~ Pergamon S0005-7967(96)00017-4

Behav. Res. Ther. Vol. 34, No. 5/6, pp. 483-488, 1996 Copyright © 1996 Elsevier Science Ltd

Printed in Great Britain. All rights reserved 0005-7967/96 $15.00 + 0.00

Evaluative decision latencies mediated by induced affective states

D I R K H E R M A N S * , J A N DE H O U W E R * and P A U L E E L E N

Department of Psychology, University of Leuven, Tiensestraat 102, B-3000 Leuven, Belgium

(Received 1 May 1995)

Summary--Recent priming studies (e.g. Hermans, De Houwer & Eelen, 1994, Cognition and Emotion, 8, 515-533) have demonstrated that response latencies to target stimuli are mediated by the affective relation between prime and target. The time needed to evaluate or pronounce targets is facilitated if preceded by similarly valenced primes, but is inhibited for trials on which prime and target have an opposite affective valence. These data suggest that information stored in memory is associatively linked with similarly evaluated information, through association with some general representation of goodness or badness. To investigate whether affective priming is merely one type of conventional semantic priming, or whether it is mediated by affective responses, the affective context provided by the primes was replaced in this study by the induction of an emotional state using a Musical Mood Induction procedure (Depression/Elation). Subjects had to evaluate target pictures as quickly as possible. The data revealed a significant Mood Change (More Depressed/Less Depressed/No Change) x Target Valence (Positive/Negative) interaction, indicating that emotional states can mediate evaluative response latencies to affectively valenced target stimuli. The results are interpreted in the context of a biphasic emotion theory, and are related to previous research on mood congruency effects on perceptual responses. Copyright ~3 1996 Elsevier Science Ltd.

INTRODUCTION

Traditional models have conceptualised human memory as associative networks (Anderson & Bower, 1973; Collins & Loftus, 1975). Experimental evidence for these models has largely been based on priming studies (e.g. Meyer & Schvaneveldt, 197 I). In a typical priming study, the presentation of a prime stimulus to which no overt response is required, is immediately followed by the presentation of a target stimulus, which has to be pronounced (pronunciation task), or to which a word/nonword decision is required (lexical decision task). It has been repeatedly demonstrated that response latencies on such tasks are mediated by the type of prime-target relation. Associatively (e.g. bread, butter) (Lupker, 1984) and categorically related prime-target pairs (e.g. bird, robin) (Neely, Keefe & Ross, 1989) lead to facilitated responses as compared to unrelated control pairs. To a lesser extent, similar results have been obtained for semantically but not associatively related primes and targets (e.g. bird, fish) (for a critical discussion concerning pure semantic priming, see Shelton & Martin, 1992). The results from these priming tasks, and other experimental cognitive paradigms like the Stroop task (De Houwer & Hermans, 1994; Glaser & Glaser, 1989), have helped to reveal the structure and processes of human associative memory.

