The Cortical Organization Of Visual Perception And Visuomotor Control

DF -- brain injured because of monoxide poisoning. Result was damage to her visual system, particularly "form vision". She "was unable to describe the orientation and form of any visual contour, no matter how that contour was defined."

Acting Without Perceiving

DF's deficit was perceptual, not sensory. "She simply cannot perceive shapes and forms, even though her visual system appears to have the requisite low-level sensory information at its disposal." DF cannot copy drawings, but can produce good drawings from memory. "Needless to say, when shown any of the drawings seh had done herself, whether the ones retrieved from memory or those copied from another drawing, she had no idea what they were and commented that they all looked like `squiggles'." Crucial discovery -- DF has intact visuomotor abilities. "Even though she cannot recognize a very familiar object on the basis of its visual form, she can grasp that object under visual control as accurately and as proficiently as people without normal vision."

E.g., "post the card in the slot" task. "In one task, in which she was asked to match the orientation of the card to that of the slot, DF was profoundly impaired; in another task, in which she was asked to insert the card in the slot, she was virtually normal." Why does DF's vision fracture in this way?

The Function(s) Of Vision

60% of monkey neocortex is devoted to vision. "It is commonly assumed that the function of all this visual machinery is to provide some sort of internal representation of the external world to serve as the perceptual foundation for thought and action. ... In this chapter, I will challenge this assumption and present evidence suggesting that there is no single multipurpose representation and that visual input is transformed in different ways for different purposes." Specifically, Goodale wants to dissociate perceptual experience from visuomotor action.

Motor Control: The Origins Of Visual Systems

"Vision did not evolve to enable organisms to `see'; it evolved to control their movements with respect to distal stimuli." E.g. in the frog there exist distinct neural pathways for prey-catching and for moving around barriers. "In fact, accumulating evidence from studies in both frog and toad suggest that there are at least five separate visual-motor modules, each responsible for a different kind of visually guided behavior and each having distinct input and output pathways." Similar modularity exists in mammals, as shown in studies of rodents.

Perception And Action In The Primate Brain

Human vision is more complex, and therefore must be more flexible. We *do* have representationl systems. "Perceptual systems are clearly very different from the simple visuomotor modules of amphibia. They are not linked to specific motor outputs, but to cognitive systems involving memory, semantics, planning, and communication." But execution of action may still be mediated by visuomotor modules! (NB: This is going to be Goodale's major contribution to the modularity issue.)

The Emergence Of Cortical Visual Systems

E.g., saccadic eye movements are carried out by superior colliculus. "There is evidence that in the primate brain the outputs from thse basic visuomotor modules and the transfomrations they perform on incoming sensory information can be modified by higher-order input from the cerebral cortex." E.g., saccadic eye movements also involve input from the brain stem. "The primitive input-output visuomotor pathways of the early vertebrates, which were little more than simple servomechanisms, have been upgraded to much more subtle control mechanisms capable of mediating the visual guidance of the complex motor outputs that primates are capable of producing." But, they are still modular! So, Goodale views cortical expansion of the primate brain as reflecting the emergence of perceptual systems and separate action systems.

Two Streams Of Processing In Primate Cerebral Cortex

Ungerleider and Mishkin argued in the early 1980s for the separation of "what" and "where" streams in the brain. The old view was that the ventral stream of processing was for object identity, and that the dorsal stream was for object location. Goodale proposes a new interpretation. "The critical difference between the two streams lies no so much in the nature of the visual information they receive but in the transformations they perform upon that information." Ventral stream is perceptual identification, but for Goodale the dorsal stream provides sensorimotor transformations. "Notice that in this account information about object attributes, such as size, shape, orientation, and spatial location, are processed by both streams but that the nature of that processing is very different."

Why are separate streams necessary? Because different types of outputs are required --> object-based vs. viewer-based. "Perception tends to be concerned with the enduring characteristics of objects (and their interrelations), so that they can be recognized when they are encountered again in different visual contexts or from different vantage points. To generate these long-term representations, perceptual mechanisms must be object-based." In contrast, "underlying visuomotor transformations have to be viewer-centered; in other words, bot the location of the object and its orientation and motion must be encoded relatve to the observer." So, the two systems have similar inputs, but have dramatically different output requirements. "As a consequence of these differences in processing requirements, two cortical visual systems have evolved: a ventral stream for visual perception and a dorsal stream for the visual control of skilled actions."

