7 Principles of goal-based priming

211

Seven Principles of Goal Activation:A Systematic Approach to DistinguishingGoal Priming From Priming of Non-GoalConstructs

Jens FrsterJacobs University Bremen and University of Amsterdam

Nira LibermanTel Aviv University

Ronald S. FriedmanUniversity at Albany, State University of New York

moderated by self-regulatory processes (see, e.g.,Gollwitzer, 1990; Sorrentino & Higgins, 1986; Srull &Wyer, 1989). Affect, motivation, goals, standards, andneeds appear to influence all stages of the information-processing sequence, namely, perception, representation,and a variety of mental operations, such as inference gen-eration and memory, thereby altering behavior, emotion,judgment, and problem solving (see Frster & Denzler,2006, in press; Gollwitzer & Moskowitz, 1996). Inrecent years, social cognition researchers also have madetremendous progress in understanding the extent towhich judgment and behavior are guided without inten-tion or awareness (for reviews, see Dijksterhuis & Bargh,2001; Frster & Liberman, 2007). Most notably, demon-strations of what has been described as automatic goalpursuit or implicit volition (Bargh, Gollwitzer, Lee-Chai,Barndollar, & Trtschel, 2001; Chartrand & Bargh,1996; Moskowitz, Gollwitzer, Wasel, & Schaal, 1999;

Authors Note: This research was supported by two grants (Fo-392/8-2

and Fo-244/6-3) from the Deutsche Forschungsgemeinschaft to Jens

Frster. Special thanks go to Konstantin Mihov for editing the article.

Correspondence concerning this article should be addressed to Jens

Frster, Jacobs University Bremen, School of Humanities and Social

Sciences, PO Box 750561, D-28725 Bremen, Germany; e-mail: j.foerster

@iu-bremen.de.

PSPR, Vol. 11 No. 3, August 2007 211-233DOI: 10.1177/1088868307303029

2007 by the Society for Personality and Social Psychology, Inc.

Countless studies have recently purported to demon-strate effects of goal priming; however, it is difficult tomuster unambiguous support for the claims of thesestudies because of the lack of clear criteria for deter-mining whether goals, as opposed to alternative vari-eties of mental representations, have indeed beenactivated. Therefore, the authors offer theoretical guide-lines that may help distinguish between semantic, pro-cedural, and goal priming. Seven principles that arehallmarks of self-regulatory processes are proposed:Goal-priming effects (a) involve value, (b) involvepostattainment decrements in motivation, (c) involvegradients as a function of distance to the goal, (d) areproportional to the product of expectancy and value, (e)involve inhibition of conflicting goals, (f) involve self-control, and (g) are moderated by equifinality and mul-tifinality. How these principles might help distinguishbetween automatic activation of goals and primingeffects that do not involve goals is discussed.

Keywords: goal priming; semantic priming; proceduralpriming; automaticity; unconsciousness

From its earliest beginnings, but most prominently dur-ing the past decade, research in social cognition hasdemonstrated that both information processing andsocial behavior are quite substantially, if not inescapably,

2007 Society for Personality and Social Psychology, Inc.. All rights reserved. Not for commercial use or unauthorized distribution. at Universiteit van Amsterdam SAGE on July 11, 2008 http://psr.sagepub.comDownloaded from

Shah & Kruglanski, 2002) have challenged the field,revealing the remarkable capacity of the mental system toefficiently adapt to the environment in the absence ofattentional capacity, consciousness, or volition. However,although auto-motive researchers (Bargh, 1990; Barghet al., 2001; Dijksterhuis & Bargh, 2001) have adducedcompelling evidence that goals are mentally representedand may be unconsciously activated and pursued, the ques-tion remains as to how implicit self-regulatory processes, asthey have essentially described them, may be conceptu-ally and/or empirically distinguished from other auto-matic, activation-based phenomena, including semanticand procedural priming (for discussion of this issue, seeKruglanski et al., 2002; Moskowitz, Li, & Kirk, 2004).This question will be the focus of this article.

As discovered decades ago, priming of semantic con-cepts increases their accessibility, facilitating process-ing of related constructs (see Neely, 1977). Forexample, after reading the word lamp, people maybecome faster at reading the word light than withoutpriming. Increased semantic activation due to primingalso may engender perceptual assimilation. For instance,if participants are exposed to aggression-related con-cepts by unscrambling sentences such as leg her breakhe (Srull & Wyer, 1979), they may use this activatedknowledge in rendering judgments of ambiguouslyaggressive target individuals, such that these targetscome to be viewed as more hostile (for a review, seeFrster & Liberman, 2007; Higgins, 1996; Wyer,2004; for a meta-analysis, see DeCoster & Claypool,2004).

As alluded to above, priming has also been posited toactivate goals, representations of desired end-states,thereby eliciting action consistent with goal attainment.For example, after processing words related to achieve-ment, people have been found to become more compet-itive (Bargh et al., 2001). However, regardless ofwhether goals have indeed been activated, goal-relevantprimes may enhance the accessibility of goal-relatedknowledge constructs. For example, a shoe advertise-ment may trigger the goal to buy new footwear, pro-moting enhanced accessibility of constructs related tothis goal and thereby engendering increased efficiency atdetecting shoe stores. However, the question remains:How can we determine whether processing the adver-tisement has activated a goal or merely enhanced theaccessibility of semantic constructs related to this goal?Restated, did the advertisement ultimately increaseshoe-store detection by triggering a goal to find newshoes or by simply rendering the concept of shoes moreactive in long-term memory? In this article, we outlinesome principles of goal pursuit that may help to answersuch questions. First, however, we briefly review someof the key findings that have been marshaled in support

of the notion that goals are automatically activated, thetheoretical precepts that have been advanced to accountfor these effects, and the means that have been taken todistinguish between effects of semantic versus goal-related accessibility.

EVIDENCE FOR AUTOMATIC GOALACTIVATION AND PURSUIT

The empirical finding that may have been most sem-inal in advancing the notion of automatic goal activa-tion is Bargh, Chen, and Burrowss (1996, Experiment2) striking demonstration that participants primed withconcepts related to the elderly stereotype (e.g., Florida,old, lonely) walked more slowly down a hallway com-pared to those primed with words unrelated to age(e.g., thirsty, clean, private). In another influentialexperiment, Bargh et al. (1996, Experiment 1) showedthat priming the trait constructs of rude versus polite(using different trait-related words) resulted in partici-pants interrupting an experimenter sooner versus later,respectively. The two effects were based on consciouslyactivated knowledge that had an unintended influenceon further behavior. Yet another pivotal finding wasreported by Bargh et al. (1996, Experiment 3), whodiscovered that subliminal priming of a young, AfricanAmerican male face led participants to respond withmore hostility (a stereotypic behavior associated withthis social category) to provocation by the experi-menter, as compared to priming with a Caucasian face.This experiment shows that even though social infor-mation such as a face is not consciously perceived, con-structs associated with the stimulus can unintentionallyand unconsciously drive behavior with minimal or nodemand for controlled processing (see Bargh, 1994).

These effects have been replicated and extended, andit is now accepted as common knowledge that exposureto specific trait constructs, actual behaviors, or socialgroup members (whose stereotypes contain trait andbehavioral constructs) can result in the nonconsciousexpression of the activated behaviors (for review, seeDijksterhuis & Bargh, 2001): Although there are excep-tions (e.g., Dijksterhuis et al., 1998; Mussweiler &Frster, 2000), priming a trait, a stereotype, or a behav-ior results in an increased likelihood that individualswill perform the primed behavior. We will refer to theseeffects as behavioral assimilation effects (for a recentreview, see Wheeler, DeMarree, & Petty, in press).Assimilation occurs if a subsequent behavior reflects theimplication of the prime, whereas contrast or uncon-scious anchoring occurs if the subsequent behaviorreflects the opposite (see DeCoster & Claypool, 2004;Frster & Liberman, 2007).

212 PERSONALITY AND SOCIAL PSYCHOLOGY REVIEW


Assimilation effects have not only been shown to alterbehaviors but also to alter performance level on speedtests, intelligence tests, memory tests, creativity tests, andflexibility tests. For example, priming elderly increasesresponse latencies on lexical decision tasks (Dijksterhuis,Spears, & Lpinasse, 2001; Kawakami, Young, &Dovidio, 2002) and decreases performance on memorytasks (Dijksterhuis, Aarts, Bargh, & van Knippenberg,2000; Dijksterhuis, Bargh, & Miedema, 2000; but seeLevy, 1996) and priming professor leads to more cor-rect responses to general knowledge questions, whereaspriming soccer hooligan leads to less correct responses(Dijksterhuis & van Knippenberg, 1998). Primingpolitician results in participants producing longeressays than those not primed (with long-windedness asan associated trait of politicians; Dijksterhuis & vanKnippenberg, 2000), and priming punk leads to highercreativity than priming engineer (Frster, Friedman,Butterbach, & Sassenberg, 2005).