Recently, in a series of priming studies (Bargh, Chaiken, Govender & Pratto, 1992; Fazio, Sanbonmatsu, Powell & Kardes, 1986; Hermans, De Houwer & Eelen, 1994) significant priming effects were demonstrated using prime-target pairs for which the affective relation was manipulated. It was observed that the time needed to evaluate targets (words or pictures) as either 'positive' or 'negative', or to pronounce target words were mediated by the affective valence of the prime that immediately [stimulus onset asynchrony (SOA) 300 msec] preceded the target. On trials for which both prime and target shared the same affective valence (e.g. corpse-jealous), response latencies were facilitated. Affectively incongruent prime-target pairs (e.g. cemetery-reliable) on the other hand, led to relatively slower response latencies. The affective priming effect has been demonstrated using different types of stimulus materials and priming procedures, and several aspects of its automatic character are now well established or under research (Bargh, 1994; Greenwald, Klinger & Liu, 1989; Hermans et al., 1994). Of critical importance however is that the stimulus pairs were not selected to be associatively or categorically related. From the point of view of associative network models, these data seem to indicate that there exist two large memory networks; one which interconnects all positively valenced memory representations, and a similar network that contains all negatively valenced representations (Bargh, 1988). It is not yet clear how such affective networks should structurally be conceptualised. A possible, and rather parsimonious architecture would be that similarly valenced representations are associated by their shared evaluation (Isen, 1984). More precisely, this could imply the existence of two central units in the associative network, which respectively represent 'General Positive Affect' and 'General Negative Affect'. Hence, a specific memory representation acquires its evaluative meaning through a direct association to one of both general affect nodes. These representation-evaluation associations are supposed to be very strong (De Houwer & Hermans, 1994). Activation of a memory concept will automatically lead to the immediate activation of its related affect node. When activated above a threshold, the general affect node transmits excitation to other similarly valenced concepts. Consequently, if a target stimulus is presented in the context of a stimulus--the prime---with a similar affective connotation, this spreading of activation will facilitate a response towards the target stimulus. Additionally, to account for the inhibition effects for affectively incongruent prime-target pairs, it has to be assumed that both affective networks are linked by an inhibitory association.

*Research assistant for the National Fund for Scientific Research (Belgium).

483

484 SHORTER COMMUNICATIONS

This explanation fits perfectly within the more general cognitive-representational emotion theory proposed by Bower (1981, 1987). In fact, Bower (1991) provided an analogous explanation for the affective priming effect in the context of an amendment of his earlier network model. Also, Hill and Kemp-Wheeler (1989) who observed a facilitation of lexical decisions for negative target words which were preceded by a negative prime, discussed their results in the context of Bower's associative network theory. These authors however also raised the critical question, to what extent the affective priming phenomenon can be attributed to a truly affective/emotional process. According to the authors, it could well be argued that the affective priming effect is merely one type of the conventional semantic priming effect, as evaluation is a major constituent of the semantic meaning of a stimulus (Osgood, Suci & Tannenbaum, 1957). This 'cold' explanation of the affective priming effects clearly differs from an affective/emotional account according to which facilitation (or inhibition) of the responses towards the target stimuli is mediated by the activation of genuine, although short-lived emotional responses. This would mean that primes elicit an emotional response against which similarly valenced targets are more easily processed; an emotional context which hampers the processing of affectively incongruent targets. According to the latter view, replacing the emotional context, temporary elicited by the prime, by a longer living emotional state (mood) should lead to a similar pattern of facilitation (or inhibition) of evaluative responses towards target stimuli with a valence congruent (or incongruent) with the emotional state of the S.

Accordingly, the primary aim of this study was to test the hypothesis that an induced mood state also primes affective congruent responses to target stimuli. A similar procedure was used as in Hermans et al. (1994, Experiment 1). Subjects had to evaluate complex visual target stimuli as quickly as possible, but the pictorial primes were now replaced by an emotional context which was introduced through a mood induction procedure. A Musical Mood Induction Procedure (MMIP: Sutherland, Newman & Rachman, 1982) was chosen over the traditional Velten Mood Induction Procedure (VMIP: Velten, 1968) for two reasons. First, the VMIP has been criticized for demand characteristics (Polivy & Doyle, 1980) as most researchers using the VMIP instructed the Ss to "try to feel the mood suggested by the statements". Pignatiello, Camp, and Rasar (1986) however have demonstrated that the MMIP is an effective alternative to the VMIP, even if each single reference to the mood inducing characteristics of the procedure is omitted. Second, a VMIP induces some kind of 'cognitive priming'. Shorter response latencies towards the word ' threat ' for example, might not only be due to a changed mood state, but also to a higher accessibility of the ' threat '-concept in memory if this or a related word was in one of the VMIP statements. It was reasoned that because of its nonverbal nature, a MMIP would not lead to such direct cognitive priming.

