When Deepika first kissed Saif, he trembled with excitement and his face turned away slightly. But Deepika was quick and adjusted her swoop: her lips softly met his square and centre.
A collar-ripper, and no doubt you want to know what happened next. But no, dear reader: this is New APPS, and I am afraid the ‘S’ stands for Science. My interest is in something no less interesting but (sadly) much less heart-racing. Visually guided movement. For had Deepika (like Saif) ecstatically shut her eyes while closing in for the smooch (instead of peeking, as you can clearly see), she would have managed only a sisterly kiss on Saif’s cheek.
(Warning: [very] slightly salacious material below the fold.)
So much, so obvious. But as all empirically informed philosophers of perception are aware, cognitive scientists have discovered over the last half century that conscious feedback from vision is not what was at work here. Conscious vision is concerned with movement in the external world. But since head and eyes are constantly moving, the retinal image is very jittery. The brain therefore inserts an image stabilizer. Since the conscious image is stabilized, small and sudden movements are often edited out, and the guidance of on-line course corrections have to be handled by a different system.
Luckily, there are several visual systems devoted to the control of behaviour. Some non-cortical systems control autonomic movements such as eye saccades and the like. But the system that has attracted the attention of philosophers most of all is the so-called “dorsal stream.” As I recounted a while ago in a post on “one-shot induction” in neuroscience, this vision for action data-stream came to prominence when David Milner and Mel Goodale examined a patient, DF, who, while taking a bath one day, collapsed of carbon monoxide asphyxiation due to a poorly installed water heater. When she was revived, DF seemed to be blind. But she soon found that she could perform visually guided tasks. Today, she can clear a table and load the dishwasher. She navigates a crowded airport with little difficulty. But when she goes to check in, she cannot see the airline agent’s face.
Milner and Goodale posited a visual data stream that goes upward from the primary visual cortex to the parietal lobe of the brain. They hypothesized that this data stream is dedicated to the control of bodily movement. In DF, this “dorsal stream” was spared. What the asphyxiation destroyed was a different data-stream, the so-called ventral stream, which descends to the temporal lobe.
In work published over the last seven years or so, I have argued that dorsal stream vision (which I call “motion-guiding vision”, both so I can stay in the realm of functional description, and so as not to be hostage to the changing trends of cognitive neuroscience) gives subjects a feeling of engagement with their material surroundings. And because it allows them to interact with objects in their surroundings, it introduces a “demonstrative” element into vision. Last week, when I wrote on episodic memory, Berit Brogaard asked what distinguished such memory (which carries a feeling of pastness) from live vision (which carries a feeling of presence). Well, one of the things that distinguishes the two is the demonstrative character of the latter, brought to it by the involvement of the dorsal stream.
Milner and Goodale’s work followed on from work that had been building since the 1960s, or even earlier. Lawrence Weiskrantz, David Ingle, and Gerald Schneider were pioneers of the many-visual-pathways idea; Ungerleider and Mishkin had proposed “what” and “where” pathways (see above); and so on. But Milner and Goodale’s model gradually became the most influential, though it is still far from universally accepted. They were the ones to emphasize the vision-for-“perception”/vision-for-action distinction, and to propose that dorsal stream vision codes egocentrically, while ventral stream vision codes (more) allocentrically.
Milner and Goodale’s model has stood up to attacks pretty well, but in October’s Trends in Cognitive Sciences, Edward de Haan and Alan Cowey launch a critique from an angle that does not affect the points made above. De Haan and Cowey are critical of the notion that visual data might be sequentially processed in cortical pathways. A part of their argument relies on an evolutionary argument. The visual system could not have evolved “in one sweep”, they (rightly) say:
The view that the different, functionally distinct maps in the visual cortex are organised in a hierarchical manner within two different and more or less linear visual routes assumes a much more elaborate Darwinistic step compared to the view that different visual abilities evolved independently.
I am not persuaded by this argument. Evolution does have to be “modular,” that is, it does have to build elaborate systems with minimal innovation from pre-existing independently functioning bits, but I can’t see why this is incompatible with data-streams.
A second argument that de Haan and Cowey advance is that a surprising number of cognitive faculties show elaborate structure: colour perception is “high-level”, for example. What they mean is that colour perception depends on a number of subordinate processing tasks. Thus, they argue that there is a variety of “agnosias” (or specific deficits) associated with colour. One patient is able to see and match colour, for example, but “unable to identify colours or match the correct colour to objects (red fire engine, green grass, etc.)”
I find this second argument much more persuasive. The idea is that processing for relatively simple features such as colours shows an interdependency of many different processes. This is suggestive more of a “patchwork” than a pathway.
No doubt Milner and Goodale will respond in an upcoming issue of TICS. They are not ones to take a challenge lying down.