Respondents' Moods As a Biasing Factor in Surveys: an Experimental Study

Morten Heide, Rogaland Research
Kjell Gronhaug, Norwegian School of Economics and Business Administration
ABSTRACT - This paper reports an experiment conducted to estimate the impact of subjects' mood states on their evaluations. Four experimental groups were induced by either a very negative, mildly negative, neutral or positive mood-inducing film. As hypothesized, systematic differences were found between the groups in their evaluations. Factual knowledge about the evaluation object was, contrary to expected, not found to modify mood effects. This negative finding could, however, be explained by a strong stereotypical impression of the product (a distant travel destination) held by the respondents.
[ to cite ]:
Morten Heide and Kjell Gronhaug (1991) ,"Respondents' Moods As a Biasing Factor in Surveys: an Experimental Study", in NA - Advances in Consumer Research Volume 18, eds. Rebecca H. Holman and Michael R. Solomon, Provo, UT : Association for Consumer Research, Pages: 566-575.

Advances in Consumer Research Volume 18, 1991      Pages 566-575

RESPONDENTS' MOODS AS A BIASING FACTOR IN SURVEYS: AN EXPERIMENTAL STUDY

Morten Heide, Rogaland Research

Kjell Gronhaug, Norwegian School of Economics and Business Administration

ABSTRACT -

This paper reports an experiment conducted to estimate the impact of subjects' mood states on their evaluations. Four experimental groups were induced by either a very negative, mildly negative, neutral or positive mood-inducing film. As hypothesized, systematic differences were found between the groups in their evaluations. Factual knowledge about the evaluation object was, contrary to expected, not found to modify mood effects. This negative finding could, however, be explained by a strong stereotypical impression of the product (a distant travel destination) held by the respondents.

INTRODUCTION

Response effects are factors that make the reported response differ from the actual or true value of the variables being investigated (Bradburn and Sudman 1974). Survey researchers have long been aware of the potentially disturbing impact response effects may have on self-report data. The sources of response effects have traditionally been viewed as uncontrollable environmental influences external to the individuals from which data are being collected. In recent years, however, researchers have started to look at internal, and especially psychological characteristics that may be sources of response effects (Peterson and Sauber 1983). Cognitive psychologists have found that an individual's temporary mood state, that is the transient feeling state perceived by the individual, may influence the evaluations s/he makes.

Several studies have investigated mood effects on evaluations, and he results indicate asymmetric effects of positive and negative moods. This implies that while a positive mood tends to give more favorable evaluations, the opposite will not always be true for negative moods (Isen 1984). It has been suggested that a negative mood will lead to a more pessimistic view of the world which in turn will result in less favorable evaluations, and some studies do indeed confirm this characterization. For example, Isen and Shalker (1982) found that subjects rated ambiguous slides as less pleasant when they were in a negative mood. Other studies (Griffitt 1970, Veitch and Griffitt 1976) indicate that being in a negative mood state may result in less favorable conceptions of others, and higher estimates of the likelihood of various types of disasters and unpleasant events occurring (Johnson and Tversky 1983). On the other hand, some studies do not report symmetric effects of positive and negative moods. Masters and Furman (1976) found that a positive mood was associated with higher expectations among children, but not more pessimistic expectations among children in the negative mood condition. Schwarz and Clore (1983) in their study of judgements of well-being, discovered that the negative impact of bad moods was eliminated when subjects were induced to attribute their present feelings to transient external sources irrelevant to the evaluation of their lives. Subjects who were in a good mood, on the other hand, were not affected by misattribution manipulations. Studies of person-perception judgements (Forgas et. al. 1984, Forgas and Bower 1987) clearly indicate that negative mood effects appear to be less pronounced and may depend on a variety of contextual factors. Thus, based on previous research we can conclude that while the effects of positive mood on evaluations seem to be consistently in a mood-congruent direction, the effects of negative moods are found to be more diffuse.