METHOD

S u b j e c t s

Forty psychology students (17 males, 23 females) volunteered to participate in the experiment in partial fulfilment of undergraduate course requirements. All had normal or corrected-to-normal vision.

S t i m u l i

A set of 100 colour pictures and 100 identical colour slides were used throughout the entire experiment. Stimuli were selected in order to obtain a very wide range of content as well as affective value (e.g. mutilated faces, flowers, a basket). During the target presentation phase, 42 × 46 cm slides were presented at a viewing distance of about 3 m. For the Elated condition, the first and second movement from the Spring (Vivaldi's Four Seasons) were selected for the MMIP. For the Depressed condition, excerpts from Dvorak's Ninth Symphony were used. Selection was based on the material used by Peeters (1990). Both passages were recorded repeatedly on separate 45-min tapes.

A p p a r a t u s

The experiment was carried out in a dimly lit room. Slides were backprojected on a translucent glass projection screen, which separated the experimenter's room from the Ss room. A random-access slide projector ( K O D A K Carousel S-RA 2000), equipped with a Compur electronic m3 shutter, was installed in the former part. Response latencies were recorded by way of a microphone activated voice key attached to an XT IBM compatible computer. This microcomputer controlled both slide presentation and exposure duration. For the presentation of the music, a TEAC four-track tape recorder and AKAI amplifier were used, together with two two-way loudspeakers and a Sennheiser HD 520 II headphone.

The mood questionnaire consisted of four scales. For each scale there were four items: Anxiety (afraid, nervous, tense, calm), Depression (forlorn, gloomy, unhappy, discouraged), Elation (active, alive, fine, healthy), and Hostility (angry, cruel, disagreeable, tender). These items were selected from the Van den Bergh (1985) translation of the M A A C L (Today Form: Zuckerman & Lubin, 1965). For each item, Ss had to indicate to what extent the item represented their p r e s e n t mood state on a l l -point Visual Analogue Scale (0 = 'not at all', to 10 = 'very much').

Procedure

Half of the Ss were assigned randomly to either the Elated or the Depressed condition. Upon entering the room, Ss were asked to fill out a questionnaire as part of a larger reliability study by a colleague of the experimenter. This was done to avoid possible demand effects, by diverting the attention from the actual purpose of the experiment. All Ss volunteered to complete the mood questionnaire. Next, Ss were asked to evaluate the colour pictures on a 21-category scale ( - 100 = very unp leasan t , 0 = neutra l , + I00 = v e r y p l e a s a n t ) . It was stressed that Ss had to rely on the first and immediate reaction towards the picture as a whole. Subsequently, the experimenter handed out an instruction sheet which essentially read that we were interested in the influence of external stimuli on cognitive tasks.

The S was seated in front of the glass projection screen. The experimenter explained that a set of 24 slides would be presented four times. Each slide had to be evaluated as quickly as possible by expressing aloud either 'Positive', 'Negative', or 'Neutral ' in the microphone. Both before and during the slide presentation, music would be presented to which Ss had to listen carefully. Next, the S placed the headphone in the correct position, the tape was started, and the volume was adjusted to a comfortable level. The light was dimmed, and Ss listened to the music for 5 min. Meanwhile, the experimenter used the S 's evaluative ratings to select the eight most negatively, and the eight most positively rated stimuli, together with eight neutral pictures (rating = 0). The corresponding slides were placed in the projector, and the presentation, which consisted of 104 trials, was initiated. The first eight trials were practice trials, and were followed by four blocks of test trials. Test blocks were separated by a short rest of approximately I min. For each block, the order of the 24 test trials was randomized. Each slide disappeared following the S ' s response. The inter trial interval was always 7 see.

SHORTER COMMUNICATIONS 485

After the slide presentation, Ss had to fill out the mood questionnaire again as the second part of the reliability study. It was stressed to rate how their mood was at that moment , regardless of their answers at the first rating. After completion, the tape was stopped and the experimenter inquired about the S ' s assumptions about the effect of the music on the evaluative response latencies. No S made any reference towards the mood inducing function of the music. Finally, the S was thanked and dismissed.

