TA: Flavia Filimon 2/19/2002

The Aperture Problem

For PATTERN MOTION

- several cortical areas can perceive motion (e.g. V1 layer 4b, MT, MSTd)

- V1 (layer 4b) has the aperture problem, whereas MT solves the aperture problem

- The problem is caused by the lack of an overall view in each cell. MT neurons combine information across space in order to solve it.

- V1 cells have small receptive fields, whereas M cells have larger receptive fields (since they receive input from all the other visual processing stages preceding MT).

- à V1 cells: sharp orientation tuning curve; MT cells: broad orientation tuning curve

- V1 cell receptive fields are like little straws. Each cell sees the world like through a small straw aperture, and is trying to make sense of it from the limited information it receives - about a tiny part of the visual field.

- Hence, when a compound object (containing edges of different orientations – e.g a square tilted on one of its corners) moves across the visual field, each/most of the little straws / V1 cells will be deceived by the contour of the edge apparently moving in one direction (the ”local” direction) which is not the total (“pattern”) direction of movement.

Specifically, V1 only detects the perpendicular component of motion (perpendicular to the object contour perceived through the straw).

- V1 is not capable of detecting the true pattern motion – averaging of all local responses will give the wrong answer.

- Evidence for this – comes from the tuning curves of V1 neurons, where responses are largest to patterns that are not moving in the cells’ preferred direction globally, yet whose local motions appear consistent with a horizontal motion.

How MT solves the problem:

- MT combines information from several V1 neurons – i.e. it gets inputs from several local motion detectors. Given a certain pattern motion, several local motion vectors will be compatible with it. An MT cell receives several overlapping local pattern inputs; the local pattern motion that “gets the most votes” (that most local motion inputs are consistent with) is most likely to be the true pattern motion.

- However, MT has its own aperture problem: MT cannot detect rotation, dilation, shrinking, sheering, spiral movements à composite motion. (e.g. rotating a paper around its center, motion of different spots on a balloon that is being inflated, bringing a piece of paper closer to one’s eyes, etc). MT can only detect translation.

- This can be tested by , e.g., presenting a stimulus (to an MT neuron’s receptive field) that is rotating clockwise, note the cell’s response, then rotate that stimulus counterclockwise. If the cell does not respond differentially, it probably cannot distinguish between the two types of movement. Similarly: expansion versus shrinking.

MSTd: middle superior temporal dorsal area

- MSTd cells have even larger receptive fields (compared to V1 and MT cells).

- MSTd combines inputs from MT neurons.

- Evidence for its composite motion sensitivity: different directions of rotation or dilation versus shrinking have different effects on the cell’s response.