Perceived Risk As a Mediator in Repetition-Affect Relationships

ABSTRACT - This paper reports on two experiments that assess the relevance of the functional exposure hypothesis for the explanation of consumer repetition-affect relationships. The hypothesis suggests that these relationships are less positive under conditions-of low and high perceived risk than under conditions of intermediate (medium) risk. In spite of operational problems concerning perceived risk levels, some su?porting evidence is obtained.

Citation:

Theo B.C. Poiesz (1986) ,"Perceived Risk As a Mediator in Repetition-Affect Relationships", in NA - Advances in Consumer Research Volume 13, eds. Richard J. Lutz, Provo, UT : Association for Consumer Research, Pages: 387-391.

Advances in Consumer Research Volume 13, 1986      Pages 387-391

PERCEIVED RISK AS A MEDIATOR IN REPETITION-AFFECT RELATIONSHIPS

Theo B.C. Poiesz, Tilburg University, The Netherlands

ABSTRACT -

This paper reports on two experiments that assess the relevance of the functional exposure hypothesis for the explanation of consumer repetition-affect relationships. The hypothesis suggests that these relationships are less positive under conditions-of low and high perceived risk than under conditions of intermediate (medium) risk. In spite of operational problems concerning perceived risk levels, some su?porting evidence is obtained.

INTRODUCTION

Repetition of marketing stimuli, such as brandnames, is a particular feature of the consumer's environment. For a variety of reasons, including differences between individual marketers' communication intensities, the level of repetition varies greatly over brandnames. Repetition has been shown to play a multiple role in consumer behavior. Repetition enhances learning and information processing (see Sawyer 1981, 1977, 1974). It also interacts with other factors relevant for the explanation of consumer behavior (see Belch 1982). Repetition, as an autonomous variable, has been linked with consumer affect in several attempts. The most widely known attempt was formulated and documented by Zajonc (1968). In his mere exposure hypothesis, mere repeated exposure to a stimulus is taken as a sufficient condition for the enhancement of positive affect toward that stimulus. However, research aimed at the assessment of the hypothesis' validity produced conflicting results. Vanbeselaere (1984), after discussing the available evidence, concluded that '(...) the exposure phenomenon remains a phenomenon in search for an explanation'. Poiesz (1983, 1981) introduced the 'functional exposure hypothesis' (FE-hypothesis) as a possible alternative to Zajonc's (1968) hypothesis. A series of both social psychological and consumer psychological experiments, reported in Poiesz (1983), produced supporting results. In the first part of this paper, the FE-hypothesis is described and a brief summary is presented of the social psychological experiments. The second part deals with two consumer psychological studies. Finally, in part three, possible implications for the status of the exposure hypotheses are discussed.

THE FE-HYPOTHESIS: SOME SOCIAL PSYCHOLOGICAL EVIDENCE

For a correct interpretation of exposure effects, it is necessary to first describe the experimental conditions under which they typically tend to be obtained. In a laboratory situation, subjects are requested to carefully watch the presented stimuli. Exposure may or may not be preceded by task instructions. In the case of instructions, they are usually very implicit. Stimuli tend to be unknown words, names or symbols that are exposed at different frequencies. Following exposure, subjects rate the stimuli on a Likert affect scale. Poiesz (1983) suggested that repeatedly exposed stimuli will be associated with more positive affect only and to the extent that these may be considered capable of reducing the subjects' subjective uncertainty, a state conceived as determined by incongruity or by the inability to predict the future, combined with apprehension concerning the consequences (Kagan 1972). At first sight, the FE-hypothesis seems very similar to the 'uncertainty reduction' hypothesis that Sawyer (1981) considers tenable for the explanation of exposure effects: the response competition hypothesis (Matlin 1970; Harrison 1968), the optimal arousal hypothesis (Berlyne 1971, 1960), and the two-factor theory (Stang 1973; Berlyne 1970). The FE-hypothesis is similar to these hypotheses in that it takes uncertainty reduction as effectuated by familiar stimuli to be the responsible factor for the relatively positive affect towards these stimuli. Therefore, all these hypotheses could actually be labeled 'functional exposure' hypotheses. However, the uncertainty reduction hypotheses take functionality of repetition in a more narrow sense than the FE-hypothesis. They depart from uncertainty and repetition functionality as resulting from person and/or object characteristics. The FE-hypothesis, on the other hand, considers more explicitly the interaction of person, object and situational characteristics for the assessment of repetition functionality.