Other researchers have demonstrated effects of primingon performance. For example, priming helpful facilitateshelpful behaviors (Macrae & Johnston, 1998), primingconformity increases consensus (Epley & Gilovich,1999), priming intelligent leads to more correctresponses on general knowledge questions (compared topriming stupid; Dijksterhuis & van Knippenberg, 1998),priming creativity enhances creative thinking (Sassenberg,Kessler, & Mummendey, 2007), and priming flexibilityreduces perseverance on the Wisconsin Card-Sorting task(Hassin & Bargh, 2005), to name a few.

Finally, exposing participants to actual behaviorsalso has led to similar effects, in that the observedbehavior tends to be imitated (Chartrand & Bargh,1999; Lakin & Chartrand, 2003). Mimicry of an inter-action partners behavior (e.g., facial touching or footshaking) has been shown to occur without awareness ofhaving observed the partner behavior at issue or onesown expression of the behavior. More recently, Aarts,Gollwitzer, and Hassin (2004) argued that people maynot only automatically mimic concrete behaviors butmay also adopt and pursue a goal that is implied byanother persons behavior (given that this goal issocially acceptable). Their research on goal contagionshows that if participants are presented with behavioraldescriptions implying a certain goal, they may initiatethis behavior if an opportunity is provided to do so,even without being aware of the influence.

WHY DOES PRIMING AFFECTBEHAVIOR AND PERFORMANCE?

To account for behavioral assimilation effects,researchers have introduced both nonmotivational models

stating that semantic or associative links between con-cepts and behavior exist and automatically elicit actionand motivational models that involve the notion of goalsto predict automatic goal pursuit.

One common nonmotivational explanation for auto-matic behavioral assimilation is based on the assump-tion of a perceptionbehavior link (Bargh & Barndollar,1996; Bargh et al., 1996). This link is described as anunintentional mechanism whereby action is directlyassociated with perception so that perceiving a stimulusmay have a direct effect on behavior. Jamess (1890)principle of ideo-motor action is perhaps the first theoryto suggest that actions are directly connected to mentalrepresentations of the same actions. According toJames, the mere act of thinking about a behaviorincreases the likelihood that the behavior will be carriedout, given an appropriate situation for the behavior tobe expressed. Notably, the source of the activation isirrelevant for the ensuing behaviorboth externalpriming (e.g., seeing someone act in the manner at issue)and an internal thought about a behavior may result incarrying out a corresponding behavior.

This account resembles classic explanations forsemantic priming (Collins & Loftus, 1975). Presumably,semantic priming occurs because activation spreads in thesemantic (or associative) network and increases the acti-vation potential of connected constructs, thereby makingtheir activation more likely. According to the notion of aperceptionbehavior link, unconscious behavioral prim-ing occurs in much the same way, only some of the con-structs in the semantic or associative network arebehaviors rather than constructs. Activation spreadsfrom the prime to the behavior and makes it more likelyto be carried out than if it had not been primed. As withsemantic priming, this principle predicts that the moreoften and the more recently a concept has been activated,the more likely it will drive enactment of the correspond-ing behavior given that the behavior is applicable to thesituation (see Wheeler et al., in press).

To summarize, the perceptionbehavior link andrelated accounts of automatic social behavior (Kawakamiet al., 2002) posit direct activation of behavioral repre-sentations by the act of perception: Perceptual inputsare translated automatically into corresponding behav-ioral outputs (Dijksterhuis & Bargh, 2001, p. 1).However, there are alternative accounts. One much-vaunted possibility is that a prime activates a goal to per-form a certain task (Bargh, 1990, 1997). Inasmuch asgoals can be represented in memory (see Kruglanski,1996; Moskowitz et al., 2004; Shah, Kruglanski, &Friedman, 2003), it is possible that, for example, the ver-bal prime aggressive not only activates semantic asso-ciates of aggression that eventually lead to aggressivebehavior but also or alternatively activates the goal to

Frster et al. / GOAL AND NON-GOAL PRIMING 213


aggress against a certain person, thereby eliciting auto-matic goal pursuit.1 This is because certain features maybe frequently and habitually linked to certain goals(Bargh, 1997). In a similar vein, people (Aarts et al.,2004; Fitzsimon & Bargh, 2003; Shah, 2003) or envi-ronments can remind one unconsciously or consciouslyof ones goals (Aarts & Dijksterhuis, 2003). The afore-mentioned examples refer to situations in which theprime and the goal are connected via a history of frequentco-occurrence. It is important to note, however, that suchconnections also may be established by a single act ofvolition, that is, the formation of an implementationintention (Gollwitzer, 1990, 1993; Gollwitzer & Bayer,1999). For example, when a student decides that at theend of a biology class she will go to the computing centerto write an e-mail to her mother, biology class may acti-vate the goal to write such an e-mail to her mother auto-matically, with little or no awareness.2

Aside from this, it is also possible that a prime activatesa procedure. For example, Mussweiler and Frster (2000)primed women and men with sex-related constructs andfound cognitive accessibility of aggression-related actionsto be enhanced for both genders compared to anunprimed control group (Mussweiler & Frster, 2000,Experiment 1). However, in another experiment, anincrease of overtly aggressive behavior after sex primingwas only observed for men (Mussweiler & Frster, 2000,Experiment 3). The authors argued that procedural prim-ing took place: Because men frequently act aggressively insexual situations, the procedures involved in this behaviorare more likely elicited when they are mentally simulatingsex (due to situational exposure to sex-related concepts).Because women are usually the victims of sexual aggres-sion, for them, activation of a cognitive procedure todetect cues of aggression is more likely than activation ofaggressive behavioral scripts (as found in another study;Mussweiler & Frster, 2000, Experiment 4). Thus, eventhough the semantic activation of aggression occurred forboth genders, the procedures elicited by the primes weredifferent for men versus women (cf. Schooler, 2002; Smith& Branscombe, 1987). Assuming the results of this studyindeed reflect dissociation between perception and behav-ior, they highlight the difficulty of distinguishing whethera given manipulation has primed a semantic concept, agoal, or a procedural routine.

SEMANTIC PRIMING, PROCEDURAL PRIMING,OR GOALS?

How, then, can we tell whether a prime activated agoal as opposed to only producing semantic priming?Let us first emphasize that these are different outcomes:Not every effect of priming activates a goal. For

example, it is difficult to see how goals might beinvolved in the effect of priming lamp on subsequent(faster) readings of light.3 Two seemingly obvious cri-teria to distinguish between goal priming and semanticpriming seem to us rather problematic: using the factthat the effect is behavioral as an indicator for primedgoals, and using the fact that the manipulationsinvolved motivational, as opposed to only semantic,content as an indicator of primed goals.

First and foremost, the first account is misleadingbecause some internalized cognitive operations, such asjudgment or memory, may well involve goals (e.g., fair-ness vs. accuracy goals, impression formation vs. mem-orization goals). Second, theories such as the ideomotortheory suggest that behavior may result from sheer con-ceptual priming. For instance, the results of the afore-mentioned study by Bargh et al. (1996), in which elderlyprimed participants walked slower, may be seen asresulting from the operation of the perceptionbehaviorlink, a process that in no way requires the involvementof goals. According to this account, semantic conceptsare directly linked to motor representations. Thus, theconcept of slow, which was presumably activated by theelderly category, could directly lead to slow walking. Ifthis is true, one cannot tell from a behavioral measurethat a goal was activated.

The phenomenon of procedural priming introducesanother difficulty with using behavioral outcomes as ahallmark of activated goals because it presumablyinvolves goalless actions (see Frster & Liberman, 2007).Procedural priming refers to priming of proceduresatthe priming phase, people perform or are led to men-tally simulate an action and then (facilitating) carryovereffects on the performance of related actions are exam-ined. For example, solving analytical reasoning prob-lems may activate procedures used in such tasks,thereby transiently facilitating the ability to solve relatedproblems (Schooler, 2002). Notably, experimentaldemonstration of procedural priming does not requirelearning of novel procedures. Rather, the action per-formed at the priming stage is typically overlearned tosuch an extent that a few more instances of practicemake virtually no difference (e.g., adding numbers,inferring traits from behavior).

The distinction between semantic priming and proce-dural priming is best exemplified by a study by Smith andBranscombe (1987), who replicated a classic experimenton the effect of priming on disambiguation of the behav-ior of a target person (Srull & Wyer, 1979). In additionto the original priming group, in which participantsunscrambled sentences describing hostile behaviors (e.g.,leg her break he), they added a procedural primingcondition, in which participants were presented with thesame sentences in an unscrambled form and were asked



to choose the matching trait (e.g., hostile). Inferring traitsfrom behavior was, in fact, the procedure people neededto employ when judging the target person in the secondphase. After either 15 s or 15 min, participants rated theaggressiveness of a description of a vaguely aggressivebehavior. Priming traits (i.e., unscrambling sentences)had an assimilative effect only after a short delay, butprocedural priming continued to have this effect after alonger delay. It seems that participants practiced the pro-cedure of extracting traits from behavior and that makingthis procedural knowledge accessible facilitated makingsuch inferences at the test stage.

