Evaluative conditioning depends on higher order encoding processes

Abstract

Evaluative conditioning (EC) is commonly conceived as stimulus-driven associative learning. Here, we show that internally generated encoding activities mediate EC effects: Neutral conditioned stimuli (CS) faces were paired with positive and negative unconditioned stimuli (US) faces. Depending on the encoding task (Is CS a friend vs. enemy of US?), Experiment 1 yielded either normal EC effects (CS adopting US valence) or a reversal. This pattern was conditional on the degree to which encoding judgements affirmed friend or enemy encoding schemes. Experiments 2a and 2b replicated these findings with more clearly valenced US faces and controlling for demand effects. Experiment 3 demonstrated unconditional encoding effects when participants generated friend or enemy relations between CS and US faces. Explicitly stated friend or enemy relations in Experiment 4 left EC effects unaffected. Together, these findings testify to the importance of higher order cognitive processes in conditioning, much in line with recent evidence on the crucial role of conditioning awareness.

Keywords:

• Encoding schemes
• Conditioning awareness
• Associative learning
• Generation processes

Acknowledgements

The research underlying the present paper was supported by grants from the Deutsche Forschungsgemeinschaft to the first (FI 294/23-1) and second author (UN 273/1-1).

The authors wish to thank Soraya Haarmann, Susanne Germann, and Yue Zhang for their help in conducting the first experiment.

Notes

¹Kremer (1978) showed that when compounded with two previously conditioned CSs, a neutral CS can acquire conditioned inhibitory properties, causing an inverse conditioning effect.

²Indeed, the notion of “preparedness” could be expanded to denote all kinds of top-down influences that constrain the ease with which particular associations can be learned.

³Field and Davey's criticism is based on the possibility that CS–US pairings highlight CS features that are similar to the US, and thus, CSs take on US valence, independent of any associative learning proper. Randomly pairing CS–US faces makes this explanation highly unlikely (see also De Houwer et al., 2001).

⁴This selection procedure is standard practice and does not undermine the validity of the obtained findings (e.g., Walther, 2002).

⁵US post-test ratings were independent of how likely a US face was a friend/enemy of a CS face. Interestingly, US post-test minus pre-test differences reveal a regressive fading-out of facial valence. Positive faces became somewhat less positive over time (reflected in negative difference scores), whereas negative faces became somewhat less negative (positive difference scores). The corresponding valence main effect tendency, F(2, 50)=2.43, p<.10, becomes significant when neutral faces are excluded, F(1, 50)=6.64, p<.05.

⁶Sixteen and fifteen participants in Experiments 3 and 4 did not meet the selection criteria in the pre-test.