Methods of Cognitive Neuroscience
Given the vast complexity of human perceptual and cognitive processes on one hand, and that of the human brain on the other (with approximately 1011 neurons with each neuron having on the average ~7000 synaptic connections with other neurons), it is clear that developments in research methodology play a pivotal role in the progress of cognitive neuroscience towards it’s ultimate goal of elucidating the neural basis of perceptual, cognitive and emotional functions in both health and disease. Fortunately, progress in non-invasive functional neuroimaging methods that allow one to record brain activity during presentation of various stimuli and execution of tasks in healthy volunteers has been very rapid. Today, modern imaging technology, along with sophisticated signal analysis algorithms, provides possibilities for scientists that one could only have dreamt of a few decades ago, such as imaging patterns of brain activity that can be associated with specific thoughts.
The extent that the neuroimaging methods can be utilized depends largely on theoretical advances in cognitive neuroscience. This is because theories and models of the neural basis of perceptual and cognitive functions makes it possible for the cognitive neuroscientists to design experiments that test specific hypotheses. As an example of hypothesis testing, recall the early experiments conducted by Pierre Flourens described. He set forth to critically examine hypotheses that he could formulate based on the theory of phrenology (i.e., that various mental functions are strictly localized in the brain). Since he did not find any apparent relationships between site of lesion and behavioral deficits, his results argued against phrenology and he proposed the aggregate field model of brain function.
As another simple example of a research hypothesis one could assume (i.e., hypothesize) that there are differences in height between men and women with men being taller. Then, by acquiring a large enough sample of male and female volunteers and measuring their heights, one would be able to test this hypothesis by comparing height distributions between the male and female samples. There are simple statistical tests that enable one to estimate probability at which the two samples differ. Typically in science the probability that two samples are similar has to be less than 0.05 (i.e., less than a five percent chance) before it can be concluded that the samples significantly differ from one another.
Hypotheses are very important in cognitive neuroscience. Without a solid theoretical framework and specific hypotheses, recording of brain activity would be more or less random and the results that are obtained would be difficult to interpret meaningfully (save for specific data mining approaches that find consistencies across huge datasets – but even in the case of such studies, interpretation of the emerging patterns of analysis results requires theoretical knowledge of cognitive processes and how the human brain functions). It is also important to note that given the complexity of data obtained using the modern neuroimaging methods, it is quite typical that there are several alternative explanations to a given set of findings. One can then test between the alternative explanations by formulating specific hypotheses and running further experiments. The outcome of the hypothesis testing then advances the theory/model either by validating or disproving it. In the latter case, one has to amend the theory/model to account for the discrepancy. This is the way that science progresses.
Importantly, none of the cognitive neuroscience methods that are available today can provide all the necessary information to solve how the brain works. Rather, each of the methods should be seen as providing complementary information that helps solving the big puzzle. One highly useful viewpoint to the complementary nature of the methods that are used in the
study of brain function is presented in Figure 2-1 below. As can be seen, each of the methods is alone either temporally or
spatially limited, but when information across the methods is combined, a relatively high spatiotemporal accuracy is achieved.