the properties of chemicalcompounds, the spectrum of sunlight; others are probabilistic). Natural selection not only selects genes, but alsoincludes or excludes (selects) environmental factors as part of this inheritance by selecting for genes that specifydevelopmental programs that either involve various factors in development or render them irrelevant. Wehypothesize that what is genetically specified should be a first order set of developmental programs (distributed inthe gross structure of the early brain): programs that bootstrap their way uphill to greater levels of functionalorganization through programmed endogenous interactions, and interactions with evolutionarily targetted environmentalinvariants that supply missing information. Because the functions of the brain's network of neural devices arecomputational rather than physical or chemical, the developmental programs in the brain must necessarily useevolved computational criteriaadaptive targetsto guide the process of wiring to the final levels of functionalprecision. That is, each functionally distinct evolved neural unit must contain feedback-driven developmentalprograms that test the unit (including expected inputs and computed outputs) against its own proprietary set ofcomputational benchmarks (adaptive targets). This system can detect the degree to which its computationsconform to its required functional specifications, and use this error term as input to feedback-driven adjustmentsuntil the target coordination is achieved. Robustly specified adaptive targets allow the developing brain to anti-entropically countervail against environmental perturbations, genetic noise (except in the specification of theadaptive targets), and insults, so that the necessary evolved computational interrelationships are exactinglyachieved, even though substantial individual differences in the physical layout may emerge in the course of wiringaround an insult. On this view, observed neural plasticity is not a mechanism, but an effect of a diverse set ofadaptively highly targeted, hill-climbing mechanisms. The intuitively appealing idea of global plasticity or generalflexibility is an insufficient way of describing the principles involved because flexibility just means the capacity tovary, and there are vastly more potential variants that are wrong than are functional. The hard computationalproblem for developmental programs is the identification, out of the vast space of possible configurations, of thattiny subset that create, improve or restore function. Heterogeneous sets of adaptive targets are the necessarysolution required to build the species' diverse complement of neurocognitive mechanisms, because each mustcompute different functional outputs according to their own divergent proprietary evolved criteria of functional
Showing the most recent 10 out of 36 publications