Srull (1984) raises the important issue of familiarity with regard to mood effects on evaluations. For instance when asked for their opinion about a low familiarity object, most people will act in accordance with a computational model. This means that they will use some form of a reference frame and "compute" their answer in accordance with this. For example, imagine that you are asked the following question: "Is Buick Regal a luxury automobile?". A computational model implies that you will compare Buick Regal with your references for "luxury automobile", and figure out an answer. For high familiarity objects it will be different, as the answer to the question has already been determined and stored in memory. The subject will use a retrieval model, which implies that s/he just has to retrieve the answer from the memory. Srull (1984) argues that mood states will be of greatest importance when dealing with low familiarity objects. For such objects the person has to "compute" an evaluation, and, as discussed above, such an evaluation may very well be biased by the present mood state. In contrast, when asked about an object with which s/he is already highly familiar, the person will often already have made an evaluation, and thus be more or less immune to the effects of temporary mood states. The importance of familiarity has been tested empirically, and the findings clearly support Srull's hypothesis (Srull 1983). For low familiarity subjects, the evaluations were biased in a mood-congruent direction, in other words, the subjects who were in a negative mood rated the products lower than the control group and vice versa for subjects in a good mood. The high familiarity subjects, on the other hand, were not affected by their temporary mood states.

The complexity of the evaluation task is another important issue. Previous research has shown that faced with a complex evaluation task, people will often employ some form of a simplifying strategy (Isen et. al. 1982). One such strategy may be to rely on an availability heuristic (Tversky and Kahneman 1973). Research has shown that mood can serve as a retrieval cue, which implies that mood-congruent material will have a memory advantage. If an availability heuristic is employed, this memory advantage may cause a mood-biased evaluation. Another strategy may be to simplify the evaluation task by using the informative function of mood states (Schwarz and Clore 1983). This function implies that the individuals use their perceived affective reactions as a basis for their evaluations, rather than computing a judgement on the basis of recalled features of the target object (Schwarz and Clore 1988). Whatever strategy is used, chances are that the severity of the mood-biases will increase in accordance with the complexity of the evaluation task.

An intuitive implication from the above literature review is that since mood states may influence people's evaluations, mood may influence people's responses and thus be a source of response effects.

Below are reported our hypotheses, target object, and research method.

HYPOTHESES

The preceding discussion suggests the following study hypotheses:

H1: When subjects are in a positive mood, we expect the evaluations to be systematically higher then when in a neutral mood, while no certain direction is hypothesized for the mood-biases among respondents holding negative moods.

The rationale for hypotheses H1 lies in the asymmetry between the effects of positive and negative moods. As discussed above, studies of mood effects on evaluations generally report mood-congruent findings for positive moods, while more diverse effects have been obtained for negative moods.

H2: We expect to find a negative correlation between mood effects and level of knowledge about the target object.

The subjects' level of knowledge about the object to be evaluated is predicted to be an important determinant of the severity of mood-biases. High level of knowledge indicates that the subject is familiar with the target object and is therefore expected to be a safeguard against mood effects, while subjects low in knowledge are anticipated to be more biased by their temporary mood states in the evaluations they make about the target object. A negative correlation between mood effects and level of knowledge is thus hypothesized.

TARGET OBJECT

Most studies of mood effects on product evaluations have focused on high familiarity products. Isen et. al. (1978) for instance, asked their subjects to evaluate products they owned. Here a low familiarity object,"Norway as a travel destination",was selected as the target object. This particular travel destination was expected to be a low familiarity "product" as previous information indicates that the familiarity with Norway, both in general and as a travel destination is fairly limited among distant foreigners. By focusing on a low familiarity product, instead of a high familiarity product, our study represents a fairly new approach.

Another reason for the choice of target object, lies in its complexity. The tourist product constitutes a rather complex combination of goods and services. In the evaluation of a travel destination, a thorough evaluator has to compare and weigh a great number of product dimensions against each other, before s/he can reach an overall judgement. As the majority of studies of mood effects on evaluations (see Gardner 1985 for a review) have focused on rather trivial objects, the present choice of product also in this respect, represents a rather new approach for research on mood effects.