This difference between the two types of hypotheses has two implications. First, uncertainty (reduction) is taken in a broader sense by the FE-hypothesis than by the uncertainty reduction hypotheses. Second, while the other hypotheses view functionality as relating exclusively to uncertainty reduction, the FE-hypothesis allows for the possibility that uncertainty enhancement is preferred to uncertainty reduction. If so, to the extent that repetition produces uncertainty reduction, repetition is dysfunctional and will be associated with negative affect. That the two implications are not trivial ones will be clarified with the help of a few examples, in which the two types of hypotheses will be distinguished. According to the response competition explanation, the exposure of a novel stimulus elicits different, sometimes antagonistic response tendencies. By repeated exposure response competition and the associated tension is reduced, resulting in more positive affect towards the more frequent stimuli. However, this explanation focuses on novelty per se, thus ignoring the possibility that novelty may have a different psychological meaning in different circumstances. For example, for an experimental subject a novel task related stimulus is psychologically basically different from the same stimulus encountered outside of the laboratory as one of the many novel stimuli. It is clear that in the first case, the stimulus is much more likely to generate uncertainly and aversive tension (i.c. apprehension) than in the second case. Consequently, repetition in the first case is more likely to be functional and to lead to positive affect than in the second case. A second example may be offered in relation to arousal interpretations. In these, stimulus complexity is hypothesized to be related to exposure effects: the repeated exposure of a complex stimulus reduces its arousal potential to the preferred level, resulting in positive affect. In some studies, complexity was operationalized in mere physical terms: number of bits and redundancy (e.g. Vanbeselaere 1980). Again, the same physically very complex stimulus may or may not be associated with aversive tension (a surplus of arousal potential) depending upon the person and the situation: a complex visual pattern presented as task material in a laboratory is psychologically incomparable to the same pattern introduced as a painting in a museum. In the former as opposed to the latter case, uncertainty and apprehension are more likely to be generated. And, even though stimulus complexity is the same in both cases, only in the former case repetition may be expected to reduce uncertainty/to be functional and to result in affect enhancement. This latter example may also be turned around. If a person has been deprived of external stimulation for some time, the first exposure to a complex stimulus will be more functional than the repetition of that stimulus, which would result in a negative repetition affect relationship. The FE-hypothesis does not view preference for familiarity and preference for novelty (Berlyne 1970, 1960) to be in theoretical conflict: these apparent opposites are seen as having a common denominator: repetition functionality.

It must be noted that also arousal and two factor explanations discuss negative repetition-affect relationships. However, these explanations view negative relationships as resulting from repetition overdoses. At lower repetition levels, positive relationships are predicted. The examples above suggest that the interaction of person, object, and situational characteristics should be considered for the adequate assessment of repetition functionality as a basis for positive repetition-affect relationships. In summary, the FE-hypothesis does not provide a strict alternative for the uncertainty reduction hypotheses. It supplements them in that it takes the potential function or psychological significance of repetition to be broader than that of uncertainty reduction alone; mere object related uncertainty reduction is not taken as a necessary condition for positive repetition-affect relationships to occur. Extending this, it is expected that with the FE-hypothesis also nonpositive relationships can be predicted more accurately.

For unknown complex stimuli the uncertainty hypotheses would predict repetition to lead to more positive affect. The FE-hypothesis would make the same prediction but only for conditions under which repetition can be considered functional. For example, in a task situation in which subjects are apprehensive with regard to the quality of their performance. In some initial social psychological experiments in which paralogs were repeatedly exposed (Poiesz 1983) subjective uncertainty/apprehension and repetition functionality was expected to take place

- in task conditions in which the quality of S's performance is at stake (as opposed to non-task conditions or to conditions in which performance quality is irrelevant, stimulus type held constant);

- in task conditions in which subjects are informed of an intermediate probability (i.c. 75%) of success (as opposed to conditions with a very low or very high probability of success, stimulus type held constant).

