Psychonomic Bulletin & Review 2004, 11 (2), 326-331 Loading working memory enhances affective priming MARK ROTTEVEEL and R. HANS PHAF University of Amsterdam, Amsterdam, The Netherlands Stronger affectivepriming (Murphy & Zajonc, 1993)with suboptimal (i.e., reduced consciousness) than with optimal (i.e., full consciousness) prime presentation suggests that nonconscious processes form an important part of emotions. Merikle and Joordens (1997) have argued that both impoverished presenta- tion and divided attention can produce suboptimal conditions and result in parallel effects. We manipu- lated attention by means of a concurrent working memory load while keeping presentation duration con- stant, as participantsevaluatedJapanese ideographs that were preceded by happy, neutral, or angry faces (affectivepriming) and male or female faces (nonaffective priming). In contrast to nonaffective priming, affective priming was larger with divided attention than with focused attention. It is concluded that ma- nipulations of stimulus quality and of attention can both be used to probe the distinction between con- scious and nonconscious processes and that the highest chances of obtaining the pattern of stronger priming with suboptimal presentation than with optimal presentation occur in the affective domain. To investigatewhat it means whether or not a stimulus 2001). Merikle and Joordens (1997) have argued that di- is processed consciously, Baars and McGovern (1993; vided attention and impoverished (i.e., short, masked) see also Merikle, 1992) proposed the general method of presentation can both be used to produce suboptimalcon- experimentally manipulating level of consciousness and ditions,so that both shouldhave correspondingeffects. To contrasting the effects of conscious (optimal) and less our knowledge,however, this correspondencehas not been conscious (suboptimal) conditions (i.e., a contrastive explicitly studied in the affective domain. We have inves- analysis). This method can be applied to the study of tigated whether stronger suboptimal affective priming many independent variables and has revealed a number than optimal affective priming (Murphy & Zajonc, 1993) of qualitative differences (e.g., Cheesman & Merikle, also occurs when suboptimal conditions are achieved by 1984; Greenwald, 1992; Groeger, 1984, 1988; Jacoby & dividing attention instead of through impoverished pre- Whitehouse,1989)which suggest that conscious and non- sentation and whether this contrasts with nonaffective conscious processes do not always have the same form. priming. In addition, the finding of stronger effects in less con- Murphy and Zajonc (1993) investigatedpriming in af- scious than in fully conscious conditions contradicts the fective and nonaffectiveevaluationof unfamiliar Chinese view that suboptimal processing constitutes only diluted ideographs. Affectively valenced (i.e., happy or angry conscious processing (i.e., with some trials inadvertently faces) and control (polygons, faces with neutral expres- processed consciously). This type of result has primarily sions) stimuli were presented in either a suboptimal been found in the affective domain (Bornstein, 1989; (4 msec, or about 10 msec effectively; see Winkielman, Fulcher & Hammerl, 2001; Janssen, Everaerd, Spiering, Zajonc, & Schwarz, 1997), or an optimal (1,000 msec) & Janssen, 2000; Murphy & Zajonc, 1993; Rotteveel, de fashion by means of a projection tachistoscope. Because Groot, Geutskens, & Phaf, 2001; Stapel, Koomen, & the absence of an effect in a (direct) task measuring con- Ruys, 2002), whereas weaker suboptimaleffects than op- scious effects cannot guarantee that all conscious pro- timal effects are usually obtained in other, more nonaf- cessing has been exhaustively excluded (Cheesman & fective, domains (e.g., Lubke, Kerssens, Phaf, & Sebel, Merikle, 1984), the cautious terminology of suboptimal 1999; Maxfield, 1997; Murphy & Zajonc, 1993, Experi- and optimal, instead of subliminaland supraliminal, pre- ments 3–6; but see De Fockert, Rees, Frith, & Lavie, sentation was adopted by Murphy and Zajonc. In the first two experiments, happy and angry faces led to reliable, valence-congruent, shifts in ideograph evaluation only The authors are grateful to Julie Boland, Phil Merikle, and an anony- with suboptimal prime presentation. The stronger subop- mous reviewer for very helpful comments. We thank Susannede Boer and timal as opposed to optimal results preclude the possibil- Marisa Stoffers for their assistance in performing the experiments. R.H.P. ity that suboptimal effects are caused only by conscious is in the Department of Psychonomics; M.R. is in the Department of Psy- processing “leaking through.” Nonaffective priming of chonomics and the Department of Social