Skills and theoretical neuroscience

The approach developed here is complementary to the most common research strategies in neurosciences.

Theoretical, computational, or systems neuroscience looks at a subset of skills that might be realized as hardwired functions in the brain, and that can be localized. These might be specific functions in visual perception etc., or they might be general shared resource functions that are important for skills, such as episodic memory.

One experimental paradigm correlates tasks with brain activitiesn as measured by functional MRI. There are several limitations to that approach:

• Brains of distinct individuals in the same species are similar but less that identical, which restricts mapping functions by precise location.
  • A simple analogy is to investigate New York to find out about Toronto and other cities.
• fMRI has limited resolution, both in space and time. Localization is roughly within 2mm and therefore includes a lot of neurons. Time resulution suffers from a delay and is integrated over seconds rather than at the msec required for single neuron firing. (Note: the data derives from oxygen-using consequences of naural activity.)
• A simple analogy is to use traffic and parking for a city block over time intervals to investigate business and residential use of the area.

Another neuroscience strategy, that is closer to the neural level, is to build models out of idealized neurons and neural networks to see whether the function can be simulated. An example is neural networks that exhibit learning.

• A limiting factor is complexity, so that the scope of functions to be explored tends to be narrow, such as focusing on a specific perception - recognition problem.

Both of the strategies above assume a largely hardwired brain-information-processing architecture, rather than the stored-instruction architecture suggested by von Neumann et al after the ENIAC, and mostly used for modern computing.

The approach developed here is independent of assumptions about hardware and software implementation of information processing. It takes a systems approach, with emphasis on the interfaces with the skeleton (for action - output), and with input interfaces with perception through gravity and vision, where the 'real-world' mechanisms are known.

Another paradigm in neuroscience focuses the chemistry of the brain. This approach has even less of a location and time resolution.