However, reductionism in all aspects of science -- particularly in the study of human cognition -- can suffer from an inappropriate emphasis on the process of reducing phenomena into component pieces, without the essential and complementary process of using those pieces to reconstruct the larger phenomenon. We refer to this latter process as reconstructionism. It is simply not enough to say that the brain is made of neurons; one must explain how billions of neurons interacting with each other produce human cognition. Teitelbaum67 argued for a similar complementarity of scientific processes -- analysis and synthesis -- in the study of physiological psychology. Analysis entails dissecting and simplifying a system to understand its essential elements; synthesis entails combining elements and understanding their interactions.
The computational approach to cognitive neuroscience becomes critically important in reconstructionism: it is very difficult to use verbal arguments to reconstruct human cognition (or any other complex phenomenon) from the action of a large number of interacting components. Instead, we can implement the behavior of these components in a computer program and test whether they are indeed capable of reproducing the desired phenomena. Such simulations are crucial to developing our understanding of how neurons produce cognition. This is especially true when there are emergent phenomena that arise from these interactions without obviously being present in the behavior of individual elements (neurons) -- where the whole is greater than the sum of its parts. The importance of reconstructionism is often overlooked in all areas of science, not just cognitive neuroscience, and the process has really only recently become feasible with the advent of relatively affordable fast computers.
Figure: Illustration of the importance of reconstructionism -- it is not enough to say that the system is composed of components (e.g., two gears as in a), one must also show how these components interact to produce overall behaviors. In b, the two gears interact to produce changes in rotational speed and torque -- these effects emerge from the interaction, and are not a property of each component individually.
Figure 1.1 shows a simple illustration of the importance of reconstructionism in understanding how systems behave. Here, it is not sufficient to say that the system is composed of two components (the two gears shown in panel a). Instead, one must also specify that the gears interact as shown in panel b, because it is only through this interaction that the important ``behavioral'' properties of changes in rotational speed and torque can emerge. For example, if the smaller gear drives the larger gear, this achieves a decrease in rotational speed and an increase in torque. However, if this same driving gear were to interact with a gear that was even smaller than it, it would produce the opposite effect. This is essentially what it means for the behavior to emerge from the interaction between the two gears, because it is clearly not a property of the individual gears in isolation. Similarly, cognition is an emergent phenomenon of the interactions of billions of neurons. It is not sufficient to say that the cognitive system is composed of billions of neurons; we must instead specify how these neurons interact to produce cognition.