Functional Exposure and Consumer Behavior: Introduction an Alternative Hypotheses For Frequency-Affect Relationships

Theo Poiesz, Tilburg University, The Netherlands
ABSTRACT - Consumers are frequently confronted with product-related stimuli. According to Zajonc (1968), frequent exposure to stimuli renters these stimuli more attractive in the eyes of the perceiver. However, the "mere exposure" hypothesis cannot give an overall-explanation of observed frequency-affect relationships, neither can alternative hypotheses presented subsequently.
[ to cite ]:
Theo Poiesz (1982) ,"Functional Exposure and Consumer Behavior: Introduction an Alternative Hypotheses For Frequency-Affect Relationships", in NA - Advances in Consumer Research Volume 09, eds. Andrew Mitchell, Ann Abor, MI : Association for Consumer Research, Pages: 435-438.

Advances in Consumer Research Volume 9, 1982      Pages 435-438

FUNCTIONAL EXPOSURE AND CONSUMER BEHAVIOR: INTRODUCTION AN ALTERNATIVE HYPOTHESES FOR FREQUENCY-AFFECT RELATIONSHIPS

Theo Poiesz, Tilburg University, The Netherlands

[The author thanks Ruud Drabbe for his assistance and Bob Chestnut for his helpful comments on an earlier draft.]

ABSTRACT -

Consumers are frequently confronted with product-related stimuli. According to Zajonc (1968), frequent exposure to stimuli renters these stimuli more attractive in the eyes of the perceiver. However, the "mere exposure" hypothesis cannot give an overall-explanation of observed frequency-affect relationships, neither can alternative hypotheses presented subsequently.

The attempt here is to join the different explanations of exposure-effects into one hypothesis. This hypothesis, focusing upon the functionality of exposure, finds support in Experiment I. Experiment II, an exploratory consumer-oriented study, suggests that this alternative hypothesis is possibly of relevance to consumer frequency-affect relationships. Results ant research implications are discussed.

INTRODUCTION

This paper focuses upon frequency-affect relationships in the context of laboratory consumer behavior. It does so in three stages. In the first stage, procedures and findings of mere exposure studies are discussed. It is hypothesized that functionality of exposure is critical for obtaining positive frequency-affect relationships. In the second, an exploratory laboratory study on the new hypothesis is described. Finally, a consumer-oriented version of this study is described. This laboratory experiment is the first in a sequence of studies with the objective of approximating real life consumer behavior. Its results are not as yet generalizable to consumer affect outside the laboratory, but are suggestive of future research.

Mere exposure research

Zajonc's (1968) mere exposure (ME) hypothesis maintains that merely repeating the exposure of a particular stimulus will enhance the liking of that stimulus. As consumers frequently interact with products and product-related stimuli, a phenomenon as suggested by this hypothesis might be relevant for the explanation of consumer affect. Zajonc's (1968) paper was followed by numerous studies on the conditions under which positive frequency-affect (FA) relationships would or would not be observed (see Harrison, 1977). This resulted in the formulation of a number of hypotheses. The major hypotheses concern: the ME-effect as experimental artifact (Burgess and Sales, 1971), response competition (Harrison, 1968; Matlin, 1970), expectancy arousal (Crandall, 1970), two factor (positive habituation and tedium) theories (Harrison and Crandall, 1972), semantic satiation and generation (Grush, 1976) and arousal formulations (e.g. Berlyne, 1970). Instead of discussing pro- and counterarguments for each of the hypotheses, we can refer to Stang (1974), who summarized FA-research by statistically analyzing past experiments as to their conditions and types of FA-relationships observed. He found two major factors: I. the size of the time-interval between exposure and rating - a longer interval increases the likelihood of a positive FA-relationship; 2. the type of stimulus - a positive FA-relationship is more likely with paralogs, ideographs and portraits than with abstract patterns. The major FA-hypotheses cannot account for this finding consistently. A possible explanation is suggested, however, by the nature of the procedures of FA-studies. Typically in these studies, pre-exposure instructions merely require S to pay close attention to the stimuli that will be shown. S is not informed about the subsequent task. It is assumed here that under such ambiguous circumstances, S will infer the nature of the future task from the characteristics of the experimental situation. One such characteristic is the type of stimulus exposed. It seems reasonable to assume that if paralogs or ideographs are presented, S expects a "performance-task" (like a memory-test). On the other hand, if abstract stimuli like modern paintings are presented, S is likely to expect a "nonperformance- (like/dislike evaluation-) task". It is hypothesized that if stimuli are presented that suggest a performance-task, the exposure-frequency determines S's expectation to do well on this task. As a consequence, frequently exposed, ant therefore 'helpful" or instrumental stimuli will be evaluated more positively than less frequent stimuli. Stang's (1974) second factor fits in consistently with this interpretation. S's (confidence in his) ability to do well on the task (to correctly identify or memorize the stimuli) is likely to be negatively related to the size of the interval between exposure ant rating. The most frequently exposed stimuli are the most helpful or instrumental ant will, therefore, receive the most positive affect-ratings. This suggests that the critical factor may be that of the overall functional meaning of exposure-frequency. If functionality is fount to be a key factor, it would be more appropriate to speak of "functional exposure" rather than "mere exposure".

