*Reliability and *validity of *experiment in the neurobiology of learning and memory.

摘要

The concept of reliability has been defined traditionally by philosophers of science as a feature that an experiment has when it can be used to arrive at true descriptive or explanatory claims about phenomena. In contrast, philosophers of science typically take the concept of validity to correspond roughly to that of generalizability, which is defined as a feature that a descriptive or explanatory claim has when it is based on laboratory data but is applicable to phenomena beyond those effects under study in the laboratory. Philosophical accounts of experiment typically treat of the reliability of scientific experiment and the validity of descriptive or explanatory claims independently. On my account of experiment, however, these two issues are intimately linked. I show by appeal to case studies from the contemporary neurobiology of learning and memory that measures taken to guarantee the reliability of experiment often result in a decrease in the validity of those scientific claims that are made on the basis of such experiments and, furthermore, that strategies employed to increase validity often decrease reliability. Yet, since reliability and validity are both desirable goals of scientific experiments, and, on my account, competing aims, a tension ensues. I focus on two types of neurobiological experiments as case studies to illustrate this tension: (1) organism-level learning experiments and (2) synaptic-level plasticity experiments. I argue that the express commitment to the reliability of experimental processes in neurobiology has resulted in the invalidity of mechanistic claims about learning and plasticity made on the basis of data obtained from such experiments. The positive component of the dissertation consists in specific proposals that I offer as guidelines for resolving this tension in the context of experimental design.