RESULTS

In approximately 7.5% of the test trials, the voice key was not appropriately activated. The data of these trials, together with the data of the practice trials were excluded from all analyses. In addition, latencies over 1500 msec, or shorter than 250 msec were excluded to reduce the effect of outlier responses. For each S, mean response latencies were calculated for each of the three levels of the affective valence of the stimuli (Positive, Negative, Neutral). These data were analyzed in a 2 (lnduced Mood: Elated/Depressed) x 3 (Target Valence: Positive/Negative/Neutral) analysis of variance with repeated measures for the last variable. The expected Induced Mood x Target Valence interaction failed to reach significance [F(2, 76) = 1.543; NS]. There was also no effect of Induced Mood [F(l, 38) < I; NS]. There was however a reliable effect of Target Valence [F(2, 76) = 12.7; P < 0.00002]. Tukey a posteriori contrasts revealed longer response latencies for neutral stimuli (M = 919 msec), as compared to positive (M = 833 msec) and negative stimuli (M = 853 msec), which did not differ.

To test for the effects of the MMIP, difference scores (Moment 2 - Moment l) of the mood questionnaire were analyzed separate for each of the four subscales with Induced Mood as a between Ss variable. For the Elation, Anxiety, and Hostility subscales these differences were not significant [F(I, 38) < I]. Only the Depression scale showed an effect in the predicted direction [F(1, 38) = 8.79; P < 0.001]. This effect was however very small. For the Elated group, mean depression scores became only 0.45 points more positive on the 1 l-point VAS scale. For the Depressed group, mean depression scores changed 1.25 points. This effect might also simply reflect a regression to the mean as initial depression scores (Moment 1) differed significantly for both conditions [t(38) = 2.98; P = 0.005; MElafion = 11.3; MDepression = 4.65]. A similar baseline difference was however absent for the other subscales of the mood questionnaire.

Hence, it seemed that the mood induction procedure had not been effective. Nevertheless, inspection of the data displayed large shifts in mood state for individual Ss of both conditions in the expected, as well as in the opposite direction. Due to the absence of any instructions concerning the mood inducing properties of the music and its desired direction, Ss might have experienced mood changes in unintended directions. For example, for some Ss the slow, grim music of Dvorak might have been quite relaxing.

Consequently, it was decided to use the Depression difference scores (Moment 2 - Moment 1) to split the sample into three groups: More Depressed (N = 13), No Change (N = 14), and Less Depressed (N = 13)*. A subsequent 3 (Mood Change: More Depressed/No Change/Less Depressed) x 2 (Target Valence: Positive/Negative)t analysis of variance now revealed a significant Mood Change x Target Valence interaction [F(2, 37) = 3.66; P < 0.05]. As shown in Fig. l, evaluative response latencies were shorter for positive stimuli than for negative stimuli if Ss experienced a positive mood change. An opposite pattern is observed if the change in mood state was negative. No differences are observed for the No Change group. In addition, there was a main effect of Mood Change [F(2, 37)= 6.08; P <0.01]. The More Depressed group (M = 781 msec) was generally faster than the Less Depressed group (M = 903 msec) (Tukey HSD). Differences concerning the No Change group (M = 845 msec) were not significant. Finally, there was no effect of Target Valence [F( l, 37) = I. 14; NS].

