In this blog post, we will examine Frank Keil’s virtual animal transformation experiment to explore how appearance and underlying essence create differences in category judgments. This will help us easily understand the cognitive principles behind how we classify objects.
How do people determine what a new object is when they encounter it? This directly relates to the problem of category judgment—identifying which category the object belongs to. This category judgment process has primarily been explained by two theories: the similarity-based approach and the description-based approach.
First, the similarity-based approach posits that the category judgment of a new object is based on perceptual similarity between that object and mental representations stored in memory. This approach is further divided into the prototype model and the exemplar model, depending on the form of the mental representations used in category judgment. The prototype model utilizes an abstract construct, a single prototype, formed by averaging the common attributes shared by instances belonging to a category. In contrast, the exemplar model employs specific instances encountered in the past, stored as mental representations, as exemplars. The prototypicality effect, where typical instances are judged much faster than atypical ones in category judgment, aligns well with the prototype model’s explanation. Conversely, the phenomenon where typicality varies depending on context is much more appropriately explained by the exemplar model, which utilizes accumulated exemplars. However, this similarity-based approach has the limitation of failing to provide a principle for selecting which attribute among multiple perceptual attributes should be chosen as the criterion for category judgment.
In contrast, the explanation-based approach posits that people connect instances to specific explanatory structures based on implicit theories, rules, and causal relationships concerning categories. According to this approach, category judgments are not merely comparisons between mental representations and instances; they are based on the underlying essence that unites instances into a single category. From this perspective, if the similarity-based approach is correct, similarity judgments comparing the perceptual similarity between a specific category and an instance, and the subsequent category judgments based on this, should naturally align. Conversely, if the explanation-based approach is correct, similarity judgments and category judgments need not necessarily align. Of course, in actual situations, the underlying essence often determines perceptual attributes, making the two judgment processes appear identical.
Frank Keil, who supported the explanatory approach, devised an experiment using the transformation of a fictional animal to prove the hypothesis that similarity judgments and category judgments are not the same process. In this experiment, subjects were asked to read a text describing the fictional animal undergoing physical transformations, and then separately evaluate how similar the animal was to certain categories (similarity judgment) and which category it actually belonged to (category judgment).
The text used in the experiment consisted of two parts. The first part was designed so that participants could easily judge the fictional animal as belonging to the category of birds. The second part presented information that, for a specific reason, the fictional animal had undergone a physical transformation resembling an insect. When constructing this second part, researchers divided the text into two types based on whether the cause of the physical change was accidental environmental conditions or a natural maturation process, like a tadpole turning into a frog. Accordingly, text “A,” presenting the former case, stated: “The animal Solp had two legs and wings with feathers. … However, after exposure to chemical waste, Solph developed six legs and transparent membrane-like wings, but later gave birth to offspring resembling the original Solph.“ The text ”B,“ presenting the latter case, stated: ”Dun is called Solph in its youth. Solph had two legs and feathered wings. …… After several months, Solph became Dun, and Dun developed six legs and transparent membrane-like wings.”
Frank Keil added another condition here. Participants were divided into four groups: a control group that read only the first part of each text, and an experimental group that read both parts. After reading, participants responded to the questions “To which is the solp more similar, a bird or an insect?” and “Where does the solp belong, among birds or insects?” by selecting a point on a scale where birds scored 10 points and insects scored 1 point.
The results showed that the control groups for both text “A” and text ‘B’ assigned an average score of 9.5 for both similarity judgment and category judgment. However, the experimental group that read text “A” assigned an average score of 3.8 for similarity judgment and an average score of 6.5 for category judgment, revealing significantly different judgments. The experimental group that read text “B” similarly assigned an average of 7.6 points for similarity judgments and 5.2 points for category judgments, confirming the same pattern. These experimental results clearly demonstrate that category judgments are more influenced by changes in underlying essence than by changes in appearance, whereas similarity judgments are more affected by changes in appearance than by changes in underlying essence.
Frank Keil’s research thus demonstrates that similarity judgments and category judgments, while appearing similar on the surface, can actually operate according to distinct principles. This provides a significant theoretical foundation for categorization research in contemporary cognitive psychology and cognitive science, highlighting the necessity of addressing not only perceptual information but also causal explanations, underlying essence, and conceptual knowledge.
