mood efects on evaluative judgment

CO G N IT IO N AN D EMO TIO N, 1998, 12 (6), 783± 805 Effects of M ood on Evaluative Judgements: In¯ uence of Reduced Processing Capacity and Mood Salience Matthias Siemer Free University Berlin, Germany Rainer Reisenzein University of Bielefeld, Germany Schwarz and Clore (1983) proposed that the effects of mood on evaluative judgem ents are due to people’s use of a ``feeling heuristic’ ’ . Results of the present study suggest that this heuristic is particularly likely to be used under conditions of reduced processing capacity, induced by time pressure and com peting task dema nds, as both factors intensi® ed the effects of moo d on evaluative judgements. In add ition, previous ® ndings that increasing the salience of a judgem ent-irrelevant cause disrupts the effects of mood on evaluative judgem ents were replicated. A ll of these effects were, however, obtained only when mood was salient to the participants, suggesting that to be effective, mood must exceed a threshold of salience. Taken together, the ® ndings further support the hypothesis that at least in som e situations, the effects of moods on evaluative judgem ents are based on a controlled inference strategy, rather than on automatic priming effects. INTRODUCTION Positive and negative m oods have been shown to have a var iety of effects on cognitive processes. Particularly well established are the effects of mood on evaluative judgem ents, including satisfaction with consum er go ods (Isen, Shalker, Clark , & C arp, 19 78), satisfaction with life in general (e.g. Schwarz & C lore, 19 83), evaluatio n of others (e.g. Berkowitz & Troccoli, 199 0; E rber, 199 1; F iedler, Pam pe, & Scherf, 1986 ), health-related cognitions (Salovey & Birnbau m, 1989) , an d evaluation of activities (C unningham, 1988 ; for reviews see Clore, Schwarz, Request s for reprints should be sent to Matthias Siemer, D epartme nt of Psycholog y (WE 08 ), Free U niversity Berlin, 10195 Berlin, Habelsch werdter Allee 45, Germ any; e-mail: masie mer@ zedat.fu-berlin .de. q 1998 Psychology Press Ltd 784 SIEMER AND REISENZEIN & C onway, 199 4; Fo rga s, 1994, 1995; Isen, 1984 ; M orris, 1989; Schwarz, 19 90). A prom inent exp lanation of these ® ndings focuses on the effects of m oods on the recall of va lenced material from m em ory (e.g. Bower, 19 91; Isen, 198 4). This exp lanation rests on the following assumptions: (1) the evalu ative judgem ents in question are based on an integration of recalled positive an d negative featu res of the eva luated object; (2) individuals typ ically do not retrieve all potentially judgement-relevan t inform atio n from m emory, but base their judgem ent on a subset of this informatio n that is m ost accessible (cf. Bodenhausen & Wyer, 198 7; Sherm an & Corty, 1984; Wyer & Srull, 1986 ); an d (3) m ood-congruent m em ory (e.g. Blaney, 1986 ; M orris, 198 9) cau ses m ood-congruent inform atio n to be over-represented in this subset. These assum ptions entail that eva luations should be biased in a m ood-congru ent m anner. Further enhancing this effect, inform atio n that com es to m ind ® rst, or that can be easily retrieved, m ay be m ore heavily weigh ted in the judgement process (see e.g. K ahnem an , Slovic, & Tversky, 1982) . Finally, at least in the case of com plex an d extended social judgem ent processes, the increased accessibility of m ood-congruent concepts an d associations could bias the subsequent interpretation of the judgem ents through this route. This ``mood-congruent retrieval’ ’ or ``m ood-priming’ ’ explan ation of the effect of m ood on evaluative judgem ents has been questioned by Schwa rz and Clore (1983; see also C lore et al., 19 94; Schw arz, 19 90, p. 52 9), who proposed an alternative explanation that ``focuses on the informative function of affective states in controlled inference processes, rather than on the au tom atic process of m ood-congru ent retrieval’ ’ . M ore speci® cally, Schwarz and Clore proposed: (1) W hen m aking evaluative judgements, individuals often do not use presum ably effortful an alytic judgem ent strategiesÐ such as recalling the var ious positive and negative attributes of the object and then forming a sum m ary judgem entÐ but rely on a sim plifying judgem ent m ethod called the ``feeling heuristic’ ’ : They base their judgements primarily on their perceived affective reactions to the object. That is, individ uals ask them selves, in effect: ``How do I feel about the object?’ ’ and base their evaluative judgem ent on the answer to this question. (2) Becau se moods, other than emotions, are global in character (e.g. Frijda, 1993; M orris, 19 89)Ð that is, they are not directed at speci® c objects (e.g. A rm onJones, 1991; Lorm an d, 1985; O atley & Johnson-Laird, 1987) Ð people who are in a positive or negative m ood may m istake their preexisting feeling at the tim e of judgement as an affective reaction to the judged object. In other wo rds, they may misattribute their m ood partly or EFFECTS OF MOOD ON EVALUATIVE JUDGEMENTS 785 wholly to the object. (3) As a consequence, their evaluative judgem ents 1 are in¯ uenced in a mood-congruent m an ner. During recent years, evidence has accumulated that favo urs this ``feelingheuristic’ ’ explanatio n of the effect of m ood on evaluative judgem ents, at least for situations in which people rely on judgem ental heuristics rather than on m ore elab orate, extensive judgem ent strategies (cf. Fo rga s, 1994 , 1995 ). Probably the m ost important supportive ® ndings (see C lore et al., 1994 and Schwarz, 1990 for m ore detailed review s) are the following two. First, it has been found that the effect of negative mood on evalu ative judgem ents can be reduced or elim inated if the trueÐ or a plausible alternativeÐ cau se of the m ood (e.g. the weather) is m ade salient to the participan ts (Keltner, Locke, & Audrain, 1993; Schwarz & Clore, 19 83; Schwarz, Servay, & Kum pf, 198 5). Second, the effect of m ood on judgements seem s to be independent, to a large degree, of the speci® c content of the m aterial used to induce the mood (Clore, Schwarz, & K irsch, 19 83, quoted in Schwarz, 1990 ; Johnson & Tversky, 1983 ). These ® ndings are in accord with feelingheuristic theory, because according to this theory, the effect of affective states on evaluative judgem ents depends prim arily on their informational valu e for judgin g the target; whereas the content of the m aterial used to induce the affects should be irrelevan t as long as it does not discredit that inform ational va lue. In contrast, according to the theory of mood-congruent retrieval, the key factor in mood effects on evaluative judgements is the content of the retrieved inform ation, whereas the cau sal attribution of the m ood should be irrelevant. However, as pointed out by Forgas (1994 , 1995) , the two exp lan ations of the effects of m oods on eva luative judgem ents are not necessarily contradictory, becau se they may apply under different circum stances. Indeed, this assum ption is a central tenet of an integrative m odel, the so-called ``affect infusion m odel’ ’ , proposed by this author. A ccording to Fo rgas, the m oodcongruent retrieval m odel applies in those situations where people engage in substantive and elaborate processing, whereas the feeling-heuristic model applies wh enever subjects are not m otivated or able to engage in extensive processing and therefore sim plify the task by using the feeling heuristic. It m ay be noted that this theory of the effects of m ood on evaluative judgem ents shows a strong form al analogy to the theory of excitation transfer (Zillmann, 1978), whic h is itself a specialisation of Schachter’s (1964) two-factor theory of emotion s to the case of sequentially occurrin g states of arousal (Reisenzein 1983). According to excitation transfer theory, em otional reaction s are intensi® ed by misattributed residual physiolog ical arousal stemm ing from a prior inducer (cf. e.g. Cantor, Zillmann, & Bryant, 1975; Reisenzein & Gattinger, 1982 ). Althoug h excitation transfer theory deals with the effects of residua l arousal on emotions, whereas Schwarz ’s (1987) theory deals with the effects of moods on