Everyone knows what behavior is but they just don't agree on it

Abstract

Studying "behavior" lies at the heart of many disciplines. Nevertheless, academics rarely provide an explicit definition of what "behavior" actually is. What range of definitions do people use, and how does that vary across disciplines? To answer these questions, we have developed a survey to probe what constitutes "behavior." We find that academics adopt different definitions of behavior according to their academic discipline, animal model that they work with, and level of academic seniority. Using hierarchical clustering, we identify at least six distinct types of "behavior" which are used in seven distinct operational archetypes of "behavior." Individual respondents have clear consistent definitions of behavior, but these definitions are not consistent across the population. Our study is a call for academics to clarify what they mean by "behavior" wherever they study it, with the hope that this will foster interdisciplinary studies that will improve our understanding of behavioral phenomena.

Keywords: Biological sciences; Ecology; Neuroscience; Zoology.

Graphical abstract

Figure 1

Definitions of behavior are varied but internally consistent (A) The five questions with the most disagreement between respondents as defined by the response entropy (n = 455). (B) Responses are predictable with a regression model. Gray dots represent the performance at predicting answers on held-out data using 5-fold cross-validation, black diamond is mean. Mean prediction accuracy across responses is 63%. The dashed line represents chance performance. (C–F) We used a factor analysis (multiple correspondence analysis (MCA)) to reveal how groups responded on similar questions. Plotting responses on the two largest factors, colored by (C) scientific field, (D) neuroscience specialty, (E) model organism used, and (F) academic seniority. All data shown are the mean, and error bars are SEM.

Figure 2

Survey responses reveal clustered definitions of behavior Hierarchical clustering on the 455 responses reveals six clusters in the questions and seven clusters in the responses. The six categories of questions, labeled 1–6, are unique definitions of behavior, and the seven categories of responses, labeled A–G, are behavioral archetypes. See Figures 3 and 4.

Figure 3

Behavioral definitions revealed by clustering (A–F) Example questions for each type of behavior (left). Illustration representing that type of behavior (right). These definitions are (A) reflex, (B) actions are behavior, (C) we must understand the mind of an animal to identify its behavior, (D) motor or sensorimotor, (E) non-animal behaviors, (F) learning and memory/cognition.

Figure 4

Validation and consistency of behavioral archetypes (A) The mean “Yes” response in each category and (B) its relative (log-valued) difference from the overall average allow us to understand what the definitions are composed of.

Figure 5

No one cares about behaviorism and cognitivism Questions were labeled “behaviorist,” “cognitivist,” or “neither.” Responses were scored to either totally agree (+1) or totally disagree (−1) with each category. Respondents did not agree with behaviorism or cognitivism, but used definitions that were a mix of the two. Circles represent mean and error bars are +/− SEM.

Figure 6

Validation and consistency of behavioral archetypes (A) Behavioral archetypes are built out of the responses to different definitions. To illustrate the archetypes, we use the relative responses (Figures 4, S5, and S6). For each archetype, we color the definitions by whether they are required (pale pink), excluded (brown), included (purple), or sometimes included (gray).

Figure 7

Different groups have distinct distribution of definitions (A) Academic fields show different propensities to use each category of “behavior.” This is also true of (B) neuroscience specialties, (C) model organism used, and (D) academic seniority. See STAR Methods for definitions of each field.