METHOD

An experimental approach was preferred to examine the research problem.

Experimental Stimuli

Four mood states were induced, i.e. a very negative, a mildly negative, a neutral, or a positive mood. As discussed above, the effects of negative moods are found to be more diverse than for positive moods, which is the reason why two negative mood groups were included. The mood states were induced by exposing subjects to approx. 15 minutes of videotaped films.

(1) For the very negative mood induction, the film Night and Fog was used. This is a documentary film depicting Nazi concentration camps. With its authentic scenes of torture and genocide, the film was expected to have a very strong negative mood inducing effect.

(2) The second film, But Jack was a Good Driver was used for the mildly negative mood induction The theme here is teen-suicide. Although the film was intended to induce a negative mood, the mood inducing properties were not anticipated to be as powerful as those of Night and Fog.

(3) Neutral mood was induced by Forces and Moments, an instructional film which illustrates the way the effects of forces are used in engineering.

(4) Positive mood was induced by a 15-minutes segment of The Best of Candid Camera. This film shows people who are tricked into different awkward situations, without knowing that they are being filmed. As this is a recognized humorous film, it was expected to have a strong positive mood inducing effect.

Groups and Subjects

Four groups were used in the actual experiment. Each group was exposed to one of the films noted above. Subjects were 34 male and 31 female students at a major U.S. university. All subjects were in their junior- or senior-year and received course-credit for their participation. Male-female composition varied somewhat in the different groups with between 41% and 69% men in each group.

Procedure

Subjects were randomly assigned to one of the four conditions, all with roughly 16 participants. All sessions were conducted using the same classroom setting. After the subjects had been seated they were informed that they would be asked to complete a questionnaire, where most of the questions concerned Norway as a travel destination. However, first the subjects were asked to watch a film. They were told that the film should be used in another study, next term, but before it was decided whether or not to use the film it was important to learn people's general reaction to the film. The subjects were instructed not to memorize anything from the film, but just to watch il. The purpose of this procedure was to legitimize the mood induction procedure, and to prevent the subjects from suspecting any connection between the film and the evaluation task. The procedure has previously been used with success by Isen et. al. (1987). When talking with the subjects after they had completed the questionnaire, it was learned that they had no suspicion that the film had anything to do with the evaluation task.

Measurements

Nowlis' (1965) Mood Adjective Check List was used in the present study. This list is perhaps the most widely used mood measure (Peterson and Sauber 1983), and it has been employed in numerous studies in the last couple of decades (see e.g. Samuel 1980, Stone and Neal 1984, Hedges et. al. 1985). The list consists of 35 adjectives, where the respondent is asked to indicate to what extent each of them describes his/her current mood. A varimax-rotated factor analysis of the subjects' answers revealed two major factors, explaining 38% of the total variance. The factors could easily be interpreted as Negative affect and Positive affect, and for each subject a mood score was computed as follows:

Mood score = Factor score on the Positive affect factor - Factor score on the Negative affect factor

The subjects also completed an evaluation of 26 different aspects of the target object. For all evaluations, the following response-scale was used:

very poor    -3    -2    -1     0    1    2    3      very good

The majority of the evaluation-aspects were sampled from a previous tourist survey [The tourist survey was sponsored by the Tourist Department of the Royal Norwegian Ministry of Transportation, and was carried out in the summer of 1987. Approximately 1800 questionnaires were returned from foreign tourists on vacation in different parts of Norway.] conducted in Norway, which was done because of two reasons. First, since these variables had already been used in the survey, a comprehensive pretesting of the variables was superfluous. Second, using the sa ne variables gave us an opportunity to compare the evaluations made by the subjects with those made by tourists that had actually been to Norway. As we shall see later, this type of comparison proved to be very useful. The evaluation-aspects were related to four different domains; 1) nature, peace and quiet aspects, 2)vacation activities, 3)general aspects, 4)overall evaluation of Norway as a travel destination. The various aspects are listed in the appendix.

