Special Session Summary New Research on Associative Learning

Nader T. Tavassoli, MIT Sloan School of Management
[ to cite ]:
Nader T. Tavassoli (2002) ,"Special Session Summary New Research on Associative Learning", in NA - Advances in Consumer Research Volume 29, eds. Susan M. Broniarczyk and Kent Nakamoto, Valdosta, GA : Association for Consumer Research, Pages: 284-286.

Advances in Consumer Research Volume 29, 2002     Pages 284-286



Nader T. Tavassoli, MIT Sloan School of Management

This session focused on the acquisition of "likes and dislikes" via associative learning. The presentations addressed questions such as: Under what circumstances do cues such as brand names or logos acquire evaluative significance? How do multiple cues interact in this process? What is the role of consumers’ processing goals? In their answers to these questions, the presentations point to multiple associative learning systems which interact in complex ways with consumers’ goals (motivation, focus of attention), stimulus characteristics (verbal labels, product ingredients), and the learning context (presence of other product cues, contextual affect).

Van Osselaer, Janiszewski, and da Cunha examine the interaction of multiple product cues, each of which can become a predictive cue for brand quality. They find low-effort learning is based simply on the number of times consumers encountered a cue in the presence of the brand. Memory for the stimulus information appears to be stored as an inseparable whole under this type of learning. Motivated consumers, in contrast, process product cues in an interactive fashion. This type of adaptive learning results in more elemental stimulus representations in memory, where stimuli are stored in terms of their constituent parts.

Raymond and Tavassoli examine basic associative or automatic learning systes in the tradition of the mere exposure paradigm. Specifically, they examine whether the mere co-occurrence of a neutral stimulus and an affective cue is sufficient, or whether affect is bound to a neutral stimulus (e.g., logo or brand name) only when attention is focused on this stimulus, but not when consumers ignore the stimulus. Their results suggest that affective cues can affect the evaluation of neutral information such as a logo, but that this sometimes results in an evaluation that contrasts the affective context. They further suggest that "mere ignoring" of a stimulus (perceptual inhibition) leads to the reverse effect of mere exposure (perceptual fluency), such that an ignored stimulus is less liked after exposure.

Vanhouche, Warlop, and Baeyens examine whether a brand name acquires evaluative significance from an actual consumption experience in the presence of other cues (e.g., smell or color). The authors find that a pleasant or aversive taste experience can indeed affect the liking or disliking of an associated brand name but that a brand name, unlike a color or scent, does not affect the actual consumption experience (liking of a drink) after having been associated with a pleasant or aversive taste. Instead, the brand name acts in a discriminatory manner by heightening the perceptual significance of other physical cues (e.g., scents) and amplifies their acquired valence.



Stijn M. J. van Osselaer, University of Chicago

Chris Janiszewski, University of Florida

Marcus V. M. da Cunha Jr., University of Florida

Recently, van Osselaer and Janiszewski (2001) demonstrated that consumers have two, qualitatively different associative learning systems available that allow them to make predictions about product performance. The Human Associative Memory (HAM) system is a relatively unfocused, backward-looking, and less effortful process in which stimulus information gets stored for later retrieval (Anderson & Bower, 1973). Learning in this system is cue-independent. That is, learning of the relationship between one cue (e.g., a family brand name) and a dimension of product performance (e.g., taste) is independent of the presence of other cues (e.g., a sub-brand name). The other, adaptive learning system is more focused and forward-lookingCit tries to create a prediction rule that allows the system to predict a specific dimension of product performance on the next occasion. This system requires that a dimension of performance be the focus of prediction during learning and assumes cue-performance associations change only to the extent the expected performance of the product does not match the experienced performance of the product. Learning in this system is cue-interactive. That is, learning of the relationship between one cue and a dimension of product performance is influenced by the presence of other cues and the strength of those other cues’ associations with the same dimension of product performance.

