This article was downloaded by: [University of La Laguna Vicerrectorado]On: 07 October 2011, At: 02:04Publisher: Psychology PressInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: MortimerHouse, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Cognitive PsychologyPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/pecp21

Counterfactual sentences activate embodiedmeaning: An action–sentence compatibility effectstudyManuel de Vega a & Mabel Urrutia aa University of La Laguna, Canary Islands, Spain

Available online: 07 Oct 2011

To cite this article: Manuel de Vega & Mabel Urrutia (2011): Counterfactual sentences activate embodied meaning:An action–sentence compatibility effect study, Journal of Cognitive Psychology, DOI:10.1080/20445911.2011.590471

To link to this article: http://dx.doi.org/10.1080/20445911.2011.590471

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Any substantial orsystematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distributionin any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that thecontents will be complete or accurate or up to date. The accuracy of any instructions, formulae, anddrug doses should be independently verified with primary sources. The publisher shall not be liable forany loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever causedarising directly or indirectly in connection with or arising out of the use of this material.

Counterfactual sentences activate embodied meaning:An action!sentence compatibility effect study

Manuel de Vega1 and Mabel Urrutia1

1University of La Laguna, Canary Islands, Spain

Recent evidence suggests that understanding factual action-related sentences involves embodiedsimulations. But, what happens with counterfactual sentences that describe hypothetical events in thepast? This study demonstrates that even in this case embodied simulations of actions take place.Participants listened to factual or counterfactual sentences describing a transfer away from or towardsthem. After the transfer verb (e.g., gave) was received, either a motion cue (Exp. 1) or a static cue (Exp.2) prompted participants to move their finger towards or away from them to press a button. Fingermotion was initially interfered with in cases involving a concurrent matching sentence (e.g., transferaway-motion away), suggesting that counterfactual meaning involves a motor simulation or ‘‘resonance’’.The temporal course of this resonance differs slightly between factual and counterfactual sentences.

Keywords: Action!sentence compatibility effect; Comprehension; Counterfactuals; Embodiment.

People not only communicate factual informa-tion, they are also capable of producing andunderstanding counterfactual sentences like ‘‘IfMary had bought that lottery ticket, she wouldhave won a prize worth millions’’. These sen-tences involve a sort of semantic dead end: Theydescribe a sequence of events in the past or in aparallel world that never occurred. Their psycho-logical functions in reasoning, causal attribution,or emotion activation have been described else-where (e.g., Byrne, 2005; Roese, 2005).

From a semantic perspective, counterfactualsseem to involve a dual meaning. Thus, under-standing the counterfactual example above in-volves the representation of two possible states ofthe world: A ‘‘real’’ situation in which Mary didnot buy the ticket or win the lottery (not-p & not-q), and a situation in which Mary bought theticket and won the lottery (p & q). The notion of

dual meaning in counterfactuals has been empiri-cally investigated by psychologists of reasoning(e.g., Byrne, 2005; Kahneman & Tversky, 1982;Thompson & Byrne, 2002). Most of these studiesused offline methods in which participants firstread a factual conditional (e.g., ‘‘If Mike went toCalgary, then Barbara went to Edmonton’’) or acounterfactual conditional (e.g., ‘‘If Mike hadgone to Calgary, then Barbara would have goneto Edmonton’’), and were then asked to choosefrom among several sentences those whichwere more consistent with or were implied bythe conditional sentence’s meaning. The resultstypically confirm that participants tend to exclu-sively choose a factual interpretation (p & q) forfactual conditionals, but they choose both factual(p & q) and negation (not-p & not-q) for counter-factual conditionals. Moreover, moving beyondthis choice-task methodology, Santamarıa, Espino,

Correspondence should be addressed to Manuel de Vega, Facultad de Psicologıa, Campus de Guajara, 38205 La Laguna, CanaryIslands, Spain. E-mail: mdevega@ull.es

This research was supported by the Spanish Ministry of Science and Innovation (Grant SEJ2007-66916), the Canary Agency forResearch, Innovation and Information Society (Project NEUROCOG), and the European Regional Development Fund to MDV.We are grateful to Victor Romeo and Yurena Morera for their help in programming the experiments in E-Prime, and analysing thedata, respectively.

JOURNAL OF COGNITIVE PSYCHOLOGY, 2011, iFirst, 1!12

# 2011 Psychology Press, an imprint of the Taylor & Francis Group, an Informa businesshttp://www.psypress.com/ecp http://dx.doi.org/10.1080/20445911.2011.590471

Dow

nloa

ded

by [U

nive

rsity

of L

a La

guna

Vic

erre

ctor

ado]

at 0

2:04

07

Oct

ober

201

1

and Byrne (2005) gave participants counterfactualconditionals such as ‘‘If there had been roses thenthere would have been lilies’’ as primes to under-stand target sentences like ‘‘There were roses andthere were lilies’’ or ‘‘There were no roses andthere were no lilies’’. The same procedure wasused with factual conditional primes such as ‘‘Ifthere are roses, there are lilies’’. When targetreading times were collected, factual sentencesprimed only the affirmative target ‘‘There wereroses and there were lilies’’, whereas counter-factual sentences primed both the affirmative andthe negative targets, suggesting that readers hadgenerated a double representation.

Only recently have researchers in languagecomprehension paid attention to counterfactuals,exploring the temporal course of their dual mean-ing activation (de Vega, Urrutia, & Riffo, 2007;Ferguson & Sanford, 2008; Stewart, Haigh, &Kidd, 2009; Urrutia, de Vega, & Bastiaansen,2011). Thus, de Vega et al. (2007) presentedreaders with short stories describing a factualstate in the first paragraph, followed by additionalinformation that was either factual or counter-factual. For instance:

1. Initial situation: John was still in the officesitting in front of the computer. He started totype a report that his boss had asked him for.

