Evaluative Conditioning

I am interested in the acquisition of likes and dislikes through classical conditioning procedures. Specifically, past work has shown that if you present a subjectively neutral stimulus alongside either a liked or disliked stimulus, then that neutral stimulus will acquire the value of the stimulus with which it was paired. As such, Evaluative conditioning (EC) can be defined as the transfer of affect from one stimulus to another by contiguously pairing the two stimuli in a classical conditioning paradigm. Usually, an affectively neutral stimulus (a conditioned stimulus, CS) is paired with either a liked or disliked stimulus (the unconditioned stimulus, UCS), resulting in the CS acquiring the same valence as the UCS with which it was paired (Levey and Martin, 1975). Although EC is a paradigmatic example of classical conditioning, it is unlike conventional autonomic conditioning in humans because (1) it appears to occur without subjects possessing awareness of the contingencies involved (Baeyens, Eelen & Van den Bergh, 1990a) and (2) seems to be resistant to extinction (Baeyens, Crombez, Van den Bergh & Eelen, 1988). This appears to make EC qualitatively distinct from more traditional human autonomic conditioning (see Davey, 1994).

Methodological Issues: A typical visual EC paradigm uses pictures (such as human faces) as stimuli in a three-stage procedure. In the first stage, subjects rate the pictures along a 21 point scale ranging from -100 (dislike) through zero (neutral) to +100 (like). At the end of this stage, the experimenter selects the three most liked stimuli, the three most disliked stimuli and three neutral stimuli (stimuli with ratings between -10 and +10) to act as the UCSs, and a further 9 neutrally-rated faces to act as the CSs. This resulted in nine CS-UCS pairings: 3 Neutral-Like (N-L); 3 Neutral-Dislike (N-D); and 3 Neutral-Neutral (N-N). In the second stage, the nine CS-UCS pairs are presented several times in semi-randomised order. Finally, subjects re-rate all of the CSs and UCSs using the same like-dislike scale as in the first stage. The standard findings in such an experiment are that, after conditioning, CSs paired with liked UCSs are rated more positively whereas CSs paired with disliked UCSs are rated more negatively.

An important concern regarding this paradigm is the absence of between-subject control conditions. The main contention stems from the observation by Shanks and Dickinson (1990) that in the typical EC paradigm CSs and UCSs are selected by subjects and, in some studies, paired on the basis of perceptual similarity. In short, Shanks and Dickinson suggest that this CS-UCS assignment procedure may interact with the effects of stimulus exposure to create the illusion of conditioning. They argue that true conditioning effects can be isolated only if the pairing of a particular CS with a particular UCS is counterbalanced across subjects. If this criterion is not met then it is possible that any conditioning-like effects could be due to nonassociative factors arising from the paradigm. In visual EC paradigms, CS-UCS combinations are typically not balanced because they are selected on the basis of a subject's ratings of the stimulus set. Therefore, the influence of nonassociative factors (such as stimulus similarity) cannot be ruled out. Field and Davey (1998) suggest that to demonstrate that EC effects are the result of specific CS-UCS associations, rather than the result of biases in the way in which stimuli are selected and paired together, a necessary condition must be fulfilled: experimental effects should not be shown in a condition where all associations are eliminated. If conditioning-like effects are found in such a condition, then experimental results must be due to nonassociative factors.

There has been considerable debate over what constitutes an adequate between-group control (See Davey, 1994; Baeyens & De Houwer, 1995). Field (1996, 1997) has proposed the Block/sub-block control (BSB control), which eliminates all CS-UCS associations, holds presentational factors constant and controls for effects resulting from stimulus selection procedures. Field (1997) also recommends the use of a no-treatment condition in which subjects are exposed to no stimulus presentations. A no-treatment control provides a situation in which all associations are eliminated and so it acts as a good gauge of subjects' expectancies when they come to re-rate the stimuli after conditioning. In addition, it controls for the possibility that effects are due to the stimulus selection procedure (because subjects are not exposed to any CSs or UCSs during the conditioning stage of the experiment). However, unlike the BSB control, it does not control for the effects of exposure. Although the use of the BSB control alone does allow conclusions to be drawn about the associative nature of conditioning, it does not allow conclusions to be drawn about the role of presentation. By using both a no-treatment condition, and a BSB control condition, effects due to exposure to the stimuli can be dissociated from effects due directly to the stimulus selection procedure and subjects' experimental expectancies.

Interestingly, when these control conditions are employed, the same effects are usually observed in both the experimental and control conditions indicating that the so-called conditioning effects found in EC research are not due to associative learning (e.g. Shanks & Dickinson, 1990; Field & Davey, 1997, 1999). One such nonassociative factor appears to be similarity between the CS and UCS, which can facilitate EC-like effects (Field & Davey, 1999). Interestingly, stimulus similarity is a known confound of other learning processes such as classification learning (Wattenmaker, Nakamura & Medin, 1988) and incidental learning (Cock, Berry & Gaffan, 1994). Current research in our EC group is looking into whether true conditioning effects can be established in a paradigm in which these nonassoiative artefacts are eliminated.

Contingency Awareness: There is also considerable debate over the role of contingency awareness in EC and whether early findings that conditioning can be established without contingency awareness are merely an artefact of the paradigm used (see Field & Davey, 1998 for a review). Field (2000a) has raised several concerns over the way in which awareness is measured and Field (2000b, 2001) has made a distinction between the possible role of awareness in ecologically-relevant conditioning episodes compared to non-ecologically-relevant episodes. Current work is exploring the role of awareness when counterbalanced CS-UCS stimulus allocations are employed , and we have some evidence that attention may moderate the relationship between attention, awareness and learning (e.g. Field & Moore, 2005).

Summary: My main interests are now:

1. Whether genuine evaluative conditioning effects can be elicited using ecologically-relevant stimuli,
2. The distinction between ecologically-relevant and -irrelevant conditioning episodes
3. The role of conceptual similarity between the CS and UCS
4. The role of conscious awareness of contingencies in EC
5. Whether evaluative conditioning can be found in children
6. Whether EC can be found using physiological Measures and ERP (this is what Carina Ugland is doing her D.Phil. on)

Grants:

• 2000-2001: £37,089 to Andy Field from Unilever

Students

• Carina Ugland (D.Phil. commenced 2006)

Current Collaborators

• Dr. Ben Dyson (University of Sussex)