Notably, procedural priming also may involve fairlygeneric cognitive processes, such as processing styles,because processes elicited in one task may carry over toanother task. Of relevance here is Schoolers (2002;Schooler, Fiore, & Brandimonte, 1997) theory of pro-cessing shifts, which assumes that a cognitive procedurethat is activated in the course of engaging in one taskremains active and is transferred to subsequent tasks.Transfer-appropriate processing shifts are said to resultwhen the residually activated procedures are beneficial forsubsequent processing, whereas transfer-inappropriateshifts are said to result when the procedures at handimpair subsequent processing. For example, Macraeand Lewis (2002) found that face recognition wasenhanced following a global processing task and wasimpeded following a local processing task. This was pre-sumably the case because the encoding of faces profitsfrom holistic processing. Procedural priming, likesemantic priming, does not necessarily induce goal pur-suit. For example, in Macrae and Lewiss (2002) experi-ment, it was probably not the case that global processinginstilled in participants a particular goal. Rather, itfacilitated performing certain mental operations (e.g., toprocess information in a more holistic way) by virtue ofmaking them more accessible.

Can we decide that goals have been activated simplybecause the manipulation clearly involved motivationalcontent, such as the use of verbs implying goal-directedaction or intention? For example, if we tell participantsto search for the word aggress among other words, thismay indeed activate the goal of acting aggressively.However, again, if as the result of this manipulation thesame participants exhibit an enhanced accessibility ofconstructs related to aggression, are we to attributeenhanced accessibility to semantic priming (i.e., merelymentioning the word aggression) or to an active goal(i.e., to seek out stimuli related to current concerns)?Likewise, are individuals primed with the word aggressmore likely to find the experimenter impolite and there-fore aggress against him because they perceive anaggressive response to be more appropriate or, in con-trast, are aggressive means to pursue ones goal more

accessible (interrupting vs. waiting and suffering)? It isnot easy to find manipulation checks in the literaturethat certify that goals have indeed been activated; more-over, it is unclear as to what shape such checks shouldtake (for a similar challenge, see Wheeler & Petty,2001). Explicit checks are problematic, not onlybecause their validity is called into question by the factthat individuals lack introspective insight into theirmotives (Nisbett & Wilson, 1977) but because adminis-tration of such checks following implicit or subliminalgoal manipulations will bring goals to consciousness,undermining the ability to draw conclusions regardingthe automaticity of goal pursuit. Furthermore, in manycases, asking questions about motivation (e.g., Howmotivated were you to complete the task?) seems redun-dant to the participants and may violate conversationalnorms (Schwarz, 1994), thus undermining the diagnos-ticity of self-reports.

Thus, it seems difficult to conclude from the sheernature of the dependent or independent variable whetheran effect was driven by semantic priming, perceptionbehavior links, procedural priming, or goal pursuit. It isespecially difficult to infer goals from the independentvariable in cases of subliminal priming or unconsciousgoal activation. And it becomes even more obvious thatmanipulation checks in these cases would exclude self-reports because people do not even know that concepts(goals, procedures, or concepts) were activated. Webelieve that effects of goal activation are qualitativelydifferent from those of semantic or procedural primingand suggest a systematic means of distinguishing theseeffects. Some researchers have begun to address suchissues (Bargh et al., 2001; Moskowitz et al., 2004;Wheeler & Petty, 2001) and discussed how differenttypes of priming may differentially influence behavior.In the discussion that follows, we will integrate andexpand on their ideas. We propose to do so by examin-ing the processes of goal pursuit. In the remainder ofthis article, we will examine seven principles derivedfrom research on motivation that may be useful in dis-tinguishing goal-priming effects from non-goal-primingeffects. We are by no means suggesting that these prin-ciples are exhaustive. However, we hope that they willprove useful for researchers, particularly those studyingautomatic goal activation.

SEVEN PRINCIPLES OF GOAL ACTIVATION

Goal activation may be identified by a set of featuresthat derive from and often cut across numerous theories ofself-regulation. In short, we suggest that these features maydistinguish goal-priming effects from effects of semantic orprocedural priming. These principles are as follows:



1. Goal-priming effects involve value.2. Goal-priming effects involve postattainment decrements

in motivation.3. Goal-priming effects involve gradients as a function of

distance to the goal.4. Goal-priming effects are proportional to the product of

expectancy and value.5. Goal-priming effects involve inhibition of conflicting

goals.6. Goal-priming effects involve the operation of self- control

processes.7. Goal-priming effects are moderated by equifinality and

multifinality.

We now turn to discussing these principles and how theymay help distinguish between goal-related effects, thosereflecting goal priming, and effects that do not necessar-ily involve goals. We will summarize studies that exem-plify such principles, although we admit, sometimesexisting studies do not provide the best example for suchan effect. When existing studies may be interpreted interms of semantic priming, we propose possible exten-sions of these studies that in our view would serve to offermore conclusive evidence for the principle.

Goal-Priming Effects Involve Value

In the present context, we wish to discuss three aspectsof the connection between goals and value. First, goalschange valuethey make people assign positive value tothings that are conducive for goal achievement and nega-tive value to things that are detrimental for goal achieve-ment. Second, goals involve emotion and energymobilization. People feel happy, satisfied, and/or relievedwhen they achieve a goal; they feel frustrated, tense,and/or depressed when they fail to achieve the goal; andthey feel energized, eager, or vigilant in the process ofstriving toward the goal. Third, goals are sensitive tochanges in valuethey become stronger or weaker ifvalue increases or decreases, respectively. Let us examinehow each of these aspects of goals may help us distin-guish between semantic priming and goal priming.

Goals affect value. Most generally, people want toattain positive end states and avoid negative end states.But what is good and what is bad varies tremendouslyfrom person to person and, more important for our pur-poses (and even more dramatically), from situation tosituation. For example, when we need to find a screw-driver to assemble furniture, the value of that objectincreases relative to a state in which no such goal exists.We believe that priming a goal should increase the valueof the end state and the value of objects and actions thatare perceived as conducive to goal attainment. Priminga semantic construct, on the other hand, would not nec-essarily have the same effect. For example, if priming

furniture activated the goal to assemble furniture,then the value of a screwdriver may increase, but whenit only activated the semantic construct of furniture, nosuch effect on value would be observed.

Of importance, goals have the power to constructand reconstruct value in a way that does not necessarilyreflect preexisting, overlearned associations betweenconstructs and evaluations (see Ach, 1910; Custers &Aarts, 2005; Lewin, 1951).4 For example, if our onlychance to get a screwdriver and therefore assemble ournew furniture is to borrow one from a neighbor, anexcitatory link might be created between neighborand furniture, a link that did not exist in the associa-tive network before that situation. Moreover, positivevalue would be added to the concept neighbor (or theassociation between neighbor and a mental representa-tion coding for evaluative positivity would be strength-ened; Bargh, 1997). More generally, a goal pursuitsystem is capable of structuring and restructuring theassociative network in line with the actors goals andthe situational affordances, a point to which we willreturn when discussing other principles as well.

It is true that semantic priming may affect value viadisambiguation. For example, priming aggressionmay enhance perceptions of another person as aggres-sive, thus also making these impressions more nega-tive compared to a nonpriming condition. As anotherexample, priming thirsty may enhance the positivevalue of the word water because it will make it morelikely to be disambiguated as drinking water ratherthan, for example, flood water. Change of value dueto disambiguation is not indicative of a primed goal.Thus, researchers that use change in value of goal-related constructs to demonstrate primed goals maybe advised to make sure that such changes were notmerely the result of the primes effects on the way thetarget word is interpreted (e.g., by making the targetwords nonambiguous).

Many studies have shown that primed goals andprimed needs affect the value of target constructs, butin these studies, associative links and disambiguationwere not ruled out. For example, priming the goal ofstudying enhanced the value of academic concepts(e.g., library, college) and reduced the value ofnonacademic temptations (e.g., television, chat;Fishbach, Zhang, & Trope, 2007). Thirst enhancedthe value of water as assessed in implicit measures(Ferguson & Bargh, 2004). In both examples, valuewas assessed implicitly via affective priming. In bothstudies, however, the prime and the target constructswere semantically associated and the target waspotentially ambiguous in value; therefore, these stud-ies do not rule out the possibility that value was in factaffected by semantic priming.



It is also useful to note that instrumental valuechanges in the course of goal pursuit. In our example,the neighbor would increase in positive value before hehelps us with the tool, but his instrumental value woulddecrease sharply after he gives us the tool. Water wouldincrease in positive value as long as the goal to drinkexists but will lose positive value once thirst isquenched. Such patterns of value change would thusserve as a further indication of an active goal.