2. Factual continuation: Because he had en-ough time, he went to the cafe to drink abeer.

3. Counterfactual continuation: If he had hadenough time, he would have gone to the cafeto drink a beer.

After reading the story, participants were asked toverify a test probe that either belonged to thebeginning of the story (type), or to the subordi-nate clause shared by the factual and counter-factual version of the story (drink). Immediatelyafter reading the critical sentence (2 or 2?),readers verified the first test probe more slowlyin the factual stories than in the counterfactualstories, whereas they verified the second probeequally rapidly in both types of stories. Theinterpretation of these results is quite straightfor-ward. In the factual stories, the situation model isupdated and the attention focus moves on to thenew factual content, whereas in the counter-factual version, updating is cancelled and theattention focus goes back to the initial factualinformation. However, the new information con-veyed by the counterfactual sentence is also

accessible for a while, clearly indicating that itsalternative meaning (p & q) is also represented.

In the same vein, Ferguson and Sanford (2008)used the eye-movement technique to demonstratethat, after reading a counterfactual sentenceconflicting with real-world knowledge, partici-pants cannot avoid representing the real-worldsituation. In their study, Ferguson and Sanfordused conditional sentences either with a counter-factual (‘‘If cats were vegetarians they would becheaper for owners to look after’’), or a factualmeaning (‘‘If cats are hungry they usually pestertheir owners until they get fed’’). Immediatelyafter reading one of these conditional sentences,participants received a continuation sentenceconsistent with their real-world knowledge (‘‘Fa-milies could feed their cat a bowl of fish and itwould gobble it down happily’’), or consistentwith the counterfactual world (‘‘Families couldfeed their cat a bowl of carrots and it wouldgobble it down happily’’). In spite of the fact thatthe counterfactual sentence establishes a hy-pothetical scenario in which cats are vegetarians,participants do not immediately suppress theirreal-world knowledge. Thus, the first-pass readingtime*a measure of initial cognitive load*forreal-world continuation showed a similarly slowreading after the counterfactual context (whichprovided a consistent setting) as after the real-world context. In other words, real-world viola-tions were not neutralised by a counterfactualcontext.

In sum, theoretical proposals, offline data, andonline data converge to show that readers ofcounterfactual sentences activate, at least mo-mentarily, a realistic scenario in which the coun-terfactual events did not occur, and an alternativescenario that considers the counterfactual events‘‘as if’’ they had happened. This paper goes onestep further, exploring how the meanings ofcounterfactuals are represented. Specifically,does the ‘‘as if’’ meaning involve embodiedsimulations of actions, or merely a disembodiedsymbolic representation? Briefly, the symbolistapproach assumes that the meaning of wordsconsists of mental symbols, which are abstract,arbitrary, and amodal. For instance, the word‘‘cat’’ would activate an abstract symbol in yourmind, which is connected to other equally abstractsymbols, none of which have any similarity toyour visual, auditory, tactile, or motor experienceof cats. By contrast, the embodiment approach tomeaning considers that language is grounded inthe world (see de Vega, Glenberg, & Graesser,

2 DE VEGA AND URRUTIA

Dow

nloa

ded

by [U

nive

rsity

of L

a La

guna

Vic

erre

ctor

ado]

at 0

2:04

07

Oct

ober

201

1

2008, for a debate between the two approaches).According to the embodiment view, meaningconsists of mental simulations of the objects,events and situations to which words refer (e.g.,Barsalou, Santos, Simmons, & Wilson, 2008;Glenberg & Kaschak, 2002; Pulvermuller, Hauk,Nikulin, & Ilmoniemi, 2005; Taylor & Zwaan,2008, 2009). For instance, the word ‘‘grasping’’could briefly activate in your mind a motorsimulation of the grasping action. A plausibleneuronal mechanism for embodied meaning ismotor resonance; that is, the partial reactivationof the motor cortex elicited by action-relatedwords (Taylor & Zwaan, 2008, 2009; Zwaan &Taylor, 2006). Mixed approaches also exist, whichassume that both shallow linguistic representa-tions and/or deeper embodied representationscould be activated by action-related language(e.g., Barsalou et al., 2008; Taylor & Zwaan,2009). These mixed approaches are relevanthere. Thus, it might be possible that factualsentences elicit embodied representations as de-monstrated elsewhere, whereas counterfactuals,given their hypothetical character, would notactivate any motor event.

Recently, an intriguing phenomenon has pro-vided support for theories of embodied cognition:the interaction between language comprehensionand action. While understanding sentences de-scribing motor events, participants were askedto perform a motor task designed to match ormismatch the meaning of the sentences. In somecases, a facilitating Action!sentence Compatibil-ity Effect (ACE) was reported; that is, themeaning!action matching conditions producedfaster responses than the mismatching conditions(Borreggine & Kaschak, 2006; Glenberg &Kaschak, 2002; Kaschak & Borreggine, 2008;Santana & de Vega, 2011; Zwaan & Taylor,2006). Thus, Glenberg and Kaschak (2002) askedpeople to judge the sense of sentences describinga transfer motion towards or away from them(e.g., ‘‘Andy delivered the pizza to you’’ or ‘‘Youdelivered the pizza to Andy’’) or of nonsensesentences. For some participants, the ‘‘yes’’ re-sponse involved a hand motion towards theirbody, and the ‘‘no’’ response involved a motionaway from their body; for other participants, itwas the other way around. The judgement timefor sensible sentences was faster for the matchingconditions (e.g., sentences describing a transfertowards the person, responding ‘‘yes’’ towards theperson) than for the mismatching conditions.Other studies have found interfering rather than

facilitating ACE for the matching conditions(Boulenger et al., 2006; Buccino et al., 2005; deVega, Moreno, & Castillo, 2011). For instance,Buccino et al. (2005) gave participants hand-action sentences (e.g., ‘‘He took the cup’’), foot-action sentences (e.g., ‘‘He kicked the ball’’), orabstract sentences, and they were to give aresponse only if the sentence described an action,and refrain from responding otherwise. Handresponses were faster for foot-action than hand-action sentences, and foot responses were fasterfor hand-action than foot-action sentences.

