In order to determine whether moving visual stimuli were sufficient to induce the emergence of direction-selective responses, the animals were exposed to two "training" stimuli consisting of grating patterns which drifted back and forth across the visual field perpendicular to the orientation of the grating in opposite directions. These stimuli were presented to the ferrets for 5 seconds at a time, with intervals of 10 seconds, for a period of 20 minutes. Subsequently, the activity of primary visual cortical neurons was observed whilst these stimuli were presented again.   

For the first 8-10 hours of visual stimulation after this motion training, no changes were observed in the functional properties of visual cortical neurons. Most neurons were highly responsive to the orientation of the stimuli, but their selectivity to the direction in which the stimuli moved was very weak. Later on, small groups of cells with a preference for one of the two training stimuli began to emerge. With time, these responses progressively increased, so that each group became highly tuned to one or the other training stimuli (see above figure). The number of neurons selective for each orientation was also found to increase with time.

To test whether it was the motion of the training stimuli that induced these changes in activity, the researchers flashed identical gratings in the ferrets' visual fields for brief periods of time. This "flash training" elicited responses in the same cortical neurons, but the responses did not increase with time. Gratings which moved in eight directions that differed from those in the training elicited little response or none at all. This confirmed that the observed emergence of orientation selectivity was indeed due to exposure to the training stimuli.

Closer examination of the responses of individual pyramidal neurons in layer 2/3 of the cortex revealed that the preferred direction of motion of each changed over time, so that it became more like the preferences of its neighbours. Prior to training, most of the cells exhibited uncertain or moderate orientation preferences. Upon presentation of the training stimuli, however, the responses of most neurons became more certain, and the neurons segregated into small domains with a preference for one direction or the other.

Other interesting functional changes were also observed. Some neurons maintained their initial moderate preference for one direction of movement and later increased their response to it, while others reversed their orientation preference during training. If, for example, a neuron was surrounded by cells with a preference for the opposite direction, it was likely to reverse its own preference so that it matched that of its neighbours. On the other hand, a neuron surrounded by others with the same preference was unlikely to change its own preference during training. This suggests that the functinal grouping of neurons occurs because of some kind of interaction between neighbouring cells during motion training.

These experiments show that early experience of moving visual stimuli has a strong and relatively rapid effect on the functional properties of neurons in the primary visual cortex. Initially, the ferret primary visual cortex contains an array of neurons with weak direction preferences, possibly because of light entering through the closed eye lids. The two training stimuli used, which consisted of gratings moving in opposite directions, transformed this array into two highly ordered columns, each containing neurons with a highly selective preference for one of the directions of stimulus motion. The study supports the widely-held belief that sensory experience is essential for proper visual development, but adds some fascinating details of how it does so. It also raises the question of exactly how visual cortical neurons interact with each other during their selection of direction preference.

Précis of Neuroconstructivism: How the Brain Constructs Cognition

Another important implication is the central role of developmental trajectories in the interpretation
of adult cognition. There is no teleology involved in development; mature, normative cognition is an outcome of development, not a pre-specified target (Thomas & Karmiloff-Smith 2003).

I agree we are not in a newtonian universe anymore, but in a dynamical universe. By reintroducing the aristotelian final causality as teleology you are still in a essentialist world view, were things have internal purpose. We should reduce the phenomena of teleology not as a internal finality but as the result of fractal or dynamical constraints that give rise to attractors. The funny thing is that this dynamical systems frame of analysis is offering us the mathematical tools for reducing functionnal neuroscience to physics by general dynamical process like self-organization. I'm really encouraging everyone to read the Scholarpedia article on "Self-organization" to get a grip on how these circular causal process EXPLAIN AWAY the teleology or final causality that Matt want to reintroduce.

You should google that : "Three Fallacies of Teleology" and read the little blog article. My level of english is just too poor to continue the debate, but I believe you're attributing goal-directed behavior to things that are not goal-directed.