evaluative judgement, the two theories share the assumpt ion that the predict ed effects are m ediated by casusa l attributions (or m isattributions) of, respect ively, arousal or mood. 1 786 SIEMER AND REISENZEIN The experiment reported in this article is a further empirical test of the feeling heuristic exp lan ation of the effects of m oods on evalu ative judgem ents under circumstances where people are likely to engage in heuristic processing. The study had three go als. The ® rst go al was to test a further prediction that has been derived from the feeling heuristic model but has not yet been empirically tested. Schwa rz (1990) assum ed that the use of the feeling heuristic is a controlled rather than an au tom atic process (in the sense of Shiffrin & Schneider, 197 7; see also Bargh, 1994), and that one of the m ain reaso ns for using the feeling heuristic instead of an an alytic judgement strategy is that doing so simpli® es the judgem ent task . If so, one should exp ect that the feeling heuristic is increasingly relied on under conditions that lim it the availab le central processing capacity and that, as a result, the effects of moods on evaluative judgem ents should be enhanced under such circumstan ces (Schwarz, 199 0; an analogous prediction can be derived from the affect infusion model for heuristic processing situations; see Forgas, 1995) . N ote that this is a strongly discrim inative prediction of the feelingheuristic m odel; that is, one that sets it off clearly from the m oodcongru ent retrieval m odel (for a given typ e of processing situation). The reason is that, according to the latter m odel, m ood effects should be most pronounced when judgements are m ad e in an elab orated, extensive m anner that require a high amount of processing resources, becau se of the greater am ount of mood-congruent m aterial that is accessed (cf. Forgas, 19 95). In the present study, this prediction wa s tested by using two stan dard m ethods of reducing ava ilable processing capacity: tim e pressure and a com peting task . It was expected that the effects of m oods on evaluative judgem ents wo uld be enhanc ed under theses circumstances. The second goal of the present study was to further clarify the role of salience of mood in effects of m ood on evaluative judgem ents (m oods are the m ore salient, the more they are at the focus of a person’s attention or aw areness). Research on the effects of cue salience in other dom ains has dem onstrated that salient cues are particularly in¯ uential in determining judgem ents (e.g. Sherm an & Corty, 1984). Extrapolating from these findings, Schwa rz (1987, 1990) proposed that salience of mood also plays an im portan t role in m ood effects on evaluative judgem ents. Speci® cally, Schwa rz suggested that, the m ore salient m ood is relative to other cues potentially relevan t to the judgements in question, the m ore heavily weigh ted is the information that it provides in the judgem ent process and therefore, the m ore pronounced is its judgem ental effect. In fact, according to Schwarz (e.g. 19 87, p. 91) other sources of information m ay even be entirely ign ored in the presence of a salient m ood. Some, alb eit rather indirect, support for these sugge stions was provided by Strack , Schwarz, an d G schneidinger (198 5), who found that participants relied m ore strongly EFFECTS OF MOOD ON EVALUATIVE JUDGEMENTS 787 on their m ood state in evaluatin g their current life satisfaction if they were asked to report a past positive or negative life event in an emotionally involving style (a procedure that presumably m ade their mood salient to them ). On the other han d, results of a more recent study by Keltner et al. (1993, study 4) sugge st that high salience of mood may under certain conditions reduce rath er than enhance m ood’s in¯ uence on evaluative judgem ents. In one exp erim ental condition of this study, the participan ts were put into a negative m ood and were then asked to re¯ ect on their current emotions an d their cau ses, and to lab el the em otions. In this condition, the effect of negative m ood on evaluative judgem ents was found to be reduced (relative to a control condition) to the sam e degree as in two other conditions, in wh ich the participan ts were induced to attribute their m oods to an alternative, judgement-irreleva nt cau se. The interpretation of these ® ndings is, however, am biguous, because in the ``salien t mood’ ’ condition of the Keltner et al. study, not only the m ood but also its caus e was m ad e salient. That is, as m entioned, the participants were not only asked to lab el their current em otions, but also to re¯ ect on their causes. It may be safely assum ed that m ost participants cam e to regard the just-completed m oodinduction procedure as an im portant cause of their current mood. The elimination of m ood effects in these experim ental conditions could therefore again have been due to the cau sal attribution of the mood, as was presumably the case in the explicit attribution conditions and in the earlier studies by Schwa rz an d Clore (1983) and Schwa rz et al. (1985). To disentangle the effects of mood salience and m ood attribution, in the present study an attem pt was m ade to man ipulate these two factors independently of each other. U nravellin g these effects is im portant because, if it were found that salience of m ood is alone suf® cient for eliminating mood’s in¯ uence on judgem ents, this could provid e for an alternative explan ation of the ® nding that increasing the salience of an alternative cau se elim inates the effects of moods on evalu ative judgem ents (e.g. Schwarz & C lore, 1983; Schwarz et al., 1985) . One need only assum e that in these exp erim ents, making people awa re of a plausible cause of their m ood also made them aw are of their negative m ood itself. According to this alternative explanation, then, the active ingredient of the attributional m anipulatio ns used in previous exp eriments was their attention-directing effect, rath er than the cau sal inform atio n that they provided (see also M orris, 19 89, for a sim ilar view ). Finally, and conjointly with the second goal of the study, our third aim was to replicate and clarify one of the central ® ndings advanced by Schwarz (1990) in support of the feeling-h euristic model: Namely, that m aking an alternative cau se of the mood salient to participants reduces or eliminates the effects of m oods on eva luative judgements (Schwarz & Clore, 1983; Schwarz et al., 1985; see also Keltner et al., 1993) . 788 SIEMER AND REISENZEIN In sum, the following m ain hypotheses were tested (they were assum ed to hold at least for evaluative judgem ents of the type exam ined in the present experim entÐ judgem ents of life satisfactionÐ that presum ably do not elicit an extensive processing style): 1. T he effect of mood on evaluative judgements is enhanced under conditions of time pressure and com peting task dem an ds. 2. T he effect of m ood on eva luative judgements is enhanced if m ood is made salient to the participants. 