Within this explanation, Stang's (1974) two factors can be interpreted as determinants of uncertainty as experienced by S. Conceptually, uncertainty is taken here as consisting of two components: 1. the lack of a clear course of action, and 2. the concern with the consequences of this lack.

In Experiment I the mere exposure hypothesis is tested against the functional exposure hypothesis: if higher exposure frequencies are experienced as (more) functional by S, a (more) positive FA-relationship will be observed. In accordance with the above, it is assumed that functionality of frequent exposure can be manipulated: I. directly, by instructions that indicate to S to what extent higher frequencies can be functional, 2. indirectly, by instructions that affect S's uncertainty. These can be of two kinds: 2a. instructions as to the availability of a clear course of action, 2b. instructions regarding the personal consequences of a lack of a clear course of action.

In more operational terms:

1. If, prior to exposure, S is informed that the task will be very hard/very easy, higher exposure-frequencies are not functional. The task ahead is expected to be hard/ easy anyway, irrespective of prior (exposure) conditions.

2a.If, prior to exposure, S is made to expect a nonperformance (like/dislike) task in which any answer is acceptable, exposure-frequency is irrelevant.

2b. If S's interest in the outcome of a performance-task is low, functionality/exposure-frequency is irrelevant.

The instructions of Conditions I, 2 and 3 of Experiment I correspond to possibilities 1, 2a and 2b, respectively. For the functional exposure hypothesis-to be corroborated three conditions must be met:

1. the observation of non-positive FA-relationships in Conditions 1, 2 and 3.

2. the observation of positive FA-relationships in two control conditions: condition 4: a condition in which S is made to expect a performance-task, and Condition 5: a conventional ME-condition:- no pre-exposure task-information; paralogs presented as stimuli.

3. in Condition 5 Ss expect a performance-task rather than a nonperformance-task. Experiment I tests the hypotheses reflecting these three conditions

EXPERIMENT I

Method

Ninety students from Tilburg University participated in the study. They had signed up for two consecutive studies (of relevance for instructions in Condition 3). Ss were divided evenly between conditions. One suspicious subject was replaced.

The stimuli were six paralogs of three syllables each. Examples: HEWONAT, ZODEGIN. Prior to the experiment, stimuli had been rated as to their neutrality by non-participants. Presentation was on slides. Stimuli were exposed at frequencies 0, I, 3, 6, 10 or 11, one stimulus per frequency. Stimuli were rotated over frequencies. Affect-ratings followed immediately after exposure. To maximize contrast between frequency-levels, the order of frequencies in the rating-phase was fixed (6, 0, 15, I, 10, 3). In both exposure- and rating-phase, stimuli were exposed for 2 seconds. Inter-exposure intervals were 4 seconds. Some additional remarks must be made with regard to the five conditions:

Condition 1: hard/easy task. For one half of the Ss the task was described as very hard; for the other half it was described as very easy.

Condition 2: nonperformance task. The instruction was: we are starting a new phase in our research. All we want you to do is to give your opinion as to the quality of the slides.

Condition 3: low interest (in task-outcome). Ss had signed up for participating in two consecutive experiments. E told them that the first one had been canceled ant asked them to wait until the start of the second. In the interval, E aroused Ss' interest in the canceled study. He offered them the opportunity to participate "as if". He made clear that data obviously would not be checked. Then, Ss received the Condition 5 (ME) instruction.

Condition 4: performance task. Instruction: try to get as high a score as possible on the task that you will receive after the slide-presentation.

