Comparative Psychology
The Misuse of Anthropomorphism
in Comparative Psychology
A philosophical analysis examines two assumptions in Dwyer and Burgess’ interpretation of rat causal learning research. Drawing on Fitzpatrick, Keeley, Papineau and Heyes, and Blaisdell, the argument shows that Morgan’s Canon should not function as a methodological criterion and that the evidence supports causal inference in rats.
The Setup
Dwyer and Burgess (2011)
“Available evidence does not licence the conclusion that non human animals display evidence of human like cognitive processes,” including causal reasoning, sensitivity to the absence of stimuli, and the relationship between effort and reward. They argue accounts of animal behaviour in terms of higher order cognitive functions should only be accepted when simpler explanations are unavailable.
What they invoke
Morgan’s Canon, which prefers lower-level associative processes whenever possible. Higher cognitive explanations are only accepted when associative mechanisms have been ruled out.
The standard practice
What Cameron Buckner calls Standard Practice in animal psychology: methodological conservatism justified by the alleged fallacy of anthropomorphism.
Two Load-Bearing Assumptions
The Dwyer and Burgess argument rests on two claims that, once examined, do not survive the objections raised by Fitzpatrick, Keeley, and others.
Assumption 1
Lower explanations should win by default
Animal psychologists should follow Morgan's Canon and favor lower, associative explanations whenever possible. Higher cognitive explanations need to overcome a strong default in the other direction.
Assumption 2
Avoiding anthropomorphism justifies the canon
The rationale for the canon is that it protects research from anthropomorphism, where human-like traits are mistakenly attributed to non-human animals.
Objections to the Canon
Five objections, drawn from Fitzpatrick and Keeley, undermine the use of Morgan’s Canon as a methodological criterion.
Simplicity is not a criterion of truth
FitzpatrickCustance and others concede they have no empirical reason to prefer one explanation to another, and Morgan's Canon is then invoked as the tiebreaker. The simplicity of an explanation is no necessary criterion of its truth, and there is no reason for the canon to trump all other forms of simplicity in every circumstance.
Evolutionary constraints favor generosity
FitzpatrickEvolvability constraints, phylogenetic constraints, brute facts of physiology, and fitness considerations shape what cognitive processes are plausible in a species. If an animal occupies a niche meeting the conditions for a sophisticated process, or its close relatives possess that process, attribution should lean generous rather than conservative.
Clever Hans does not support the canon
HistoricalThe Clever Hans case had nothing to do with ascribing human-like abilities to a non-human animal. Its issue was reliance on anecdotal evidence instead of controlled experiments. Morgan's Canon, as it existed prior to the case, did not prevent it. Avoiding anthropomorphism does not justify methodological conservatism.
Replacing one bias with another
KeeleyWithout sufficient independent justification, accepting cognitively unsophisticated explanations that are not well supported by the data is just as much an error as crude anthropomorphism. Shared traits between humans and non-humans are not inherently anthropomorphic; they can be understood as mammalomorphic or primatomorphic.
The canon limits research advancement
FitzpatrickWithout willingness to entertain hypotheses involving more sophisticated cognitive processes, subtle and surprising capacities such as the honeybee dance language go undiscovered. The Scaling Up Strategy, which always starts from the least cognitively sophisticated explanation, would have prevented important findings.
The Blaisdell Evidence
Blaisdell (2006) tested whether rats could differentiate between observing an effect and generating it through intervention. The findings remain compelling once the canon is set aside.
What Blaisdell found
Rats made correct inferences for instrumental actions on the basis of purely observational learning. They correctly differentiated between common cause models, causal chains, and direct causal links.
The results contradict the view that causal learning in rats is solely driven by associative mechanisms and are consistent with causal Bayes net theories.
What Dwyer and Burgess respond
Response competition between lever pressing and food magazine responding might account for the observed pattern. Replication of the theoretically vital interaction failed.
They know of no existing experiment where their response competition hypothesis can be directly tested.
Why the response does not invalidate the finding
Response competition does not directly invalidate the claim that rats might be capable of causal reasoning.
Differences in replication conditions could contribute to the divergent findings.
The existence of an alternative explanation does not mean the original conclusion is wrong; both can be evaluated on their merits.
Defining rational processes as the capacity to derive predictions for interventions after observational learning gives a core competency not reducible to associative learning.
Conclusion and Takeaways
Neutral evidentialism
Higher and lower explanations should be evaluated equally. Asymmetric burden of proof produces methodological bias rather than clarity.
Causal inference need not be human-like
Tying rational processes to a human standard introduces a fallacy. Inference in rats can be inference without being human inference.
The evidence is compelling
Alternative processes could explain the behavior, but the Blaisdell study provides strong, well-supported reasons to think rats are capable of causal inference.
Key Takeaway
Morgan’s Canon should not be used as a criterion for evaluating evidence, nor should it override empirical findings. Under neutral evidentialism, the Blaisdell study provides compelling support that rats are capable of causal inference.