Repetition-affect relationships were significantly more positive in the apprehension conditions than in the other conditions, supporting the functional exposure interpretation. Additionally, in these other conditions, repetition-affect relationships were usually nonpositive (not significantly different from 'slope zero', depicting the absence of any relationship). The obtained results were replicated in several laboratory studies.

TWO CONSUMER PSYCHOLOGICAL EXPERIMENTS

In essence, the step from a social psychological experiment to a consumer psychological experiment is not necessarily a large one. To adult subjects, the consumer's role is a familiar one; in prepurchase situations the concept of perceived risk can be taken as the equivalent of the concept of subjective uncertainty as both contain an unpredictability and an apprehension component; and unknown nares can easily be presented as (unknown) brandnames. In the following, two experiments will be briefly described in which the validity of a functional exposure interpretation in a consumer laboratory situation was assessed.

In line with this interpretation, it is hypothesized that positive repetition-affect relationships will be observed only if subjects experience an intermediate level of perceived risk. At this level, to be interpreted as located between the levels of high and low perceived risk, repetition is expected to be functional by its uncertainty reducing effect. The functionality of higher exposure frequencies at the other two levels is limited for two different reasons respectively. If perceived risk is low, so is the need for information, that is, any information. So stimulus familiarity is bound to be irrelevant/ nonfunctional as well. If perceived risk is high, the need for issue relevant (intrinsic product/brand quality or price) information is high. Here, the exposure frequency of stimuli is unlikely to serve as the single basis for the risky decision (unless no further information can be expected). Therefore, in general, if perceived risk is high, the functionality of higher exposure frequencies is low. In conclusion, we expect less positive repetition-affect relationships under conditions of low and high perceived risk than under conditions of intermediate (medium) perceived risk. Note that an uncertainty reduction hypothesis would predict the relationships to become more positive with risk.

EXPERIMENT 1

Subjects (Ss). 48 persons, 20 men and 28 women of various ages participated in the experiment. Prior to the experiment, they were told that they would receive the standard compensation worth approximately $ 3.00 for their cooperation. Ss were divided evenly over the three conditions. The data of 3 Ss were discarded because of incompleteness. These Ss were replaced.

Stimuli. The stimuli ('brandnames') employed were the presumably unknown names of French hamlets that a group of 15 non-subjects rated affectively neutral (TRESSON, ENCAUSSE, LE DORAT, PREGONDE, VERGONS and MARCILLAT). Stimuli were presented 0, 1, 3, 6, 9 and 12 times, and were rotated over frequency levels. A particular stimulus did not follow itself in the exposure sequence. Stimuli were exposed for 2 seconds each. Interexposure intervals were 4 seconds.

Procedure. The experiment took place in the Laboratory of the Psychology Department, Tilburg University. Ss were situated vis-a-vis a Philips monitor. All Ss could clearly see the stimuli. Conditions were run simultaneously, with a maximum of 8 Ss participating at the same time.

Instructions. Ss will be confronted with one of three types of instructions, each aimed at producing a particular level of perceived risk. We may expect Ss to consider decisions to be more difficult with increasing risk. Also, with increasing risk Ss are assumed to reduce their perceived chance of being able to select the brand they would prefer most. Therefore, subjective ratings of 'difficulty of choice' and 'chance of preferred brand' will serve as manipulation checks on perceived risk.

Condition 1: Low Risk (LR). Instructions: This study is on the evaluation of roller pens. The Consumers' Guide wrote (in fact, it did not - note added in the present text) about these pens: "All brands of roller pens have passed our test excellently. There are quality differences but a bad buy is definitely impossible". The value of these pens is between 2 and 3 dollars a piece. You may select one of them later-on and keep it. It is not possible to try them out first because they are wrapped. However, if you, after selecting a pen, consider another pen somewhat more attractive, then we will simply change pens around. So, you will receive the pen that you prefer most. In a few moments we will acquaint you with the brandnames. These will appear on the monitor.