Condition 5: ME-condition. Instruction: please watch the slides that will be presented carefully.

Following exposure, Ss received an "in-between-question" the affect-rating. Stimuli had to be rated on a 9-point scale (this word makes an extremely favorable (=9)/unfavorable (-I) impression upon me). Prior to this rating-task, Ss in Condition 5 were asked to indicate their perceived chance of having to do a performance-task ant the perceived chance of a nonperformance-task (chances adding to 100%). It was made clear how both types should be interpreted. After the rating-phase, Ss received a bogus task reflecting initial instructions. Finally, they were questioned for suspicion. (It was decided to refrain from manipulation checks. These either would have increased the exposure rating interval, thus confounding manipulations, or would reflect the joint effects of the uncertainty and frequency-manipulation);

Results

The-mean scores per exposure-frequency are presented in Figure I (for the Nonfunctionality-Conditions I, 2 ant 3) ant in Figure 2 (for the Functionality-Conditions 4 and 5).

FIGURE 1

MEAN AFFECT PER FREQUENCY, NONFUNCTIONALITY-CONDITIONS

FIGURE 2

MEAN AFFECT PER FREQUENCY, FUNCTIONALITY-CONDITIONS

Inspection of Figures 1 ant 2 suggests the sensitivity of exposure effects for the type of task-instruction given. Differences between conditions seem in accordance with those suggested by the functional exposure hypothesis. This is most apparent when comparing Conditions 2 (nonperformance-task) ant 5 (ME-condition). In the latter condition, 14 out of 18 Ss expected a performance- rather than a nonperformance-task. Only one S tit not expect a performance-task at all (chance 0%). The mean perceived performance-task chance was 66.3%. This finding supports the suggested interpretation of ME experimental procedures. As it is always possible, by simply increasing exposure-frequencies to obtain curvilinear effects, emphasis will be here upon the linear components of the FA-relationships. This restriction toes not affect the hypotheses. Furthermore, because hypotheses refer to differences between FA-relationships, interaction-effects rather than main effects will be our main point of interest. An analysis of variance (repeated measures on exposure-frequency) indicated a nonsignificant interaction effect of conditions x frequency: F (20,425) - 1.10, which is not surprising as we predicted the presence of interactions between the Nonfunctionality and Functionality-conditions ant the absence of interactions within each of these condition-groupings. Of the simple interactions that were predicted only the one in the Condition-combination 3 (nonperformance-task) ant 5 (ME-condition) is significant: F (5,170) s 2.57, p< .05. The interactions in the Condition-combinations I (hard/easy task) and S (ME-condition), ant 3 (low interest) ant 5 tent in the right direction: F (5,170) = 1.65 ant 1.52, respectively. Simple interactions within the Functionality- and Nonfunctionality-groupings are nonsignificant, as predicted. Although irrelevant here, main effects for conditions ant exposure-frequency were significant: F(4,85) = 2.64, p < .05 ant F(5,425) = 4.93, p <.01, respectively. Assuming psychologically equal intervals on the frequency-continuum, an analysis of linear trend was performed to find out whether results reflect differences in slopes of the best-fitting straight lines to the profiles of the conditions. The overall interaction, F(4,85) - 2.39, is not significant. Again, simple interaction effects were calculated. As expected, within the Nonfunctionality- and Functionality-groupings interactions were nonsignificant. Simple interactions between conditions showed significant effects for the Condition-combinations I (hart/easy task) and 5 (ME): F(1,34) = 4.39, p< .05; 2 (nonperformance-task) ant 5: F(1,34) = 11.89, p < .01; and 3 (low interest) and 5: F(1,34) = 4.21, p < .05. Interactions between Conditions I, 2, 3 ant 4 (performance-task) were nonsignificant, however. As Figures 1 ant 2 seem to suggest a turning-point in the FA-relationship between frequency-levels 10 and 15, ant as it is legitimate to focus upon linear components (curvilinearity can always be obtained by increasing frequency-levels), an additional analysis of linear trend was performed, now excluding ratings on the 15-frequency paralogs. The overall conditions x frequency interaction proved significant: F(4,85) - 2.77, p <.05. The general pattern seems to support a functional exposure interpretation: simple interactions in Condition-combinations 2 (nonperformance-task) ant 4 (performance-task), I (hard/easy task) ant 5 (ME), and 2 (nonperformance-task) and 5 (ME) proved significant: F's (1,34) were, respectively, 4.47 (p < .05), 5.63 (p < .05) and 8.14 (p < .01). Simple interactions in Condition-combinations I (hart/easy task) and 4 (performance-task), ant 3 (low interest) and 5 (ME) tent in the right direction: F's (1,34 were, respectively, 3.02 (p < .10) ant 3.89 (p < .10). Finally, the simple in combination 3 (low interest) ant 4 (performance) shows a very slight trend in the right direction: F(1,34) - 1.48. As predicted, simple interactions within the Nonfunctionality- ant Functionality-groupings were nonsignificant (all F-values < 1.0).