3. T he effect of mood on evaluative judgements is reduced or elim inated if a judgem ent-irreleva nt cau se of the m ood (here: the mood-induction procedure) is m ade salient to the participan ts. METHOD Participants Subjects were 80 (50 fem ale, m ean age 26.4) introductory psychology students at the Free U niversity Berlin, who participated in partial ful® lment of their study requirem ents. Data from three participants were lost becau se of equipm ent malfunction, reducing the ® nal sample size to 77. Design A lthough the described hypotheses called for speci® c com parison of particular experim ental conditions (see Results for details), for reaso ns of subject economy, an d because we wanted to explore possible additional interactions between the factors of interest, a ® ve-factorial split-plot design consisting of three between-subjects factors an d two within-subjects faco rs was used. All factors had two levels. Between-subjects factors were mood (positive vs. negative), m ood attribution (in¯ uenced by m ak ing the m oodinduction procedure, a judgem ent-irrelevant m ood cause, salient vs. nonsalient), an d salience of moo d (high vs. low). W ithin-subjects factors were tim e pressure (present vs. absent) an d secondary task (present vs. absent). The subjects were ran dom ly assign ed to the exp erim ental treatm ents, with 10 or 9 (due to the m issing data) subjects in each cell. The main dependent variab le was the m ean of a set of evaluative judgements related to satisfaction with different life areas. We selected these judgem ents because they can, on the one hand, be m ad e in an analytical, elaborate fashion (by retrieving from m emory, and integrating, releva nt inform ation about the respective life areas that were to be eva luated), but are at the sam e tim e likely candidates for sim pli® cation by a feeling heuristic, either EFFECTS OF MOOD ON EVALUATIVE JUDGEMENTS 789 spontaneously or under additional processing demands (cf. Schwarz, 1987 ). Altogether, 48 satisfaction item s were used. They were selected from a pool of 10 0 such item s m odelled after similar ones used by Borg (1987) as those items that, according to the results of pretesting, correlated 2 relatively strongly with simultaneously m easu red mood. In addition, for exp loratory purpose, the latencies of the judgem ents were m easu red (see Results for details). Procedure The participants were tested individually. They were inform ed that they would participate in two independent experiments that for econom ic reaso ns were conducted in tan dem. The ® rst exp erim ent allegedly concerned the m ental representation of au tobiographical memories. In truth, it served to induce positive versus negative m ood (Factor 1), as well as to m anipulate the perceived cause of the mood (Factor 2). M ood was induced through a com bination of two m ethods found to be effective in previous investigations: Recallin g personal m em ories of positive versus negative events (e.g. A bele, 1990 ; Schwarz & Clore, 19 83), and listening to happy versus sad music (e.g. Kenealy, 198 8; Parrott & Sab ini, 1990) . It wa s though t that the combinatio n of both m ethods would be particularly effective, becau se Spies, H esse, G errards-Hesse, and U ef® ng (1991) found this to be the case for the com bination of music with the read ing of selfreferenced statem ents (i.e. the technique developed by Velten, 196 8). The participants were asked to revisualise, as concretely as possible, a very positive or a very negative life exp erience and to w rite it dow n while at the sam e tim e listening to happy versus sad music (care was taken to assure that the participan ts were convin ced of the an onym ity of their reports). Attribution of mood (Factor 2) wa s man ipulated as part of the exp lanation give n for the simultaneous music presentation. In the m ood m isattribution (or nons alient irrelevant caus e) cond ition the participants were only told that the exp erim enter was interested in the effects of different kinds of music on autobiographical memories. It was expected that in this condition, participants would m isattribute part of their m ood to their considering the various life areas that were presented to them as part of the evalu ative judgem ent task. In the control condition (or salient irrelevan t cause condition), the participan ts were informed that happy versus sad music has been found to put people into a positive versus negative m ood, which in turn would help them to recall positive versus negative life experiences. It was exp ected that making the true but judgem ent-irrelevant mood cause salient 2 A copy of the satisfaction item s is available from the ® rst author on request. 790 SIEMER AND REISENZEIN wo uld prevent mood m isattribution from occurring (cf. Schwarz & Clore, 19 83). Post-experim ental interviews indicated that these cover stories were accepted by all participants; also, all but three subjects suspected no relation between the ``two exp erim ents’ ’ , an d the three who did had erroneous hypotheses. The music consisted of a selection of pieces developed and tested for the purpose of m ood induction by Pign atiello, Cam p, and Rasa (1986) . The m ood-induction procedure took ab out 15 m inutes. Im m ediately after the mood induction, the third between-subjects factor (m oo d salience) was manipulated. Subjects in the high m ood salience condition were asked to com plete a mood questionnaire consisting of scales for happiness, sad ness/depression, an xiety, and anger (seven item s each). This m ethod of increasing m ood salience, which wa s previously used and found to be effective by Rothkopf an d Blaney (1991), has face valid ity: A sking som eone to introspect an d to report on his/her m ood seem s to be the most direct possible method of getting the person to focus attention on his/her current m ood and to m ake it an object of thought; hence this m ethod m akes the m ood salient alm ost by de® nition. At the same tim e, the m ood scale also served as a man ipulatio n check for the effectiveness of the mood-induction procedure. Subjects in the low mood salience cond ition did not receive the mood questionnaire. Im m ediately afterward, the alleged ``second’ ’ experiment was conducted. It was presented as a study of the effects of different answer conditions on judgem ents related to life satisfaction. Participan ts were seated in front of a personal com puter. After three practice item s, the 48 experimental satisfaction item s (e.g. ``How satis® ed are yo u with yo ur social relationships?’ ’ ) were individ ually presented on the computer screen. Below each item , a 6-p oint rating scale an chored in the lab els not at all satis® ed (1), to very satis® ed (6) was shown. T he participants had to an swer the item s by entering a num ber from 1 to 6. To control possible effects of individ ual differences in reading speed, as well as differences in item length, all item s were presented in slow reading speed (60m sec per letter). The subjects were also inform ed that additional instructions would be presented on the computer screen during the experim ent, but to prevent possible effects of prior know ledge of the task -dem and conditions on the results, these instructions were not speci® ed further. Time pressure (Factor 4) and com peting task (Factor 5) were manipulated w ithin subjects by presenting 12 of the 48 items each under the four possible combinations of these two factors. The resulting four experimental conditions were presented in a ® xed order that re¯ ected increased taxing of cognitive resources. That is, the ® rst 12 items were presented under the notim e pressure/ no com peting task condition. They were followed by 12 m ore items that were presented under conditions of a com peting task (on ly). Fo llow ing Logan (1979 ), an instruction appeared on the screen asking EFFECTS OF MOOD ON EVALUATIVE JUDGEMENTS 791 participants to m em orise the subsequently appearing num ber. A 5-digit num ber was then presented for 5 seconds, followed by the 12 judgem ent item s; subsequently, the number was to be recalled, but no performance feedback was give n. Prior to the third batch of 12 items, tim e pressure was m anipulated (tim e pressure on ly conditions). The participants were: (a) instructed to answer as quickly as possible from now on; an d (b) a maxim um response time w indow was de® ned individually for each subject. The tim e window was de® ned as 80% of each subject’s average response tim e on the ® rst 24 item s, but in no case less than 2sec. O ne second before the m axim um response time was reach ed, the participants were warned both optically an d aco ustically; if no response occurred within the response window, a second tone was sounded, a m essage appeared on the screen that the maxim um response tim e had been exceeded an d the next item was presented. The ® nal 12 satisfaction item s were to be answered under conditions of both time pressure an d secondary task conditions (time pressure plus competing task cond ition ). W hich of the 48 items appeared under the four experim ental conditions was determined separately for each subject by randomisation. The described order of task demands was chosen to allow us to clearly distinguish the effects of cognitive load from any potential effects of changes in m