface. Acute experiments have shown a rough correspondence between the spatial positions of RFs of neurones encountered along a single dorsal-ventral penetration of the colliculus, regardless of the modality of the effective stimulus, suggesting that auditory, visual and somatosensory maps might be in register. However, in these conditions the head-centred auditory system and the retinotopic visual system are aligned because the eyes are in the primary orbital position. Moreover, other data have suggested that the primate SC is organized in motor, not sensory, coordinates, although in the cat, eye position was found to have no effect on auditory receptive fields. We therefore sought here to determine what happens to the registration of the auditory and visual maps in the alert, behaving animal. Monkeys, with heads fixed, were trained to make delayed saccadic eye movements to auditory or visual targets from one of three initial fixation points while the activity of single neurones was recorded extracellularly. We found that the auditory receptive fields shifted with changes in eye position, allowing the auditory and visual maps to remain in register.

Receptive fields of solitario-parabrachial relay neurons responsive to natural stimulation of the oral cavity in rats.

The receptive field (RF) of 67 taste and 85 mechanoreceptive neurons in the solitary tract nucleus (NTS) were located in the oral cavity in albino rats. All of the taste and most (62.4%) of the mechanoreceptive neurons examined had an RF on the ipsilateral side of the tongue and/or the palate. Regardless of whether they were solitario-parabrachial relay (SP) neurons or non-SP neurons, RFs of taste neurons were found on the anterior as well as the posterior tongue. But there were some differences in the RF distribution between the SP and non-SP mechanoreceptive neurons. Most of the mechanoreceptive SP neurons (9 of 11) had an RF on the tongue, while ca. half of the mechanoreceptive non-SP neurons (43 of 79) had an RF on the tongue and palate, but the rest had an RF on other tissue. Most of the neurons studied had a small restricted RF, but complex RFs, e.g. two separate RFs on the tongue, were found in a relatively small number of neurons (four taste and five mechanoreceptive neurons). An inhibitory RF, usually in a remote place from the excitatory RFs, was found in four mechanoreceptive neurons but no inhibitory RFs for taste neurons. Electrical stimulation of the epithelium in the RF with a low current of short duration evoked a few spikes in most units. Two of the three units, giving rise to a vigorous response to taste stimulation, but having single restricted RFs on the anterior tongue, produced a train of spikes lasting more than 20 ms in response to electrical stimulation of the RF.(ABSTRACT TRUNCATED AT 250 WORDS)

Responses of primate SI cortical neurons to noxious stimuli.

Recordings were made from single SI cortical neurons in the anesthetized macaque monkey. Each isolated cortical neuron was tested for responses to a standard series of mechanical stimuli. The stimuli included brushing the skin, pressure, and pinch. The majority of cortical neurons responded with the greatest discharge frequency to brushing the receptive field, but neurons were found in areas 3b and 1 that responded maximally to pinching the receptive field. A total of 68 cortical nociceptive neurons were examined in 10 animals. Cortical neurons that responded maximally to pinching the skin were also tested for responses to graded noxious heat pulses (from 35 to 43, 45, 47, and 50 degrees C). If the neuron failed to respond or only responded to 50 degrees C, the receptive field was also heated to temperatures of 53 and 55 degrees C. Fifty-six of the total population of nociceptive neurons were tested for responses to the complete series of noxious heat pulses: 46 (80%) exhibited a progressive increase in the discharge frequency as a function of stimulus intensity, and the spontaneous activity of two (4%) was inhibited. One population of cortical nociceptive neurons possessed restricted, contralateral receptive fields. These cells encoded the intensity of noxious mechanical and thermal stimulation. Sensitization of primary afferent nociceptors was reflected in the responses of SI cortical nociceptive neurons when the ascending series of noxious thermal stimulation was repeated. The population of cortical nociceptive neurons with restricted receptive fields exhibited no adaptation in the response during noxious heat pulses of 47 and 50 degrees C. At higher temperatures the response often continued to increase during the stimulus. The other population of cortical nociceptive neurons was found to have restricted, low-threshold receptive fields on the contralateral hindlimb and, in addition, could be activated only by intense pinching or noxious thermal stimuli delivered on any portion of the body. The stimulus-response functions obtained from noxious thermal stimulation of the contralateral hindlimb were not different from cortical nociceptive neurons with small receptive fields. However, nociceptive neurons with large receptive fields exhibited a consistent adaptation during a noxious heat pulse of 47 and 50 degrees C. Based on the response characteristics of these two populations of cortical nociceptive neurons, we conclude that neurons with small receptive fields possess the ability to provide information about the localization, the intensity, and the temporal attributes of a noxious stimulus.4+.

Reorganization of raccoon somatosensory cortex following removal of the fifth digit.