Several factors could contribute to the discre-pancy in the ACE results. The experiments differin the complexity of the linguistic materials(single words, single-clause sentences, or double-clause sentences), the semantic task (sensibilityjudgements, categorical judgements, lexical deci-sion), the response paradigm (choice, go/no-go,self-paced reading), and the temporal overlapbetween the linguistic stimulus and the motorresponse. Concerning the latter issue, de Vegaet al. (2011) manipulated the temporal overlapbetween the comprehension process and theproduction of the motor response. The motoraction accompanying the transfer sentences was asimple directional response to an apparent mo-tion cue. This psychophysical response does notrequire the burden of a semantic judgement; andthe apparent motion cue could be placed atspecific times in the sentence without the needto process the whole sentence. De Vega et al. usedtransfer sentences, and collected the motor re-sponses at several intervals following the transferverb, rather than at the end of the sentence. Forshort verb!action intervals (100!200 ms) theyfound interference in the matching conditions,whereas when the motion cue was given 350 msafter the verb onset, there was facilitation in thematching conditions, indicating priming betweenmeaning and action. The early interference couldbe consequence of the competition for motorcortex resources between language-based motor‘‘resonance’’ and action performance. In otherwords, understanding action-related sentenceswould partially activate the same motor regionsas performing actions. This claim is consistentwith neurological evidence that demonstrates thatneural resonance of sensory-motor areas reachmaximum peak 200 ms after action verbs onset(Pulvermuller et al., 2005).

This research uses the aforementionedparadigm to analyse whether counterfactual sen-tences describing transfer events produce similar

EMBODIMENT IN COUNTERFACTUALS 3

Dow

nloa

ded

by [U

nive

rsity

of L

a La

guna

Vic

erre

ctor

ado]

at 0

2:04

07

Oct

ober

201

1

interfering or facilitating ACE to their factualcounterparts. The rationale is that the ‘‘as if’’meaning of counterfactuals is similar to a factualmeaning. Therefore, understanding action-relatedcounterfactuals may trigger a simulation of theactions or a motor resonance resulting in obser-vable ACE. Another possibility is that, unlikefactual sentences, counterfactuals do not activateany embodied representation and, consequently,ACE would not be observed. This might occur ifthe two alternative meanings of the counter-factual (‘‘not-p & not-q’’ and ‘‘p & q’’) arerepresented in a relatively shallow (e.g., proposi-tional) format, without the need to activate anymotor representation (Barsalou et al., 2008;Taylor & Zwaan, 2009). Experiment 1 manipu-lated the interval between the transfer verb onsetand the motion cue (100 or 200 ms). With theseshort intervals we expect to replicate both withfactual and counterfactual sentences the interfer-ence pattern in the matching conditions pre-viously reported with factual sentences, andwhich is considered a marker of automaticresonance processes (de Vega et al., 2011). Inaddition, Experiment 2 was run as a control taskin which a static cue (a colour change) substitutedthe visual motion cue. If the ACE were observedeven without the presence of a motion cue, thenthis could only be attributed to the participant’smotor response alone and not to the perceptualprocessing of the visual motion cue.

EXPERIMENT 1

Method

Participants. Ninety-four students of Psychol-ogy (including 74 women) from the University ofLa Laguna received credits for their voluntaryparticipation in the experiment. All were nativespeakers of Spanish.

Materials. Ninety-six two-clause sentences werecreated, each of which was preceded by a one-word title. The first clause introduced the agent ofthe transfer event, and the second clause de-scribed the transfer action away from or towardsthe agent. A factual and a counterfactual versionof each sentence were written (see Table 1),sharing the same critical word in which the ACEwas tested (the past participle ‘‘prestado’’ in oneexample of Table 1). Factuals and counterfactuals

differ in syntactic structure, mood, and tense(both in Spanish and English). A stricter match-ing between structures would have been accom-plished by using a conditional form in factuals(e.g., ‘‘If my brother was generous he lent me theHarry Potter novel without hesitation’’), ratherthan the causal construction employed here.However, the Spanish conditional (e.g., ‘‘Si mihermano fuera generoso me prestarıa la novela deHarry Potter’’) expresses a wish or intentionrather than a factual event, and the critical worddiffers in the counterfactual and factual (e.g.,‘‘prestado’’ vs. ‘‘prestarıa’’, respectively). Noticethat the agent in the sentences was described inthe first person (I, me), although the details of thesituations were fictitious (e.g., a specific partici-pant might not have a brother, or might not beinterested in reading Harry Potter). In spite ofthis, we expect that the participants take theagent’s perspective to understand the transferevents, as previous ACE’s experiments haveshown (Glenberg & Kaschak, 2002). Given therelative freedom of Spanish word order, it waspossible to construct the sentences such that thetransferred object and the direction of the trans-fer were mentioned before the transfer verb.Eight training sentences and an additional set of48 filler sentences (24 factual and 24 counter-factual) using nontransfer verbs (e.g., knowing,remembering, etc.) were also included.

Procedure and design. A computer keyboardwas modified to meet the needs of the experi-ment. The function keys on the right part of thekeyboard were removed, except for K5, K2, andK8, which were assigned as the resting-key, thetowards-key and the away-key, respectively, asillustrated in Figure 1A. The structure of eachtrial, illustrated in Figure 1B, was as follows: Thetrial started with the word ‘‘DIANA’’ (TARGET)presented auditorily as a prompt to press theresting key. Then, two convergent lines resem-bling a road in perspective appeared on thescreen, and immediately afterwards, the sentencewas presented auditorily, except for the transferverb. Immediately after finishing the last spokenword, the transfer verb was presented visually inthe middle of the screen within the ‘‘road’’,remaining still on the screen for either 100 or200 ms, and then apparently ‘‘jumped’’ away fromor towards the participant according to theanimation sequence shown in Figure 1B. Thisapparent motion was a cue for the participants torelease the resting-key and move their hand in the

4 DE VEGA AND URRUTIA

Dow

nloa

ded

by [U

nive

rsity

of L

a La

guna

Vic

erre

ctor

ado]

at 0

2:04

07

Oct

ober

201

1

same direction to press the corresponding away-or towards-key. Five hundred milliseconds afterthe keypress response, the end of sentence waspresented auditorily. At the offset of the auditorysentence, participants were given a visual wordprobe and asked to decide whether or not it hadappeared in the sentence. They used their lefthand middle and index fingers to press the yes/noresponse keys (numbers 1 and 2 on the keyboard),respectively. Finally, in 30% of trials, there was amemory question related to the sentence content.Participants were instructed to understand eachsentence and perform two tasks speedily: re-

sponding to the test probe (and, eventually, tothe memory question), and moving their finger topress the away- or towards-key when they wereprompted by the visual animation. Then theyreceived eight training sentences, followed by the96 experimental sentences (12 in each of the eightconditions resulting from crossing sentence type,transfer direction, and motion direction), whichwere randomly mixed with the 48 filler sentences.