ood intensity across the exp erim ent. Becau se cognitive load increased during the exp erim ents, the feeling heuristic hypothesis predicts a parallel increase of m ood effects on evaluative judgem ents in the course of the exp erim ent. In contrast, mood intensity, and thus any effects of intensity on eva luative judgem ents, should if anything decrease in the course of the exp erim ent. The present test of the cognitive load hypothesis is thus conservative. However, note that the total judgem ent task lasted only from 8 to 10 minutes an d was thus well within the duration of m ood effects usually assum ed in previous research (see e.g. Isen, 1984) . At the end of the exp erim ent, the participants were tested for suspicion and were carefully debriefed, an d the notes of their autobiographical m em ories were returned to them . RESULTS Effectiveness of Mood Induction The effectiveness of the m ood m anipulation wa s checked by com paring the 20 participants in the positive m ood condition who had completed the mood questionnaire as part of the m anipulation of mood salience with the 20 participants in the corresponding negative mood condition. The four scales, of the m ood questionnaire (happiness, sadness/depression, 792 SIEMER AND REISENZEIN an xiety, and anger) had internal consistencies (Cronbach’s alpha) ranging from .90 (an xiety) to .96 (positive and negative m ood). A s expected, subjects in the positive m ood condition were signi® cantly happier (M = 5.35) than those in the negative m ood condition [M = 2.63, t(37) = 4.0, P < .001 (onetailed)], corresponding to an effect size of d = 1.35; whereas subjects in the negative mood condition were signi® cantly sadder (M = 3.61) than those in positive m ood condition [M = 1.46 , t(37) = 3.43, P < .001 (one-tailed)], d = 3 1.13 . There were no signi® can t differences between the two mood groups with regard to an xiety [F(1,37 ) < 1], and an ger [F(1,37) = 1.8, P > .18]. H ence, the m ood induction was not only effective, but also speci® c, inasmuch as it did not affect an ger an d fear. In ad dition, there was no signi® cant interaction between m ood condition and causal information on the happiness and sadness/depression subscales [Fs(1,35 ) < 1], indicating that the different instructions concerning the reason for playing the music did not affect the effectiveness of the m ood-induction procedure. Therefore, any effects of mood attribution cannot be explained by differences in m ood (cf. Schwarz & C lore, 19 83). The hypotheses concerning m ood effects on evaluative judgem ents formulated in the introduction were tested by m eans of focused contrasts (i.e. com parisons of m ean s or com binations of means) w ithin the context of the ® ve-factorial design (for details see e.g. Rosenthal & Rosnow, 198 5; Ro snow & Rosenthal, 198 9). This procedure, rather than the m ore standard overall an alysis of va riance, was chosen for two m ain reasons. First, in contrast to AN OVA , it allows one to test directed m ain an d interaction effects (H orton, 19 78), which have a higher power than their undirected Because we m anipula ted m ood salience by m eans of a m ood question naire, it is clear that we could not use the questio nnaire as a m anipulat ion check in the low salience group. N onetheless, one can be fairly certain that the m ood m anipulation was also effective in the latter group: Because the m ood induction procedur e was the sam e in both mood salience groups and subjects were randoml y assig ned to the groups, there are no systematic differen ces in the empiric al distribut ions of subject s variables between groups (i.e. the two groups can be rega rded as sam ples from the sam e populati on). Consequ ently, any differences between the groups with rega rd to the effects of the m ood-indu ction procedu re must be due to the random assignm ent itself (``errors of random isation’ ’ , cf. Siem er, 1997). Speci® cally, lack of effectiveness of the m ood induction in the low salience condition must have been due to the chance allocation of a dispropo rtiona lly hig h num ber of subjects to this condition who were insensitive to the m anipulat ion. Based on conventi onal power analysis (Cohen , 1988), the likelihood for this to happen is equal to b , the probability of a non signi® cant difference in the low m ood salience group given the obtaine d effect siz e in the high mood salience group. As noted, this effect siz e (for the happy m ood sca le) is d = 1.35 in the sam ple. Using this sam ple statistic as an estimate of the population effect size, the probabili ty of obtainin g a signi® cant differen ce in a sam ple of N = 40, that is the power (1± b ), is .96 for an a of .01 (cf. Faul & Erdfelder, 1992 for comput ational details). Apart from this consideration, m ood also ha d a signi® cant effect on processi ng latency in the low sa lience group. 3 EFFECTS OF MOOD ON EVALUATIVE JUDGEMENTS 793 AN OVA counterparts. A ccordingly, all directional a-p riori hypotheses could be tested using one-tailed tests. Second, the use of a-priori contrasts reduces the num ber of independent statistical tests, an d thus the danger of typ e I errors. In the present case, the standard approach would have m eant a ® ve-factorial A NOVA with 31 independent statistical tests, which m akes the possibility of type I errors a serious problem . In sum , although being slightly nonconventional, the use of focused contrasts provided for greater statistical power (by allowing one-sided tests), and reduced the probab ility of type I errors (by restricting the number of statistical tests to those being of theoretical im portance). (In addition, this approach also allows to m atch precisely the statistical hypotheses tested to the theoretical predictions.) H owever, in order to not overlook ad ditional effects for which no explicit hypotheses were stated, we did additionally conduct an explo ratory ® vefactorial A NOVA . Replication of Previous Findings We expected, in accord with previous ® ndings: (a) that m ood wo uld in¯ uence the life-satisfaction judgem ents, that is, they would be more positive under positive than negative mood; and (b) that this effect wo uld be reduced or eliminated if the judgem ent-irrelevant cause of the m ood (the m ood-induction procedure) is mad e salient to the participan ts (this is hypothesis 3 stated in the Introduction; cf. Schwarz & Clore, 1983 ; Schwarz et al., 198 5). Exp ressed in terms of statistical effects, this hypothesis amounts to an interaction between mood (positive/negative) an d m ood attribution (m isattribution vs. no misattribution, or low vs. high salience of the m ood cause), such that the difference between positive and negative m ood groups is less pronounced if the m ood cau se is made salient to the participants. We thought initially that the conditions best comparable to those used in previous studies would be the low m ood salience conditions; therefore, the two hypotheses were ® rst tested for the latter conditions. Unexpectedly, the predictions were not con® rm ed for these conditions (cf. F ig 1); Evaluatio ns under positive mood were not signi® cantly m ore positive (M = 3.56 ) than under negative m ood [M = 3.6, t(69) < 1], and the exp ected interaction between m ood an d salience of the cau se wa s not sign i® can t [t 4 969) < 1]. Support for the predictions wa s obtained, however, in the high As m entione d, this t-test formally correspo nds to the AN OVA test of the interaction between mood (positive/negative) and m ood attribution (low/hig h salie nce of mood cause); however, in constras t to the interaction F-test, the t-test is directional (see Rosentha l & Rosnow, 198 5, for the com putation al details). An analogo us com ment pertain s to the other ``interaction’ ’ t-tests reported later. 