The organization of part of the primary somatosensory cortex was examined in anesthetized raccoons at 2, 8, or 16 weeks after the normal peripheral input to this region of cortex had been removed by amputation of the fifth digit. Electrophysiological recordings were made in and around the cortical area representing the fifth digit. Eight intact animals were used to verify that this specific area could be accurately localized on the basis of the sulci and to determine the normal response characteristics of this area. The results from nine animals with the fifth digit removed provided evidence for a gradual reorganization of the cortical area which had been functionally denervated. At 2 weeks postamputation the field was almost totally unresponsive to sensory input. At 8 weeks many sites were responsive to high intensity stimulation of rather extensive regions of the hand. At 16 weeks the cells fired more readily to peripheral stimulation than at 8 weeks and tended to have smaller, more restricted receptive fields. The location of receptive fields in this latter group suggested that the fifth digit area was taken over primarily by input from the fourth digit. The time course of this reorganization is suggestive of extensive anatomical changes either within the cortex itself or at subcortical levels.

Properties of velocity-mechanosensitive neurons of the cat ventrobasal thalamic nucleus with special reference to the concept of convergence

Neurons in the ventrobasal (VB) thalamic nucleus of lightly anesthetized cats were studied in order to analyze their discharge properties in response to controlled mechanical stimuli. Properties of the vast majority of the neuronal population largely resemble those of peripheral sensitive mechanoreceptors in their response to the velocity and, to a more limited extent, the amplitude component, of skin and hair displacement within restricted receptive fields. Detailed examination reveals some ‘complex’ characteristics suggestive of central integration in about 11% of VB neurons. Complex properties appear to indicate convergent input reflecting receptive field organization, the variety of velocity-related discharges and patterns of inhibition. The rarity of isolated ‘surround’ inhibition and the definition of what constitutes ‘lemniscal’ properties are discussed in the context of these and other related findings.

Sensory Processing in the Brain and Evoked Potentials

This chapter discusses sensory processing in the brain and evoked potentials (EPs). Simple cells have relatively restricted receptive fields, in which elongated excitatory and inhibitory zones flanked one another and gave rise to a strong preference for appropriately oriented bars or edges of luminance moving across them. Complex cells, which also had well-defined orientation preferences, are responsive over a wider area and are presumed to depend on simple cells for their inputs. Simple and complex cells are thought of as serially related “feature detectors”, each of which signified by its firing one particular geometrical detail of the figure of excitation on the retina. Models of pattern recognition are canvassed in which these simple and complex cells served as the first stages in a hierarchic pattern classifier leading to unique specificity of single cells for the most complex patterns. Simple cells are unresponsive to all forms of visual noise presented alone, except when individual blobs of texture are large enough to function as luminance stimuli in their own right.

Short-latency peripheral inputs to thalamic neurones projecting to the motor cortex in the monkey.

One hundred seventy-five neurones in the n.ventroposterior lateralis (VPL) and n.ventralis lateralis (VL) in the thalamus of anaesthetised monkeys have been tested antidromically for projection to the cortex and for somatosensory input from the contralateral arm. Using bipolar stimulation of the cortical surface, 113 thalamic neurones were successfully identified as antidromically driven from the hand area of the postcentral gyrus (48 neurones) or from the hand area of the precentral gyrus (65 neurones). All but one of these 113 neurones could only be antidromically discharged from the postcentral cortex or from the precentral cortex, and not from both. Most had antidromic latencies between 0.5 and 1.5 ms. Twenty-five/sixty-five precentrally projecting neurones and 45/48 postcentrally projecting neurones were activated by stimulation of the contralateral median or radial nerves. Both groups responded at short latency (4--8 ms) and many were activated by low-threshold shocks (0.8--1.3 T) and had restricted receptive fields on the hand. Precentrally projecting neurones responded most powerfully to joint movement or deep pressure, and some of these neurones were also responsive to cutaneous stimuli. Precentrally projecting neurones with peripheral inputs were all found in the oral subdivision of the VPL (the VPLO). The properties of these neurones suggest that they may be partly responsible for rapid somatosensory input to the motor cortex.

Response properties of units in the lateral geniculate nucleus of the domestic chick ( Gallus domesticus)

One hundred and four visual units of the (ventral) lateral geniculate nucleus (GL) of chicks (Gallus domesticus) were studied with extracellular microelectrodes. Most were extremely responsive to moving targets even at low speeds (0.1 degrees/sec). They were classified as follows. (1) movement-sensitive units with uniform, restricted receptive fields (47%). (2) Movement-sensitive units with small receptive fields possessing regions responding to differing modes of stimulation ('center-periphery units'). (3) Movement-sensitive neurons with uniform, wide receptive fields, larger than 75 degrees x 65 degrees (15%). (4) Movement-insensitive cells, responsive only to illumination changes (7%). (5) Units with indeterminate receptive fields, a poor visual response or a poor response to movement (14%). The firing of neurons in classes I-III was mostly an accurate function of target speed; half of them also provided accurate monitoring of the beginning, course and termination of target movement (tonic units). In view of the above, the high proportion of directionally selective cells (31%) and the large size of many of its receptive fields, the GL is considered as a candidate structure for optomotor integration but the evidence is otherwise scant. There is, however, an apparent contribution to its unit properties from both retinal ganglion cells and neurons from different tectal layers. Whatever its role, it is unlikely to be the functional homologue to the mammalian ventral lateral geniculate as the unit properties of two structures differ substantially.