What evidence supports this view? Consider dorsal stream first. For example, activity of cells in parietal cortex is behavior dependent (nb: extraretinal influences). Processing of motion is a key component of this stream. Also there are links from PP to premotor regions of frontal cortex.

Ventral stream gets most of its input from V1 -- magnocellular and parvocellular streams equally represented. IT cells behave as if they deliver object-centered coding (NB: Look back to Biederman's chapter!)

"We can now begin to understand the reasons for the remarkable dissociations in the abilities of DF." I.e., DF's ventral stream was compromised, but not her dorsal stream.

The Neuropsychology Of Perception And Action

Visual agnosia -- two types, apperceptive and associative. "According to Lissauer, an associative agnosic would be able to copy a drawing even though he or she many not be able to identify it, while an apperceptive agnosic would be unable to copy or identify the drawing." Goodale suggests that Lissauer's notion can be mapped onto Goodale's notion of separate streams.

DF is an apperceptive agnosic. But, for some tasks, she is remarkably sensitive to object shape. "DF's visuomotor systems work just fine; it is her perception of the slot's orientation that is the problem." Simialr results when DF's perception of object size was examined. She couldn't do same/different judgments of visual size, but grasp during reaching was sytematically and normally sensitive to object size. Similar results on shpae -- DF is unable to make shape judgements, but grasp is appropriately sensitive to shape changes. "In summary, then, the brain damage that DF suffered as a consequence of anoxia appears to have interrupted the normal flwo fo shape an contour information into her perceptual systems without affecting the processing of shape and contour information by her visuomotor control systems."

MRI studies of DF show her brain is damaged in ventrolateral regions of occipital cortex, and that V1 is intact. Goodale is cautious aobut interpreting such anatomical evidence, because the brain damage is diffuse, and so is hard to interpret.

Patients VK and RV

Now Goodale explores the double dissocation. Optic ataxia: "They are unable to use visual information to reach out and grasp objects in the hemifield contralateral to the lesion. At the same time, they often have no difficulty recognizing or describing objects that are presented in that part of the visual field." Eg. VK -- objects can be identified, but grasp is not systematically adjusted to visual properties. Similar results found with patient KV. "Observations such as these suggest that it is not only the spatial location of the object that is inaccessible for controlling movement in pateints with dorsal stream lesions butthe inrinsic characteristics of the object as well." (NB: This last statement is against the old what/where distinction between the two anatomical streams.) Bottom line -- double dissociation established.

Limits On Visuomotor Processing

DF has trouble with more complex versions of the slot task, such as when a T shape is used -- she is "unable to combine the orientaiton of the stem and the top of the T appropriately." But she is good at identifying grasp points, so..."one visuomotor system, that controlling the posture and placement of thefingers during grasping, appears to be sensitive to relatively complex shape information while another, the system controlling the rotation of the wrist (when the hand is holding an object) is not."

Contours defined by more complex, Gestalt properties could *not* be used by DF in the slot task! Visuomotor channels rely on luminance differences!

Recent studies have examined graspoing trajectories to remembered objects. Normal subjects can do this, but not DF, even if the required memory is only 2 seconds duration.

Some Remaining Issues And Puzzles

Not much evidence to show that "spatial vision" is mediated by PP. What about visual neglect? "The area most commonly implicated in visual neglect following damage to the posterior portion of the brain is a region at the junction of the occipital, temporal, and parietal lobes -- an aread that is several centimeters below the area typically associated with optical ataxia." In other words, posterior parietal cortex serves many complicated functions

Selection of "grasp goal" is likely not done by visuomotor modules, so the two streams have to be able to interact. This involves attention, and "the problem of how semantic information about hte goal object is used to modulate the action itself. The kinematics of a particular action often reflect the function of the goal object to whicht hat action is directed as well as its size, shape, orientation, and location."

Is ventral stream associated with consciousness? Only speculation is possible! Goodale does not want to claim that it is the seat of consciousness.

Pearl Street | "An Invitation To Cognitive Science" Home Page | Dawson Home Page |