Pieces of Mind: The Proper Domain of Psychological Predicates by Carrie Figdor (review)

The objective of Pieces of Mind is to defend the literalist claim that the best interpretation of the extension of psychological predicates to other domains is a literal one. The literal interpretation justifies the use of terminology such as plants “decide,” bacteria communicate “linguistically,” and neurons “prefer.” A literal extension of psychological predicates would provide a unified account across a wide range of new discoveries in the biological sciences.

In chapter 2 on qualitative analogy the author discusses several cases in which psychological predicates are extended to nonhuman domains. One case is the signaling capacity of plants, which have many components found in the animal neural system including voltage-gated channels and hormones that play a qualitatively similar role to neural chemicals found in animals. The result, according to the author, is that analogies between plants and animals can now be conceptualized within an information-processing framework that launched cognitive science. Another case is communication among bacteria. Communication allows colonies to develop collective memory, generate common knowledge, organize themselves, and learn from experience.

In chapter 3 on quantitative analogy Figdor identifies similarity relations based on mathematical models and equations. She extends the Lotka-Volterra model of population dynamics to the decision-making of fruit flies. She next extends a drift-diffusion model of human decision making to the anticipating of neurons. The extent to which such quantitative analogies reveal semantic similarities is a complex issue that is advanced by the author’s arguments. For a different perspective see my “Constraints on the Abstraction of Solutions” (1989).

Three chapters are devoted to defending the literal view against semantic alternatives that the author labels the nonsense view, the metaphor view, and the technical view. The metaphor view is the most applicable to cognitive science because of the many studies that have investigated it. Studies in cognitive science typically involve relationships among terms, which Figdor argues is consistent with literalism. Consider the statement “resonator neurons prefer inputs having frequencies that resonate with the frequency of other oscillations.” Use of the term “prefer” is not a similarity mapping between neurons and humans but a mapping between what neurons do and what humans do.

I found myself in agreement with the many examples that Figdor uses to support the literal interpretation of psychological terms. But she has selected the low-hanging fruit—the clear cases—to support her position. In my collaboration with ontologists to build cognitive ontologies we have had to make many decisions regarding which cognitive terms generalize across different domains and which terms do not. Our efforts begin with definitions, and here is where I have a problem with Figdor’s analysis. In her concluding summary she states, “Literalism agrees that psychological concepts are anchored in human experience and allows that we can still gain insight from phenomenology but these concepts were never defined within a closed formal system.” The extension of psychological terms should begin with their definitions, but it is difficult to find definitions in Pieces of Mind. Definitions are where the rubber hits the road.

One challenge is that words often have multiple definitions. In our article “A Framework for Constructing Cognition Ontologies Using WordNet, FrameNet, and SUMO” (2015), Adam Pease and I began by deciding which definitions of a word are relevant for cognitive theories. For instance, WordNet has six definitions of “attention,” but we selected only two as theoretically important: (1) attention is the faculty or power of mental concentration, and (2) attention is the process whereby a person concentrates on some features of the environment to the (relative) exclusion of others.

A second challenge is that tools in the information sciences may have different definitions of the same word. Fortunately, some tools have coordinated definitions, such as for “seeing” and “looking” in WordNet, FrameNet, and the Suggested Upper Merged Ontology (SUMO). Each defines seeing as perceiving by sight and looking as seeing with attention. They all make a corresponding distinction between...