In a recent study by Ferguson and Bargh (2004,Experiment 1), participants worked on a word creationgame. Whereas some were asked to perform wellbecause they would get points for each solution (theachievement goal condition), others were not (the con-trol condition). All participants completed an implicitevaluation measure with primes related to the goal (e.g.,compete, achieve, win) and primes unrelated to thegoal. In addition, some completed the implicit evalua-tion measure before finishing the word game (the unful-filled goal condition), whereas some had to completethe game (the fulfilled goal condition). Results showedthat implicit evaluations of goal-related stimuli (i.e.,slower responses to negative targets after presentationof goal primes) were more favorable in the goal condi-tion only if the goal has not been fulfilled. Thus, goalfulfillment renders formerly favorably valued objectsless favorable because they lose utility. Notably, inFerguson and Barghs study, the goal was not primedbut was induced explicitly. We are not aware ofresearch that has demonstrated a similar effect withprimed goals. If demonstrated, however, such a patternof change in value would be indicative of a primed goalrather than a primed semantic construct.

Goals involve emotion. Closely related to value areemotions. People feel frustrated, angry, or disappointedif goal attainment is blocked and they feel happy,relieved, or satisfied if they attain the goal. If a primeactivates such emotions, then most likely the primesinduced a goal. Further support for an active goalwould be obtained if emotions follow the course thatwould have been predicted by active goalstension,eagerness, and vigilance before fulfillment; happinessand relief after fulfillment; disappointment and anxietyafter failure. To the best of our knowledge, no suchstudies have been conducted. In fact, we can only spec-ulate that these emotions, which typically accompanypursuit of conscious goals (Carver & Scheier, 1990;Friedman, 1999), also characterize pursuit of goals thatwere activated via priming. If this were to be demon-strated, however, the elicitation of emotions that char-acterize goal pursuit would demonstrate goal priming asopposed to priming of semantic constructs. Notably, itis possible that emotions bring the activated goal to

awareness. Thus, if people are happy or disappointedbecause they fulfilled or failed to attain a certain goalthat was unintentionally activated, they may infer thata goal was activated. To the best of our knowledge,there is no research examining such inference processes.It also would be interesting to see whether these infer-ences are invariably correct.

Goals are affected by value. People pursue thingsthat they value, and the extent of their motivationdepends on the value of the end state. Therefore, if aneffect of priming may be shown to depend on the valueof the prime, then it is likely that a goal, rather than asemantic construct, has been primed.

Aarts et al. (2004) showed effects of prime value ongoal priming. In one experiment, they primed partici-pants with descriptions of people who worked for moneyor on a voluntary basis. Subsequently, they were told thatif time would allow it, they could earn cash by perform-ing a task following the one on which they were currentlyworking. The time participants took to finish working onthe current task served as an indicator as to how stronglyparticipants were motivated to earn cash. In an ostensiblyunrelated task, the researchers also measured how muchparticipants needed money. They found that participantswho read about someone with the implied goal ofmaking money were faster to finish the task in time to dothe money task, but this only happened for those partici-pants who, by self-report, needed money. The resultsshow that goal priming only takes place if the primedbehavior is valued by the participants (in the presentexperiment, if they needed the money).

Conceptually similar results were demonstrated byStrahan, Spencer, and Zanna (2002), who showedeffects of subliminal priming on behavior and evalua-tion in a state of need but not when needs were absent.In their studies, some participants were made thirsty,whereas other participants were made nonthirsty beforebeing primed with names of drinks. Subliminal primingof thirst-related words led to more beverage consump-tion only for those participants who were primed withthirst-related words and who were thirsty compared toall other participants. In a second study, only thirstyparticipants were tested. They were primed with wordsrelated to thirst or with words unrelated to thirst andthen evaluated a thirst-quenching drink and a less thirst-quenching drink that was said to be electrolyte restor-ing. Thirst-primed participants, more than those notprimed by thirst, assigned positive value to the thirst-quenching drink but not to the other drink.

Similarly, Shah (2003) found that goals can beprimed by subliminally exposing participants to namesof significant others. He showed that commitment,goal accessibility, and goal pursuit were enhanced when



participants were primed with a significant other thatvalued this goal (e.g., their mother valued achievement),whereas the opposite was true when the significantother did not like the goal (e.g., their friend preferredaffiliation goals over achievement). The priming effectwas moderated by liking or closeness to the significantother who valued the goal. For example, if people likedtheir mother and the mother valued achievement, thenparticipants worked harder on a task and were morecommitted after mother priming.

Another way to increase the value of end states is byintroducing discrepancies (Higgins, 2006; Lewin,1951). In a recent experiment, Moskowitz, Li, and Kirk(2007) reminded participants of their failure to be anegalitarian person. On the basis of earlier research, theyassumed that this would enhance the value of showingegalitarian behavior (Moskowitz et al., 1999). Thoseparticipants showed in subsequent tasks less implicitstereotyping and a higher activation of egalitarian con-cepts when exposed to stimuli that are related to astereotyped group. It is possible to contend that thereminder of ones failures to be egalitarian increased thevalue of showing egalitarianism, and a goal to be egali-tarian was activated when exposed to a cue regarding astereotyped group. For our purposes, it is interesting toexamine the possibility of inferring a primed goal if itaffects the value of the end state to a greater extent aftera discrepancy to the goal has been activated than in theabsence of such a discrepancy.

Cesario, Plaks, and Higgins (2006) proposed thatupon perception of a social group member, the per-ceivers motivational system begins to prepare for aninteraction with the primed target. The specific outputsof this preparation depend not only on characteristics ofthe target (Is the target threatening? Is the targetrude?) but also on ones evaluation of the target (DoI like this target? Do I want to approach or avoid thistarget?) and the context of the interaction. Cesarioet al. (2006) primed participants with the elderly stereo-type or the youth stereotype and examined their walk-ing speed. In addition, they assessed participantsattitudes toward the elderly and toward youth with animplicit attitude measure (Fazio, Jackson, Dunton, &Williams, 1995). The results showed that for partici-pants primed with elderly, as positivity of attitudestoward the elderly increased, walking speed decreased,whereas for participants primed with youth, theopposite occurredpositivity of attitudes toward youthwas associated with increased walking speed.

A critic may argue that associative networks varied forpeople who do versus do not like the elderly. Morespecifically, it may be argued that for the former, but notfor the latter, walking slower was associated with (andtriggered by) the elderly prime. Would similar results

emerge if the value of the primed group were manipu-lated rather than assessed? A recent study by Epstude andFrster (2007) seems to provide an affirmative answer.

In one of their studies, Epstude and Frster (2007)asked participants to sort cards that could be categorizedonly by either shape or color. Prior to this task, a mini-mal group manipulation was administered (Swinney,2000; Turner, 1978), in which all participants com-pleted an alleged personality test, which was used toassign them to groups. Half of the participants receivedfeedback that they had a concave personality, whereasothers were told that they had a convex personality (cf.Otten & Wentura, 1999). Before participants startedsorting the cards, they were asked to look at exampleson the computer of how other participants sorted thecards to get an idea about the task. Half of the pho-tographs on the screen showed sorting based on color,whereas the other half showed sorting on the basis ofshape. However, before the presentation of the pictures,participants were subliminally primed with the conceptsconvex or concave. For half of the participants,convex always co-occurred with shape sorting and con-cave with color sorting; for the other half of the partic-ipants, convex always co-occurred with color sortingand concave with shape sorting. Thus, in some condi-tions, shapes were associated with the ingroup andcolor with the outgroup; in others, shapes were associ-ated with the outgroup and color with the ingroup.

The results of this study showed that participantswere more likely to sort the cards according to the cri-teria of their ingroup. For example, if convex was sub-liminally primed before shapes and participants thoughtthey were in the convex group, then they would tend tosort the cards by shape as well. This was the case despitethe fact that group membership was minimal, the prim-ing of group membership before participants receivedalleged examples of other participants was subliminal,and participants did not report any relationshipbetween the personality tests and the sorting task.Notably, the effects in Epstude and Frster (2007) weremoderated by identification with the ingroup. Morespecifically, after they had received feedback about theirperformance in the so-called personality test, partici-pants were asked how much they identified with theconvex or concave personality groups. Results showedthat the more participants identified with their respec-tive group, the more they imitated group behavior.

In sum, and most pertinent to the issue at hand, itseems that people automatically imitate behavior orstart actions upon goal priming only if the primed goalhas value to them. Value can be derived from the situa-tion, as in the above-mentioned case, where ingroupmembers ostensibly shared it. The studies show that thiscan happen even if the association of a certain behavior



with group membership has been experimentallymanipulated (rather than preexisting) and uncon-sciously perceived.