A similar Mood Change x Target Valence interaction was observed when Ss were divided in a More Elated (N = 13), No Change (N = 14), and Less Elated group (N = 13), according to the difference scores (Moment 2 - Moment I) for the Elation subscale [F(2, 37) = 3.46; P < 0.05]. For the More Elated group, evaluation latencies were faster for positive stimuli (M = 857 msec) than for negative stimuli (M = 946 msec). For the Less Elated group this effect was reversed (Mpo~ai, ¢ = 845 msec; Mne~iw = 818 msec). There was no difference for the No Change group (Mpositi~e = 801 msec; Mne,~tiv~ = 800 msec). Similarly, there was also a main effect of Mood Change [F(2, 37) = 4.23; P < 0.05]. The No Change group (M = 902 msec) was generally slower than the More Elated group (M = 800 msec) (Tukey HSD). Differences concerning the Less Elated group (M = 831 msec) were not significant. Also, there was no effect of Target Valence [F(I, 37) = 1.18; NS].

No Mood Change x Target Valence interactions were observed when Ss were classified according to their difference scores for the Hostility subscale [F(2, 37)= 2.77; NS], or the Anxiety subscale [F(2, 37 )= 1.61; NS].

DISCUSSION

First of all, it should be noted that in the final analyses, mood induction condition was no longer a factor. As the musical mood induction procedure did not markedly affect the self-reported mood, the sample of Ss was subdivided in three groups,

*It was decided to use Mood Change scores to subdivide the sample, over mood scores at Moment 2, as there was a strong correlation between depression scores at Moments I and 2 [Pearson r (40) = 0.601; P < 0.00005]. A significant interaction between Moment 2 depression scores and Target Valence might be a simple reflection of a chronically high activation of negative concepts in habitually more depressed persons, and would give no indication whatsoever of the momentary impact of the presented music. This is however not the case using the Mood Change scores. A Mood Change (More Depressed/No Change/Less Depressed) x Moment of rating (depression scores on Moment 1/Moment 2) revealed a significant interaction [F(2,37) = 47.69; P < 0.00001]. A posteriori contrasts (Tukey HSD) show that each of the three groups change (or do not change) in the expected direction, but moreover, Ss from the More Depressed group were significantly less depressed than Ss from the Less Depressed group at Moment I. So a possible Mood Change × Target Valence interaction can not be attributed to a chronic activation of concepts. [At Moment 2 however there was a very strong difference in the opposite direction. So this change in moodscores can not be attributed to a regression to the mean.]

~Neutral stimuli were included in the design to create a baseline against which changed evaluation latencies for positive and negative stimuli could be weighted. The Target Valence main effect discussed before however indicated that response latencies for neutral stimuli were considerably slower as compared to positive and negative slides. This might indicate that these stimuli were ambiguous rather than neutral. As they lost their function as a baseline, neutral stimuli were discarded from further analyses.

486 SHORTER COMMUNICATIONS

Evaluation Latenolee (ms) 9 5 0 ...................................................................................................................................................................................................

9 0 0 ............................................... " ' " ' ' " ' " ' " " ~ . . . - . . - - - - . - . . - . - . - - . .................................................

/ - 8 6 0 ...................................................................................................... :,::::: ................................ :::::::.: .................................................

8 0 0 ............................................... ~ - . . . . . . . - . - - . - . - . - . . . . . . . . . . . . . . . ..................................................

7 5 0 J I Posit ive Negat ive

Stimulus Value

• Lees Depressed ......... No Change " " More Depressed

Fig. 1. Mean evaluative response latencies for the More Depressed, No Change, and Less Depressed groups, as a function of the affective value of the target (Positive or Negative).

based on changes in mood from Moment l to Moment 2, thus leading to a differential design. Although not unimportant with respect to the interpretation of the data, it has to be noted that differential designs are not uncommon in this type of research (e.g. MacLeod, Tata& Mathews, 1987; Powell & Hemsley, 1984). A possible explanation for the ineffectiveness of the MMIP might be the absence of additional suggestions on how to change mood (e.g. to remember a past sad event), which are present in the standard MMIP. A recent study by Lenton and Martin (1991) for example, indicates that such additional instructions might be of major importance for the effectiveness of a MMIP.

