Elemental representations of stimuli in associative learning.

Abstract

This paper reviews evidence and theories concerning the nature of stimulus representations in Pavlovian conditioning. It focuses on the elemental approach developed in Stimulus Sampling Theory (Atkinson & Estes, 1963;Bush & Mosteller, 1951b) and extended by McLaren and Mackintosh (2000;2002), and contrasts this with models that that invoke notions of configural representations that uniquely code for different patterns of stimulus inputs (e.g., Pearce, 1987Pearce, , 1994Wagner & Brandon, 2001). The paper then presents a new elemental model that emphasizes interactions between stimulus elements. This model is shown to explain a range of behavioral findings, including those (e.g., negative patterning and biconditional discriminations) traditionally thought beyond the explanatory capabilities of elemental models. Moreover, the model offers a ready explanation for recent findings reported by Rescorla (2000;2001;2002b) concerning the way that stimuli with different conditioning histories acquire associative strength when conditioned in compound.

Response rate and reinforcement rate in Pavlovian conditioning.

“…Rescorla and Wagner (1972) greatly extended the explanatory scope of this simple description by proposing that learning on any trial is proportional to the difference between  and the sum of what has been learned about all CSs present on the current trial (V). This simple error-correction rule, or an operation that is functionally equivalent, has been incorporated into numerous more elaborate models of learning (Desmond & Moore, 1988;Grossberg & Schmajuk, 1989;Harris, 2006;McLaren & Mackintosh, 2000;Pearce, 1994;Sutton & Barto, 1981;Wagner, 1981;Wagner & Brandon, 2001). …”

Symmetrical generalization decrements: Configural stimulus processing in human contingency learning

“…Stimuli with common components will therefore activate the same associations. Modern elemental models, such as the replaced elements model (REM; Wagner, 2003) and Harris's (2006) model, assume flexible nonlinearity, which means that a compound need not activate exactly the same associations as its components. 2 f In the following studies, we compared the application of configural models and of elemental models to two different generalization tasks.…”

Generalization decrements: Further support for flexibility in stimulus processing

“…This indicates that stimulus processing and representation differed between the experiments. As described above, asymmetrical decrements are in line with elemental models (Harris, 2006;Wagner, 2008), whereas the previously observed symmetry is captured only by models based on configural principles (Pearce, 1994). The experiments thereby provide further empirical support for the assumption that the representation of the cues in associative learning is flexible and, instead of contrasting configural and elemental accounts, research needs to shift its focus to the variables controlling the nature of representation and processing.…”

“…Pearce's (1994) configural model predicts symmetrical decrements (i.e., the amount of decrement produced by testing ABX after training with AB will be the same as that produced in testing with EF after training with EFG). On the other hand, both elemental theories (Harris, 2006;Wagner, 2008) predict asymmetrical generalization decrements: Removal of components always results in a larger decrement than adding does. This prediction of asymmetry remains unaffected by the supposed variability in the components' representations.…”

“…On the other hand, there is considerable evidence that humans as well as animals can successfully handle discrimination problems that cannot be solved in such a purely elemental manner. Therefore, more recent elemental models like the replaced elements model (REM; Wagner, 2008) and the model of Harris (2006) assume, first, that a component presented within a compound may not activate exactly the same entities and associations as when presented on its own and, second, that the degree of overlap between these two instances of the component's representation is somewhat variable.…”