Discussion

In all, the pattern of results seems to justify the conclusion that functionality of higher exposure-frequencies may indeed be the necessary condition for positive frequency-affect relationships to occur. However, the mechanism through which functionality is supposed to be effective was not assessed directly Furthermore, it is not clear why there is no clear confirmation of hypotheses in condition-combinations involving Condition 4 (performance-task). The post hoc explanation that might be suggested is that the performance-task instruction, by emphasizing a high score, may have changed the nature of this condition towards that of Condition l: the very hart task instruction. If so, perceived functionality of the higher exposure frequencies was depressed relative to the ones in ME-condition 5.

EXPERIMENT II

After having focused, in Experiment I, on exposure-effects in abstract, Experiment II attempts to explore a more consumer-oriented task. In this study, perceived risk is taken as the consumer behavior counterpart of uncertainty as conceptualized ant manipulated in Experiment I. Conceptually, perceived risk is also taken as consisting of two components: I. the lack of brand information (or, the lack of a clear course of action) ant 2. the consequences of this. It is hypothesized, in accordance with the functional exposure hypothesis, that a positive FA-relationship will be observed under conditions of perceived risk and that the relationship will be nonpositive when perceived risk is low. Next to exposure frequency (within Ss) and perceived risk, position of risk-instruction relative to exposure (before/after) is included as a between subjects factor for exploratory reasons.

Method

Forty-four housewives participated in this "foot-test" study. They were divided evenly over the four cells. The stimuli were of the same type of stimuli as exposed in the previous experiment, but were labeled as "brandnames". Words, two per frequency-level, were presented on slides. Basically, the procedure is similar to that of Experiment 1, with the exception that frequencies were 0, I, 3, 6 and 10. Manipulation of the risk-factor is through the descriptive of product-types. In the two risk-conditions, Conditions I and 3, products are described as African food-products. Ss are informed that the demand for these in European countries is low, due to their unusual "swamp- ant jungle-ingredients". The instructions were chat Ss were to test several brands of such a product. In the two low risk conditions, Conditions 2 ant 4, the product is described as a successful foreign product exported to the Netherlands. Ss were informed that this product had received very positive ratings by 196 out of 200 housewives. Instructions were that they (Ss) also were to test several brands of this product. In Conditions 3 ant 4 risk-instructions were given after the exposure-phase. In order to avoid a conventional ME-setting in these latter conditions, Ss received, prior to exposure, the instruction that at first example-slides would be presented to familiarize them with the procedure. In the rating-phase, Ss were required to guess the quality of each of the brands. Slides were shown again once. Possible scores ran from I to 10 (from extremely inferior to extremely superior quality, respectively). Finally, Ss received some foot-items to taste. ManipuLations were not checked for reasons indicated in the description of Experiment I.

Results

Figure 3 and Figure 4 show the mean affect (quality) ratings at the various exposure frequencies, Figure 3 for the two risk conditions and Figure 4 for the low-risk conditions. Position of instruction seems to be a more dominant factor than product-risk as manipulated here. An analysis of variance (repeated measures on exposure-frequency) showed that the predicted risk x exposure frequency interaction is not significant: F(4,160) - .32.