The subsequent analyses revealed a Mood Change x Target Valence interaction demonstrating that mood states can mediate evaluative response latencies towards affectively loaden target stimuli. This indicates that mood states can function as an affective context against which target stimuli are processed, in a similar way as prime stimuli do in the affective priming effect. This result is compatible with the idea that in the standard affective priming studies (Bargh etaL, 1992; Fazio et al., 1986; Hermans et al., 1994), the observed facilitation or inhibition of response latencies towards affectively valenced target stimuli might be due to the activation of congruent or incongruent emotional responses.

According to Bower's (1991) amended network model, both prime stimuli and moods are capable of influencing the speed of affective judgements because of the activation of the positive or negative general affect nodes (which are memory representations of emotional states). The activation of the general positive affect node for example, will spread activation to all concepts with a positive valence, thereby raising the activation level of each of them. Simultaneously, the general negative affect node will be inhibited (Bower, 1981). It is assumed that responses to subsequently presented target stimuli will be relatively biased because of these relatively augmented or inhibited activation levels for positively or negatively valenced concepts.

There is however at least one theoretical consideration which suggests that a spreading of activation model might not be sufficient to account for the affective priming effect. Since the number of positively or negatively valenced concepts in memory is large if not infinite, and because the quantity of activation is limited, it is implausible that the activation of one concept would be sufficient to activate all similarly valenced concepts, and inhibit all concepts with an opposite affect, to the degree that this would facilitate or hinder subsequent responding towards one of these concepts (Hermans et al., 1994). This is known as cue overload or fanning effect (Anderson & Bower, 1973), and poses a general problem for all models that use the notion of spreading of activation to account for the role of affect and emotion in memory (Isen, 1984; Niedenthal, Setterlund & Jones, 1994). A more parsimonious alternative would be to adopt the idea of emotions as affective-motivational systems. According to this view emotions can be conceptualized as general response dispositions or action tendencies (e.g. Frijda, 1986; Lang, 1984), i.e. a positive emotional state is characterized by an approach tendency and a negative emotional state by a tendency to withdraw. These two basic affective-motivational systems selectively sensitise for congruent emotion-eliciting material. In this view, any behaviour that was previously motivated by one of the drive systems (aversive or appetitive) is particularly potentiated, when an evoking stimulus occurs during a subsequent, independent activation of that same system (Lang, 1994). On the other hand, a reciprocal inhibition between both systems is postulated (Lang, Bradley & Cuthbert, 1990). In this context, it was demonstrated that even a primitive aversive/defensive reflex such as the startle response is augmented or diminished by the affective context provided by the presentation of positive or negative pictures (Vrana, Spence & Lang, 1988), or the induction of emotion via emotional imagery (Vrana & Lang, 1990). A similar argumentation could be made for the results observed in the affective priming studies discussed before, and the mood induction study presented here. The affective priming context (positive or negative external stimuli or induced mood) can either match or differ from the affective response disposition activated by the target, consequently sensitising or inhibiting responses toward the subsequent target.

Taken together, it is argued here that the data from the affective priming studies and the mood induction study reported here are most parsimoniously explained by assuming two stages in the affective/evaluative processing of stimuli. At the first

SHORTER COMMUNICATIONS 487

stage the stimulus activates its related concept in semantic memory, thereby immediately activating the related affect node with which it is strongly associated (De Houwer & Hermans, 1994). This spreading of activation entails a crude stimulus evaluation as either good or bad, and activates one of both opponent (appetitive or aversive) affective-motivational systems (stage two). Similarly, a mood induction procedure can mobilize one of both systems. Processing of subsequently presented target stimuli will be sensitised if congruent with the prevalent affective/motivational state, but inhibited if there is a mismatch. Of major importance is that no spreading of activation from one memory node to all other similarly valenced concepts has be to postulated, thus avoiding the problem of cue overload.

