ABSTRACT
Wherever one stands on the issues of how a thought experiment is executed and how it provides novel insights into the real world, clearly thought experiments utilize the human capacity for imaginative thinking. A simple exemplar of that capacity is one that Herbert Simon liked to provide audiences in his talks on mental imagery. He would ask the question: “How many windows are in your house?” His expectation, overwhelmingly confirmed, was that you would envision each room and simulate walking through your house until you arrived at an answer. At the end, you had gained information about your house that you are unlikely to have been aware of previously. This exemplar shares the imaginative dimension of thought experiments, though the capacity equally enables humans to entertain riding on a flying pig. Importantly, unlike the flights of fancy humans are capable of, thought experiments in science are developed to provide novel insight into real-world phenomena or to confirm or disconfirm hypotheses. Considerable metacognitive control is exercised over their development, likely requiring numerous iterations even though they are mostly recorded in finished form. Several hypotheses have been advanced as to how an experiment only in thought can provide novel data applicable to the real world or “promote basic conceptual reform” (Kuhn 1977, 291). Surprisingly, thought experiments per se, and the various empirical hypotheses offered about them, have not been investigated in the experimental literature in cognitive science. I say “per se” because, as will be advanced here, there is literature in experimental psychology that can be recruited to examine the cognitive basis of thought experimenting, which, on a hypothesis that Nenad and I introduced independently (Nersessian 1991, 1992; Miščević 1992), also provides insight into how they are executed, promote conceptual reform, and provide novel real-world insights in the absence of new data. Our joint hypothesis was that thought experimenting is a form of mental modeling. I have since extended that hypothesis to incorporate a wider range of empirical investigations (Nersessian 2002, 2008). In Creating Scientific Concepts (Nersessian 2008), I argued that thought experimenting is a 310form of simulative model-based reasoning, where inferences are made through constructing and manipulating models, whether conceptual, physical, or computational (see also Chandrasekharan, Nersessian, and Subramanian 2013; Chandrasekharan and Nersessian 2015). Providing the cognitive basis of this form of reasoning requires a synthesis of research including on various kinds of mental models (logical reasoning, discourse models, situation models), mental imagery, mental spatial simulation, mental animation, and embodied mental representation (perceptual-motor, perception-based).
