John W. Donahoe

Essentialism and Selectionism in Cognitive Science and Behavior Analysis

Contingencies of selection, be they phylogenetic or ontogenetic, merely set boundaries on units; they do not provide blueprints. Thus, variability is fundamental to all products of selection. Skinner, by characterizing the units of analysis in behavior as generic in nature, established his science squarely within the selectionist paradigm, thereby avoiding the tendency, common throughout psychology, to slip into essentialist analyses. The distinction between essentialism and selectionism is refined in this article, and prominent examples of essentialism in linguistics, theories of memory, theories of representation, associationism, and even in behavior analysis are identified. Recent trends in cognitive science--specifically, research on adaptive networks--is amenable to a selectionist interpretation, suggesting the possibility of future fruitful interactions with behavior analysis.

Attenuation of blocking by a change in US locus

A classical conditioning procedure was instituted in which the locus of the US was changed coincidentally with the onset of the compound conditioning phase of the Kamin blocking design. The nictitating membrane of the rabbit served as the conditioning preparation. Three groups of nine rabbits each were employed: a conventional blocking group in which the US was unchanged during the simple and compound conditioning phases, a control group that received only the compound conditioning phase, and an experimental group given both simple and compound conditioning but in which the US was administered to the contralateral paraorbital region during the compound conditioning phase. Postconditioning presentations of the elements of the compound CS indicated that, relative to the control group, responding during the new element was markedly reduced in the conventional blocking group but was strongly conditioned in the group in which the US locus was changed. Implications of these findings for theoretical analyses of conditioning are discussed.

Pavlovian Conditioning: The CS-UR Relation.

A new Pavlovian procedure used fluid-elicited throat-movement responses of the pigeon (N=66) to study the effects on conditioning of the temporal relation of the conditioned stimulus (CS) to the unconditioned stimulus-unconditioned response (US-UR). Because the throat-movement response has a substantial latency and duration, the relation of the CS to the US and UR could be independently evaluated. Four experiments indicated that, operationally, the relation of the CS to the UR--not to the US--is critical for conditioning in this preparation. The conventional emphasis on CS-US relations is based on procedures that confound the occurrence of the US with the UR and that foster generalization decrement between training and testing. The authors indicate how several conditioning phenomena may be reinterpreted in terms of CS-UR relations.

Behavior analysis and neuroscience

Darwin proposed a powerful functional principle-natural selection-to interpret phylogenetic diversity and complexity. Nevertheless, some 70 years elapsed before even biologists embraced his account. The triumph of natural selection required two additional factors: (a) biological mechanisms that implemented the functional principle (i.e. genetics); and (b) quantitative procedures that traced its cumulative effects (i.e. population genetics). Thorndike and, later, Skinner proposed a functional principle-selection by reinforcement-that interpreted ontogenetic diversity and complexity. This principle has been substantially refined by subsequent experimental and theoretical work and now provides an equally powerful functional account. However, a purely functional principle has once again not persuaded most scientists (apart from the behavior-analytic minority) that complex behavior can be understood by a selection principle. If the history of ontogeny recapitulates the history of phylogeny, the biological mechanisms that implement selection by reinforcement must be discovered and quantitative techniques that trace its effects must be devised. In short, the triumph of reinforcement may await the integration of behavior analysis with the neurosciences and the development of coordinated quantitative procedures. This article identifies specific interrelations with neuroscience that form the basis for a behavior-analytically faithful interpretation of reinforcement using neural-network techniques. Implications for the architecture of networks and for the learning algorithm are emphasized.

CHAPTER 18. SELECTION NETWORKS: SIMULATION OF PLASTICITY THROUGH REINFORCEMENT LEARNING

ABSTRACT Evolution through natural selection has addressed the problem of modifying synapses throughout large networks of neurons by exploiting diffusely projecting neuromodulatory systems. When pre- and postsynaptic neurons are coactive, synaptic efficacies increase or decrease dependent upon whether the neuromodulator dopamine is simultaneously present or absent. Salient characteristics of this process can be simulated with selection networks, artificial neural networks whose architecture instantiates a processing system whose connection weights are modified by a scalar reinforcing signal. This arrangement resolves both the temporal paradox and the binding paradox, twin challenges to any attempt to interpret complex behavior by means of neural networks. Further, by exploiting an emergent property of selection networks—acquired reinforcement—critical aspects of imagining, thinking, and language acquisition can also be interpreted.