Visual and visual-vestibular responses of frog cerebellar neurons

Micropipette recordings were obtained from single neurons of the frog cerebellum to monocular and binocular visual stimuli. 64% of the visually activated cerebellar neurons could be identified as Purkinje cells. Two major groups of cerebellar visual neurons could be distinguished: The neurons of the first group have restricted receptive fields within the visual field of one or both eyes and respond to diffuse illumination or to small contrast stimuli moving through the receptive field. The neurons of the second group have no restricted receptive fields and respond best to a moving patterned surround, stimulating one or both eyes. Among these neurons four classes could be distinguished by binocular surround stimulation: Neurons activated by horizontal movement to the right (C1 neurons), neurons activated by movement to the left (C2 neurons), neurons activated by movement in both directions (C3 neurons) and neurons which are inhibited by movement in both directions (C4 neurons). The contribution of each eye to the binocular response revealed that the signals from both retinae interact either in a Homodirectional or a Heterodirectional manner. Stimulation of the receptors in the horizontal semicircular canal (horizontal turning movement) leads to a change in the activity of cerebellar visual neurons. Depending on the directionality of the neurons, an enhancement or a reduction in the neuronal activity can be achieved by combining visual and vestibular stimulation. Electrical optic nerve stimulation revealed two latency ranges for the activation elicited by the mossy fiber as well as by the climbing fiber system.

The Afferent Discharge Elicited by Vibrotactile Stimulation

When skin is mechanically stimulated, differentreceptor systems will be activated depending upon the type of skin stimulated and stimulus characteristics such as rate and amplitude of displacement and repetition frequency. The general properties of the primary mechano-receptive units are (for the rapidly adapting intracutaneous receptors) low thresholds, restricted receptive field, relatively good rate sensitivity (in terms of repetitive impulse discharge), and poor capacity of tracking on high-frequency repetitive stimulation; (for the Pacinian corpuscles) high critical slope, very low displacement threshold, diffuse receptive field, poor rate sensitivity, and good frequency-following capacity; (for the tactile spots in hairy skin) low displacement threshold, small receptive field, good rate sensitivity, and good frequency-following capacity. When tactile displays are designed, the stimulus parameters will have to be chosen according to the psychophysical functions as well as to the receptor properties, in order to obtain a device that is optimal from a discriminatory point of view.

An electrophysiological analysis of cutaneous mechanoreceptors in a snake

1. 1. A systematic analysis of mechanoreceptors in skin of the snake Pseudechis porthyriacus revealed several response patterns. Some units adapted rapidly to an applied stimulus and others gave a maintained discharge to prolonged stimulation. A few receptors we encountered which had an intermediate adaptation rate. 2. 2. The majority of rapidly adapting receptors had a relatively high mechanical threshold, were insensitive to high-frequency vibrational stimuli and had restricted receptive fields that overlapped extensively. 3. 3. The slowly adapting receptors were of low mechanical threshold making it difficult to map their receptive fields. Their response pattern could be modified by altering skin temperature.

Response of cells to restricted visual stimuli in an association area of cat cerebral cortex

Cellular responses to restricted light patterns (circular spots or vertical slits, 0.5–3.0 degrees in diameter) were studied in the anterior part of middle suprasylvian gyrus of cats anesthetized with chloralose. “On”, “off”, or “on-off”, phasic responses were observed over almost the entire visual field when stimulus intensities were two or more log units brighter than the background. When near-threshold stimuli were employed, restricted receptive fields (3–30 degrees in diameter) were located near the center of the visual field. The most extensive receptive field mapping was done near the horizontal meridian and the type of response often changed as the stimulus was moved to different parts of the field. Cells were influenced by both eyes and summation was observed during binocular stimulation. Many of the cells also responded to auditory click stimuli. After surgical ablation or undercutting of ipsilateral visual cortical areas and section of the corpus callosum, cells rarely responded to restricted light stimuli although diffuse light stimuli were still effective. Transient depression of visual cortex by the application of KCl also eliminated cell responses to restricted spots of light; responses returned with recovery of the depressed cortical areas. The present data indicate that neurons in this association area respond to restricted visual stimuli in a specific manner, and this sensory-specific visual input is dependent upon participation of visual cortex. It is suggested that this cortical locus plays a role in complex sensory mechanisms involving intermodality, sensory-specific interactions.