Notes on constructing Ontologies from Ethograms
On the surface, ethograms serve to describe behavior. However,
in the process of describing, they define terms, at least within the limited
context of one observer during one study. Therefore, as terms label
descriptions, so ethograms provide an set of term-definition pairings,
and thus can be viewed as an ontology of the behavior described.
Construction Overview
The construction approach I have taken is:
-
Construct a list of the behaviors described in the ethogram. A second
list of things that were not behaviors is also started at this point, to
be added to as the process proceeds.
-
Determine the part-of and sequence relations that hold among the behaviors
-
Place the behaviors into a type hierarchy. For this ontology, I used
complexity of the behavior to divide actions among three levels
-
scripts - most complex sequences of behaviors, vary with circumstance,
may be interrupted, terminated or completed fortuitously
-
action patterns - a simpler action or sequence, sometimes intended to be
species characteristic
-
description units - the smallest descriptive unit of action described in
the ethogram.
-
Behaviors are then classified among multiple dimensions, yielding a lattice.
A common error in traditional ethograms is to describe behavior only in
terms of its function (e.g., the crow ate the egg), rather than providing
details of what actions constituted the eating "script." Using an
ontology allows description and multiple categorizations (e.g., body parts
used, function, sensory modality, ontogenetic stage etc.) The upper-level
I constructed explicitly represents Tinbergen's four causes as subclasses
of explanation.
-
Add concepts for the other "things" - the animals themselves, body parts,
the substrate and medium where the behavior occurs, etc.
-
Add slots (relations) and type restrictions for associating behavior actions
with the other concepts in the ontology
-
Leave individuals (instances) for individual animals, events, etc.
A Note on Temporal Sequencing
Consider an apparently identical action appearing as part of two different
sequences. Should different concept nodes be created for the action
in each sequence in which it occurs? I have taken the affirmative
answer to this question in this ontology, as I consider it the more conservative,
safer choice, though it arguably introduces a distortion of the original
author's intent in the ethogram.
All ontologies on this site and content of this page are copyright 2001
Peter E. Midford (pmidford@u.arizona.edu).
Last update of this page: 21 Dec 2001