Following the evaluation task, several questions were included-to measure the subject's level of knowledge about the target object by using indicators such as an estimate of the population in Norway, name of the capital of Norway, three Norwegian cities, and countries bordering Norway. Furthermore, the subjects were asked to name the three most important industries, i.e. the industries with highest annual sales and pick out the Norwegians from a list of ten famous persons. The answers to the questions were used to compute a knowledge score as a measure of how familiar the subjects were with the target object. The last section of the questionnaire was comprised of questions about the film (the answers to these questions proved to be very useful in a later study where the same films were employed). Also information about gender was gathered.

ANALYSIS AND RESULTS

Manipulation Check

Table 1 presents the mean mood scores in the four groups. As we can see from the table, the observations correspond with the expected effects. The film, Night and Fog appears to have had a powerful negative effect on the subjects' mood, while the negative effect of But Jack was a Good Driver seems to be more moderate. In accordance with our expectations Forces and Moments does not seem to have any mood inducing properties, while a positive mood is induced by The Best of Candid Camera.

TABLE 1

MOOD SCORES IN THE FOUR EXPERIMENTAL GROUPS

An analysis of variance (ANOVA) was performed on the data. The analysis contrasting the four different film conditions indicated a significant effect of the mood manipulations, F(3,60) = 46.54, p < .001. Subsequent one-tailed t-tests revealed that the subjects in the Candid Camera group after they had watched the film, were in a significantly more positive mood than was the case for the neutral control group, t(29) = 3.77, p < .001. Subjects in the Night and Fog group had a significantly more negative mood, t (30) = 9.14, p < .001. As for the effect of the film "But Jack was a Good Driver", the effect on mood although not significant (t (29) = 1.32, p= .097) was in the predicted direction. Thus, the manipulation check data confirm the expected mood induction properties of the four films.

Mood Effects on Evaluations

It was hypothesized that the evaluations would differ in the various groups. For positive mood, a mood-congruent bias in evaluations was hypothesized, which implies that more positive evaluations in the positive mood group than in the neutral control group were expected. Because Of the inconsistent effects of negative moods, reported in past research, no specific direction for the mood-biases was hypothesized for the two groups exposed to negative mood induction.

In our first analysis we computed the grand-mean of all evaluations for each subject. By looking at the grand-mean random fluctuations in the individual variables should be eliminated, and any true differences between the groups should thus be easier to detect. An analysis Of variance (ANOVA) contrasting the four groups revealed a significant difference between groups with regard to grand-mean evaluations, Fl 1,,61) = 2.998, p < .05. To investigate the directions of the mood-biases, the evaluations in the positive and the two negative mood groups were compared with those in the neutral control group. A graphical illustration of the first comparison is presented in figure 1.

On the X-axis we find the various aspects of Norway which the subjects were asked to evaluate. On the Y-axis the group-mean ratings in the positive and the neutral groups are shown. A visual inspection of figure 1 indicates that the evaluations seem to be generally higher in the positive mood group. Of the 26 evaluations, 20 are highest in the positive group, while only four are higher in the neutral group. For the remaining two evaluations, the group-means are identical in the two groups. Because of the relatively low number of participants (16 in each group), only one of the evaluations (#1) shows a significant difference (p < .05) between the two groups. Computing the grand-mean of all evaluations, we found this figure to be 1.28 in the positive mood group as opposed to .95 in the neutral control group. A one-tailed t-test established that the difference between the two groups was significant, t(30) = 2.25, p < .05. The results support hypothesis H1 concerning a mood-congruent evaluation-bias in the positive mood group.

Another way of testing the hypothesis, is to look at the evaluations as an approximate binomial series. The results from running a non-parametric sign test show that the probability that at least 20 out of 24 evaluations (i.e. the 24 evaluations where the group-means differ) are highest in the positive mood groups is less than .001 (one-tailed) if the two groups were drawn from the same population. In other words, it is not very likely that the group-differences were obtained by chance. In a binomial series, there is a requirement that the variables should be independent of one another. In the experiment, some of the evaluation-variables are correlated, and this violates the independence-assumption. To what extent this will influence the binomial probabilities reported above, is difficult to say, but it is likely that it will increase the probability somewhat.