A repeated-measures factorial design was per-formed: 2 Sentence type (factual/counterfact-ual)"2 Transfer direction (away/towards)"2Motion direction (away/towards)"2 Motion

TABLE 1Experiments 1 and 2: Examples of materials in Spanish and the approximate translation to English (the word order was normalised

to make the English version grammatical)

Towards me Factual Novela. Como mi hermano es generoso la novela de Harry Potter me la ha prestado sin vacilar(Novel. Given that my brother is generous he has lent me the Harry Potter novel withouthesitation)

Counterfactual Novela. Si mi hermano hubiese sido generoso, la novela de Harry Potter me la habrıa prestado sinvacilar(Novel. If my brother had been generous he would have lent me the Harry Potter novel withouthesitation)

Probe Hermano [Brother]Question ¿Me ha prestado mi hermano El senor de los anillos? [Has my brother lent me The Lord of the

Rings?]

Away fromme

Factual Baloncesto. Como en ese momento yo estaba lejos de la canasta, el balon se lo he pasado a otrojugadorBasketball. Given that I was far away from the basket at that moment I’ve passed the ball toanother player

Counterfactual Baloncesto. Si en ese momento yo hubiera estado lejos de la canasta, el balon se lo habrıa pasado aotro jugadorBasketball. If I had been far away from the basket in that moment I would have passed the ball toanother player

Probe gol [goal]

Figure 1. (A) Response keys display for the finger motion task, and (B) temporal sequence of a trial (approximate translation:‘‘Given that my brother is generous he has lent me the Harry Potter novel without hesitation’’/NOVEL). All words were presentedauditorily except the transfer verb (prestado/lent) that was presented visually, and ‘‘moved’’ away or towards oneself.

EMBODIMENT IN COUNTERFACTUALS 5

Dow

nloa

ded

by [U

nive

rsity

of L

a La

guna

Vic

erre

ctor

ado]

at 0

2:04

07

Oct

ober

201

1

delay (100/200 ms). All factors were manipulatedwithin-participants except the delay: Half of theparticipants were randomly assigned to eachmotion delay. The motion direction applied tothe transfer verbs was counterbalanced amongparticipants. The dependent variables were: (1)the releasing time of the resting-key, (2) the timeto press the corresponding direction key, (3) thelatency of response for the word probe, (4) thedirection response errors, and (5) the probeerrors.

Results

Five participants were discarded from the databecause of their poor performance in the memoryquestions (more than 30% errors). Wrong re-sponses either in the motor task (less than 1%), orthe probe task (7.5%) were also discarded fromthe reaction time analyses. We will report only onthe results of the key-pressing time and the wordprobe time, because these were the only depen-dent measures sensitive to the experimentalmanipulations.

Key-pressing time

Towards-responses were faster (M#228 ms)than away-responses (M#267 ms), F(1, 87)#23,MSE#18830.571, pB.0001. The most importanteffect, however, was a four-way interaction: Sent-ence type"Transfer direction"Motion direc-tion"Motion delay, F(1, 87)#7.64, MSE#3965.158, pB.007. To interpret this complexinteraction, we ran separate analyses for factualand counterfactual sentences:

Factual sentences. A three-way Transfer"Motion"Delay interaction was found, F(1,87)#4.13, MSE#23655.33, pB.045, suggestingthat the ACE was qualified by the verb!actiontemporal overlapping. Specifically, for the 100 msdelay, there was no ACE, Transfer"Motion, F (1,44)B1, whereas for the 200 ms delay, the ACEwas significant, Transfer"Motion, F(1, 43)#4.31,MSE#11138.229, pB.04. As Table 2 shows, atthis delay matching conditions produced slowerresponses than mismatching conditions, indicatingthe presence of interfering ACE.

Counterfactual sentences. Here too, there was athree-way Transfer"Motion"Delay interaction,F(1, 87)#3.87, MSE#8517, pB.05. In this case,however, the ACE was only significant for the 100ms delay, Transfer"Motion, F(1, 44)#5.86,MSE#494.363, pB.02; that is, the matchingconditions produced slower responses than themismatching conditions, again indicating the pre-sence of an interfering ACE (Table 2). Thepairwise contrasts showed longer latencies for‘‘away transfer-away motion’’ than for ‘‘towardstransfer-away motion’’, t(44)#2.94, pB.005. Bycontrast, the ACE was not significant for the 200ms delay, Transfer"Motion, F(1, 43)#1.48,MSE#3950.950, p#.23.

Probe reaction times

The only significant effect here for the wholeset of data was a main effect of sentence type, F(1,87)#4.16, MSE#7655.164, pB.04. Specifically,probe identification was faster for factual(M#1019 ms, SD#208) than for counterfactualsentences (M#1032 ms, SD#212).

TABLE 2Experiment 1: Mean key-pressing times (in milliseconds) and standard deviations (in parentheses) as a function of type of

sentence, transfer direction, motion direction, and delay

Delay

100 200

Sentence type Transfer Away motion Towards motion Away motion Towards motion

Factual Away 235 (87) 210 (87) 314 (215) 231 (118)Towards 237 (88) 212 (90) 276 (137) 259 (133)ACE 2 $2 !20 !28

Counterfactual Away 247 (103) 207 (86) 287 (133) 250 (144)Towards 242 (102) 219 (87) 299 (173) 239 (128)ACE !5 !12 12 11

The ACE was the subtraction between pairs of matching!mismatching conditions.