Goal-Priming Effects Involve PostattainmentDecrements in Motivation

With its roots in Lewins (1926; see also Lewin, 1951)field theory, the Zeigarnik effect was demonstratedwhereby people remember interrupted tasks better thancompleted tasks (Zeigarnik, 1927; for reviews, seeButterfield, 1964; Heckhausen, 1991; Wicklund &Gollwitzer, 1982). Presumably, this occurs becauseinterruption preserves the goal to complete the task andmaintains the goal-related tension, whereas task com-pletion releases the tension and reduces memory for thetask. More recently, this logic has been applied to socialcognitive research on accessibility as well. Libermanand Frster (2005) argued that although Lewin andZeigarnik did not formulate their findings in terms ofaccessibility, the free recall measure they employedcould be interpreted in this manner (see also Tulving &Patterson, 1968). The Zeigarnik effect could then sug-gest that goals enhance the accessibility of goal-relatedconstructs, and goal fulfillment reduces this accessibility.

The most direct evidence for postfulfillment inhibitioncomes from Marsh, Hicks, and Bink (1998), who used alexical decision task to examine the accessibility ofintended actions both prior to and after goal completion.In this lexical decision task, participants had to indicateas quickly and as accurately as they could whether astring of letters was a word or a nonword. Faster lexicaldecisions on semantic associates of a construct indicate ahigher accessibility of the construct (Neely, 1991). First,replicating an earlier study by Goschke and Kuhl (1993),they found that before completion, accessibility ofintended actions was enhanced relative to unintended(i.e., to-be-observed) actions. Second, it also was foundthat after performing the action, accessibility of the per-formed action was inhibited and in fact lower than thatof observed actions, a result that is in line with the notionof postfulfillment inhibition. Marsh and his colleagues(1998) conducted their studies within the framework ofprospective memory, and the goals they examined placeda considerable burden on working memorytheyinvolved memorizing (and, in the goal condition, alsoexecuting) a series of actions (e.g., distribute the cutlery,sharpen the pencil, pour the water, spray the aerosol,stack the articles). It could be, then, that inhibition aftercompletion is only characteristic of situations in whichmemory is overloaded and the system seeks to refreshworking memory.

Recently, Frster, Liberman, and Higgins (2005)obtained similar findings with less memory-taxing goals.

In a series of studies, participants looked through a seriesof pictures with the goal of finding and reporting to theexperimenter once they had found a picture of glassessequentially followed by a picture of scissors. They weretold that there would be one and only one such combi-nation in the entire set of stimuli. Participants in the con-trol, no-goal condition looked through the same pictureswithout being instructed to find a target. Specifically,glasses and scissors were mentioned to them as examplesof stimuli that they might see but they did not receive anyspecial instructions about these stimuli. Participants inboth groups looked through four blocks of pictures.Pictures of glasses and pictures of scissors appeared anumber of times among the stimuli but the target combi-nation appeared only once, in the third block. Thus,Blocks 1 and 2 constituted the prefulfillment stage andBlocks 3 and 4 constituted the postfulfillment stage.

After each block of pictures, participants performeda lexical decision task similar to the one mentionedabove. Words related to glasses, words unrelated toglasses, and nonwords were presented randomly, andparticipants had to indicate as quickly and as accuratelyas they could whether a string of letters was a word ora nonword. In this way, accessibility of goal-related andgoal-unrelated constructs was examined both beforeand after goal fulfillment, and it was compared to theaccessibility of the same constructs in a control, no-goalcondition. The results showed that, as predicted, in thegoal condition, accessibility of goal-related words wasenhanced prior to goal fulfillment. Furthermore, itincreased from Block 1 to Block 2. Inhibition of goal-related constructs was evident immediately after fulfill-ment, in Block 3. Notably, postfulfillment inhibitionwas evident both in comparing the postfulfillment stageto the prefulfillment stage and in comparing the goalcondition to the control, no-goal condition. It is inter-esting to note that postfulfillment inhibition was par-tially released later, in Block 4, indicating that suchinhibition is eventually released (for a detailed discus-sion on this effect, see Liberman & Frster, 2005).

We argue that diminution of accessibility (i.e., inhi-bition) after goal completion is another mechanism thatdistinguishes goal priming and semantic priming. Fromthe perspective of semantic priming models (Frster &Liberman, 2007; Higgins, 1996; Wyer, 2004), goalattainment merely constitutes activation of a goal-related construct, and therefore, it should increaserather than decrease accessibility. To illustrate, in theexperiments by Frster et al. (2005), participants had tomention the glassesscissors combination to the experi-menter as soon as they saw it. According to traditionalmodels of semantic priming, ongoing consideration ofand search for these target objects should continuallyincrease their activation levels, thereby increasing their



accessibility to consciousness even after goal fulfillment(Higgins, 1996). Thus, the finding of postfulfillmentinhibition, as opposed to increased or merely persistentactivation, may distinguish accessibility from goalsfrom accessibility that is due to semantic priming. Forexample, if a researcher wishes to investigate effectsfrom subliminal activation of goals, then she mightmeasure accessibility of the goal after the goal has beencompleted. She should expect a decline of accessibilityrelative to baseline after goal fulfillment.

Indeed, Liberman and Frster (2000) used this specificfeature of inhibition after fulfillment to advance a moti-vational theory of postsuppressional rebound. They rea-soned that people may activate the goal to engage in thesuppressed action while they are trying to suppress it.People may infer this goal from suppression difficulty andfailure to suppress (Frster & Liberman, 2001). Forexample, they may think, If it is so difficult for me notto think of the cake, then it must be the case that I reallywant to eat it. In line with the principle of goal activa-tion, Liberman and Frster (2000) reasoned that once agoal is activated, it renders goal-related informationaccessible. Furthermore, they assumed that goal-relatedaccessibility is different from accessibility due to semanticpriming in that goal-related accessibility should decreaseupon goal fulfillment. In an experiment, Liberman andFrster (2000) provided to some participants such anopportunity to attain the goal that was presumably acti-vated during suppression. More specifically, they askedAmerican undergraduates to describe a day in the life ofan African American person who was presented in a pho-tograph. Participants in the suppression condition wereasked to avoid writing anything racist, whereas partici-pants in the no-suppression condition did not receive anyspecific instructions. Afterward, half of the participantsin each condition were assigned to the expression condi-tion, in which they had to describe a day in the life ofanother African American person while taking a perspec-tive of a racist writer. Participants in the no-expressioncondition did not write another story.

In terms of predictions, the authors reasoned that if agoal has been activated by the first suppression phase,then expressing the racist thoughts should constitute goalfulfillment and accessibility of racist thoughts shouldplummet. If, however, suppression only activated racistthought by way of semantic priming, writing a raciststory should serve to further enhance such accessibility.The authors measured accessibility of racist stereotypesby presenting to participants a description of an ambigu-ously aggressive person and asking them to evaluate him(see Srull & Wyer, 1979). High accessibility of raciststereotypes is reflected in higher aggressiveness ratings ofthe person because it has been shown that in Americanundergraduates, the African American stereotype is

related to aggression and thus renders aggressive contentaccessible (see, e.g., Devine, 1989). Replicating thephenomenon of postsuppressional rebound, in the no-expression group, participants that suppressed racism(and were given no opportunity to express it in the sec-ond phase of the study) rated the person as more aggres-sive than did participants who did not undergosuppression. More important, participants who first sup-pressed racism and then expressed it in the second phaseof the study rated the ambiguously aggressive person asless aggressive than the group that did not express racism.Thus, it seems that suppressing racist thoughts had acti-vated the goal to think those thoughts, a goal that wasfulfilled and thus deactivated by further expression ofthese thoughts (for reviews of this research, see Frster &Liberman, 2004, 2005). Such findings strongly point tothe utility of examining patterns of activation and inhibi-tion following presumed goal completion as a means ofdetecting the operation of goal priming (for another appli-cation, see Liberman, Frster, & Higgins, 2007). In addi-tion, the findings on suppression of thoughts illustratethat goals that are not made explicit can produce activa-tion and postfulfillment inhibition because participants inthe experiments were not aware of the fact that the goalto be racist was activated or attained. Notably, in thesestudies, participants answers during thorough debriefingreflected that they neither knew that they were pursuing agoal nor that they had attained it (see Bargh, 1994).

It is interesting to conceptualize failure as a specificcase of nonfulfillment and predict that accessibility ofgoal-related constructs would be higher after failurethan after success. Moskowitz (2002) asked partici-pants to remember either an event of athletic failure oran event of athletic success. He reasoned that the failurecondition would enhance accessibility of constructsrelated to athleticism. Using a Stroop-like measure ofaccessibility, he found an enhanced accessibility of ath-letic words after imagining a failure but not after imag-ining success. Accessibility of semantic constructs due topriming should not show a similar sensitivity to successversus failure, and therefore, those effects may serve todistinguish semantic priming from priming of goals.

