Generality of the Summation Effect in Human Causal Learning

Abstract

Considerable research has examined the contrasting predictions of the elemental and configural association theories proposed by Rescorla and Wagner (1972) and Pearce (1987), respectively. One simple method to distinguish between these approaches is the summation test, in which the associative strength attributed to a novel compound of two separately trained cues is examined. Under common assumptions, the configural view predicts that the strength of the compound will approximate to the average strength of its components, whereas the elemental approach predicts that the strength of the compound will be greater than the strength of either component. Different studies have produced mixed outcomes. In studies of human causal learning, Collins and Shanks (2006) suggested that the observation of summation is encouraged by training, in which different stimuli are associated with different submaximal outcomes, and by testing, in which the alternative outcomes can be scaled. The reported experiments further pursued this reasoning. In Experiment 1, summation was more substantial when the participants were trained with outcomes identified as submaximal than when trained with simple categorical (presence/absence) outcomes. Experiments 2 and 3 demonstrated that summation can also be obtained with categorical outcomes during training, if the participants are encouraged by instruction or the character of training to rate the separately trained components with submaximal ratings. The results are interpreted in terms of apparent performance constraints in evaluations of the contrasting theoretical predictions concerning summation.

Explaining compound generalization in associative and causal learning through rational principles of dimensional generalization.

“…However, in our model the outcome is produced by latent causes, whereas in the Kalman filter the outcome is directly caused by observable stimuli and the only latent variables are the weights connecting stimuli to outcomes. The generative process we assume for the outcome allows our model to explain the summation effect (e.g., Kehoe et al, 1994; Collins & Shanks, 2006; Rescorla & Coldwell, 1995; Soto et al, 2009; Whitlow & Wagner, 1972), which cannot be explained by latent cause models in which the outcome is assumed to be a binary, non-additive variable (Courville, 2006; Courville, Daw, & Touretzky, 2002). …”

“…This could be the case for some causal learning experiments, in which participants are directly instructed to learn causal relations between stimuli and outcomes. For example, humans learning causal relations between foods and allergy show a summation effect (Collins & Shanks, 2006; Soto et al, 2009). Our model can explain this result only if it is assumed that different foods vary along separable dimensions, but some models (e.g., the Kalman filter) that assume a stimulus-generates-outcome structure can explain summation in this experiment without resorting to assumptions about separability.…”

“…Furthermore, because in Courville et al's (2002) model the outcome is a binary variable, the only thing that can be estimated during compound generalization is probability of the outcome's occurrence. As such, the model cannot explain summation for two 100% reliable stimuli, that is, the observation that sometimes animals respond more to a compound of two stimuli that each reliably predict an outcome than to each stimulus alone (e.g., Kehoe, A. J. Horne, P. S. Horne, & Macrae, 1994; Collins & Shanks, 2006; Rescorla, 1997; Rescorla & Coldwell, 1995; Soto, Vogel, Castillo & Wagner, 2009; Whitlow & Wagner, 1972). As Courville et al (2002) admit, it seems likely that the summation effect is concerned with the estimation of expected outcome magnitude and not just its probability.…”

Three Ways That Non-associative Knowledge May Affect Associative Learning Processes

“…Thus the process that provides a means of generalization is assumed to automatically produce an expectation of the outcome based on some combination of the associative strengths of the cues present. This assumption is based partly on direct evidence of summation in human and animal learning (see for example Myers et al, 2001; Pearce, 2002; Soto et al, 2009; Thorwart et al, 2016) but also on the fact that it is necessary for the associative account of the blocking effect and that the blocking effect is found in diverse and various circumstances and paradigms, indicating that the additivity rule is in fact the default mode by which our learning system operates.…”

Generalidad de la Hipótesis de la Flexibilidad de Procesamiento en el Aprendizaje Humano: Del Condicionamiento al Aprendizaje Predictivo

“…Si se supone una representación altamente sensible al contexto, donde AB tiene muy poco en común con A y B, se esperaría que el nivel de respuesta a AB sea igual o inferior a A y B. Por el contrario, si se supone una representación con poca sensibilidad contextual, se espera que mucho de lo aprendido acerca de A y B se transfiera a AB, obteniendo una respuesta mayor al compuesto que a sus elementos sumados. Los hallazgos con este procedimiento han sido controversiales, obteniéndose tanto sumatoria (Collins & Shanks, 2006;Cornejo, Castillo, Saavedra & Vogel, 2010;Díaz et al, 2009;Rescorla, 1997;Soto, Vogel, Castillo & Wagner, 2009;Whitlow & Wagner, 1972) como su ausencia (Aydin & Pearce, 1995;Rescorla & Coldwell, 1995).…”