The selectionist account of behavior is actually a focused discussion of operant selection. To this end, the authors essentially exclude stimuli from their analysis. This exclusion is inconsistent with the importance placed on environmental interaction in their general account. Further, this exclusion limits the generality of their account by missing important sources of stimulus-elicited behavior (e.g., classical conditioning).

Many analogies exist between the process of evolution by natural selection and of learning by reinforcement and punishment. A full extension of the evolutionary analogy to learning to include analogues of the fitness, genotype, development, environmental influences, and phenotype concepts makes possible a single theory of the learning process able to encompass all of the elementary procedures known to yield learning.

Hull et al. rightly point out the special character of selection as a causal mode, but ironically they seem to force selection back into traditional causal modes by decomposing it into replication, variation, and environmental interaction. Many processes are selective, and a taxonomy of a broad range of kinds of selection may be preferable to narrowing the applicability of the term.

In their account of learning and behavior, the authors define an interactor as emitted behavior that operates on the environment, which excludes Pavlovian learning. A unified neural-network account of the operant-Pavlovian dichotomy favors interpreting neurons as interactors and synaptic efficacies as replicators. The latter interpretation implies that single-synapse change is inherently Lamarckian.

The authors note that theoretical revision is likely to be required to refine and clarify a general theory of selection so that it is capable of accounting for selection processes across different subject matters. Some aspects of conceptual language from the biological realm work across all three domains. However, terminological revision may be required where concepts of “replication” and “information” pose particular problems for the general theory as that is applied to operant research.

Hull et al. see responses and properties of responses as units of selection in behavioral change. However, this perspective cannot account for goal-directed behavior in which organisms employ variable means to reliably attain intended consequences. An alternative perspective is offered in which the intended consequences (goals) of behavior serve as the units of selection in behavior change.

Selection in behavior analysis fits the criteria of replication, variation and interaction proposed by the authors except for information under replication. If information requires physical structure, behavior analysis does not fit that model because functional analysis may provide parallels between behavior, neurology, and biochemistry but not sequencing. The three sciences are not unified by the model but another is available.

A profound problem in viewing operant learning as selection appears to be the identification of replicators. Given the lack of consensus on what constitutes the appropriate unit of analysis for behavior, there may be multiple levels at which the metaphor of selection may be usefully applied. A final difficulty: The elements of selection in the evolution of species are objects. In behavior, they are events.

Hull et al. argue that information and replication are both essential ingredients in any selection process. But both information and replication are found in only some selection processes, and should not be included in abstract descriptions of selection intended to help researchers discover and describe selection processes in new domains.

This commentary takes issue with the idea that replication is a “fundamental element” in natural selection. Such an assumption is based on a traditional, mechanistic view of evolution. A symbiogenetic theory of evolution offers an alternative to traditional theories, emphasizing reproduction and qualitative development rather than replication and quantitative development. The issues raised by the symbiogenetic alternative may be extended to discussions of behavioral development.

Operant conditioning is not a selection process. According to Hull et al., selection processes require entities that reproduce to form lineages. However, since operant behaviors do not reproduce, operant conditioning is not a selection process.

Natural selection in the sense of Darwin always means physical propagation (positive case) or disappearance (negative case) of living organisms due to differential reproduction. If one concentrates on this simple materialist principle, one arrives at a much better method of discerning true selection processes from largely nonrandom processes of internal rearrangement (somatic mutations) and reorganisation (operant learning).

Darwin's theory of natural selection is as applicable to the analysis of the behavior of organisms as it is to their origins. Skinner's theoretical writings have guided operant psychologists in this area. The behavioral account of selection by Donahoe and Palmer (1994) is positively compared to the points on operant selection made by Hull et al. The “general account of selection” was found to be useful.

Comparing and sharing selectionist terms and concepts from disparate domains can aid understanding in each domain. But constraints of interpretive language will make this difficult – such as the bipolar constraint of interpretive language when addressed to intrinsically tripolar phenomena. Hull et al. acknowledge that some key terms in their account remain problematic; the term, “information,” probably needs to be replaced.

The generic account of selection proposed by Hull et al. readily fits operant learning where, by comparison with natural selection, the process is well understood but little is known about the mechanism. Objections within psychology, that operant learning ignores internal processes, fail to recognise the general significance of behaviour-environment interactions. Variation within operant response classes requires further investigation.

Seeking parallels among disciplines can have both risks and benefits. Finding parallels may be a vacuous exercise in categorization, generating no new insights. And pointing to analogous functions may cause us to treat them as homologous. Hull et al. have provided a basis for the generation of insights in different selectionist areas, without confusing analogy with homology.

Commentaries surrounding Skinner (1984) were re-examined and applied to Hull et al. Hull et al. were found to address many of these concerns by paying attention to neuroscience, by providing some discussion of the origins of behavior, and by forwarding a deterministic account that may prove as revolutionary as that of Copernicus and Darwin.

Selectionist models date back to Empedocles in Ancient Greece. The novelty of Darwinian selection is that it is able to produce adaptively valuable things without being sensitive to adaptive value. Darwin achieved this result by a restriction of environmental feedback to the replicative process. Immune system selection definitely does not respect this restriction, and it is doubtful whether operant learning does.

The idea that similar selective processes operate in gene-based evolution, immunology, and operant psychology provides an intuitively appealing metaphor. This idea also isolates questions that operant psychologists should ask and makes some empirical predictions. However, the idea currently lacks the detail needed to precisely separate it from some plausible alternatives. This sort of thinking is the kind that operant psychologists should do if operant theorizing is to survive the competition among ideas.

Behavioral theorists such as B. F. Skinner have argued that the thesis of selection by consequences applies to behavior just as much as to morphology. This commentary specifically examines certain respects in which the thesis of “selection by consequences” applies to the development of ontogenic operant behavior.