FIGURE 3

MEAN AFFECT PER FREQUENCY, RISK CONDITIONS

FIGURE 4

MEAN AFFECT PER FREQUENCY, LOW RISK CONDITIONS

The instruction-position x exposure-frequency interaction is significant: F(4,160) = 2.63, p <.05. The (between Ss) interaction risk x instruction-position almost reaches significance: F(1,40) = 3.84, .10< p <.05, suggesting that the pattern of affect ratings in the two risk groups depends upon the position of the risk-instruction. Main effects were significant for risk, F(1,40) = 5.59, p <.05, ant exposure-frequency, F(4,160) = 9.07, p <.01. These, however, are not of direct relevance for the hypothesis. The results do not confirm the hypothesis: there was no significant risk x exposure-frequency interaction. An analysis of linear trend (assuming psychologically equal intervals on the frequency-continuum) indicated no interaction between Condition I (pre-exposure instruction, risk) ant Condition 2 (pre-exposure instruction, low risk): F = 0.00 - the slopes of these conditions to not differ. Therefore, if the risk-factor would have any effect at all, it would have to be in the Condition-combination 3 (post-exposure instruction, risk) ant 4 (post-exposure instruction, low risk). The simple interaction effect was F(4,80) = 2.72, p <.05. However, an analysis of linear trend indicated that this outcome tit not reflect differences in slopes of the best-fitting straight lines in the two conditions: F(1,20) = 0.88, n.s. Only when the 5-level frequency-factor was reduced to a 2-level factor (exposed vs. nonexposed brandnames), the risk x exposure interaction was significant: F(1,20) = 6.19, p <.05.

Discussion

With regard to the outcomes of Experiment II, there seem to be (at least) two questions. The first one is: why is risk as manipulated not successful, especially when risk-instructions are given prior to exposure? Ant second: why does position of instruction relative to exposure have an influence on the FA-relationship? The answer to the first question may be that instructions in Condition 2 (pre-exposure instruction, low risk) made Ss expect, in fact, a risky task: the judgment of foreign products before their export to the Netherlands possibly contains two elements of risk: S's responsibility ant the product's origin. With regard to the second question, it may be that the relaxing pre-exposure instructions depressed the effect of the risk-manipulations, thus depressing also the functionality of the higher exposure-frequencies relative to the ones in Conditions I and 2. Another possibility is that the contrast between the relaxing pre-exposure instructions ant the more arousing post-exposure instructions may have affected the credibility of the latter.

General discussion

The hypotheses ant results of Experiment I suggested a curvilinear relationship between subjective uncertainty ant the likelihood of a positive frequency-affect relationship, with most positive relationships at intermediate -uncertainty-levels. This makes the operationalization of subjective uncertainty extremely difficult ant may have hampered the finding of results in Experiment II. With additional information provided by manipulation-checks it might have been possible to correctly interpret the failure of the risk-manipulation ant the effect of position of instruction. Therefore, future studies will contain control-conditions for checking manipulations. However, for reasons indicated earlier, such checks should be the exclusive purpose of such conditions. In general, outcomes of both experiments are suggestive of future studies on the effects of functional exposure ant on its possible relevance to consumer affect.

REFERENCES

Berlyne, D.E. (1970), "Novelty, Complexity ant Hedonic Value", Perception ant Psychophysics, 8, 275-286.

Burgess, T.D.G. ant Sales, S.M. (1971), "Attitudinal Effects of Here Exposure: A Re-evaluation", Journal of Experimental Social Psychology, 7, 461-472.

Crandall, J.E. (1970), "Predictive Value ant Confirmability of Traits as Determinants of Judged Trait Importance, Journal of Personality, 38, 77-90.

Grush, J.E. (1976), "Attitude Formation and Mere Exposure Phenomena: A Non-artifactual Explanation of Empirical Findings", Journal of Personality ant Social Psychology, 33, 28 1-29.

Harrison, A.A. (1977), "Mere Exposure". In L. Berkowitz, Ed., Advances in Experimental Social Psychology, Vol. 10, Academic Press, New York, 39-83.

Harrison, A.A. (1968), "Response Competition, Frequency, Exploratory Behavior ant Liking, Journal of Personality and Social Psychology, 9, 363-368.

Harrison, A.A. ant Crandall, R. (1972), "Heterogeneity Homogeneity of Exposure Sequence ant the Attitudinal Effects of Exposure", Journal of Personality ant Social Psychology, 21, 234-238.

Matlin, M.A.W. (1970), "Response Competition as a Mediating Factor in the Frequency-Affect Relationship", Journal of Personality ant Social Psychology, 16, 536-552.

Stang, D.J. (1974), "Methodological Factors in Mere Exposure Research", Psychological Bulletin, 81, 12, 1014-1025.

Zajonc, R.B. (1968), "The Attitudinal Effects of Mere Exposure", Journal of Personality ant Social Psychology, Monograph Supplement, 9, 2, pt.2, 1-27.

----------------------------------------