6 DE VEGA AND URRUTIA

Dow

nloa

ded

by [U

nive

rsity

of L

a La

guna

Vic

erre

ctor

ado]

at 0

2:04

07

Oct

ober

201

1

Discussion

Some aspects of these results are remarkable.First, the ACE was obtained in both factual andcounterfactual sentences, demonstrating that sen-sory-motor processes are activated also during thecomprehension of counterfactual meaning. Sec-ond, the ACE consisted of slower motor re-sponses in the matching (interference) than inthe mismatching conditions. This interfering ACEdiffers from other ACEs reported in the literaturethat involve matching advantage (Borreggine &Kaschak, 2006; Glenberg & Kaschak, 2002;Kaschak & Borreggine, 2008; Zwaan & Taylor,2006), although it is coincidental with otherstudies that obtained similar mismatching advan-tage when the action word and the motorresponse temporally overlap (Boulenger et al.2006; Buccino et al. 2005; de Vega et al., 2011).Differences in the sentence-action delay betweenthis and other studies could be responsible offinding either facilitation or interference in thematching conditions. Thus, whereas in the typicalACE studies the motor responses were testedafter understanding the whole transfer sentencein this study the motor response was placed atseveral intervals following the transfer verb andbefore finishing the sentence. In this way, themotor action could be placed at the moment inwhich most of the transfer information (agent,object, and direction of transfer) was integrated inthe Spanish sentences. Third, the ACE was short-lived, and its time course differed in the two typesof sentences. In counterfactuals, the ACE wasestablished very early on (at about 100 ms), anddissipated rapidly (in 200 ms). By contrast, infactual sentences, the ACE emerged 200 ms afterthe verb onset. This clearly indicates that thecounterfactual dual meaning modulates the acti-vation of embodied representations, speeding upmotor resonance, as will be explained in theGeneral Discussion.

Concerning the latency of responses to theprobe, the faster responses in the factual ascompared to the counterfactual sentences couldbe related to the information suppression occur-ring for counterfactuals at the time of the probetest (de Vega et al., 2007), or simply be due to thefact that counterfactuals consume more cognitiveresources than factuals because their complexgrammar or their dual meaning. In any case, thisis not an especially relevant result for the current

hypothesis, because no ACE was observed on thismeasure.

This experiment clearly demonstrated that thecomprehension of action-related sentences ineither a factual or a counterfactual format inter-feres with the performance of a matching motorresponse. The use of a motion cue in thisexperiment was a very efficient procedure toprompt participants’ motor responses at preestab-lished intervals following the transfer verb onset.However, this technique involves a potentialconfusion, because the apparent motion itselfcould interact with the semantics of the transferverbs. Thus, some papers have reported thatunderstanding motion-related sentences is modu-lated by the watching of a visual animation thateither matches or mismatches the sentence mean-ing (Zwaan, Madden, Yaxley, & Aveyard, 2004).In this experiment, both the motor and the visualcompatibility could influence the observed ef-fects. In the next experiment, the visual animationwas substituted by the use of a static cue toprompt participants’ finger motion.

EXPERIMENT 2

This experiment was identical to the previous one,except that in each trial, the cue prompting theparticipants’ finger motion away from or towardsthem was a colour change in the critical transferverb. In addition, only the 100 ms delay wasemployed in this experiment with the purpose ofreplicating the ACE obtained for counterfactualsin the previous experiment.

Method

Participants. Seventy-eight students of Psychol-ogy (including 53 women) from the University ofLa Laguna received credits for their voluntaryparticipation in the experiment. All were nativespeakers of Spanish.

Materials, procedure, and design. The samestimuli as in Experiment 1 were employed. Likein Experiment 1, the keyboard with a resting-, anaway-, and a towards-key was employed, exceptthat the colours red and blue, rather than thearrows, were attached to the towards- and away-key. The sequence of events in each trial wasidentical to the previous one, except that the

EMBODIMENT IN COUNTERFACTUALS 7

Dow

nloa

ded

by [U

nive

rsity

of L

a La

guna

Vic

erre

ctor

ado]

at 0

2:04

07

Oct

ober

201

1

transfer verb remained still on the screen for 100ms, and then changed colour to either red or blue.For half of the participants, the colour red was theprompt to release the resting-key and move theirfinger away from them, and the colour blueindicated that they should move their fingertowards them and press the corresponding key.For the remaining participants, the colour!motionassignment was reversed. At the offset of theauditory sentence, participants were given a visualword probe and asked to decide whether or not ithad appeared in the sentence. Finally, in 30% oftrials, there was a memory question related to thesentence content.

Given the arbitrary character of the colour!motion assignment, before starting the experimen-tal session participants received 24 training trials inwhich they were given a string of symbols (@) inthe centre of the screen which, after a variableamount of time, turned red or blue as a prompt forthe participants tomove their finger in the assigneddirection and press the corresponding key.

A repeated-measures factorial design was per-formed: 2 Sentence type (factual/counterfactual)"2 Transfer direction (away/towards)"2 Motiondirection (away/towards). The same dependentvariables as in Experiment 1 were collected: (1)the releasing time of the resting-key, (2) the timeto press the corresponding direction key, (3) thelatency for the word probe, (4) the directionresponse errors, and (5) the probe errors.

Results

Nine participants were discarded from the databecause of their poor performance in the memory

questions (more than 30% errors). The same datatrimming procedure as in Experiment 1 was used.Wrong responses either in the motor task (lessthan 1%), or the probe task (7%) were alsodiscarded from the reaction time analyses. We willreport only on those measures that producedsignificant results: the key-releasing time andthe latency for the word probe. Unlike in theExperiment 1, the key-pressing time did notproduce significant results.

Key-releasing time

The Transfer"Motion interaction was signifi-cant, F(1, 68)#4.586, MSE#988.953, pB.05. Inaddition, a three-way interaction, Sentence"Transfer"Motion, F(1, 68)#7.118, MSE#869.698, pB.01, was also significant. The firstinteraction is interpreted as a general interferenceeffect, because the matching conditions produceda slower response than the mismatching condi-tions. To interpret the three-way interaction, weran separate analyses for factual and counter-factual sentences. In factual sentences, the Trans-fer"Motion interaction was absent, F(1, 68)B1.However, in counterfactual sentences, a signifi-cant Transfer"Motion interaction was found,showing again interference in matching condi-tions, F(1, 68)#10.897, MSE#978.463, pB.002(see Table 3). Post hoc planned comparisonsshowed longer latencies for ‘‘away transfer-awaymotion’’ than for ‘‘towards transfer-away mo-tion’’, t(48)#2.74, pB.042, replicating the pat-tern obtained in Experiment 1 for pressing-time.Moreover, unlike in the previous experiment,‘‘towards transfer-towards motion’’ was signifi-cantly longer than ‘‘away transfer-towards mo-tion’’, t(48)#2.73, pB.008.