Because of the widely reported asymmetry between the effects of positive and negative moods, we did not predict any mood-congruent mood-biases for negative moods. Still it will be interesting to compare the evaluations in the two negative mood groups with those in the neutral control group. A comparison between the mildly negative group and the neutral group showed that as many as 23 of the 26 were highest in the mildly negative mood group. The probability of this occurring at random is .0002 (two-tailed). The grand-mean of all evaluations was significantly higher in the mildly negative group, 1(30) = 3.50, p < .01 (two-tailed).

FIGURE 1

GROUP-MEAN EVALUATIONS FOR EACH ASPECT IN THE POSITIVE MOOD GROUP VS. THE NEUTRAL CONTROL GROUP

TABLE 2

INTERCORRELATIONS BETWEEN THE GROUP-MEAN EVALUATIONS IN THE DIFFERENT GROUPS

By comparing the very negative mood group with the neutral control group, we found that 19 of the 26 aspects were rated highest in the very negative group, while the remaining 7 were highest in the control group. The two-tailed binomial probability is .03. A comparison of grand-mean evaluations showed a higher grand-mean in the very negative mood group, but the difference was not significant, t(31) = 1.13, p > .25 (two-tailed). The findings indicate mood-incongruent evaluation-biases for both of the negative mood groups.

Response Pattern

In Figure 2 we have reported the group-means for each aspect to be evaluated. It is evident that the overall response pattern is very similar in all groups, although mood biases have been traced.

An inspection of figure 2 reveals that all groups seem to follow the same pattern. By inspecting the intercorrelations between the different group-means the following is observed:

From table 2 it can be seen that the means in the various groups are highly intercorrelated. This confirms that the response pattern is generally the same in all groups. By treating each group as a ease and each group-mean evaluation as an observation, the Kendall's coefficient of concordance was computed to get an estimate of the similarity of the ratings between the four groups. The computed coefficient of concordance (W) was .8658 (p<.0001), which means that the four groups are fairly similar in their ratings of the different aspects of the target object.

TABLE 3

VARIATIONS IN EVALUATIONS BY GROUP-DIFFERENCES AND KNOWLEDGE

TABLE 4

VARIATIONS IN EVALUATIONS BY GROUP-DIFFERENCES AND INTERACTION WITH KNOWLEDGE

The Impact of Knowledge

A negative correlation between knowledge about the target object and mood effects on evaluations was hypothesized (H2), which implies that subjects with a high level of knowledge about Norway, were expected to be less influenced by their current mood in their evaluations about Norway. To test the hypothesis an analysis of covariance was performed, using the knowledge score as a covariate to determine how much of the variance in the grand-mean evaluation could be explained by differences in knowledge about Norway.

From table 3, it is obvious that the knowledge score does not have any significant explanatory power as a covariate (F= .15, p = .703). An analysis of variance was included to determine if the interaction between group and knowledge score was able to explain a significant part of the variance in the mean of all evaluations. The output from this analysis is presented below.

The interaction term between group and knowledge score is reported on the last line. As we can see the F-value for the interaction term is only .43 and definitely not significant (p= .735). The analyses show that the knowledge score is not able to explain any significant part of the variation in the mean of all evaluations, either as a covariate or in interaction with group. Thus, our prediction that subjects high in knowledge about the target object would be less likely to be influenced by their current mood states in their evaluations, did not come true.

DISCUSSION

In the experiment, we compared one positive mood group and two negative mood groups with a neutral control group to see how the subjects' evaluations were influenced by their mood states. Subtle but consistent mood biases for both positive and negative moods were detected. The biases were in a mood-congruent direction for the positive mood group, as hypothesized. For negative mood, no direction was hypothesized, but the results indicate mood incongruent mood effects for both negative mood groups. Even though there were differences between the various groups, the overall response pattern seemed to be the same in all groups. It was also hypothesized that subjects familiar with the object, i.e. respondents high in knowledge about Norway would be less likely to be influenced by their current mood states in their evaluations. In the experiment, however, no support for this hypothesis was found. Can this negative finding be explained?