Goal completion sometimes may be achieved viaattaining a substitute goal (Lissner, 1933; Mahler, 1933;Wicklund & Gollwitzer, 1982; see also principle 7). Forexample, if one fails in an exam and if this threatens thehigh-priority goal of being intelligent, understanding atough passage of James Joyces Finnegans Wake couldpotentially compensate for failure in the exam and canthus reduce frustration and rumination. Koole, Smeets,van Knippenberg, and Dijksterhuis (1999) demonstratedthat accessibility of task-related concepts after failuremay be halted by self-affirmation of values unrelated tothe task. In a series of experiments, they let participants



fail at an ostensible intelligence task that was manipu-lated to be unsolvable. Failure increased accessibility ofintelligence-related concepts. This enhanced accessibility,however, was not evident if, after failure, participantshad a chance to affirm other subjectively important val-ues (e.g., the value of being social or being politicallyinvolved). The results demonstrate that certain values,such as demonstrating intelligence, serve a higher ordergoal of thinking positively about oneself, which can beattained by different means. After goal attainment, acces-sibility of the higher order goal may drop (also see Martin& Tesser, 1996; Tesser & Cornell, 1991; for a more crit-ical discussion of goal fulfillment of abstract goals, seeLiberman & Frster, 2005).

Substitution may be achieved by instantiating fairlybroad goals, as demonstrated by research on self-affirmation(see Martin & Tesser, 1996). According to research onself-affirmation, people strive to maintain the overallintegrity of the self, and experiencing different typesof failure (e.g., a failure to maintain consistency ofones attitudes) may be alleviated by an act ofself-affirmation (e.g., reminding oneself of ones impor-tant values; Steele, 1988). Recently, Koole and vanKnippenberg (2006) applied this logic to suppression ofstereotypes. They reasoned that suppression failures cre-ate a general sense of tension that can be reduced byself-affirmation in domains that are unrelated to the sup-pression content or the act of suppression. Consequently,any positive feedback on the integrity of the self shouldeliminate postsuppressional rebound effects because itwould stop rumination on suppression failures. This isexactly what they found: After the suppression phase,participants were either given positive (e.g., You arevery social) or neutral feedback (You are reasonablysocial) on a personality questionnaire they had to fill outbefore the suppression phase started. Those with the self-affirming feedback showed no enhanced accessibility rel-ative to a nonsuppression group (i.e., no rebound effect)compared to the neutral feedback suppression group.Presumably, the tension produced by the difficulty tosuppress a stereotype was reduced by the self-affirmingfeedback. This feedback fulfilled the goal of self-affirmationand reduced accessibility of goal-related constructs. Inthese studies, it is unlikely that participants were aware ofthe content of the goal, its activation, or attainment.

In sum, we think that inhibition after goal fulfillmentis an important principle for deciding whether a goalwas primed. Semantic priming mechanisms cannotexplain reduction in the accessibility of goal-relatedconstruct below baseline upon goal completion. Notethat in most of the experiments cited above, participantsexplicitly mentioned the goal or goal-related concepts.For example, in the goal studies by Frster et al. (2005),participants explicitly told the experimenters that they

found the combination of scissors and glasses beforethey started working on the lexical decision task.Similarly, some participants in the Liberman andFrster (2000) studies wrote racist stories, presumablyfulfilling the goal to use racist concepts that was trig-gered by suppression instructions, and thus frequentlymentioned stereotypic concepts. By the logic of seman-tic priming, these should have enhanced the accessibil-ity of goal-related constructs. Instead, goal fulfillmentresulted in inhibition, that is, reduction of accessibilitybelow baseline (see Higgins, 1996). Of course, modelsof semantic priming (e.g., connectionist models; Smith& DeCoster, 1998) can predict inhibition effects.However, these models do not specify that inhibitionshould follow goal attainment.

Turning from semantic to procedural priming, littleis known about how the latter phenomenon is affectedby completion versus interruption of the priming tasks.The mechanisms that supposedly underlie proceduralpriming (carryover effects, see Schooler, 2002), how-ever, presume that its effects should not be sensitive towhether the priming task was successfully completed,interrupted, or failed.

Goal-Priming Effects Show Gradients as a Functionof Distance to the Goal

A classic principle of goal-directed behavior is thatmotivational strength increases as distance from thegoal decreases, that is, the goal looms larger as one iscloser to it (Lewin, 1935; Miller, 1944, 1959). In alandmark study demonstrating this principle, Brown(1948) trained one group of rats to run down a shortalley to attain food when hungry. Each rat wore a har-ness connected to a recording device such that thestrength of its pull when stopped at a specific point inthe alley could be measured in grams. Brown (1948)found that the animals that stopped nearer to the foodpulled harder than those who were stopped fartheraway. Further supporting this principle, stronger moti-vation closer to the goal has been observed in severalother studies, with both animals and humans (e.g.,Frster, Grant, Idson, & Higgins, 2001; Frster,Higgins, & Idson, 1998; Gjesme, 1974; Hearst, 1960,1962; Losco & Epstein, 1977; Miller & Kraeling, 1952;Miller & Murray, 1952; Murray & Berkun, 1955;Rigby, 1954; Smith, 1965, 1969).

Generally speaking, goal gradients demonstrate thatgoal-driven behavior is situation sensitive (in this case, itis sensitive to the changing distance to the goal). Moreconcretely, however, this principle may serve to distin-guish between priming of goals and semantic priming.Specifically, accessibility from semantic priming decreasesfairly rapidly after the onset of the prime. For example,



Srull and Wyer (1979) asked participants to judge anambiguously aggressive target for 5 min, 1 hour, or 1 dayafter the priming task. After 1 hour, when the number ofprimes was low (6 out of 30 scrambled sentences wererelated to hostility), there was little or no priming effect(also see Smith & Branscombe, 1987). When the numberof primes was high, slower but still noticeable decay wasobserved. Decay changes dramatically, however, if acces-sibility stems from the activation of goals because acces-sibility from goals seems to remain as long as the goal isactive and even increase as a function of goal proximity(see Frster, Liberman, & Higgins, 2005). Most promi-nently, Goschke and Kuhl (1993) demonstrated this prin-ciple by having participants rehearse a series of actionsand then informing them that they would either performthe actions (i.e., a goal) or observe another person per-forming them (i.e., a non-goal). Using a recognition test,they found a higher accessibility of the actions in the goalcondition than in the no-goal condition, even whenrehearsal of the actions was not possible in the interven-ing time. A similar persistence of the effect of goals wasdemonstrated by Bargh and Barndollar (1996), who acti-vated achievement concerns and found that delayincreased, rather than decreased, the effects of priming.

It may be argued that such increases in accessibilityare simply due to rehearsal. For example, a personsearching for his glasses in the morning might repeat-edly think, Where are my glasses, did I use them forreading? Are they near the TV? Who could have them?thereby rendering goal-related concepts highly accessi-ble. If so, then the lack of decay in accessibility aftergoal onset could simply reflect frequent (and recent)priming. To rule out this possibility, in another study,Goschke and Kuhl (1993) endowed participants with agoal and then systematically prevented them fromrehearsing it by asking them to carry out a visuospatialtask requiring spoken responses as participants wereapproaching the goal. Nevertheless, they foundincreased accessibility the closer participants were to thegoal. Thus, it seems that decay rates from goal activa-tion are different from those of semantic priming andthat this effect cannot be reduced to simple rehearsal.

We should caution that contrary to the classic notion ofgoal gradients, motivation does not necessarily increasecloser to the goal (for a recent discussion of this topic, seeLiberman & Frster, in press); for example, our motiva-tion to write an article does not necessarily increase as weget closer to actually writing it. A careful analysis of howdistance to the goal affects value and how it affectsexpectancy may allow us to predict whether motivationwould increase or decrease closer to the goal. For example,Construal Level Theory (Liberman, Trope, & Stephan,2007; Trope & Liberman, 2003) proposes a detailed theo-retical account of how psychological distance affects value.

It suggests that the positive value of some events (e.g.,having guests staying at your place for a week) is attachedto the abstract, high-level construals of these events andthus would increase with distance. In contrast, value that isattached to a concrete, low-level construal of an event (e.g.,having a candy) would be reduced with distance. Corres-pondingly, motivation and accessibility of goal-relatedconstructs would decrease closer to a goal in the formercase but increase in the latter case.

How do goal gradients help us to distinguishbetween primed goals and primed semantic constructs?We suggest that if postpriming accessibility changeswith distance to the goal in a way that is consistent withmotivation, then we may assume that in all likelihoodthe prime activated a goal rather than only producedsemantic priming.

