The Domjan et al. target article is a valuable summary of a vital field of conditioning and learning. It lists, thumbnails, and organizes classical and recent findings into a useful and familiar structure. Perhaps it is time to consider modern developments in ethology, experimental psychology, and computer science that supersede the traditional structure.

Domjan et al. continue a consistent trend in Pavlovian conditioning, that of accounting for more behaviors while sacrificing specificity of predictions. Despite the sacrifice, their model provides a valuable framework within which social behavioral research may operate. It may also allow ethologists and evolutionary psychologists to pursue questions about which feed-forward systems should produce which behaviors in social settings.

To make possible the integration proposed by Domjan et al., psychologists first need to close the research gap between behavioral ecology and the study of Pavlovian conditioning. I suggest two strategies, namely, to adopt more behavioral ecological approaches to social behavior or to co-opt problems already addressed by behavioral ecologists that are especially well suited to the study of Pavlovian conditioning.

A convincing case is made for the importance of conditioning in social interaction, but more than Pavlovian conditioning is involved: UR (unconditioned response) modification, imprinting, Skinnerian conditioning, and other forms of behavior modification are adduced as Pavlovian. Beyond its value as an icon, control theory is not brought to bear in an informative fashion on these phenomena.

Evolutionary theory supplies a framework for integrative models of social behavior. In addition to those that include conditioning, evolutionary theory is equipped to explain the acquisition of structures designed to enable individuals to learn by observing others, create mental models of the environment, and coordinate social interactions by taking the perspectives of others.

Despite a promising introduction, Domjan et al.'s target article fails to capitalize on the concept of information intrinsic to control theory. The authors limit their application of feed-forward models to simple nondynamic cases. Their applications to social behavior are stimulus-occasioned responses. Agents might as well be dogfood! The notion of “conditioning” is generalized without warrant to explain virtually any acquired predictive capability.

The extent to which Pavlovian feed-forward mechanisms operate in primates is debatable. Monkeys and apes are long-lived, usually gregarious, and intelligent animals reliant on learned behavior. Learning occurs during play, mother-infant interactions, and grooming. We address these situations, and are hesitant to accept Domjan et al.'s reliance on Pavlovian conditioning as a major operant in primates.

Biologists recognize Pavlovian conditioning as a mechanism by which individuals can adaptively modify their social and nonsocial behavior quickly to relevant features of the natural environment. This commentary supports Domjan et al.'s point that psychologists could gain important insights by broadening the range of species and behaviors they study and by continuing to adopt a functional perspective to investigate Pavlovian conditioning and other forms of learning.

Pavlovian feed-forward mechanisms represent an individual perspective that ignores how repeated interactions between the same individuals lead to social relationships. These can determine social cues, coordinated behaviors, asymmetries between partners, and physiological and emotional states associated with social interaction.

The target article provides a useful investigative model for studying social behaviors, but it falls short of establishing a more comprehensive conceptual framework for understanding complex social inter- actions. Social behaviors such as play involve a dynamic and complex interplay between two or more organisms. Even when feed-forward mechanisms are taken into account and the model is anchored to evolutionary theory, the utility of this model is still limited by the conspicuous absence of neurobiological theory and data.

Feed-forward Pavlovian conditioning can serve as a proximate mechanism for the evolution of social behavior. Feed-forward can provide the impetus for animals to associate other individuals' presence, and cooperation with them, with the acquisition of resources, whether or not the animals are genetically related. Other social behaviors such as play and grooming may develop as conditioned stimuli in feed-forward social systems.

While feed-forward mechanisms may be ubiquitous in biological systems that form the substrates of Pavlovian conditioning, the control system proposed by Domjan, Cusato & Villarreal seems too elaborate for Pavlovian conditioning of simple skeletal muscle responses. We discuss here how the known neural substrates of classical eyeblink conditioning can be described in feed-forward terms, but argue that the monitor/comparator part of the system is not necessary and perhaps could even be detrimental to simple, nonsocial forms of Pavlovian conditioning.

Top-down models typically used to explain social behavior involve specific adaptations and higher level cognition. The Pavlovian conditioning model proposed can be extended to explain formation of dominance hierarchies and group structure, can replace a pheromonal model of reproductive suppression, and can be applied to language learning. This bottom-up approach based on general learning principles is a refreshing alternative to top-down models.

Theorists of learning, regulation, and evolution explain behavior using remarkably different concepts because of pressures toward specialization, a focus on testing simple causal theories that underconceptualize the contributions of the organism and its environment, and the absence of a working model capable of surviving in a complex environment. We add suggestions for the development and testing of such a model.

Reactions to the target article included requests for extensions and elaborations of the schema we proposed and discussions of apparent shortcomings of our approach. In general, we welcome suggestions for extension of the schema to additional kinds of social behavior and to forms of learning other than Pavlovian conditioning. Many of the requested elaborations of the schema are consistent with our approach, but some may limit its generality. Many of the apparent shortcomings that commentators discussed do not seem problematic. Our schema encourages a broad view of the behavioral consequences of Pavlovian conditioning – including learned modifications of responding to the unconditioned stimulus. Costs and benefits addressed by our schema are the long-range reproductive consequences of learning – not the immediate reinforcing consequences of particular conditioned responses. Our approach allows the evolution of learning to yield maladaptive behavior and can be extended to characterize dynamic social interactions. We clarify that ours is not a homeostatic model involving ideal set points, and we clarify and defend our application of Pavlovian concepts to the analysis of social play.

Johnson-Laird & Byrne (1991; 1993) present a theory of human deductive reasoning based on the notion of mental models. Unfortunately, the theory is incomplete. The present commentary argues that pragmatic considerations, particularly of the type discussed in Sperber and Wilson (1995), can complement the theory.

Van der Henst argues that the theory of mental models lacks a pragmatic component. He fills the gap with the notion that reasoners draw the most relevant conclusions. We agree, but argue that theories need an element of “nondeterminism.” It is often impossible to predict either what will be most relevant or which particular conclusion an individual will draw.

Shors & Matzel (1997) suggest replacing the question “Is LTP a mechanism of learning?” with “Is LTP a mechanism of arousal and attention?” However, the failure of experiments to verify the LTP-learning hypothesis may arise not because it is untrue, but because in its current guise, it is not properly testable. If so, then the LTP-attention hypothesis is untestable, as well.

The absence of a clear influence of the responses modified by new connections created by LTP on the development of these connections casts doubt on an essential role of LTP in learning and memory formation without any association with reinforcement. The evidence for the involvement of the monoaminergic systems in synaptic potentiation in the cerebral cortex during learning is adduced, and their role in reinforcement system function is discussed.

Long-term potentiation (LTP) is a long-lasting increase in synaptic efficacy that many consider the best candidate currently available for a neural mechanism of memory formation and/or storage in the mammalian brain. In our target article, LTP: What's learning got to do with it?, we concluded that there was insufficient data to warrant such a conclusion. In their commentaries, Jeffery and Zhadin raise a number of important issues that we did not raise, both for and against the hypothesis. Although we agree with a number of these issues, we maintain that there remains insufficient evidence that LTP is a memory mechanism.