Condition 2: Medium Risk (MR). The instructions are equal to those of Condition 1, except for the so-called Consumers' Guide section: "You do not run the chance of a bad buy when purchasing a roller pen. Our test indicated that there are no bad pens. Yet, it can do no harm to be somewhat careful when choosing a pen because of the differences in quality. All pens are approximately equally priced". The sentence indicating that after the selection pens may be changed around is left out here.

Condition 3: High Risk (ER). Instructions are equal to those of Condition 1 and 2, except for the Consumers' Guide section: "When choosing a roller pen the consumer must be very careful. Some brands are fair or good, others are just bad. And the bad brands are not cheaper than the good ones. In all, a very risky product". Also here the sentence about the changing of pens afterwards is left out.

Immediately following the exposure phase, instructions relating to the affect ratings were handed out. For the rating each stimulus was presented once for 2 seconds. Between presentation times were 4 seconds. The order of stimuli in the rating phase was varied so as to approach counterbalancing. The time lapse between the exposure and rating phase was about 2 minutes. Affect ratings were of two types. The first type concerns the traditional affect rating on a 7-point scale ranging from 'this brandname makes a very unfavorable impression upon me' to '(...) a very favorable impression (...)'. The second type concerns the exposure frequency of the chosen brand when subjects are requested to make a selection out of the various brands/brandnames that are available.

Results. Before being exposed to the various brandnames, SS had been requested to rate the manipulation checks 'difficulty of choice' and 'chance of preferred brand'. See Tables 1 and 2 for the respective results.

TABLE 1

"DIFFICULTY OF CHOICE" (LIKERT SCALE 1-5; 5=MOST DIFFICULT), PER CONDITION

TABLE 2

"CHANCE OF PREFERRED BRAND" (LIKERT SCALE 1-7: 7=VERY SMALL CHANCE), PER CONDITION

Differences between conditions are in the right direction. However, the differences between the conditions Low and Medium Risk are not significant for neither one of the manipulation checks. In spite of these outcomes, we will first consider the results as they relate to the hypothesis. Results will be presented for linear trends and frequency of the chosen brand. Only linear trends are considered here as these relate most directly to the hypothesis.

Linear trends. For the conditions LR, MR and HR the respective means are 4.06, 5.63 and 0.44, differences being in the expected direction. These are the mean sums of the transformed raw scores per individual (transformation by coefficients of orthogonal polynomials for a linear trend with 6 observations, coefficients being -5, -3, -1, 1,-3, and 5). The overall effect is not significant, however - F2,45 = 0.60 n.s. (analysis of linear trend (Winer 1971); a main effect in this analysis may be interpreted as equivalent to an interaction effect in an ordinary ANOVA).

Frequency of chosen brand. For analyses on this variable, absolute exposure frequencies were changed into the corresponding coefficients of orthogonal polynomials referred to earlier. Cell mean differences are in the expected direction, cell means being 1.27, 3.25 and 0.00 for the LR, MR and HR conditions, respectively. An ordinary analysis of variance indicates a significant effect for perceived risk (F2.44 = 6.42, p<.01). As hypothesized, there is no significant difference between the LR- and the LR-Condition (t(29) = 1.21, n.s.), a significant difference between the LR- and the MR-Condition (t(29) = 2.11, p<.05) and a significant difference between the MR- and the ER-Condition (t(30) = 4.21, p<.001).

Discussion. For the dependent variable 'frequency of chosen brand', but not for linear trends, the factor perceived risk did have the hypothesized effect. A possible reason for this difference may be that the consequence aspect, as one of the two aspects of uncertainty/perceived risk, is rendered more dominant in the case of choice than when Ss is required to give affect scores only.