Goal-Priming Effects Are Proportionalto the Product of Expectancy and Value

As alluded to earlier, classic theories of motivationstate that motivation increases with the expectancy ofreaching a goal and with the goals value. This basicpremise underlies expected utility theory in economicsand decision making (e.g., Edwards, 1951), Atkinsons(1964) theory of achievement motivation, the theory ofreasoned action (Fishbein & Ajzen, 1974), and Vrooms(1966) theory of motivation in organizational settings,among others. Of course, expectancy and value are notalways independent, especially in achievement situations(see Croizer, 1979; Filer, 1952; Irwin, 1953; Teevan,Burdick, & Stoddard, 1976). But even in these domains,the expectancyvalue model of motivation prevails, andit is assumed that both expectancy and value are requiredfor goal commitment and that they combine multiplica-tively (Lewin, Dembo, Festinger, & Sears, 1944; Tolman,1955; Vroom, 1964; for a review, see Feather & Newton,1982; Mitchell, 1982). For example, according to theexpectancyvalue model, an applicants motivation toapply for a job position would increase with the jobsattractiveness and the estimated chances of getting theposition. Furthermore, obviously, the expectancyvaluemodel specifies that expectancy and value combine in amultiplicative way to determine motivation. Thus, if thejob is not attractive (i.e., attractiveness approaches zero),then motivation to apply would be low no matter howlikely one is to get the job. In a similar way, if there arevirtually no chances of getting the job, motivation toapply would remain low even if the job is rather attrac-tive. Based on expectancyvalue models of motivation,we predict that both accessibility of goal-related con-structs and inhibition after goal fulfillment wouldincrease as a function of the interaction of the perceivedlikelihood and value of goal attainment.



Despite its central conceptual role in earlier work onmotivation, although the effect of goal value has receivedconsiderable attention in the social cognitive literatureregarding behavioral priming, the effect of expectancyand the multiplicative effect of expectancy and value hasbeen relatively underinvestigated. However, there havebeen some noteworthy exceptions. For example, Levy(1996) subliminally primed old and young participantswith either positive elderly stereotypic words (e.g., wis-dom) or negative elderly stereotypic words (e.g., senility).Memory performance was assessed both before and afterpriming. Assimilation effects occurred for elderly partici-pants (i.e., reduced performance after negative primesand enhanced performance after positive primes) but notfor young participants. One possible explanation of theseresults is that the primes contained expectations aboutones own performance as a member of the primed groupof the elderly. For an elderly participant, but not for ayoung participant, a prime induced a correspondingexpectancy (i.e., to perform poorly when primed withsenility and to perform well when primed with wisdom;cf. Steele & Aronson, 1995; for research showing thatstereotypic expectancies can decrease as well as enhanceperformance, see Brown & Josephs, 1999; Seibt &Frster, 2004; see reviews by Shapiro & Neuberg, inpress; Wheeler & Petty, 2001). Of importance, the lack ofa priming influence on younger participants may reflectthe fact that for this group the primes were irrelevantitis reasonable to suggest that 20-year-old participants sim-ply do not care about senile amnesia and thus the prim-ing did not have any effect. In general, these findings maybe interpreted as suggesting that expectancies can be acti-vated unconsciously and influence overt performance ifthey are relevant to the self.

Frster et al. (2005) also have recently reported resultsstrongly suggesting that behavioral priming shouldexhibit expectancy-based moderation. As discussed ear-lier, Frster et al. (2005) found that goal-related accessi-bility and postfulfillment inhibition were proportional tothe motivation to achieve a goal. Again, in their experi-ments, participants were instructed to search through apresented series of pictures for a target combination(glasses followed by scissors) and to report it to theexperimenter. As noted before, the results reflectedincreased accessibility of goal-related words before thecombination was found and inhibition after the combi-nation was found compared to a group of participantsthat only watched the pictures without the goal to findthe target combination. Moreover, in some studies(Frster et al., 2005, Experiments 4-6), the authorsmanipulated the expectancy of achieving the goal (bytelling participants that the target combination was pre-sent in 90% of the cases vs. only in 5% of the cases), thevalue of the goal (by telling participants that they would

receive 1.00 vs. only 0.05 for finding the combina-tion), or both expectancy and value. The experimentsalso included a no-goal, control group that merely sawthe pictures without the goal of finding the target combi-nation. As predicted, accessibility of goal-related wordsprior to fulfillment and inhibition after fulfillment werefound in the high-expectancy and high-value conditionsbut not in the low-expectancy and low-value conditions.The combination of high value and high expectancyenhanced the effects. Therefore, the effects of expectancyand value on goal-related accessibility and on postfulfill-ment inhibition were interactive (i.e., multiplicative) andsimilar to the effects of expectancy and value on motiva-tion (e.g., Atkinson, 1964; Fishbein & Ajzen, 1974;Vroom, 1966). To the extent that behavioral primingeffects are engendered by accessible goals, these resultssuggest that expectancies regarding goal attainmentshould play a critical role in determining the occurrenceand magnitude of such effects.

In regard to previously discussed motivational prin-ciples, semantic priming is not expected to be sensitiveto the interaction of expectancy and value. For example,the extent to which light semantically primes lampshould not vary as a function of the interaction betweenthe desirability of turning on the light and theexpectancy (e.g., feasibility) of doing so. Theories ofsemantic priming or procedural priming as well as theperceptionbehavior link are silent with respect to theinteraction of expectancy and value.

Let us now turn to three more principles that havebeen suggested as uniquely characteristic of cognitive rep-resentations of goals (Kruglanski et al., 2002), namely,goal shielding, involvement of self-control processes, andmulti- and equifinality.

Goal-Priming Effects Involve Inhibitionof Conflicting Goals

While pursuing a goal, one may be sidetracked byother goals. We have all experienced going to anotherroom to get a nail clipper (or wristwatch, glasses,husband, etc.) and then forgetting about the originalgoal and finding ourselves pursuing another goal (e.g.,once in the kitchen, we might start emptying the dish-washer). This effect, which was recently attributed todeficits in working memory (Baddeley, 2001; Kane &Engle, 2003), can very generally be interpreted as a lossof concentration or loss of ones focus on the originalgoal. Alternative goals, which may be triggered by cuesboth from the outside (e.g., the kitchen one went tolooking for the nail clipper) and from the inside (e.g.,remembering an unfinished paper), can distract usfrom pursuing the original goal. Very often, however,we seem to pursue our goals to completion, despite the



distracting environment and our simultaneous engage-ment in many other pressing tasks. Obviously, in orderto enable successful completion of one goal, the self-regulatory system needs to put other simultaneouslyactive goals on hold. Could that system operate out ofawareness and automatically shift the priority of acti-vated goals? Goal Systems Theory (GST; Kruglanskiet al., 2002; Shah, Kruglanski, & Friedman, 2002) pro-vides a positive answer to that question.

According to GST, goals form a hierarchical networkwith superordinate goals associatively linked to bothsubordinate goals (i.e., means) as well as competing orcomplementary superordinate goals. These networksmay or may not correspond, either partly or fully, tosemantic networks. For example, the goal of buyinggroceries may be subordinate to a higher level goal ofpreparing a party and superordinate to the means ofdriving a car to the supermarket, but this would by nomeans suggest that the semantic constructs of gro-ceries, party, and car need be analogously linkedwith one another, if at all. GST generates the predictionthat means should activate the goals they subserve andthat goals should activate the means that subserve them.Indeed, Kruglanski et al. (2002) have shown, using sub-liminal priming, that lexical decision times for goal-related means were faster after priming goals and lexicaldecisions for goals were faster after means were primed.Moreover, increased accessibility of the goal due topriming of means was associated with increased task per-sistence and better performance (Shah & Kruglanski,2003), reflecting the existence of a direct relationshipbetween success in goal pursuit and the strength of theconnection between means and goals.

Beyond these basic tenets, GST also predicts the exis-tence of inhibitory links between competing goals, aphenomenon labeled goal shielding. Goal shieldingoccurs when mental representations of alternativepotentially competing goals are inhibited by the mereactivation of a focal goal. To demonstrate the existenceof this phenomenon, Shah, Friedman, and Kruglanski(2002) subliminally primed participants with wordsrepresenting goals that they intended to pursue (e.g.,studying, reading, running, skiing) in a lexical decisionparadigm. Specifically, following suboptimal exposureto either these goal primes or control primes (e.g., rep-resenting goals they did not intend to pursue), partici-pants were asked to make speeded responses to targetsthat represented either competing goals or controlgoals. As predicted, it was found that when goals wereprimed, reaction times to competing goals were slowed,relative to reaction times to control goals, suggestingthat the mere activation of a goal automatically sup-presses the accessibility of goals that stand to interfere

with its pursuit. Critically, this shielding effect was onlyobtained for primed goals to which participants werehighly committed and for which protection from inter-ference would therefore be most imperative. Moreover,when target stimuli represented goals that would notstand to compete with focal goal attainment, the shield-ing effect was eradicated, again reflecting its presumedfunctional utility in self-regulation.

