A Neural Signature of Divisive Normalization at the Level of Multisensory Integration in Primate Cortex

Abstract

Studies of multisensory integration by single neurons have traditionally emphasized empirical principles that describe nonlinear interactions between inputs from two sensory modalities. We previously proposed that many of these empirical principles could be explained by a divisive normalization mechanism operating in brain regions where multisensory integration occurs. This normalization model makes a critical diagnostic prediction: a non-preferred sensory input from one modality, which activates the neuron on its own, should suppress the response to a preferred input from another modality. We tested this prediction by recording from neurons in macaque area MSTd that integrate visual and vestibular cues regarding self-motion. We show that many MSTd neurons exhibit the diagnostic form of cross-modal suppression, whereas unisensory neurons in area MT do not. The normalization model also fits population responses better than a model based on subtractive inhibition. These findings provide strong support for a divisive normalization mechanism in multisensory integration.

From multisensory integration in peripersonal space to bodily self‐consciousness: from statistical regularities to statistical inference

“…This theoretical standpoint implies that there is nothing special in the computation leading to bodily self‐consciousness above and beyond an accrued history of congruent, body‐related, and highly multisensory experiences. Indeed, the dependence on synchrony and spatial co‐occurrence in many body‐related experiments is in line with the principles of multisensory integration, which are in turn in line with divisive normalization and a PPC leading to statistical inference (see above). Similarly, the reliance on PPS is in many regards similar to the spatial rule of multisensory integration, which can be explained as above.…”

“…Lastly, the divisive normalization interpretation brings forward a critical prediction; namely, that a nonoptimal stimulus for one modality which activates a neuron when presented in isolation should be able to suppress the response to a near‐optimal stimulus of the other modality when the cues are combined. Recordings in MSTd have recently confirmed this prediction, further highlighting populations of neurons as machines of (multisensory) statistical inference.…”

“…More specifically, this model uses a series of low‐pass filters and simple mathematical operations (summation, multiplication, and convolution) to resolve the correspondence problem, and then pools across spatially aligned multisensory correlation detectors in order to perform MLE. Interestingly, this last step in the model involves a divisive normalization step, reminiscent of physiological work postulating this same mechanism as culprit in neural reweighting as a function of stimuli reliability …”

Normalization models of cue combination in touch

“…http://dx.doi.org/10.1101/283218 doi: bioRxiv preprint first posted online Mar. 15, 2018; cue combination (29), which also involve normalization (30,31). To set the weights to the target and distractor cues, spatial attention could either directly modulate the activity representing the cues or attention could change the weighting of the connections between the populations signaling this information (32).…”

Emergence of opposite neurons in a firing-rate model of multisensory integration

“…Congruent (C) and opposite (O) neurons in this multisensory area receives input from two unisensory areas S1 and S2. An example of such a unisensory area would be the middle temporal area (MT), which is known to be the main visual input to MSTd [16]. A schematic of the model is shown in Fig.…”