An important question concerns the lack of a difference in both manipulation checks between the so-called Low Risk and Medium Risk Condition. The scores on the manipulation checks of the Ss in the former condition suggest that these Ss did experience an intermediate level of risk. Therefore, the outcomes of the manipulation checks leave us a second option for the analyses on the dependent variables. This option is to distinguish two instead of three levels of perceived risk, considering the insignificance of the LR and MR manipulation check differences. When the LR condition is combined with the MR condition on the basis of the manipulation checks, the effect of risk on the linear trends is not significant at the customary levels. However, a trend is observed: LR + MR vs. ER : t(46) = 1.44, 0.05 p<.0.10. As could be expected on the basis of the results presented earlier, for the dependent variable frequency of the chosen brand, the combination of the LR and the MR condition differs significantly from the ER condition: t(54) = 2.77, p<.01.

Perceived risk may be assumed to be positively and linearly related to depth of processing ar.d the various risk instructions may have resulted in depth of processing differences. These differences can not be associated with repetition-affect relationship differences, however, given the obtained results: the repetition-affect relationships do not increase positively and linearly with risk.

EXPERIMENT 2

In reference to the manipulation problems in Experiment 1, the main purpose of the present experiment is to attempt to generate low levels of perceived risk. In Experiment 2 the experimental situation and procedure will be equal to the ones in that Experiment 1. It is assumed that perceived risk may be lowered by attenuating its component 'consequence(s)'. For example, by eliminating S's possibility to make a choice, thus rendering choice outcome completely beyond his/her control. Alternatively, by giving S the expectation of an opportunity to actually try out the alternatives before making a selection, perceived risk may be assumed to be low relative to the situation in which there is no such expectation.

In combination with the Medium and Low Risk instructions of Experiment 1, these two alternative risk reducing possibilities may be worked out ia the following experimental conditions:

1. Low Risk/Choice

2. Medium Risk/No choice

3. No Risk: Low Risk instruction; No choice.

Together with the Medium Risk Condition of Experiment 1 (Medium Risk instruction; Choice), these conditions form a complete 2x2 design.

It is hypothesized that in the no-choice conditions frequency-affect relationships and frequencies of the chosen brand [In each condition, Ss will be requested to make a selection anyway, but in such a way that the consequence component is not introduced.], will be nonpositive/small, that is, approximating (slope) zero.

Because of the requirement to make a choice, the dependent variables will assume higher values in the Low Risk/Choice Condition. However, these are expected to be lower as compared to the corresponding values in the LR- and MR-Condition of Experiment 1.

Subjects. Subjects were 16 men and 32 women of various ages. Their names were drawn from the same participants pool as used for obtaining the names of participants in Experiment 1. (Of course, no person participated in both experiments). For their cooperation the compensation worth about $ 3.00 was announced. Ss were divided evenly over the three conditions: 16 Ss each. No data had to be discarded.

Stimuli and Procedure. See Experiment 1. For instructions see below.

Condition 1. Low Risk/Choice (LR/Choice). This study is on the evaluation of roller pens. The Consumers' Guide wrote about these roller pens - the same as we employ in this study: 'All brands of roller pens have passed our test excellently. A bad buy is definitely impossible'. The value of the pens is between 2 and 3 dollars a piece. Later-on we give you 6 pens of different brands. Then, you shouLd try out these pens and make a selection out of them. You may keep the selected brand. You will have ample time for trying out the pens. In a few minutes we will acquaint you with the brandnames. These will appear on the monitor.

Condition 2. Medium Risk/No choice (MR/No choice). Up to the so-called Consumers' Guide section, instructions are equal to those of Condition 1. Then: 'You do not run the chance of a bad buy when purchasing a roller pen. Our test indicated that there are no bad roller pens. Yet, it can do no harm to be somewhat careful choosing a pen because of the differences in quality. By the way, all pens are approximately equally priced'. The value of the pens is between 2 and 3 dollars a piece. Later-on we give you one of these pens. After you have tried out this pen for a little while, we would like you to answer some very simple (written) questions about it. You may keep the concerning pen. It is not possible at this moment to indicate which brand that will be as the pens will be distributed completely arbitrarily. In a few minutes we will acquaint you with the brandnames. These will appear on the monitor. (As manipulation-checks on risk would be awkward here and possibly counteracting manipulations, they are left out).