Of most pertinence to the current discussion, as theypresently stand, models of semantic priming are inca-pable of accounting for inhibitory effects of this ilk. Ifonly semantic or associative priming were at play in theparadigm devised by Shah and his colleagues, then thesubliminally primed, goal-related constructs (e.g., playingfootball) should have, if anything, facilitated responsetimes to competing goal constructs (e.g., baseball) by dintof shared semantic category membership or because bothprime and target goal constructs were associatively linkedby the very way of their competitive relationship.Moreover, traditional spreading activation (as well asconnectionist) models would be hard-pressed to explainwhy goal commitment enhanced the inhibition effect. Itmight be argued that words representing committed goalswere more emotionally engrossing, thereby capturinggreater attention and slowing responses to target words.However, this account would have difficulty explainingwhy the shielding effect was reduced when targets repre-sented goals that were noncompetitive with the focalgoal. Moreover, given the subliminal priming proceduresmost prominently employed, it is unlikely that partici-pants were distracted to an extent sufficient to reliablyretard their responses. Shah, Friedman, and Kruglanski(2002) also soundly replicated the goal-shielding effectusing supraliminal primes, arguing against the possibilitythat participants were slowed in their responses to targetstimuli by their attempts to discern the identities of themasked goal primes.

Upon consideration, there are numerous instances ofinhibitory cognitive processing that may be seen asinvolving suppression of cognitive material that standsto interfere with goal-directed behavior. For instance,in the area of psycholinguistics, it has been found thathomonyms initially activate all available semanticmeanings (e.g., the word bank activates both the mean-ing of a financial institution and the meaning of ariverside). However, once a context is determined, thecontext-irrelevant meanings become inhibited (Swinney,Prather, & Love, 2000). Here, the inhibition thatultimately results in such procedures may be seen ascrucial to satisfying the goal of semantically disam-biguating the sentence at handto the extent thatalternative meanings were to remain accessible, confu-sion would linger.



Likewise, in the retrieval-induced, forgetting paradigm(Anderson & Bjork, 1994), during a study phase, partici-pants encode several categorized lists, each composed ofseveral exemplars in a categoryexemplar format (e.g.,fruitorange). After the study phase, participants engagein directed retrieval practice for half of the items from halfof the studied categories. In this phase, they completecategory-plus-exemplar stem cue tests (e.g., fruitor___).After a retention interval, a final and unexpectedcategory-cued recall test is administered in which partici-pants are cued with each category name and are asked toreproduce any exemplar of that category that they remem-ber from any point in the experiment. It is typically foundthat performance on practiced items improves relativeto the baseline condition (i.e., items from unpracticedcategories), whereas performance on the unpracticeditems from practiced categories falls below this baseline.Anderson and Bjork (1994) have explained these effectsby suggesting that during the retrieval practice phase, thepracticed items compete with the unpracticed items fromthe same category and thus inhibit them (see alsoAnderson & Neely, 1996; Anderson & Spellman, 1995).Again, this may be reconstrued, in motivational terms, assuggesting that material is cognitively suppressed thatstands to interfere with attaining the goal of retrieving thepracticed items.

However, it is also possible that phenomena of thissort do not actually entail activation of mentally repre-sented goals. Perceptual systems may simply be hard-wired to inhibit information that stands to underminestable identification of a given percept. So, for instance,upon encountering the homograph bank, when a partic-ular interpretation is more probable, the brain suppressesactivation of less probable alternatives, irrespective ofwhether any stimulus interpretation or word com-prehension goal has been activated or has ever beenrepresented. A kindred example would be perception ofambiguous figures such as the familiar Necker cubethe perceptual system simply uses available cues to com-pute the most likely configuration for the stimulus andsuppresses activation of information that would pro-mote a conflicting perceptual organization. This processmay not, and to our knowledge has never been pre-sumed to, entail activation of a mentally representedgoal to disambiguate the sensory image. Therefore,although it is certainly possible that goal activation maycontribute to a wide variety of inhibitory phenomenonfor which motivational accounts have not previouslybeen invoked, the existence of said contribution remainscontroversial and cries out for empirical scrutiny. Assuch, for the sake of parsimony, we propose that inhi-bition of goal-incompatible material can only be confi-dently ascribed to goal priming when the mere

activation of a goal construct representing a desired endstate to which an individual is highly committed auto-matically shields itself via suppression of the accessibil-ity of competing goal constructs.

Goal-Priming Effects Involve Self-Control

Implementation of many abstract, long-term goalsoften is impeded by the necessity of overcoming concrete,momentary temptations. For instance, individuals often-times are tempted to overeat when trying to maintain adiet, watch TV when trying to read a paper, smoke whenattempting to quit, or more generally take a smaller butsooner reward at the expense of a larger one later (Ainslie,1992; Fujita, Trope, Liberman, & Levin-Sagi, 2006;Loewenstein, 1996; Metcalfe & Mischel, 1999; Mischel,1974; Mischel, Shoda, & Peake, 1988; Rachlin, 1995,1996, 1997; Shoda, Mischel, & Peake, 1990; Trope &Fishbach, 2000). Such self-control dilemmas may beviewed as conflicts between the behavioral implications ofa low-level construal (specific, local view) of a situationand the behavioral implications of a high-level (i.e., gen-eral, abstract) construal of a situation (Fujita et al., 2006).

Given this definition, goal shielding, as discussedabove, may be understood as a special case of self-control inasmuch as the nonfocal, background goal maybe thought of as following from a lower level construalof the situation than that following from the focal, pri-mary goal (Trope & Liberman, 2000). In that sense, thebackground goals or action tendencies may be concep-tualized as temptations that one is trying to overcome topursue the more important, focal goal. Nevertheless, webelieve that self-control deserves some special attentionin view of its vital importance in peoples lives. Unlikethe more general case of distracting goals, which may beimposed by situations and therefore unforeseeable (e.g.,one does not know that the goal to get clippers from thekitchen would be subverted by a competing goal to havea cup of coffee), classic self-control situations tend toinvolve a specific distracting goal that may be known inadvance. For example, people know that their attemptsto diet could be sabotaged by a dessert in a restaurantor that TV interferes with studying or that fear and lazi-ness might prevent them from getting an HIV test.

Knowing in advance the specific impediment to goalattainment allows people to preemptively employ self-control strategies that aim at counteracting the potentialtemptations. A person might try to eliminate the tempta-tions, to detract from their attractiveness, or to bolster theattractiveness of the higher level goal (Fishbach &Trope, 2001). For example, research shows that peoplemay impose on themselves penalties for failing to actaccording to their long-term outcomes (Ainslie, 1975;



Becker, 1960). These self-imposed penalties may thenserve as deterrents against failure to act according to thehigher level goal. People also may opt to change futurechoice situations by making rewards contingent on act-ing according to ones long-term interests. Thus, insteadof receiving a reward unconditionally, people may pre-fer to receive it only if they act according to their long-term interests. For example, one may prefer to receive abonus for actually completing an HIV test than formerely agreeing to do it. In making penalties andrewards contingent on performing an activity, peopleprecommit themselves to the higher level goal(Brickman, 1987), often making their decision to actaccording to their higher level goals irreversible (Ainslie,1975; Green & Rachlin, 1996; Rachlin & Green, 1972;Schelling, 1978; Strotz, 1956; Thaler, 1994; Thaler &Shefrin, 1981). Other self-control strategies involve bol-stering the value of the higher level goals (Mischel,1984), for example, by linking their attainment toimportant self-standards. Failure to choose according tolong-term outcomes is then construed as a violation ofones central values and a threat to ones sense of self-worth and self-determination.

It seems that most of these strategies involve con-scious processes. Moreover, research on ego-depletionhas demonstrated that self-control takes up extensivecognitive resources and, in that sense, cannot be auto-matic. Indeed, self-control decreases as a function offatigue, concurrent activities, and negative affect(Baumeister, Bratslavsky, Muraven, & Tice, 1998;Muraven, Tice, & Baumeister, 1998; see also Fishbach& Trope, 2001). It is still possible, however, that someof these self-control strategies would operate in anunconscious manner. Could self-control processes comeinto play in the activation and inhibition of goal repre-sentations and thereby in automatic behavioral prim-ing? Could self-control strategies themselves beactivated outside of awareness?

Recent research by Fishbach, Friedman, and Kruglanski(2003) provides a positive answer to both of these ques-tions. Based on GST, Fishbach et al. (2003) argued thatwhen situationally enticed to pursue short-term goalsrepresenting temptations (e.g., use of illegal drugs,engagement in premarital sex), people activate superor-dinate goals (e.g., academic achievement, adhering toreligious mores) to aid self-control. To test their hypoth-esis, the authors subliminally primed temptation-relatedwords (e.g., drugs, sex) and found that participantswere significantly faster to make lexical decisions aboutwords related to the higher order goals with which thesetemptations stood to interfere (e.g., grades, bible) thanwhen primed with control words associated with temp-tations that were irrelevant to these goals. Attesting tothe automaticity (in this case, efficiency; see Bargh,

1994) of the effect, temptationgoal activation also wasshown to occur under high cogni

7 Principles of goal-based priming

Documents

Transcript of 7 Principles of goal-based priming