4 794 SIEMER AND REISENZEIN m ood salience conditions: In these conditions, the judgements were sign i® cantly m ore positive in the positive mood condition (M = 3.79) than in the negative mood condition [M = 3.31, t(69) = 3.0 4, P < .00 1 (one-tailed), d = 1.12]; an d there was a m argin ally signi® cant interaction between m ood 5 an d salience of the cause [t (35) = 1.59, P < .07 (one-tailed), d = .54]. As shown in Fig 1, this interaction had the predicted form , that is, there wa s a smaller difference between positive and negative m ood in the high salient cau se condition (M = 0.27) than in the low salient cau se condition (M = 0.68 ). A dditional planned comparisons within the two m ood conditions showed a sign i® cant difference between the low and high salient cause for the negative m ood group [M s = 3.10 vs. 3.52, t(69) = 2.13, P < .05 (onetailed), d = 1.07], but not for the positive m ood gro up [t(1,69 ) < 1]. If one assumes that both the positive and the negative m ood inductions were effective in changing m ood aw ay (in opposite directions) from the preinduction baseline (as is sugg ested, e.g. by the ® ndings of Abele, 199 0), the form of the obtained interaction (Fig. 1) could be taken to indicate that a salient cau se elim inated only the effects of negative but not of positive m ood. Although in the ab sence of a no-mood-induction control gro up, this inference must rem ain speculative, it would be in accord with previous 6 ® ndings (Schwarz & Clore, 1983; Schwarz et al., 1985) . As already stated, the obtained results were in one respect counter to expectations, nam ely, inasm uch as we exp ected to obtain them preferably in the low m ood salience conditions, because these conditions seemed to be best com parable to the conditions prevailing in previous studies in wh ich m ood effects on eva luative judgem ents have been obtained. This expectation was based on the fact that an explicit man ipulatio n of m ood salience was hardly if ever used in these studies. H owever, as detailed in the D iscussion, a re-exam ination of the previous studies sugge sts that in most, m ood was probably salient to the participan ts. Therefore, it is the high m ood salience conditions of the present experiment that correspond m ost closely to these earlier studies. In any case, becau se standard m ood effects on evaluative judgem ents were obtain ed only in the high salience m ood conditions, the predicted moderating in¯ uence on this effect of time pressure an d com peting task demands can also be expected for this condition only. A ccordingly, the processing load hypothesis was tested separately for the high and low m ood Accordin g to Cohen (1988), d = .20 is considered to be a sm all effect; d = .50 a medium effect; and d = .80 a large effect in experimental contexts. 6 Accordin g to Schwarz (1987, 1990), this asymm etry of the effects of causal inform ation is due to the fact that positive affective states are viewed as requirin g less expla nation than negative ones (see also Liu, Karasawa, & Weiner, 19 92; Weiner, 19 85). It is assumed that, if no explanation for one’s feelings is sought to beg in with, directing attention to a potentia l cause of the feelings is unlikely to show any effect. 5 EFFECTS OF MOOD ON EVALUATIVE JUDGEMENTS 795 FIG. 1. Satisfaction judgem ents under conditi ons of low salience of mood (LSM ), high salience of m ood (H SM ), low salience of m ood cause (L SC), and hig h sa lience of mood cause (H SC). salience conditions, with the expectation that it wo uld be con® rm ed at least for the high m ood salience conditions. Effects of Reduced Processing Capacity According to hypothesis 1 (see the Introduction), the effect of m ood on evaluative judgem ents is enhan ced under conditions of reduced processing capacity. In term s of statistical effects, this hypothesis amounts to an interaction between m ood an d processing load, such that the difference between positive and negative m ood is high er given time pressure or secondary task . For the high m ood salience condition, this prediction wa s con® rmed, both for time pressure [t(69) = 1.92, P < .05 (one-tailed), d = .76], andÐ although only m argin ally sign i® can tÐ for secondary task [t(69) = 1.59 , P < .06 (one-tailed), d = .44]. A s shown in Fig 2, the mean differences between evaluative judgements in the two mood conditions were higher give n tim e pressure (M = .31 ) than given no tim e pressure (M = .14), and 796 FIG. 2 Effects of secondary task dem and and time pressure on satisfaction-judgem ent differences between happy and sad mood group under conditions of high salience of mood. EFFECTS OF MOOD ON EVALUATIVE JUDGEMENTS 797 higher given a secondary task (M = .28 ) than give n no secondary task (M = .18). One m ight have exp ected these effects to be m ore pronounced in the nonsalient cause conditions than in the salient cause conditions, and the results pointed indeed in this direction; however, the relevant statistical comparisons were not signi® can t. T his ® nding dovetails w ith the hypothesis that a salient cau se disrupted m ood effects only partly (in the negative m ood condition). No sign i® cant effects were obtained for the low mood salience conditions (both P > .25). As noted before, the described order of task dem an ds was used to rule out the possibility that the obtain ed effects were cau sed by a decrease of m ood intensity across the experim ent. To check whether such changes had indeed occurred, the ® rst half of the satisfaction item s w ithin each processing condition was com pared with the second half in a 2 (® rst/second) by 2 (positive/negative mood) AN OVA . A decrease of mood intensity over tim e would be re¯ ected in an interaction between these two factors to the effect that the differences between positive and negative m ood are sm aller for the second half of the item s. No such effect wa s found: B oth the interaction and the main effect of item s were nonsign i® can t, F s < 1. Effects of Mood Salience The results obtained for mood salience were already described im plicitly in the foregoing paragraphs. To sum marise: A lthough the overall directed contrast corresponding to the interaction between m ood and m ood salience was signi® cant [t(69) = 2.59, P < .01 (one-tailed), d = .62 ], indicating that m ood had stronger effects on evaluative judgem ents under conditions of high m ood salience, single contrasts revealed , as mentioned, that the ® ndings of previous studies were replicated only for the high mood salience condition. In addition, the predicted boosting effect of reduced processing capacity was also obtained only for the high mood salience condition. Although the com plete lack of effects in the low mood salience condition was unexpected, the obtained ® ndings are in accord w ith hypothesis 2 (see the Introduction), stating that the effects of mood would be enhan ced under conditions of high mood salience. To round off the statistical an alysis of the satisfaction judgements, they were subjected to a split-plot AN OVA including all ® ve exp erim ental factors with the aim of exp loring other potential effects that had not been predicted. This an alysis revealed no effects that have not already been m entioned, with the exception of a m arginally (P < .08) fourfactorial interaction between mood salience, mood cause, tim e pressure, and secondary task that we refrain from interpreting further. 