TABLE 3Experiment 2: Mean key-releasing times (in milliseconds) and standarddeviations (in parentheses) as a function of type of sentence, transfer

direction, and motion direction

Motion

Sentence type Transfer Away Towards

Factual Away 347 (107) 350 (100)Towards 345 (101) 347 (99)ACE !2 3

Counterfactual Away 353 (98) 343 (102)Towards 343 (96) 358 (100)ACE !10 !15

The ACE was the subtraction between pairs of matching!mismatchingconditions.

8 DE VEGA AND URRUTIA

Dow

nloa

ded

by [U

nive

rsity

of L

a La

guna

Vic

erre

ctor

ado]

at 0

2:04

07

Oct

ober

201

1

Probe reaction times

Again, a main effect of sentence type wasfound in this measure, as factual sentences wereresponded to faster than counterfactuals, F(1,68)#12.780, MSE#5347.440, pB.001. In addi-tion, responses to probes were faster following atowards-key (M#875 ms, SD#186) than follow-ing an away-key response (M#890 ms, SD#199), F(1, 68)#5.545, MSE#5650.100, pB.05.

Discussion

In Experiment 2, a static cue (colour change)substituted the visual motion cue of the previousexperiment. In this way visual motion effects wereprevented in the experiment, and the observedinteraction only could be characterised as action!sentence compatibility effect. Here too, the inter-action between meaning and action was one ofinterference in the matching conditions, replicat-ing the basic effect observed in Experiment 1.Unlike in the previous experiment, the intervalbetween the word onset and the static cue (colourchange) was fixed at 100 ms and thus we could notcheck the temporal course of the ACE forfactuals. Nonetheless, if we consider the resultsof Experiment 1 and Experiment 2 together, wecan observe several interesting facts. In Experi-ment 1, the ACE emerged earlier in counter-factuals (100 ms) than in factuals (200 ms).Consistently, in the present experiment, ACEwas only effective in counterfactuals at the 100ms interval. From this experiment, we cannotestablish the temporal course of ACE in factualsentences. However, by analogy with Experiment1, we could speculate that the factual ACE wasdelayed beyond 100 ms and perhaps could beobserved in a later, untested delay.

Unlike in Experiment 1, in this experiment themost sensitive measure was the key-releasing timerather than the key-pressing time. A possibleexplanation for this fact could be that encodingcolour change as a cue of directional motionmight be slower than encoding the animation cue.We checked this hypothesis by contrasting thewhole set of data from Experiment 2 with thedata from Experiment 1 corresponding to the 100ms delay condition, confirming that the key-releasing time was almost 100 ms slower in theformer than in the latter (mean colour cue#350ms; mean animation cue#252 ms), F(1,113)#31.42, MSE#67475.722, pB.0001. There-

fore, the delay in interpreting the colour cuecould make visible the ACE at the releasing-keystage of response.

GENERAL DISCUSSION

The present study explored whether the compre-hension of counterfactual sentences activatesembodied representations. Previous studies havereported action!sentence compatibility effectswhen people understand sentences describingactions. These studies employed concrete factualsentences describing transfer events (Borreggine& Kaschak, 2006; de Vega et al., 2011; Glenberg& Kaschak, 2002; Kaschak & Borreggini, 2008),hand or foot actions (Buccino et al., 2005), orhand rotating actions (Zwaan & Taylor, 2006). Bycontrast, this research compared action-relatedsentences in either counterfactual or factualformat. Counterfactuals posit a challenge to theembodied meaning approach, because they referto hypothetical rather than to real events and,therefore, constructing the costly embodied re-presentations might be superfluous. We couldconceive, for instance, that a shallow representa-tion in terms of propositions could be enough tograsp the dual meaning of counterfactuals. Thus,the sentence ‘‘If my brother had been generous hewould have lend me the novel’’ could be repre-sented as two propositions corresponding to itsdual meaning: (1) false [(brother, generous) &(brother, lend, novel, speaker)], and (2) true[(brother, generous) & (brother, lend, novel,speaker)].

However, our results show ACE in counter-factual and factual sentences alike, which provesthat they share a similar embodied representationof meaning. The observed ACE consisted of anearly interference in the meaning!action match-ing conditions that could be interpreted as agenuine motor resonance effect. As explained inthe introduction, early interference strongly sug-gests that understanding action-related sentencesrecruits similar neural resources as performingmatching actions (see also Boulenger et al., 2006;Buccino et al., 2005; de Vega et al., 2011).

But, why should counterfactuals trigger embo-died representations? One possible reason relieson their dual meaning. Counterfactuals activatenot only a realistic representation in whichthe events did not happen but also an ‘‘asif’’ simulation of the events happening, as amechanism for undoing reality and imagining

EMBODIMENT IN COUNTERFACTUALS 9

Dow

nloa

ded

by [U

nive

rsity

of L

a La

guna

Vic

erre

ctor

ado]

at 0

2:04

07

Oct

ober

201

1

the alternative course of events. The lattersimulation could be equivalent to the meaningof a factual statement and, therefore, would sharethe same embodiment characteristics. A similaridea has been recently proposed to explain hownegations activate embodied representations ofthe negated concepts (Kaup, Yaxley, Madden,Zwaan, & Ludtke, 2007). Kaup et al. (2007) builtpairs of negative sentences sharing an object, insuch a way that both of the negated situationsimplied a different shape of the same object. Forinstance, the sentence ‘‘There was no eagle in thesky’’ implies that the eagle has its wings out-stretched, whereas the sentence ‘‘There was noeagle in the nest’’ implies that the eagle has itswings folded in. Participants received each sen-tence followed by a picture that, in the experi-mental trials, depicted the critical object in one ofits shapes (e.g., the eagle with the wigs out-stretched), and their task was to verify whetherthe object in the picture was mentioned in thesentence. The readers’ responses were fasterwhen the depicted object matched rather thanmismatched the shape implicit in the negatedsituation, indicating that readers activated a visualrepresentation of the negated object.

