It is becoming increasingly clear that receptive fields are not fixed structures but dynamic (Ramachandran, 1993). In addition to these dynamic variation driven by current stimulus values, adaptation seems to play a role in changing the organization of receptive fields. So how can the brain keep up with signals from cells that do not have constant properties?
The question assumes a certain way the brain works. It assumes that the brain works in the fashion of a set of dominoes. Knowing what happens at one level, e.g., the individual domino or cell, indicates what happens at a more complex level, e.g., the whole string of dominoes or the retina or perception. There is a direct predictability from one level of analysis to the next. What if the relationship between levels is statistical instead? In other words, the events at one level of analysis, the cell level, is only statistically related to the next level of analysis (Hofstadter, 1985d). For example, the motion of a molecule does not predict the behavior of a gas except statistically. It would not make sense or be worth the time or effort to understand how a gas is behaving in a balloon from knowing the motions of all of the particles inside. Their effect is best determined statistically by looking at the behavior of the collection.
If we apply this to the situation of variable receptive fields, perhaps their input at higher levels of the visual system are not determined by individual cell patterns but statistically looking across collections of cell response over position on the retina of the receptive field and time. Thus, the interpretation of the input from the retina may be clearly statistically than on the basis of individual cells. Modeling the retina by looking at collections of the cells and, I hope eventually, over time periods may help resolve this question. One of the big questions to address is what statistic should be used to summarize the retinal input? It will probably be a multidimensional one, including both spatial and temporal dimensions.