Studies of the neural substrates of the maintenance component of spatial working memory with blood oxygenation level dependent functional magnetic resonance imaging

Spatial working memory is a psychological construct that models the ability of an organism to maintain and manipulate spatial information in the absence of stimulus energy over a brief time scale. In the current work, studies were undertaken to test hypotheses regarding the neural substrates of the maintenance component of spatial working memory (MCSWM) in normal human subjects. Blood oxygenation level dependent functional magnetic resonance imaging (fMRI) was used as a neuroimaging method to implement these hypothesis tests. As a premise for the nature of subsequent inference, it is initially argued that causal inference in neuroimaging is weak. For the purpose of grounding analysis methods, the statistical properties of fMRI data in time and space were characterized. It was found that statistical models which assume independence of observations in time yielded grossly inflated false positive rates. Next, an experimental design for fMRI was developed for the purpose of addressing families of hypotheses regarding neural mechanisms of working memory. In theory, this design allowed the testing of such hypotheses without possible confounding by the neural correlates of certain other aspects of behavior. Tests of sensitivity and specificity were performed on this experimental design. This experimental design was then implemented to test hypotheses regarding the neural substrates of the MCSWM. It was found that there were cortical regions that evidenced increases in functional activity during delays requiring the MCSWM. Such regions that did so consistently across subjects were the right frontal eye field and the right superior parietal lobule. These results were replicated in an independent sample of subjects. It was also found that the dorsal prefrontal cortex (specifically Brodmann's areas 9 and 46) displayed such functional changes in a subset of subjects. This result is consistent with models based primarily on non-human primate lesion data which posit that dorsal prefrontal cortex is involved in aspects of spatial mnemonic processing. This result is simultaneously at odds with models which posit that human dorsal prefrontal cortex is only engaged by monitoring and manipulation of information and not by maintenance alone. Other hypotheses regarding neural mechanisms of the MCSWM were tested as well.