Condition 3. No Risk: Low Risk instruction/No choice (LR/No choice). The instructions are the same as those in Condition 1 except for the part after the value information:

Later-on we give you one of these pens. After you have tried out this pen for a little while, we would like you to answer some very simple (written) questions about it. (Etcetera). (See instructions Condition 2). For the no-choice conditions, the brand-selection question was adapted by formulating it as follows: 'If you would have to make a selection yourself now, which brand would you choose?'. The rest of the procedure is equal to the one in Experiment 1.

Results. With the three conditions LR/Choice, MR/No choice and LR/No choice, it was attempted to obtain perceived risk levels lower than those generated in Experiment 1, Conditions LR and MR. The only condition for which this can be checked is the LR/Choice Condition, of which the instructions announced the request to make a selection.

In Condition LR/Choice the mean of the variable 'difficulty of choice' is 13.9 on the 5-point scale and the mean of the variable 'chance of preferred brand' is 3.50 on the 7-point scale. It was expected that these means would be lower than the respective means of Conditions LR and MR of the previous experiment. However, this is not the case. The differences are even in the wrong direction. Assuming that manipulations in the remaining two conditions of the present experiment were successful, it seems justified to adapt the hypothesis: Considering the outcomes on the manipulation checks, we should expect a repetition-affect difference between the LR/Choice Condition on the one hand and the Conditions MR/No choice and LR/No choice on the other. At the same time, no repetition-affect differences are expected to be observed between the LR/Choice Condition and the MR/Choice Condition (Experiment 1).

Linear trends. Per condition, the mean sum of the raw scores transformed by coefficients of orthogonal polynomials (linear trend) are:

LR/Choice 5.88

MR/No choice -0.06

LR/No choice -1.31

(MR/Choice, Experiment 1) 5.63

An analysis of linear trend on the 2x2 design shows a significant effect for choice (F1,60 = 6.70, p<.05), but not for type of instruction (F1,60 < 1, n.s.). Also, there is no interaction-effect (F1,60 c 1, n.s.). The mean individual (linear) trend in the choice conditions is higher than in the no-choice conditions, as predicted. Means are, respectively, 5.76 and -0.69. The means of both type of instruction levels are in the predicted direction, MR instruction: 2.78 and LR instruction: 2.28. However, as the trend analysis indicates, we should not interpret the difference as supportive of the hypothesis.

Frequency of the chosen brand. For the dependent variable frequency of chosen brand, the factor choice is highly significant (F1,58 = 13.08, p<.O01, ordinary ANOVA) and there is a significant choice x type of instruction interaction (F1,58 = 4.81, p<.05). There is no significant effect for type of risk-instruction (F1,58, p< 1, n.s.).

In the choice conditions, the mean frequency of the chosen brand is 2.31 (coefficients of orthogonal polynomials, linear trend); in the no-choice condition: -0.33. This difference is as hypothesized. When SS received an MR instruction, their mean frequency score is 1.27, while for the LR Condition, this score is 0.81. This difference is insignificant. The risk x choice interaction is significant, however.

Condition means are: 1.38, -1.00 and 0.25 for the Conditions LR/Choice, MR/No choice and LR/No choice respectively. The mean of the MR Condition (Experiment 1) was 3.25.

Discussion. In the LR/Choice Condition, the intended low risk level was not obtained. Possibly, Ss in this condition still experienced the try-out of the pens and the subsequent selection as some type of performance task (see introduction). However, this post-hoc interpretation is exactly the reverse of the initial interpretation that led to the concerning operationalization. Because it was attempted, in three of the four conditions of Experiment 2, to obtain low risk levels, no significant effect for type of instruction was to be expected. Departing from the original hypothesis that supposed low risk in all three conditions, a difference should have been observed between the condition-combination MR (Experiment 1) and LR/Choice, with the former having a significantly more positive frequency-affect relationship as measured by the two dependent variables. Now that the manipulation in the LR/Choice Condition turned out to be unsuccessful, this hypothesis had to be adapted on the basis of the manipulation checks. In the adapted version, departing from medium risk levels in both Conditions MR (Experiment 1) and LR/Choice (Experiment 2), no difference was to be expected and observed.

