In search of the ethogram


January 31, 2019

Psychology is in part, a science of behavior. Mental experience – thoughts and feelings – are also common targets of inquiry.

I recently went looking for a systematic catalogue of behaviors that should, in theory, constitute the substance of what psychologists and other behavioral scientists study. My goal was to take that catalogue as a starting point for thinking about the relationship between information available for perception and the actions (behaviors) that this information informs. The formal term for this sort of catalog is an ‘ethogram’. Here’s a sample template from the Tree of Life project. Unfortunately, I failed in my quest.

There have been attempts to create and openly share animal ethograms like the apparently stalled or inactive site associated with the EthoSource project. The Cognitive Paradigm Ontology and the Cognitive Atlas project attempt to create systematic ontologies (systematic representations of categories and concepts and their relations) of tasks used in experimental psychology. Cognitive psychology tasks clearly require very specific types of behavior from participants. Systematically organizing the nature of these tasks should lead to faster progress in understanding how they relate to one another, and ultimately, how minds and brains actually work. Still, I came up short in finding a comprehensive human ethogram.

In other contexts, I’ve called such a thing the behaviorome, partly to suggest that behavior should have a more central place in research grantmaking priorities. The PLAY project is a type of behaviorome effort. We will be creating detailed definitions for specific behaviors that our infant and adult participants produce in a small set of behavioral domains, but it will not be comprehensive by any means. For now, I’m stuck with trying to assemble my own.

One approach to an ethogram for humans could focus on biologically essential behavioral classes that we share with all other animals: ingestion (food and nutrient seeking), defense (self-protection), and reproduction. Unfortunately, this scheme fail to capture the sets of behaviors studied by the majority of psychologists. Adding in a class of communicative and affiliative behaviors helps. But even this augmented classification describes the functional purposes of behavior, not the behaviors themselves. For example, locomotor behavior, moving toward or away from some target, occurs in food seeking, predator avoidance, mate seeking, and social (affiliative) contexts.

Another way to ask the question is to observe whether physical activities that humans do for pleasure and watch others for entertainment (sports) have some latent structure we might exploit. Here, we assume that if large groups of humans engage in an activity or pay to watch others engage in it, the skills involved must be especially demanding. Indeed, professional sports represent some of humankind’s most skilled (and highly paid) actors. And while I’m still looking for a comprehensive catalog of sport types, this table shows that most of them fall into a small set of categories when focusing on the core actions that define success.

Category Sports
aiming basketball, baseball, football (passing & kicking), hocky, soccer, bowling, tennis, golf, pool/billiards, archery, …
steering slalom skiing, motorsports
performing gymnastics, diving, synchronized swimming, figure skating, …
racing running, motorsports, speed skating, swimming, cycling, paddling
fighting boxing, wrestling, …
feats of strength weightlifting

You may disagree with my categorization, but I hope you’ll agree that there are many similarities in the sorts of actions we find entertaining to watch.

Another approach to an ethogram starts with the actions that computer vision researchers find are important to extract from video. Google’s Atomic Visual Action (AVA) dataset ( involves three classes of movements, person movement, object interaction, person interaction, and has exemplars with 81 specific actions that fit into these categories. Some examples of specific behaviors tagged in the AVA dataset are crawl, dance, jump/leap, lift, open, press, hug, kiss, and listen to.

Maybe some combination of these approaches is as good a start as any. After all, if behavioral biologists don’t seem to agree about what constitutes behavior, it’s not that surprising that psychologists haven’t yet created our own list. We do need a list like this, though, don’t you think?