FIGURE 2

GROUP-MEAN EVALUATIONS FOR EACH ASPECT

In their answers to the knowledge questions, the subjects revealed very limited knowledge about the target object. The findings show that most subjects were unable to name any cities or major industries in the country. Nor had they any knowledge about the population size of Norway, its neighboring countries, neither were they able to identify famous people from the country. Consequently, out of a possible score of 17, the mean knowledge score was only 3.7 (with st. dev. = 2.3). Based on the subjects' modest knowledge-level it is tempting to classify "Norway as a travel destination" as a low familiarity product. As mood effects, according to Srull (1984), will be greatest when subjects evaluate low familiarity products, it is very surprising that knowledge was not found to be an important determinant in how sensitive the subjects were to mood effects. It was also expected that the mood-biases that were detected in the evaluations should be more severe.

There are, however, reasons to believe that we just cannot equate knowledge-level with familiarity. When looking at the evaluations made by the subjects, it is found that aspects like weather and the possibilities of having a reasonably priced vacation in Norway are rated extremely low, while possibilities for hiking and possibilities of experiencing clean and undisturbed nature are highly rated aspects. These ratings are very much in line with the general strengths and weaknesses of the Norwegian tourist product. This indicates that the subjects, even though they know little about Norway in general, nonetheless are familiar with Norway as a travel destination.

As mentioned earlier, some of the aspects that were evaluated by the subjects were sampled from a survey conducted among actual tourists in Norway (see footnote 1). To investigate further how familiar the subjects in the experiment were with Norway, a comparison with the actual tourists' evaluations was made. A graphical presentation of the comparison is found below.

Figure 3 compares the mean scores of all subjects in the experiment with the mean scores of approximately 1800 foreign tourists visiting Norway for various aspects of the travel product. It is evident that the response-patterns are very similar. The correlation between the means in the two groups is r = .704, p < .01. An intuitive interpretation of this finding is that the subjects in the experiment hold a strong stereotype about Norway as a travel destination, and that this stereotype is very much in line with the attitude of tourists that have actually visited Norway. It is not surprising that the mood effects were found to be as modest as was the case in the experiment, as the subjects could use the stereotype as basis for their evaluations, and thus were less influenced by their temporary mood states. In other words, even though the subjects' level of factual knowledge about the target object was low, they could act in accordance with a retrieval model (Srull 1984) by simply retrieving the evaluations from memory.

FIGURE 3

COMPARISON BETWEEN SUBJECTS IN THE EXPERIMENT AND ACTUAL TOURISTS

IMPLICATIONS

Our study raises several important implications. In the experiment we found consistent mood biases on evaluations. This indicates that mood may be considered as a biasing factor in surveys, and thus constitute a source of response effects. However, although the mood biases wore present, the overall response pattern was very similar in all groups. We recall that the coefficient of concordance was higher than .86, and taking the small sample sizes into account (roughly 16 subjects in each group), any higher concordance would be difficult to obtain even if the groups were drawn from the same population, simply because of sampling variance. Our conclusion is therefore that even if the subjects' evaluations seem to be influenced by their mood, the damaging impact of the mood effects seems to be rather minimal.

