Motion integration by neurons in macaque MT is local, not global

Area of Science:

• Neuroscience
• Visual Perception
• Computational Neuroscience

Background:

• Direction-selective neurons in the primary visual cortex process local feature motion.
• These local motions require integration to represent object motion, with visual cortical area MT (V5) being a candidate site.
• Previous research on motion integration in MT used stimuli that filled the receptive field, limiting the understanding of integration scale.

Purpose of the Study:

• To investigate the spatial scale of motion integration in visual cortical area MT (V5).
• To determine if MT neurons integrate motion information across their entire receptive field or within smaller local regions.

Main Methods:

• Identified two effective "patches" within the receptive field of MT neurons.
• Presented plaid stimuli with component gratings overlapping within a single patch versus separated across patches.
• Compared MT neuron responses to overlapping versus separated component gratings.

Main Results:

• MT neurons selective for pattern motion when gratings overlapped lost selectivity when separated.
• Separated gratings caused neurons to become selective for individual component motion directions.
• This indicates that pattern-direction selectivity relies on local computations within MT receptive fields.

Conclusions:

• Motion integration and pattern-direction selectivity in MT are not based on pooling inputs across the entire receptive field.
• Critical computations for pattern-direction selectivity in MT occur locally, on a scale smaller than the whole receptive field.
• This challenges the notion of MT cells simply pooling all receptive field inputs to represent a single object's motion.