Finally, it has to be noted that this study provides one of the rare demonstrations of a mood congruency effect on perceptual responses such as detection, identification, and classification, as most earlier studies have failed to observe such interaction (Clark, Teasdale, Broadbent & Martin, 1983; Gerrig & Bower, 1982; MacLeod, Tata & Mathews, 1987; Powell & Hemsley, 1984). At this moment it is still not completely clear what the crucial factors are for mood congruency at this level of processing (Bower, 1991; Niedenthal & Setterlund, 1994; Niedenthal et al., 1994).

REFERENCES

Anderson, J. R. & Bower, G. H. (1973). Human Associative Memory. Washington, DC: V. H. Winston. Bargh, J. A. (1988). Automatic information processing: implications for communication and affect. In Donohew, L., Sypher,

H. E. & Higgins, E. T. (Eds), Communication, Social Cognition, and Affect (pp. 9-32). Hillsdale, N J: Lawrence Erlbaum Associates.

Bargh, J. A. (1994). The four horsemen of automaticity: awareness, intention, efficiency, and control in social cognition. In Wyer, R. S., Jr. & Srull, T. K. (Eds), Handbook of Social Cognition (pp. 1-40). Hillsdale, NJ: Lawrence Erlbaum Associates.

Bargh, J. A., Chaiken, S., Govender, R. & Pratto, F. (1992). The generality of the attitude activation effect. Journal o[ Personality and Social Psychology, 62, 893-912.

Bower, G. H. (1981). Mood and memory. American Psychologist, 36, 129 148. Bower, G. H. (1987). Commentary on 'Mood and Memory'. Behaviour Research and Therapy, 25, 443-455. Bower, G. H. (1991), Mood congruity of social judgements. In Forgas, J. P. (Ed.), Emotion and Social Judgements

(pp. 31--53). Oxford: Pergamon Press. Clark, D. M., Teasdale, J. D., Broadbent, D. E. & Martin, M. (1983). Effect of mood on lexical decisions. Bulletin of the

Psychonomic Society, 21, 175-178. Collins, A. M. & Loftus, E. F. (1975). A spreading activation theory of semantic processing. Psychological Review, 82,

407-428. De Houwer, J. & Hermans, D. (1994). Differences in the affective processing of words and pictures. Cognition and Emotion,

8, 1-20. Fazio, R. H., Sanbonmatsu, D. M., Powell, M. C. & Kardes, F. R. (1986). On the automatic activation of attitudes. Journal

of Personality and Social Psychology, 50, 229 238. Frijda, N. H. (1986). The Emotions. Cambridge: Cambridge University Press. Gerrig, R. J. & Bower, G. H. (1982). Emotional influences on word recognition. Bulletin of the Psychonomic Society, 19,

197-200. Glaser, W. R. & Glaser, M. O. (1989). Context effects in Stroop-like word and picture processing. Journal of Experimental

Psychology: General, 118, 13-42. Greenwald, A. G., Klinger, M. R. & Liu, T. J. (1989). Unconscious processing of dichoptically masked words. Memory

and Cognition, 7, 35-47. Hermans, D., De Houwer, J. & Eelen, P. (1994). The affective priming effect: automatic activation of evaluative information

in memory. Cognition and Emotion, 8, 515 533. Hill, A. B. & Kemp-Wheeler, S. M. (1989). The influence of context on lexical decision times for emotionally aversive words.

Current Psychology: Research and Reviews, 8, 219-227. lsen, A. M. (1984). Toward understanding the role of affect in cognition. In Wyer, R. S. & Srull, T. K. (Eds), Handbook

of Social Cognition (Vol. 3, pp. 179-236). Hillsdale, N J: Lawrence Erlbaum Associates. Lang, P. J. (1984). Cognition in emotion: concept and action. In lzard, C., Kagan, J. and Zajonc, R. B. (Eds), Emotions,

Cognition, and Behavior. New York: Cambridge University Press. Lang, P. J. (1994). The motivational organization of emotions: affect-reflex connections. In Van Goozen, S., Van de Poll,

N. E. & Sergeant, J. A. (Eds), The Emotions: Essays on Emotion Theory. Hillsdale, N J: Lawrence Erlbaum Associates. Lang, P. J., Bradley, M. M. & Cuthbert, B. N. (1990). Emotion, attention, and the startle reflex. Psychological Review,

97, 377-395. Lenton, S. R. & Martin, P. R. (1991). The contribution of music vs. instructions in the Musical Mood Induction Procedure.