The second implication is that our results seem to bolster the already well-documented findings about asymmetric effects between positive and negative mood states. As hypothesized we found mood-congruent effects for the positive mood group, i.e. the evaluations were biased in an upward direction. For the two negative mood groups, the results indicated mood-incongruent effects. The evaluations in the suicide-film group were significantly higher than those in the neutral control group. For the Night and Fog group too, the evaluations, although the differences were not very significant, appeared to be more favorable than was the case for the control group. By including two instead of only one negative mood group, we have reduced the probability that the incongruent results occurred by chance only. What are the mechanisms by which these effects occurred? According to Isen (1985) there may be both a motivational and a cognitive interpretation of the asymmetric effects of positive and negative moods. The motivational interpretation is that of positive affect maintenance and negative-affect repair. Seen in relation to our study, this explanation suggests that subjects who have been exposed to the positive mood inducing film will try to retain their good feelings, while subjects in the negative film groups will try to chase "the blues" away. According to the cognitive interpretation the explanation may be that cognitive material associated with negative affect may be structured differently from that associated with positive affect. For example, depressing material may come to be less well elaborated and interconnected in the cognitive system than positive (Isen 1985). The idea of focusing on arousal instead of valence may also help to explain our findings. Our results indicate that all three affect groups yield higher evaluations than the neutral control group. It may then be that arousal and not hedonic tone is the key issue for explaining the observed asymmetry.

The third implication our study raises is the issue of familiarity. In the experiment, the subjects' factual knowledge about the target object was found to be rather marginal. We had predicted that subjects with a high level of knowledge would be less likely to be influenced by their current mood states in their evaluations, but no support was found for this hypothesis. This could have lcd us to conclude that product-familiarity was of little importance. However, by comparing the subjects' evaluations with those made by actual tourists in Norway, we were able to determine that the subjects had a clear and reasonably stereotypical impression of the target object (Norway as a travel destination). In other words, even though the subjects had little factual knowledge about Norway, their familiarity with the travel product was fairly high. This finding has the important implication that we just cannot always equate familiarity with level of knowledge, as it may occur that the stereotype and not the mood states provided most of the basis for respondents' evaluations. It may be argued that the lack of impact of knowledge resulted from a floor effect -exhibited by knowledge itself. However, because Of the clear stereotype we do not consider this alternative explanation to be very probable.

In this study we have found that mood states may bias evaluations and thus constitute a source of response effects in surveys. However, if the respondents are familiar with the object they are asked to evaluate, or if they have a clear stereotype of the object, chances are that the mood biases will be rather marginal, and thus not represent a serious threat to the validity of the survey measurements. In the current study only films have been used as mood induction stimuli. II may be argued that films trigger a number of different processes and that it is therefore difficult to be sure that it is mood change per se that causes the evaluation changes. To eliminate the possibility of a mono-method bias the study should thus be replicated using other types of mood induction procedures, preferably procedures that are fairly clean, i.e. they only manipulate mood. Basing our conclusions on only one target object (Norway as a travel destination) is problematic, because much of what we have found may result from the clear stereotypes and the low knowledge levels described above. The study should therefore be replicated using different target objects, different involvement levels, different prior knowledge levels, and so on. Furthermore, due to the asymmetry between positive- and negative moods the relationship between mood states and mood effects on evaluations appears to be a nonlinear one. In other to estimate the true relationship future studies should employ a large number of stimuli to induce moods of different intensities, and measure their effects on evaluations. Finally, to increase external validity the current study should be replicated in field-settings using random samples of the general population.

APPENDIX

REFERENCES

Bradburn, Norman M. and Seymour Sudman (1974), Response Effects in Surveys: A Review and Synthesis, Chicago: Aldine.

Forgas, Joseph P., Gordon H. Bower, and S. Krantz (1984), "The Influence of Mood on Perceptions of Social Interactions," Journal of Experimental Social Psychology, 20, 497-513.

Forgas, Joseph P. and Gordon H. Bower (1987), "Mood Effects on Person-Perception Judgements," Journal of Personality and Social Psychology, 53 (1), 53-60.

Gardner, Meryl P. (1985), "Mood States and Consumer Behavior: A Critical Review," Journal of Consumer Research, 12 (December), 281-300.

Griffitt, William (1970),"Environmental Effects on Interpersonal Affective Behavior: Ambient Temperature and Attraction," Journal of Personality and Social Psychology, 15 (3), 240-244.

Hedges, Susan M., Lina Jandorf, and Arthur A. Stone (1985),"Meeting of Daily Mood Assessments," Journal of Personality and Social Psychology, 48(2), 428-434.