Effects of delayed and trace components of a compound CS on conditioned suppression and heart rate

Following 150 classical conditioning trials with a compound CS paired with shock, barpressing and heart-rate responses to the first, (Si1)—, and second, (S2)—, occurring components and to the compound (S1, 2) were assessed using a conditioned suppression procedure. The decrease in barpressing and heart rate to S1 was greater than to S2 in the group of rats (N = 12) in which S1 and S2 overlapped during conditioning (S1 delay group). In the S1 -trace group (N = 12), in which S1 and S2 were presented successively, S2 was a more effective suppressor than S1. The presence of cardiac acceleration during conditioning but deceleration during testing was attributed to the effects of barpressing activity on heart rate.

Learning-through-Teaching: Knowledge Changes in Undergraduate Teaching Assistants.

experiments. In conducting the literature survey, for example, the students were not given a set of references, norwas it apparent what topics should be reviewed in the various abstracts. Rather, the students were required to gather information from a variety of areas In presenting the background for their experiments. In terms of methodology, the student was not given a design and procedure. Rather, they were required to improve and expand an existing methodology with the intent of conducting a more valid test of the products. In several instances this technique produced innovative (especia l ly fo r the beg inn ing student) methodologies and acquainted the student with the problems encountered in original research. Inasmuch as the results of the studies ranged from complete affirmation (White Cloud Study) to direct opposition (Pepsi and Coke taste test) of the advertisers claim, the data suggest that the students were not simply making a series of Type I or II errors, but had actually devised experiments which had the sensitivity to detect differences between products.

Pavlovian conditioning: Pigeon nictitating membrane

A new Pavlovian conditioning preparation was developed using the nictitating membrane of the restrained pigeon. Either visual or auditory stimuli served as conditioned stimuli (CSs) with an unconditioned stimulus (US) of a puff of air to the cornea. Movement of the nictitating membrane constituted the conditioned and unconditioned responses (CR and UR). Conditioning was studied with the Kamin blocking procedure. In agreement with findings from other conditioning preparations, responding to the redundant stimulus was attenuated relative to a stimulus that received the same number of CS-US pairings in a compound-conditioning procedure. Although response attenuation occurred, substantial individual variation was observed within the blocking procedure, a finding with some precedent in the experimental literature. Theoretical analysis and neural-network simulations indicate that inter-subject variation in response attenuation may result from differences in the extent to which contextual stimuli contribute to the functional CS.

Origins of the Molar-Molecular Divide

The following is necessarily a brief account of what led some in the behavioranalytic community to become enamored of the molar approach to theory, an approach that is decried in Catania’s prescient chapter. Until the 1960s, the only generally agreed-upon requirement for conditioning was contiguity of a reinforcer with a stimulus in the Pavlovian procedure or with a response in the Thorndikian procedure (or with both in the discriminated operant procedure). As but one example, “To say that a reinforcement is contingent upon a response may mean nothing more than that it follows the response. ... conditioning takes place presumably because of the temporal relation only, expressed in terms of the order and proximity of response and reinforcement” (Skinner, 1948, p. 168). Several supplemental requirements were proposed from time-to-time—such as drive reduction (Hull, 1943)—but these are not considered here. Beginning in the 1960s, findings emerged that challenged the contiguity requirement. Using the Pavlovian procedure, Leon Kamin (1968, 1969) demonstrated that a stimulus presented in a contiguous relation with a known reinforcer would not acquire control over the response if that stimulus was accompanied by another stimulus that already controlled responding as the result of prior conditioning with the same reinforcer. Prior conditioning to one stimulus blocked conditioning to the new stimulus. This finding was soon confirmed with an appropriate operant procedure (vom Saal & Jenkins, 1970). Something beyond contiguity seemed to be required. The initial explanation of the apparent failure of contiguity was provided by Robert Rescorla (1967). He proposed “a new conception of Pavlovian conditioning postulating that the contingency between CS and US, rather than the pairing of CS and US, is the important event in conditioning” (p. 71). Kamin’s new stimulus did not change the preexisting contingency of the prior stimulus with the reinforcer. Note that Rescorla’s view of contingency distinguished it from contiguity. Contiguity denoted a purely temporal relation of a stimulus (or a response) with a reinforcer, a relation that could be met on a single co-occurrence of the relevant events. Contingency required multiple observations to determine the relative frequency of one event in the presence and absence of another. A close reading of Skinner indicates that he invariably used the term contingency to denote simply a co-occurrence of events, not an “if-then” relation (Donahoe, 2006). Origins of the Molar-Molecular Divide

Conditioning preparation for the nictitating membrane of the pigeon

A procedure is described for restraining the pigeon while recording movement of the nictitating membrane. The preparation provides a means of studying classical conditioning that combines the rich sensory capacities of the pigeon with the control achieved in nictitating membrane conditioning of the rabbit.