It is useful to contrast counterfactuals andnegations, because both apparently share similarrepresentational characteristics. Thus, counterfac-tual sentences although frequently have an affir-mative format involve implicit negationscorresponding to their realistic meaning (not-p & not-q), as well as an ‘‘as if’’ simulation oftheir alternative meaning (p & q). On their side,negations in spite of having an explicit negativeoperator (e.g., not) could momentarily activate a‘‘counterfactual’’ representation of the negatedevents as if they had truly happened (Kaup,Ludtke, & Zwaan, 2006; Kaup & Zwaan, 2003),followed by the representation of the actualevents, namely a scenario with the negatedconcepts absent. In other words, it seems thatboth counterfactuals and negations share dualmeaning. Furthermore, similar mechanisms fordealing with dual meaning have been proposedfor counterfactuals and negations. For instance,counterfactual reasoning requires keeping trackof what is really true and what is imagined,although momentarily assumed to be true (Byrne,2005). In the same vein, the alternative ‘‘as if’’meaning in counterfactuals does not contribute toupdating the discourse situation model, but iskept apart and eventually suppressed (de Vegaet al., 2007; Urrutia et al., 2011), and the negated

situation in negations is not integrated into therepresentation of the described world, but isbuffered in an ‘‘auxiliary’’ representational sys-tem (Kaup et al., 2007; Kaup & Zwaan, 2003).Finally, as discussed earlier, the unrealistic mean-ing of both counterfactuals and negations couldbe associated with embodied representations.

In spite of the obvious parallel betweencounterfactuals and negations there must be alsosubstantial semantic differences between them tojustify so different linguistic constructions. Parti-cularly, why to use the complex grammar ofcounterfactuals (conditional marker, composesubjunctive tense) if you could employ thesimpler negation operator? Here we proposesome important differences. First, for negationthe representation of the negated events does notcontribute to its proper meaning, whereas forcounterfactuals the double simulation of actualevents and their imagined alternatives contributeto their meaning. We could conceive the simula-tion of the negated events as an automatic by-product of lexical processing, which is rapidlysuppressed to give place to the representation ofthe actual scenario with the negated eventsabsent. However, when you listen to a counter-factual expression you are invited to consider theactual situation but also to consider for a whilethe alternative ‘‘as if’’ situation (Byrne, 2005).This dual meaning is essential to fully understandthe counterfactual, and to accomplish its psycho-logical functions in communication and reason-ing, such as establishing causal inferences,learning from experience, or expressing the emo-tion of regret. Second, negations and counter-factuals differ in their truth values. Whereasnegative statements unambiguously shift the truthvalues of the affirmative counterparts*for in-stance if the affirmation is true the correspondingnegation is false and vice versa (e.g., Nieuwland &Kuperberg, 2008), in counterfactuals the conflict-ing truth values seem to be acceptable given thecoexistence of the two alternative meanings(Santamaria et al., 2005; Thompson & Byrne,2002). In negations the alternative scenario couldbe presupposed (e.g., an expected event) but itsrepresentation is not relevant for understandingthe statement. By contrast, the counterfactualformat explicitly instructs the addressee to repre-sent not only the real events (the implicitlynegated scenario), but also the alternative scenar-io, because both are necessary to accomplishcounterfactual functions, as explained previously.

10 DE VEGA AND URRUTIA

Dow

nloa

ded

by [U

nive

rsity

of L

a La

guna

Vic

erre

ctor

ado]

at 0

2:04

07

Oct

ober

201

1

This study showed that motor resonance occursin a narrow temporal window, which slightlydiffers in counterfactual and factual sentences.Thus, Experiment 1, which provides the mostaccurate temporal measure of ACE, revealedinterference at 100 ms and 200 ms after the verbonset for counterfactuals and factuals, respec-tively. The different timing of motor resonancein counterfactuals and factuals is quite surprisingand even counterintuitive. The results for probelatencies clearly indicate that counterfactualsinvolve more cognitive cost than factuals, eitherbecause they have a more complex syntax orbecause they involve dual meaning. Therefore, itmight be reasonable to expect that resonancewere delayed rather than speeded up in thesesentences. A tentative explanation of the ad-vanced resonance in counterfactuals requires afew assumptions. First, the timing of motorresonance in the brain is not fixed but it can bemodulated by linguistic factors such as focus (e.g.,Taylor & Zwaan, 2008) and, in this study, the dualmeaning of counterfactuals. Second, the counter-factual context is marked very early in the clausepreceding the transfer event; this means that thecounterfactual status of the sentence was alreadybeing processed when the participant listened tothe transfer verb. Third, the processing of the ‘‘asif’’ meaning (p & q) starts before that of therealistic meaning that, being a double negation(not-p & not-q), requires more elaboration.1 Withthese premises, is easy to conclude that readers ofa counterfactual context speed up the ‘‘as if’’meaning associated with the advanced motorresonance perhaps as a strategy to free cognitiveresources for elaborating the alternative meaning(the implicit double negation). By contrast, whenreaders come from a factual context they justprocess a unique meaning (p & q) and do notneed to speed up resonance. We may notice,however, that this study did not trace the wholetemporal course of motor resonance. Furtherresearch will be necessary, using a larger rangeof word!action delays, before taking definitiveconclusions on the timing of embodied meaningboth in factuals and counterfactuals.

In sum, this research has clearly demonstratedmeaning!action interference during the compre-hension of counterfactuals, extending and qualify-ing the motor resonance hypothesis to nonliteralstructures that describe possibilities rather thanfacts. In spite of the abstract meaning of counter-factuals (they refer to imaginary possibilitiesrather than real events), their meaning seems tobe embodied, at least when their content refers tohypothetical transfer events. Further studies willbe necessary to deal with the embodied repre-sentations of counterfactuals. Thus, a direct com-parison between counterfactuals and negationsseems relevant given their semantic similarity aswell as difference outlined earlier. It might be alsointeresting to test whether facilitation in matchingconditions also occurs with counterfactual mate-rials when the tasks are similar to those employedin the standard ACE paradigm.

Original manuscript received November 2010Revised manuscript received May 2011

First published online October 2011

REFERENCES

Barsalou, L., Santos, A., Simmons, W. K., & Wilson, C.D. (2008). Language and simulation in conceptualprocessing. In M. de Vega, A. Glenberg, & A.Graesser (Eds.), Symbols, and embodiment: Debateson meaning and cognition (pp. 245!284). New York,NY: Oxford University Press.