Isen, Alice M. (19842,'Toward Understanding the Role of Affect in Cognition," in Handbook of Social Cognition, eds. Robert Wyer and Thomas Srull, Hillsdale, N. J.: Erlbaum, 179-236.

Isen, Alice M. (1985),"Asymmetry of Happiness and Sadness in Effects on Memory in Normal College Students: Comment on Hasher, Rose, Zacks, Sanft, and Doren," Journal of Experimental Psychology: General, 114(3), 388-391.

Isen, Alice M., Thomas Shalker, Margaret Clark, and Lynn Karp (1978),"Positive Affect, Accessibility of Material in Memory and Behavior: A Cognitive Loop?," Journal of Personality and Social Psychology,. 36, 1-12.

Isen, Alice M. and Thomas Shalker (1982),'The Influence of Mood State on Evaluation of Positive, Neutral, and Negative Stimuli: When You 'Accentuate the Positive,' Do You 'Eliminate the Negative'?," Social Psychology Quarterly, 45, 58-63.

Isen, Alice M., Barbara Means, Robert Patrick, and Gary Nowicki (1982),"Some Factors Influencing Decision-Making Strategy and Risk Taking," in Affect and Cognition, eds. Margaret Clark and Susan Fiske, Hillsdale, N. J.: Erlbaum, 243-261.

Isen, Alice M., Kimberly Daubman, and Gary Nowicki (1987),"Positive Affect Facilitates Creative Problem Solving," Journal of Personality and Social Psychology, 52 (6), 1122-1131.

Johnson, Eric J. and Amos Tversky (1983),"Affect, Generalization, and the Perception of Risk," Journal of Personality and Social Psychology, 45 (1), 20-31.

Masters, John and Wyndol Furman (1976),"Effects of Affective States on Noncontingent Outcome Expectancies and Beliefs in Internal or External Control," Developmental Psychology, 5, 481-482.

Nowlis, Vincent (1965), "Research With the Mood Adjective Checklist," in Affect, Cognition, and Personality, eds. S. S. Tomkins and C. E. Izard, New York: Springer, 352-389.

Peterson, Robert and Matthew Sauber (1983),"A Mood Scale for Survey Research," In 1983 AMA Educator' Proceedings, eds. Patrick Murphy et. al., Chicago: American Marketing Association, 409-414.

Samuel, William (1980),"Mood and Personality Correlates of IQ by Race and Sex of Subject," Journal of Personality and Social Psychology, 38(6), 993-1004.

Schwarz, Norbert and Gerald Clore (1983), "Mood, Misattribution, and Judgements of Well-Being: Informative and Directive Functions of Affective States," Journal of Personality and Social Psychology, 45 (3), 513-523.

Schwarz, Norbert and Gerald Clore (1988), "How do I Feel About it? The Informative Function of Affective Stales," in Affect, Cognition and Social Behavior, eds. Klaus Fiedler and Joseph Forgas,. Lewiston. N. Y.: C. J. Hogrefe, 44-62.

Srull, Thomas (1983),"The Role of Prior Knowledge in the Acquisition, Retention, and Use of New Information," in Advances in Consumer Behavior, Vol. 10, eds. Richard Bagozzi and Alice Tybout, Ann Arbor, MI: Association for Consumer Research, 572-576.

Srull, Thomas (1984)," The Effects of Subjective Affective States on Memory and Judgment," in Advances in Consumer Behavior, Vol. 11, ed. Thomas Kinnear, Provo, UT: Association for Consumer Research, 530-533.

Stone, Arthur A. and John M. Neale (1984),"Effects of Severe Daily Events on Mood," Journal of Personality and Social Psychology, 46(1), 137144.

Tversky, Amos and Daniel Kahneman (1973),"Availability: A Heuristic for Judging Frequency and Probability," Cognitive Psychology, 5, 207-232.

Veitch, R. and Griffitt, W. (1976),"Good News-Bad News: Affective and Interpersonal Effects," Journal of Applied Social Psychology, 6 (1), 69-75

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