Behat, iour Research and Therapy, 29, 623~525. Lupker, S. J. (I 984). Semantic priming without association: a second look. Journal of Verbal Learning and Verbal Behavior,

23, 709 733. MacLeod, C., Tata, P. & Mathews, A. (1987). Perception of emotionally valenced information in depression. British Journal

of Clinical Psychology, 26, 67-68. Meyer, D. E. & Schvaneveldt, R. W. (1971). Facilitation in recognizing pairs of words: evidence of a dependence between

retrieval operations. Journal of Experimental Psychology, 90, 227-234. Neely, J. H., Keefe, D. E. & Ross, K. (1989). Semantic priming in the lexical decision task: roles of prospective

prime-generated expectancies and retrospective semantic matching. Journal of Experimental Psychology: Learning, Memory and Cognition, 15, 1003-1019.

Niedenthal, P. M. & Setterlund, M. B. (1994). Emotion congruence in perception. Persona6ty and Social Psychology Bulletin, 20, 401-411.

Niedenthal, P. M., Setterlund, M. B. & Jones, D. E. (1994). Emotional organization of perceptual memory. In Niedenthal, P. M. & Kitayama, S. (Eds), The Heart's Eye: Emotional Influences in Perception and Attention. San Diego: Academic Press.

Osgood, C. E., Suci, G. H. & Tannenbaum, P. H. (1957). The Measurement of Meaning. Urban, IL: University of Illinois Press.

488 SHORTER COMMUNICATIONS

Peeters, R. (1990). Geheugen en affect: een kwestie van 'overeenstemming'? [Memory and affect: a matter of consonance]. Unpublished doctoral dissertation, University of Leuven, Belgium.

Pignatiello, M. F., Camp, C. J. & Rasar, L. A. (1986). Musical mood induction: an alternative to the Velten technique. Journal of Abnormal Psychology, 95, 295-297.

Polivy, J. & Doyle, C. (1980). Laboratory induction of mood states through the reading of self-referent mood statements: affective changes or demand characteristics? Journal of Abnormal Psychology, 89, 286-290.

Powell, M. & Hemsley, D. R. (1984). Depression: a breakdown of perceptual defence? British Journal of Psychiatry, 145, 358--362.

Shelton, J. R. & Martin, R. C. (1992). How semantic is semantic priming? Journal of Experimental Psychology: Learning, Memory, and Cognition, 18, 1191-1210.

Sutherland, G., Newman, B. & Rachman, S. (1982). Experimental investigations of the relations between mood and intrusive unwanted cognitions. British Journal of Medical Psychology, .55, 127-138.

Van den Bergh, O. (1985). Weten en voelen: een paar apart. [Knowing and lceling: an odd couple]. Unpublished doctoral dissertation, University of Leuven, Belgium.

Velten, E. (1968). A laboratory task for induction of mood states. Behavioural Research and Therapy, 6, 607~17. Vrana, S. R. & Lang, P. J. (1990). Fear imagery and the startle-probe reflex. Journal of Abnormal Psychology, 99, 181-189. Vrana, S. R., Spence, E. L. & Lang, P. J. (1988). The startle probe response: a new measure of emotion? JournalofAbnormal

Psychology, 97, 487-491. Zuckerman, M. & Lubin, B. (1965). Manual for the Multiple Mood Adjective Check List. San Diego, CA: Educational and

industrial testing service.