CONCLUSION

Summing up, three general observations can be made. The first observation is that the FE-hypothesis differs from the mere exposure hypothesis and the uncertainty reduction hypotheses with regard to the prediction of repetition-affect relationships at various perceived risk levels, and that the FE-hypothesis is supported by same evidence.

The second observation is that risk Manipulations are tricky. Even though, theoretically speaking, a low level of perceived risk is not problematic, operationally it is. It is likely that the very nature of the experimental setting is a disturbing factor here.

A final observation is that a number of questions is still open, one of them concerning the level of cognizing that is required for exposure to be functional. These questions will have to be dealt with before the implications of functional exposure for consumer behavior can be discussed.

REFERENCES

Belch, G.E. (1982), "The Effects of Television Commercial Repetition on Cognitive Response and Message Acceptance," Journal of Consumer Research, 9, 56-65.

Berlyne, D.E. (1971), Aesthetics and Psychobiology, New York: Appleton-Century-Crofts.

Berlyne, D.E. (1970), "Novelty, Complexity, and Hedonic Value," Perception and Psychophysics, 8, 779-286.

Berlyne, D.E. (1960), Conflict, Arousal and Curiosity, New York: McGraw-Hill

Harrison, A.A. (1968), "Response Competition, Frequency, Exploratory Behavior, and Liking," Journal of Personality and Social Psychology, Vol.9, 4, 363-368.

Kagan, J. (1972), 'Motives and Development," Journal of Personality and Social Psychology, 22, 51-66.

Matlin, M.W. (1970), "Response Competition as a Mediating Factor in the Frequency-Affect Relationship," Journal of Personality and Social Psychology, Vol.14, 3, 536-552.

Poiesz, Th.B.C. (1983), The Relationship between Exposure Frequency and Consumer Affect; toward a Functional Interpretation, Dissertation Tilburg University, the Netherlands.

Poiesz, Th.B.C. (1981), "Functional Exposure and Consumer Behavior: Introducing an Alternative Hypothesis for Frequency-Affect Relationships," in Mitchell, A. (Ed.), Advances in Consumer Research, Vol.9, Association for Consumer Research, 435-439.

Saywer, A.G. (1981), "The Effects of Repetition and Levels of Processing on Learning and Attitudes," in Mitchell, A. (Ed.), Advances in Consumer Research, Vol.9, Association for Consumer Research, 439-444.

Sawyer, A.G. (1977), "Repetition and Affect: Recent Empirical and Theoretical Developments," in Woodside, A.G. et al. (Eds.), Consumer and Industrial Buying Behavior, New York: North-Holland, 229-243.

Sawyer, A.G. (1974), "The Effects of Repetition: Conclusions and Suggestions about Laboratory Research, in Hughes, G.D. and Ray, M.L. (Eds.), Buyer/Consumer Information Processing, Chapel Rill: University of North Carolina Press. 190-219.

Stang, D.J. (1973), "Six Theories of Repeated Exposure and Affect," Catalog of Selected Documents in Psychology, 3, 126.

Vanbeselaere, N. (1984), "Mere Exposure: a Phenomenon in Search for an Explanation," in Doise, W. and Moscovici, S. (Eds.), Current Issues in European Social Psychology, Vol.l. Cambridge: Cambridge University Press.

Vanbeselaere, N. (1980), "Mere Repeated Exposure and Experimental Stress as Determinants of Stimulus Evaluations," Motivation and Emotion, 4, 1, 61-69.

Winer, B.J. (1971), Statistical Principles in Experimental Design, 2nd Ed., New York: McGraw-Hill.

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

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