In this presentation, the authors report preliminary results from two experiments that investigate the way the two learning systems represent stimuli in memory, as sums of separate elements or as indivisible stimulus configurations. In the first experiment, focus of prediction was manipulated. Focusing on predicting product performance should activate the adaptive system. Not focusing on predicting product performance should make consumers rely on the HAM system. When consumers did not focus on predicting product performance during learning, later predictions of product performance suggested product information was stored in a more configural way than when consumers had focused during learning. In the second experiment, we used another manipulation that should affect which system dominates responses. We manipulated the direction (forward-looking versus backward-looking) of processing at the time of test (i.e., at the tie subjects made their performance predictions). As expected, results suggested that backward-looking processing led to predictions indicating a more configural memory representation than forward-looking processing.

In sum, preliminary results suggested that the two systems identified by van Osselaer and Janiszewski (2001) do not only differ in terms of learning processes, but also differ in how information is stored in memory. The HAM system seems to store stimulus information more configurally, as a whole instead of as a sum of parts. The adaptive system may store stimulus information more elementally, breaking down stimuli into their constituent parts and storing information about the parts.


Anderson, J. R., & Bower, G. H. (1973). Human Associative Memory, New York: Halstead Press.

Janiszewski, C., & van Osselaer, S. M. J. (2000). A connectionist model of brand-quality associations. Journal of Marketing Research, 37, 331-350.



Jane E. Raymond, University of Wales, Bangor

Nader T. Tavassoli, MIT Sloan School of Management

The affective priming paradigm (Murphy & Zajonc 1993) has shown that affect (such as from smiling or frowning faces) can be generated subliminally and influence the processing and evaluation of stimuli that immediately follow these affective cues. The affect generated subliminally is assumed to be free-floating and is hypothesized to become bound with the supraliminal object that is the focus of attention. Recently, this conclusion has been challenged and it has been suggested that the affect does not get bound to the object itself but simply "spills over" into the decision or evaluation at hand. This research poses a host of questions of interest to consumer research. For example, does the affect generated by a web site (content or experiences) automatically attach itself to all branded information present in the environment? Does it matter whether the branded information is actively attended to or actively ignored? Does it matter whether attention is focused on the source of the affect?

We examined these questions in a series of studies. The results suggest that an affective context can affect the evaluation of a "neutral" stimulus. However, the affect appears to only "spill over" into a decision when a stimulus is co-encountered with the affect at evaluation. Negative affective primes lower evaluations, whereas positive cues enhance evaluations of target stimuli. However, when the target stimulus is first encountered in an affective context without an explicit evaluative goalBwhether the affective context is attended to or notBthere is little evidence for this "assimilation." If anything, it appears as though the target stimulus is automatically "contrasted" to an affective context and that later evaluations of the target stimulus are affected in directions opposite to the valence of the affective context at learning. However, these results are not entirely conclusive.

One of the most interesting findings was the effect of prior exposure on the evaluation of stimuli that were selected for attention, compared to those that were not selected for attention and therefore ignored. In one experiment, participants had to push the location of a pre-specified target object in the presence of a distractor object. Studies of attention show that this task involves actively ignoring he distractor. Studies on negative priming (Tipper, 1985) suggest that this leads to perceptual inhibition the next time the distractor is encountered. In contrast, the target object benefits from perceptual fluency which has been identified as the likely cause for the mere exposure effect, when objects are liked more after having been previously encountered (Zajonc, 1968). We indeed found that "mere ignoring" leads to a decrement in subsequent liking which qualifies the mere exposure effect; mere exposure may only lead to increased liking when the object is not ignored. This has obvious implications for the processing of branded stimuli such as logos or banner ads that are often actively ignored during the goal-oriented navigation of web sites.


Murphy, S. T., & Zajonc, R. B. (1993). Affect, cognition, and awareness: Affective priming with optimal and suboptimal stimulus exposures Journal of Personality and Social Psychology, 64, 723-739.

Tipper, S. P. (1985). The negative priming effect: Inhibitory priming by ignored objects. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 37A, 571-590.

Zajonc, R. B. (1968). Attitudinal effect of mere exposure. Journal of Personality and Social Psychology, 9, 1-27.