Psychology. Correspondence about this paper shouldbe sent to M. Rotteveel, Department of Social Psy- size (Experiment 3), symmetry (Experiment 4), and gen- chology, University of Amsterdam, Roetersstraat 15, 1018 WB Amster- der (Experiment 5), however, revealed congruent shifts dam, The Netherlands (e-mail: ([email protected]). only in optimal conditions. Copyright 2004 Psychonomic Society, Inc. 326 WORKING MEMORY AND AFFECTIVE PRIMING 327 Rotteveel et al. (2001) have argued that the finding of tention in both an affective priming (i.e., with emotional stronger suboptimalas opposedto optimalaffective prim- faces as primes) and nonaffective priming (i.e., male and ing may be rather brittle and sensitive to problems asso- female faces serving as primes) task. Attention was var- ciated with impoverishedprime presentation. Eye-blinks ied by concurrently loading working memory (see Mul- or saccades may, for instance, cause participants to miss ligan & Stone, 1999). Participants were instructed to suboptimalprimes. Because one is effectively blind dur- maintain, and to reproduce at the end of the trial, a digit- ing a saccade (Morgan, 1994; Tovée, 1996), a brief stim- letter string of seven elements with divided attention, or ulus presented during the saccade cannot exert a priming an empty string (i.e., a “–”) with focused attention. The effect. The stronger suboptimalas opposed to optimal af- participants were instructed emphatically to perform the fective priming might thus be weakened only with short concurrent working memory task as accurately as possi- presentationdue to the influence of saccades. Single frame ble. They also had to judge intuitively whether Japanese presentation on a computer screen might also dilute sub- ideographs(Rotteveel et al., 2001) preceded by the faces optimal effects. The picture is built up pixel by pixel and represented something either positive or negative (i.e., line by line. The visibility of a single pixel corresponds affective task), or somethingmasculine or feminine (i.e., to the effective persistence of the specific phosphor used nonaffective task). We expected stronger affective prim- in the screen (about 4 msec, Bridgeman, 1998).This time ing and weaker nonaffectivepriming under conditionsof is generally shorter than a single screen frame, and prime divided as opposed to focused attention. presentation may sometimes be too short to allow for any processing at all. Impoverished presentation may, fur- METHOD thermore, allow for a confoundingof presentation condi- tion and instruction.Murphy and Zajonc(1993) instructed Participants A total of 96 (65 female) 1st-year psychology students (M 5 participants to ignore the faces only in optimal condi- 5 tions, but did not tell them about the suboptimal primes. 21.96 years old, SD 2.69) from the University of Amsterdam par- ticipated in the two tasks for course credit. All were right-handed The participants might have applied the controlled strat- and had normal or corrected-to-normal vision. The experiment was egy of counteractingthe prime influencesonly in optimal announced as a test of “intuitive knowledge of the Japanese lan- conditions, whereas in suboptimal conditions no effort guage.” Only participants naive with regard to eastern languages was spent to discard possible effects of the faces. Al- were enrolled. The participants were divided into two groups of 48 thoughRotteveelet al. (2001) also obtainedstronger sub- for the affective and nonaffective priming tasks, respectively. optimal as opposed to optimal priming with impover- Design ished presentation when instructions were equated, it The affective task had a 2 (focused vs. divided attention) 3 3 seems better to have some form of suboptimal presenta- (happy, neutral, or angry face) mixed factorial design, and the non- tion that allows more fully for the applicationof the same affective task had a 2 (focused vs. divided attention) 3 2 (male vs. fe- instruction. male face) mixed factorial design. Attention was manipulated be- Merikle and Joordens (1997) proposed that manipula- tween participants. The coupling of prime to target was tions of attention and of stimulus quality(i.e., in terms of counterbalanced over participants. Order of presentation was ran- domized by the computer for each participant individually. Affective presentation duration and masking) have parallel effects. and gender ratings of the ideographs served as dependent variables. They obtained equivalent effects of both manipulations As an exploratory dependent variable, the proportion of correct re- with false recognition, exclusion failure, and a variant of productions of individual string elements was calculated per condi- the Stroop-task. To achieve suboptimal processing, at- tion for each participant. tention