Prying Open the Black Box

See allHide authors and affiliations

Science  06 Oct 2006:
Vol. 314, Issue 5796, pp. 17
DOI: 10.1126/science.1135216

If the abstracts for this year's annual meeting of the society for neuroscience (later this month in Atlanta, Georgia) are any indication, collaboration between experimentalists and theoreticians is thriving. This is good news for neuroscience, given that not so long ago, neuroscientists who did experiments the old-fashioned way—recording single-cell activity in living, often behaving experimental animals—had little tolerance for theoreticians of the brain. Both groups hoped to understand how the brain works, but the theoreticians apparently wanted to achieve that goal while ignoring the complex and dynamic ways in which neurons communicate. The pervasive view among the experimentalists was that for the theoreticians, the brain may as well be a black box. Well, that sentiment is no longer pervasive. Rather, computational neuroscience (see the special section in this issue) has made great strides in the past decade, and the field is poised to resolve a spectrum of stubborn mysteries of the brain.


The sources of the old experimental/theoretical divide are understandable. Most neuroscientists trained in the 1960s, 1970s, and 1980s were comfortable with mathematical representations of basic neuronal phenomena, such as the ionic basis for resting membrane potential. They were awed by the elegant Hodgkin-Huxley equations that explained and predicted the ionic basis for the action potential (based, of course, on an ideal experimental model, the squid giant axon). Experimental neuroscience even embraced the quantal theory of neurotransmitter release by neurons. However, when the theorists began turning their attention to how the cerebellum or hippocampus works, these theories that avoided the complexities of neuronal circuitry and synaptic interactions were suddenly regarded as too speculative and not testable.

During the past decade, there has been a robust explosion of methods to characterize the functional and structural details of neurons and their interconnected cellular networks. We can finally realistically simplify the essential features of neuronal ensembles and test theories about what specific neuronal circuits may do. In other words, we can ask why specific cortical pyramidal neurons and interneurons underlie processes such as “edge detection.” A wealth of quantitative data on signaling mechanisms, ion-channel behavior, neuron populations, and morphological features has made possible the development of neuronal and synaptic models that are rigorously constrained by the biochemical and structural properties of specific types of neurons. Such models have wedged open the black box, allowing experimentalists and theoreticians to collaborate. And, as reflected by the Society for Neuroscience meeting agenda, the targets of computational neuroscience are wide-ranging, speaking not only to age-old questions about the possible information encoded in intervals of electrical activity in neurons but also to decision-making and risk/reward assessment and neurological (mainly epilepsy) and psychiatric (schizophrenia and depression) diseases. Computational neuroscience permits us to ask what structural principles explain the selective and redundant wiring diagrams of single-sense pathways and those that integrate multiple senses.

Computational neuroscience is also attempting to incorporate the burgeoning field of neuroinformatics and the growing body of electronic databases into its evolution. This is where some of the next great challenges lie. We need to develop and implement uniform, standardized modes of data presentation in neuroscience, so that data from individual research papers can be readily scanned and integrated into more comprehensive databases. This will be fundamental to developing more accurate and useful models of brain function. For many years, the U.S. National Institutes of Health funded the Human Brain Project to nurture this field, and comparable efforts were established globally. However, the funds for ongoing maintenance do not exist, and past investments are endangered.* If adequate funding over a prolonged period of time is not secured, new principles of education, psychology, and social science that are based on neuroscience are in jeopardy.

  • *M. S. Gazzaniga et al., Science 311, 176 (2006).

Navigate This Article