798 SIEMER AND REISENZEIN Analysis of Response Latencies A s m entioned, Schwarz and C lore (1983; Schwarz, 19 90) proposed that one of the main reasons for using the feeling heuristic rather than an an alytic judgem ent strategy is that the feeling heuristic sim pli® es the judgem ent task. If so, it seems not unreasonable to expect that, other factors constant, participants who use the feeling heuristic need less tim e to m ake eva luative judgem ents than those who do not. The results reported earlier could be taken to sugge st that the feeling heuristic was used m ore under conditions of high m ood salience; if so, response times should be shorter under these conditions. H owever, this hypothesis was not supported by the data: There was no sign i® cant reaction tim e (RT) difference between the high an d low mood salience condition [t(69) < 1]. A n additional, exploratory AN OVA revealed (apart from a trivial effect of tim e pressure) a sign i® cant m ain effect of m ood (positive vs. negative ) on response latency [F(1, 69) = 5.24, P < .05], quali® ed by a sign i® cant interaction between m ood and m ood salience [F(1, 69) = 8.62, P < .01]. Follow-up tests showed that participan ts in the low mood salience condition responded sign i® cantly m ore quickly when in positive than in negative m ood [F(1, 69 ) = 13.7 4, P < .001 ], whereas no difference in response tim e was obtained between positive an d negative mood for participan ts in the high mood salience condition. Based on the results of previous studies (e.g. F iedler, 1988 ; Schwarz, Bless, & Bohner, 1991 ), a possible, alth ough speculative explanation of these ® ndings could be that in the low mood salience conditions, participants in a bad m ood needed m ore time to make the judgem ents because they were m ore ``analytic’ ’ and careful than participants in a go od mood. A lternatively, participants in a bad m ood m ay have been m ore preoccupied with their feelings an d as a result m ay have had availab leÐ or m ay have been w illing to devoteÐ less processing capacity for the judgem ent task, which had to be compensated by increased processing time (cf. Klauer, Siemer, & Sto ber, 19 91). U nder conditions È of high mood salience, in contrast, both positive an d negative participants used the sim plifying feeling heuristic, which obliterated any processing time differences that existed between positive an d negative 7 m oods. N ote that we did not se lect the judgem ent task speci® cally to exam ine the impact of m ood on processin g capacity, which was not a major concern of this study. There was no interaction of m ood effects on latency with capacity effects on m ood-con gruency. 7 EFFECTS OF MOOD ON EVALUATIVE JUDGEMENTS 799 DISCUSSION Concentrating ® rst on the high m ood salience condition, we found, in accord w ith predictions: (a) that m ood affected eva luative judgem ent an d that this effect wa s disrupted by mak ing the judgem ent-irrelevant cause of the m ood salient to the participan ts; an d (b) that the effects of mood were enhan ced under conditions of tim e pressure an d competing task dem ands. H ence, a crucial piece of evidence presented by Schwarz (1987, 1990) in favour of the ``feeling-heuristic’ ’ m odel of the effects of moods on evalu ative judgement was replicated, and additional predictions derived from that m odel were con® rm ed. It may be noted that the ® ndings concerning the effects of ad ditional processing dem an ds and m ood attribution are not only dif® cult to explain by: (1) the m ood-priming hypothesis described in the Introduction. The total pattern of results also rules out potential alternative explanations of (parts of) the obtained effects; such as (2) that salience of m ood (m an ipulated by presenting a mood questionnaire) exerted its effects exclusively on the basis of a seman tic-prim ing effect, that is, through m aking the con cept of the respective m ood(s) more availab le (cf. e.g. Clore et al., 1994 ; Wyer & Srull, 1989) ; or (3) that the processing load effects, too, were due to m ood 8 m isattribution. It should be em phasized, however, that the enhancem ent effects of additional dem ands on mood-congruency were obtained in a situatio n that encourage d the use of judgem ents heuristics. The ® ndings, therefore, do not necessarily extend to situatio ns that encourage the use of substantive, elab orate judgement strategies (as, e.g. dem an ded by complex task s). In fact, results by Forgas (199 2) suggest that subjects in such circumstances show greater mood-congruency the more elaborate the judgem ents are. These results lead us to suspect that in these circum stances, a reduction of processing capacity will lead to effects contrary to those found in the present study. In fact, this prediction seems to rather directly follow from Forgas (199 4, 19 95) integrative affect-infusion model of mood effects. A s m entioned, Fo rgas assumes that, whether m ood-congru ency effects on evaluative judgem ents are based on a m ood-congru ent prim ing m echanism or on a feeling heuristic depends crucially on whether the judgem ent process is substantive and elaborate versus sim ple and heuristics-based. Consequently, the effects of a reduction of processing capacity on m oodcongruency of evaluative judgem ent should also depend on the kinds of According to this hypothes is, the mood effects on evaluative judgem ents increased over time because, as tim e passed, the particip ants attribute d their mood increasingly less to the prior inductio n. This suggetion is incom patible with the obtaine d effect of mood attribut ion and the lack of a decreas e of this effect across tim e. 8 800 SIEMER AND REISENZEIN judgem ents at issue. In situations where evaluative judgements are based on elaborate, extended reasoning processes and m ood-congruency is due to (the greater scope of) prim ed material incorporated in the judgem ent, an increase in processing load should reduc e the m ood-congruency effect because it interferes w ith the judgement process. In contrast, in cases where subjects tend to process material in a heuristic m anner to begin with, as in the present exp erim ent, an increase in processing load should further increase reliance on the feeling heuristic, and therefore the m oodcongru ency effect should be enhanc ed. If this reasoning is accepted, the affect infusion model can handle both our present ® ndings and those of Fo rga s (1992). H owever, as said , the effects of reduced processing capacity on the m oodcongru ency effect were found only under conditions of high mood salience: O nly under these conditions were the results of previous studies replicated, an d the predictions concerning the effects of increased processing load con® rm ed. Although this pattern of results attests to the effectiven ess of the m ood salience man ipulatio n, and although the enhan cement of mood effects on judgem ents implied by this result was predicted and is in line with the feeling heuristic m odel, the total ab sence of effects in the low mood salience conditions wa s unexpected and needs further discussion. The sim plest exp lan ation would be that the enhan cement of mood salience also increased its intensity. M an ipulatio n of attentional focus has been found in previous studies to intensify both positive an d negative m ood (e.g. Scheier & C arver, 1977), an d the experience of pain (Arntz, D ressen, & M eckelbach, 1991). H owever, this exp lanatio n presupposes that the mood induction used in the present study was so weak that, without the ad ditional, enhancing effect of the salience m anipulation, it remain ed ineffective. This seem s to be unlikely, give n the strong effects of the m ood induction on the mood scale docum ented for the high mood salience conditions. An alternative, and in our view m ore plausible explan ation is that m ood has an effect on evaluative judgem ents only if it exceeds a certain m inimum value, or threshold, of salience; an d that in the high mood salience conditions of the present study, as well as in previous studies in which mood effects were found, mood exceeded this threshold, whereas it remain ed below the threshold in the low salience condition. This hypothesis would certainly be in accord with the feeling-h euristic model, because the feeling heuristic is assum ed to be a controlled rather than an automatic process; therefore, a minimum degree of attention to one’s mood seem s to be required for m ood to have an effect on judgem ents. Recall that, according to the feeling-h euristic model, m oods serve as inform ation for m aking evaluative judgem ents, inform ation that is in principle processed in EFFECTS OF MOOD ON EVALUATIVE JUDGEMENTS 801 the sam e way as is any other piece of inform ation (Schwarz, 1990) ; speci® cally, infor mation that is not heeded should have no effect. Nonetheless, the question rem ains why mood was not suf® ciently salient in the low salience conditions of the present exp erim ent to be taken into consideration in the judgement process. W ith hindsight, the m ost plau sible exp lanatio n for this is in our view that, in contrast to most previous studies of a com parable natu re, the details of our experim ental procedure were such that, in the low m ood salience conditions, they strongly diverted the participants’ attention aw ay from the previously induced m ood. Recall that the mood induction an d the