Wouter Vanhouche, University of Florida

Luk Warlop, KU Leuven

Frank Baeyens, KU Leuven

In a series of seven experiments brands or scents of drinks were associated with a good or bad tasting substance, followed by taste tests of the branded or scented drinks without this substance. Prior research had shown that subjects can acquire likes or dislikes for food scents (Baeyens, Crombez, Hendrickx, & Eelen, 1995; Zellner, Rozin, Aron & Kulish, 1983), even without awareness of prior pairing with good or bad tastes. Our objective was to investigate whether mere prior association with positive or negative consumption experiences could also make brands start tasting better or worse. We contrasted the acquisition of acquired tastes with the learning of the predictive value of brands and scents before consumption (see van Osselaer and Alba 2000). If brand cues would not only predict brand quality before consumption, but also influence experienced brand quality during consumption, this might explain why in taste tests evaluations of branded products are more differentiated than those of unbranded products (Allison and Uhl 1965), and thereby increase our understanding of the origins of customer based brand equity (Keller 1998).

Subjects participated in 'a taste test of non-carbonated isotonic sports drinks’. They first received a number of learning trials, in which they consumed good (containing sugar) or bitter tasting (containing Polysorbate 20) drinks. Depending on the condition, two different brand names (e.g., Vigro vs. Fast-X) or two different neutral scents (e.g., Lemon vs. Apricot) were paired with the positive vs. negative experiences. In the test phase, transfer stimuli without the sugar or Polysorbate 20 were presented. Subjects either predicted the taste quality, or they evaluated the consumption of these drinks. They learned to predict quality with high accuracy for both types of cues, but only scents were evaluatively conditioned. That is, subjects learned to like or dislike the consumption of Lemon or Apricot scented drinks, but not of Vigro or Fat-X branded drinks.

In follow-up experiments, we systematically varied the learning and test conditions, in order to account for our failure to observe evaluative learning for brands. We tested the effects of adding a single ambiguous flavor (citric acid) to both test drinks. We induced nonsystematic scent variation or color variation to test samples or to learning samples. In each case we replicated our initial findings.

In two experiments we did find an evaluative effect for brands. In one experiment we asked subjects to indicate their liking or disliking for the brand names (brand attitude) without consumption and found an effect of prior experience. However, the absence of any dissociation between evaluations and predictions did not allow us to infer that both responses measured different constructs. In an other experiment we found that systematically combining a brand an a scent during learning produced larger differentiation in later taste evaluations than using scents only. However, we found an equivalent level of differentiation when the acquired valence of scents and brands should have been incongruous. The latter finding led to the current conclusion that brand names do not acquire evaluative valence; their function in taste evaluation may merely be to facilitate perceptual discrimination at the time of test, allowing for the behavioral manifestation of acquired valence of experience-intrinsic cues, such as scents (see Warlop, Ratneshwar and van Osselaer 2000, for a similar finding). Further research should examine on which basis we best distinguish between product cues that can acquire evaluative valence and those that merely discriminate.


Allison, R. I. & Uhl, K. (1965). Influence of beer brand identification on taste perception. Journal of Marketing Research, 1, 36-39.

Baeyens, F., Crombez, G., Hendrickx, H., & Eelen, P. (1995). Parameters of human evaluative flavor-flavor conditioning. Learning and Motivation, 26, 141-160.

Keller, K.L. (1998). Strategic Brand Management. Building, Measuring and Managing Brand Equity. Prentice Hall, New Jersey.

Van Osselaer, S. M. J., & Alba, J. (2000). Consumer learning and brand equity. Journal of Consumer Research, 27, 1-16.

Warlop, L., Ratneshwar, R., & Van Osselaer, S.M.J. (2001). On the role of trivial differentiation in learning product quality from experience. Research Report 00-31 KULeuven Department of Applied Economics.

Zellner, D.A., Rozin, P., Aron, M. & Kulish, C. (1983). Conditioned enhancement of human’s liking for flavor by pairing with sweetness. Learning and Motivation, 14, 338-350.