Borreggine, K. L., & Kaschak, M. P. (2006). Theaction!sentence compatibility effect: It’s all in thetiming. Cognitive Science, 30, 1097!1112.

Boulenger, V., Roy, A. C., Paulignan, Y., Deprez, V.,Jeannerod, M., & Nazir, T. A. (2006). Cross-talkbetween language processes and overt motor beha-vior in the first 200 msec of processing. Journal ofCognitive Neuroscience, 18, 1607!1615.

Buccino, G., Riggio, G., Melli, F., Binkofski, V., Gallese,G., & Rizzolatti, G. (2005). Listening to action-related sentences modulates the activity of themotor system: A combined TMS and behavioralstudy. Cognitive Brain Research, 24, 355!363.

Byrne, R. M. J. (2005). The rational imagination: Howpeople create alternatives to reality. Cambridge, MA:MIT Press.

De Vega, M., Glenberg, A. M., & Graesser, A. C.(2008). Symbols and embodiment: Debates on mean-ing and cognition. New York, NY: Oxford Univer-sity Press.

De Vega, M., Moreno, V., & Castillo, D. (2011). Thecomprehension of action-related language maycause interference rather than facilitation on match-ing actions. Manuscript submitted for publication.

De Vega, M., Urrutia, M., & Riffo, B. (2007). Cancel-ling updating in the comprehension of counterfactuals

1 This claim is confined to affirmative counterfactuals, likethe ones employed here. However, in negative counterfac-tuals, such as ‘‘if my brother hadn’t been generous he wouldn’thave lent me the novel’’, the ‘‘as if’’ meaning is ‘‘not-p & not-q’’, and the realistic meaning is ‘‘p & q’’. It is an openquestion, in this case, whether the ‘‘as if’’meaning is processedbefore or after the realistic meaning.

EMBODIMENT IN COUNTERFACTUALS 11

Dow

nloa

ded

by [U

nive

rsity

of L

a La

guna

Vic

erre

ctor

ado]

at 0

2:04

07

Oct

ober

201

1

embedded in narratives. Memory and Cognition,35(6), 1410!1421.

Ferguson, H. J., & Sanford, A. J. (2008). Anomalies inreal and counterfactual worlds: An eye-movementinvestigation. Journal of Memory and Language, 58,609!626.

Glenberg, A. M., & Kaschak, M. P. (2002). Groundinglanguage in action. Psychonomic Bulletin and Re-view, 9, 558!565.

Kahneman, D., & Tversky, A. (1982). The simulationheuristic. In D. Kahneman, P. Slovic, & A. Tversky(Eds.), Judgment under uncertainty: Heuristics andbiases (pp. 201!208). New York, NY: CambridgeUniversity Press.

Kaschak, M. P., & Borreggine, K. L. (2008). Temporaldynamics of the action-sentence compatibility ef-fect. Quarterly Journal of Experimental Psychology,61(6), 883!895.

Kaup, B., Ludtke, J., & Zwaan, R. A. (2006). Processingnegated sentences with contradictory predicates: Is adoor that is not open mentally closed? Journal ofPragmatics, 38, 1033!1050.

Kaup, B., Yaxley, R. H., Madden, C. J., Zwaan, R. A., &Ludtke, J. (2007). Experiential simulations of ne-gated text information. Quarterly Journal of Experi-mental Psychology, 60, 976!990.

Kaup, B., & Zwaan, R. A. (2003). Effects of negationand situational presence on the accessibility of textinformation. Journal of Experimental Psychology:Learning, Memory and Cognition, 29(3), 439!446.

Nieuwland, M. S., & Kuperberg, G. R. (2008). Whenthe truth isn’t too hard to handle: An event-relatedpotential study on the pragmatics of negation.Psychological Science, 19, 1213!1218.

Pulvermuller, F., Hauk, O., Nikulin, V. V., & Ilmoniemi,R. J. (2005). Functional links between motor andlanguage systems. European Journal of Neu-roscience, 21, 793!797.

Roese, N. (2005). If only: How to turn regret intoopportunity. New York, NY: Broadway Books.

Santamarıa, C., Espino, O., & Byrne, R. (2005).Counterfactual and semifactual conditionals primealternative possibilities. Journal of ExperimentalPsychology: Learning, Memory and Cognition, 31,1149!1154.

Santana, E., & de Vega, M. (2011). Metaphors areembodied, and so are their literal counterparts.Frontiers in Cognition, 2!90. doi:103389/fpsyg.2011.00090

Stewart, A. J., Haigh, M., & Kidd, E. (2009). Aninvestigation into the online processing of counter-factual and indicative conditionals. Quarterly Jour-nal of Experimental Psychology, 25, 1!13.

Taylor, L. J., & Zwaan, R. A. (2008). Motor resonanceand linguistic focus. Quarterly Journal of Experi-mental Psychology, 61, 896!904.

Taylor, L. J., & Zwaan, R. A. (2009). Action incognition: The case of language. Language andCognition, 1, 45!48.

Thompson, V., & Byrne, R. (2002). Reasoning counter-factually: Making inferences about things that didn’thappen. Journal of Experimental Psychology: Learn-ing, Memory and Cognition, 28, 1154!1170.

Urrutia, M., de Vega, M., & Bastiaansen, M. (2011).Understanding counterfactuals in discourse modu-lates ERP and oscillatory gamma rhythms in theEEG. Manuscript submitted for publication.

Zwaan, R. A., Madden, C. J., Yaxley, R. H., & Aveyard,M. E. (2004). Moving words: Dynamic mentalrepresentations in language comprehension. Cogni-tive Science, 28, 611!619.

Zwaan, R. A., & Taylor, L. J. (2006). Seeing, acting,understanding: Motor resonance in language com-prehension. Journal of Experimental Psychology:General, 135, 1!11.

12 DE VEGA AND URRUTIA

Dow

nloa

ded

by [U

nive

rsity

of L

a La

guna

Vic

erre

ctor

ado]

at 0

2:04

07

Oct

ober

201

1