ensuring evaluative judgement task were presented as parts of two entirely different and unconnected exp erim ents; during the ``two studies’ ’ , the participants were seated at different locations in the experimental room ; and the com puterised assessm ent of the evalu ative judgem ent appears to have been a highly attention-attracting. A reexa mination of previous studies in which m ood effects on evalu ative judgem ents were obtained sugge sts that in these studies, the participan ts’ attention was m ost likely less strongly diverted from the previously induced m oods. In fact, in som e of these studies, the participan ts’ attention seem s (inadvertently) to have been directed at their mood. Thus, in some studies, participants were asked to com plete a mood questionnaire as a m anipulation check prior to m aking the evaluative judgements (e.g. L evine, Wyer, & Schwarz, 1994) , and in other studies (e.g. Borg, 1987 ; Fiedler et al., 1986 ; Kavan agh & Bower, 198 5; M acLeod & Campbell, 1992) , the participants were explicitly instructed to try to induce in them selves a particular mood. Additional support for the hypothesis that mood m ust exceed a certain threshold of salience to be effective com es from studies by Parrott an d Sabini (199 0) an d Rothkopf an d Blaney (1991) on the effects of mood on m em ory for valen ced m aterials. Parrott and Sabini (1990) found m oodcongruent mem ory effects only if the participants them selves attem pted to induce the m ood and were aw are of the relevance of m ood to the study (if these conditions were not m et, mood-incongr uen t m emory effects were obtained). In line with our reaso ning, this could be attributed to increased m ood salience in the form er condition, particularly because the participan ts in these studies were also asked to complete a mood questionnaire after the m ood induction. Rothkopf an d Blaney (1991) found increased m ood congruency of autobiographical m emories of depressed people if the participants com pleted Beck’s D epression Inventory prior to the recall of m em ories. For m ale participan tsÐ who were assum ed to be characterised by a lower degree of chronic self-attention to their feelings than fem alesÐ the prior completion of the depression inventory was even a necessary condition for obtaining a mood-congruency effect. W hatever the exact reaso ns for the lack of effects in the low m ood salience conditions, the present results indicate that the salience of m ood 802 SIEMER AND REISENZEIN an d mood attribution need to be distinguished not only conceptually, but also in term s of their effects. Speci® cally, the obtained results speak again st the possibility, raised in the Introduction, that increasing the salience of m ood migh t be suf® cient to eliminate mood effects on evaluative judgem ents, an d that this attentional effect might have been the active ingredient of previous misattribution (e.g. Schwarz & C lore, 19 83). O n the contrary, the present ® ndings sugge st that salience of m ood m ay be a necessary precondition for both the standard mood and the m isattribution effects to occur. N onetheless, a potential lim iting condition of the effects of mood salience must be noted. N am ely, if people’s attention is directed toward their moods in such a m an ner that they are also encourage d to re¯ ect ab out the causes of their mood, then the net result could be a reduction or elim inatio n of m ood effects. The reason is that in this way, an extran eous m ood cau se (e.g. the m ood-induction procedure) could spontan eously becom e salient and thus could prevent m ood m isattribution from occurring. Incidentally, this wa s precisely the reaso n why Strack et al. (1985) refrain ed from asking subjects to com plete a mood questionnaire (as a m anipulation check) prior to the evaluative judgem ents. This suggestion wo uld also be in accord with the results of previous studies that support an analogous effect of increased salience of residual aro usal on em otions (e.g. C antor et al., 197 5; Reisenzein & Gattinger, 198 2). Although this possibility was obviously not realised in the present exp eriment, an im portan t goal of future studies should be to determ ine in more detail the conditions under which increasing the salience of m oods results in an intensi® cation versus a reduction of the effects of moods on evaluative judgem ents. For the tim e being, however, it may be concluded that the ® ndings of the present study lend further support to the sugge stion that the effects of m oods on evaluative judgem ents are not necessarily ``autom atic’ ’ , inevitable effects, but can be based on (Schwarz, 1990 , p. 53 9): ``high ly inferential strategies, which m ay result in au gm entation and discounting effects as described by the most `reaso ned’ m odels in socialcognitive research, although these inferential steps m ay not necessarily be accessible to introspection’ ’ . M anuscrip t received 7 M arch 1997 Revised manuscr ipt received 10 February 1998 REFERENCES Abele, A. (1990). D ie Erinnerung an positive und negative Lebensere ignisse. U ntersuch ungen zur stim mungsin duzieren den Wirkung und zur G estaltun g der Texte [The recollection of positive and negative life-events. Studies on mood effects and descriptions of EFFECTS OF MOOD ON EVALUATIVE JUDGEMENTS 803 positive vs. negative life-events]. Zeitschrift fu r Experim entelle und Angewandte PsychoÈ logie, 37, 181± 207. Arm on-Jones, C. (1991). Varieties of affect. New York: Harvest er Wheatsh eaf. Arntz, A., Cressen , L., & M eckelbach, H . (1991). Attention , not anxiety, in¯ uences pain. Behavior Therapy and Research, 29, 41± 50. Bargh, J.A . (199 4). The four horsema n of automa ticity: Awareness, intention, ef® ciency, and control in social cognition. In R.S. Wyer Jr. & T.K . Srull (Eds.), H andbook of socia l cognition: Vol. 1. Basic processes (pp. 1± 40) (2nd ed.). Hillsdale, NJ: Erlbaum . Berkowit z, L., & Troccoli, B.T. (199 0). Feelings, direction of attention, and expresse d evaluations of others. Cog nition and Emotion, 4, 305± 325. Blaney, P.H . (1986 ). Affect and mem ory: A review. Psychological Bulletin, 99, 229± 246. Bodenh ausen, G.V., & Wyer, R.S. (1987). Social cognition and social reality: Informat ion acquisit ion and use in the laborator y and the real world. In H .J. H ippler, N. Schwarz , & S. Sudm an (Eds.), Social inform ation processing and survey methodology (pp. 6± 41). Berlin: Springer. Borg, I. (1987). The effect of m ood on differen t types of well-being judgme nts. Archiv fu r È Psychologie, 139, 181± 188. Bower, G.H. (1991). Mood congruity of social judgmen ts. In J.P. Forgas (Ed.), Emotion and so cial judgment (pp. 31± 53). Oxford, UK : Pergam on Press. Cantor, J.R., Zillmann, D., & Bryant, J. (1975). Enhance m ent of experienced sexual arousal in respon se to erotic stimuli through m isattribution of unrela ted re sidual excitation. Journal of Personality and Social Psychology, 32, 69± 75. Clore, G.L ., Schwarz, N., & Conway, M . (19 94). Affective causes and consequ ences of social information processin g. In R.S. Wyer & T.K . Srull (Eds.), Handbook of social cognition (Vol. 1, pp. 32 3± 417) (2nd ed.). Hillsdale, NJ: Erlbaum . Clore, G.L., Schwarz , N., & K irsch, J. (1983). Generalized mood effects on evaluative judgments. Pape r presen ted at the m eeting of the M idweste rn Psyc hologic al Associa tion, Chic ago. Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). H illsdale, NJ: Erlbaum . Cunning ham , M.R . (1988). What do you do when you’re happy or blue? M ood, expectancies, and behaviora l interest . M otivation and Emotion, 12, 309± 331. Erber, R. (1991). Affective and sem antic prim ing : Effects of mood on category accessibility and inference. Journa l of Experim ental Social Psychology, 27, 480± 498. Faul, F., & Erdfelder, E. (1992). GP OW ER: A priori, post hoc, and compromise power analyses for MS -DOS [com puter program] . Bonn, FRG : Bonn Univers ity, Departm ent of Psychology. Fiedler, K . (1988). Em otional mood, cognitive style, and behavior regulation. In K . Fiedler & J.P. Forgas (E ds.), A ffect, cognition, and social behavior (pp. 100± 119). G o ttingen: H ogrefe. È Fielder, K ., Pam pe, H., & Scherf, U. (1986). M ood and m emory for tightly organiz ed social information. European Journal of Social Psychology, 16, 149± 164. Forga s, J.P. (199 2). On bad m ood and pe culia r people : Affect and person typica lity in im pression formatio n. Journal of Personality and Social Psychology, 62, 863± 875. Forgas, J.P. (1994). The role of em otion in social judgm ents: An introduc tory review and an Affect Infusion M odel (AIM ). Special Issue: Affect in socia l judgm ents and cognitio n. European Journal of Social Psychology, 24, 1± 24. Forgas, J.P. (19 95). M ood and judgme nt: The Affect Infusion M odel (AIM ). Psychological Bulletin, 117, 39± 66. Frijda, N.H. (1993). M oods, em otion episode s, and emotion s. In M . Lewis & J.M . Haviland (Eds.), H an dbook of emotions (pp. 381± 403). N ew York: G uilford Press. Hig gins, E.T. (1989). K nowledg e accessibility and activation: Subjectivity and suffering from 804 SIEMER AND REISENZEIN unconsc ious sources. In J.S. Ulem an & J.A. Bargh (Eds.), Unintended thought (pp. 75± 123). N ew York: G uilford. H orton, R.L. (1978). The general linear model. N ew York: M cG raw-H ill. Isen, A.M . (19 84). Toward understa nding the role of affect in cognitio n. In R.S. Wyer & T.S. Srull (E ds.), Handbo ok of social cognition (Vol. 3, pp. 179 ± 236). Hillsdale, N J: Erlbaum . Isen, A.M ., Shalker, T.E., C lark, M .S., & K arp, L. (1978). Positive affect, accessibility of m aterial in m emory, and behavior: A cognitive loop? Journa l of Personality and Social Psychology, 36, 1± 12 . Johnson, E.J. & Tversky, A. (198 3). Affect, generalization, and the percept ion of risk. Journal of Personality and Social Psychology, 45, 20± 31. K ahnem an, D., Slovic, P., & Tversky, A. (1982). Judgment under uncertainty: Heurist ics and biases. Cam bridge, U K: Cambrid ge Universi ty Press. K avanagh, D.J., & Bower, G.H . (1985). M ood and self-ef® cacy: Im pact of joy and sadness on perceived capabilities. Cognitive Therapy an d Research , 9, 507± 52 5. K eltner, D., Locke, K .D., & Audrai n, P.C. (1 993). The in¯ ue nce of attribu tions on the relevance of negative feelings to persona l satisfaction. Perso nality and Social Psychology Bulletin, 19, 21± 29. K enealy, P. (1988 ). Va lidation of a m usic m ood induct ion procedu re: Some preliminary ® ndings. Cognition and Emotion, 2, 41± 48. K lauer, K .C., Siemer, M ., & Sto ber, J. (1991). Stim mung und Leistungsniveau bei einfac hen È Aufgaben. [M ood and performa nce level of sim ple tasks]. Zeitschrift fu r Experim entelle È und Angewandte Psychologie, 38, 379± 393. Levine, S.R ., Wyer, R.S. Jr., & Schwarz, N. (1994). Are you what you feel? The affective and cognitive determ inants of self-jud gm ent. Specia l Issue: Affect in social judgements and cognitio n. European Journal of Social Psychology, 24, 63± 77. Liu, J.H ., Karasawa , K., & Weiner, B. (199 2). Inferenc es about the causes of positive and negative em otions. Personality and Social Psychology Bulletin, 18, 603± 615. Logan, G.D. (19 79). On the use of a concurre nt mem ory load to m easure attention and autom aticity. Journal of Experim ental Psychology: H uman Perception and Perform ance, 5, 189± 207. Lormand, E. (1985). Toward a theory of m oods. Philosophical Studies, 47 , 38 5± 407. M acLeod, C., & Cam pbell, L. (1992 ). M em ory accessibility and probability judgme nts: An experim ental evaluation of the availability heuristi c. Jo urnal of Personality and Social Psychology, 63, 890± 902. M orris, W.N. (1989). M ood. The fram e of mind. Berlin: Springe r. Oatley, K ., & Johnson- Laird, P.N. (1987). Towards a cognitive theory of em otions. Cognition and Emotion, 1, 29± 50. Parrott, W.G., & Sabini, J. (1990). M ood and m em ory under natural conditions: Evidence for m ood incongru ent recall. Journal of Personality and Social Psychology, 59, 321 ± 336. Pignatiello, M .F., Cam p, C.J., & Rasar, L.A. (19 86). M usical m ood inductio n: An alternative to the Velten mood techniq ue. Journa l of Abnorm al Psychology, 95, 295± 297. Reisenzein, R. (19 83). The Schachter theory of em otion: Two decades later. Psychological Bulletin, 94, 239 ± 264. Reisenzein, R., & Gattinger, E. (1982 ). Salience of arousal as a mediator of misattri bution of transferred excitation. Mo tivation and Emotion, 6, 315± 328. Rosentha l, R., & Rosnow, R.L. (1985). Contrast analysis: Focused compa risons in the ana lysis of variance. Cam bridge, U K: Cam bridge Universi ty Press. Rosnow, R.L., & Rosentha l, R. (1989). Statistic al procedur es and the justi® cation of knowledge in psycholo gical science. Am erican Psycholog ist, 44, 1276± 1284. Rothkopf, J.S., & Blaney, P.H. (1991). M ood congrue nt m emory: The role of affective focus and gender. Cognition and Emotion, 5, 53± 64. EFFECTS OF MOOD ON EVALUATIVE JUDGEMENTS 805 Salovey, P., & Birnbaum , D. (1989 ). In¯ uence of m ood on health-relevant cognitio ns. Journal of Personality and Social Psychology, 57, 539± 551. Schachter, S. (1964). The intera ction of cognitive and physiolog ical determinants of em otional state. In L. Berlowitz (Ed.), Advances in experim ental social psychology (Vol. 1, pp. 49± 80). New York: Academic Press. Scheier, M .F., & Ca rver, C.S. (19 77). Self-focused attention and the experience of emotion : Attraction, repulsion, elation, and depressio n. Journal of Personality and Social Psychology, 37, 62 5± 636. Schwarz , N. (1987). Stimmung als Inform ation. [Feeling as information]. H eidelberg: Springe r. Schwarz , N. (199 0). Feelings as informat ion: Inform ational and motivational function s of affective states. In E.T. H iggins & R.M . Sorrenti no (eds.), H an dbook of motivation and cognition (Vol.2, pp. 527± 561). New York: G uilford Press. Schwarz , N., Bless, H ., & Bohner, G. (1991 ). M ood and persuasi on: Affective states in¯ uence the processin g of persuas ive communications. In M. Zanna (Ed.), Advances in experim ental so cial psychology (Vol. 24, pp. 161± 199). San D iego, CA : Academic Press. Schwarz , N., & Clore, G.L . (1983). M ood, m isattributions, and judgements of wellbeing: Informat ive and directive function s of affective states. Journal of Personality and Social Psychology, 45, 513± 523. Schwarz , N., Servay, W., & Kum pf, M . (1985). Attributio n of arousal as a m ediator of the effectiveness of fear-arousing com munications. Journal of Applied Social Psychology, 15, 178± 188. Sherm an, S.J., & Corty, E. (1984). Cognitiv e heuristi cs. In R.S. Wyer & T.K . Srull (E ds.), Handboo k of social cognition (Vol. 3, pp. 189 ± 286). H illsdale, N J: Erlbaum . Shiffrin, R.M ., & Schneid er, W. (1977). Controll ed and automatic hum an informa tion processing: II. Perceptual learning , autom atic attendin g, and a general theory. Psychological Review, 84, 127± 190. Sie mer, M . (1997). W hat exactly is random about ra ndom effects? M ethods of Psychological Research , 2(2). Online http: //ww w.hsp.de/M PR//. Spies, K., H esse, F.W., G errards- Hesse, A., & Uef® ng, E. (1991). Experim entelle Induktio n emotion aler Zusta nde. Verbesser t die zusa nde. Verbesse rt die zusa tzliche Darbiet ung von È È È M usik die Wirksam keit selbstbe zogener Aussagen ? [Experimental induction of emotion al states. Does the addition of music enhance the in¯ uence of se lf-referential statem ents?] Zeitschrift fu r Experim entelle und Angewandte Psychologie, 38, 321± 342. È Strack, F., Schwarz , N., & G schneid inge r, E. (1985). H appiness and rem iniscing: The role of tim e perspec tive, m ood, and mode of thinking . Journal of Persona lity and Social Psychology, 49, 14 60± 1469. Velten, M . (19 68). A laborator y task for inductio n of mood states. Behavior Research and Therapy, 6, 473± 482 . Weiner, B. (1985 ). ``Spontane ous’ ’ causal thinkin g. Psychological Bulletin, 97 , 74± 84 . Wyer, R.S., & Srull, T.K . (198 6). Human cognitio n in its socia l context. Psychological Review, 93, 322 ± 359. Wyer, R.S., & Srull, T.K. (1989). Mem ory and cognition in its social context. H illsdale, NJ: Erlbaum . Zillmann, D. (1978). Attributio n and misattribution of excitatory reactions. In J.H. Harvey, W.I. Ickes, & R.F. K idd (E ds.), New directio ns in attribution research (Vol. 2, pp. 335± 368 ). Hillsdale, NJ: Erlbaum .