Embodiment of Abstract Concepts: Good and Bad in Right- and Left-Handers PDF

Journal of Experimental Psychology: General © 2009 American Psychological Association 2009, Vol. 138, No. 3, 351–367...

Journal of Experimental Psychology: General © 2009 American Psychological Association 2009, Vol. 138, No. 3, 351–367 0096-3445/09/$12.00 DOI: 10.1037/a0015854 Embodiment of Abstract Concepts: Good and Bad in Right- and Left-Handers Daniel Casasanto Stanford University and the Max Planck Institute for Psycholinguistics Do people with different kinds of bodies think differently? According to the body-specificity hypothesis, people who interact with their physical environments in systematically different ways should form corre- spondingly different mental representations. In a test of this hypothesis, 5 experiments investigated links between handedness and the mental representation of abstract concepts with positive or negative valence (e.g., honesty, sadness, intelligence). Mappings from spatial location to emotional valence differed between right- and left-handed participants. Right-handers tended to associate rightward space with positive ideas and leftward space with negative ideas, but left-handers showed the opposite pattern, associating rightward space with negative ideas and leftward with positive ideas. These contrasting mental metaphors for valence cannot be attributed to linguistic experience, because idioms in English associate good with right but not with left. Rather, right- and left-handers implicitly associated positive valence more strongly with the side of space on which they could act more fluently with their dominant hands. These results support the body-specificity hypothesis and provide evidence for the perceptuomotor basis of even the most abstract ideas. Keywords: abstract concepts, body-specificity hypothesis, embodied cognition, emotional valence, metaphor According to theories of embodied cognition, thoughts comprise that could not be predicted (or explained) on the basis of patterns mental simulations of bodily experiences (Barsalou, 1999; Feld- in language and culture. man, 2006; Goldstone & Barsalou, 1998; Lakoff & Johnson, 1999; It is easy to imagine ways in which specific features of our Prinz, 2002). If thinking is embodied in this sense, then people bodies might influence our mental representations of concrete with different kinds of bodies must think differently. That is, if objects and actions. For example, if thinking about objects in- concepts and word meanings are constituted in part by simulations volves mentally simulating their colors (Simmons et al., 2007), of people’s own perceptions and actions, then people with different then mental representations of apples should be qualitatively dif- bodily characteristics, who interact with their physical environ- ferent between individuals with red– green color blindness and ments in systematically different ways, should form correspond- individuals with normal vision. If thinking about actions involves ingly different mental representations. I call this proposal the mentally simulating the way we typically execute them, actions body-specificity hypothesis. that we perform with our dominant hands such as throwing a ball, The five experiments reported here tested this hypothesis and turning a key, or signing a check should have different neurocog- sought to overcome two obstacles to advancing embodied theories nitive representations in right-handed and left-handed individuals of mental representation: (1) devising experiments in which em- (Casasanto, 2008a; Longcamp, Anton, Roth, & Velay, 2005; Wil- bodied theories and their alternatives make contrasting predictions lems, Hagoort, & Casasanto, in press). and (2) determining the experiential origins of our mental repre- But how might body-specificity be relevant to the mental rep- sentations, even in abstract conceptual domains. Framing experi- resentation of more abstract concepts, like goodness and badness, mental predictions in terms of body-specificity made it possible to victory and loss, deceit and honesty? Our ability to formulate such discover a bodily basis for people’s abstract mental representations abstract ideas presents a challenge to embodied theories of con- cepts, in general: How can perceptuomotor simulations help to represent things we can never perceive with the senses or act upon Daniel Casasanto, Department of Psychology, Stanford University, and with the muscles? Like many abstract concepts, these notions carry Max Planck Institute for Psycholinguistics, Nijmegen, the Netherlands. either positive or negative emotional valence. Several lines of A portion of this research was reported at the 30th Annual Conference research have suggested important links between valence and of the Cognitive Science Society, Washington, DC, July 2008. This re- perception and action in physical space. search was supported in part by National Research Service Award Post- In English and other languages, metaphorical expressions tend doctoral Fellowship F32MH072502 and by Spanish Ministry of Education to associate positive and negative valence with the top and bottom and Science Grant SEJ2006-04732/PSIC, DGI. I am grateful to Mandana of a vertical spatial continuum (Lakoff & Johnson, 1980, 1999): a Seyfeddinipur, Rachel Staum, Merel van Rees Vellinga, and Josje Verha- happy person is “high on life,” but a sad person is “down in the gen for collecting and coding questionnaire data; to Laura Staum Casasanto and Herb Clark for invaluable discussions; and to Fernanda Ferreira for dumps”; a genius is “highly intelligent,” but an idiot has a “low comments on a draft of the article. IQ”; some research assistants are “top flight,” whereas others are Correspondence concerning this article should be addressed to Daniel at the “bottom of the barrel.” According to theories of metaphor- Casasanto, Max Planck Institute for Psycholinguistics, P.O. Box 310, 6500 ical mental representation (e.g., Lakoff & Johnson, 1999), these AH Nijmegen, the Netherlands. E-mail: [email protected] linguistic metaphors reflect mental metaphors (Casasanto, 2008b): 351 352 CASASANTO nonlinguistic associative mappings from the concrete source do- 1982), and in another participants expressed more pride in their main of space to relatively abstract target domains with positive or test performance after sitting upright during the critical phase of negative valence. Behavioral studies provide evidence that mental the experiment than after slouching (Stepper & Strack, 1993). metaphors from physical space structure our representations not On the alternative proposal, mental metaphors are established only of valenced concepts (Casasanto & Dijkstra, 2009; Meier & through experience by using linguistic metaphors. Using spatial Robinson, 2004) but also of time (Boroditsky, 2000, 2001; Casasanto words in both literal and metaphorical contexts (e.g., a high shelf, & Boroditsky, 2008), number (Dehaene, Bossini, & Giraux, 1993), a high ideal) could cause structural elements of the concrete source emotional attachment (L. E. Williams & Bargh, 2008), power (Schu- domain to be transferred to abstract target domain representations bert, 2005), and similarity (Casasanto, 2008b)— even when we’re not in the mind of the language learner via analogical processes that using any linguistic metaphors (cf. Murphy, 1996, 1997). are not necessarily “embodied” (see Boroditsky, 2000; Gentner, Mental metaphors provide a potential solution, or at least part of Bowdle, Wolff, & Boronat, 2001). One reason to consider this a solution, to the problem of representing abstract ideas via em- possibility is that associations between valence and vertical space bodied simulations. Thinking about affective states or making have been shown to generalize beyond concepts that have obvious affective judgments could involve mental simulations in both perceptuomotor correlates. For example, in one experiment, par- source and target domains. Target domain representations could ticipants were faster to judge words like polite and rude as having comprise partial reenactments of emotional states in the regions of the brain that give rise to emotional experiences. Simulating af- positive or negative valence when positive words were presented fection could involve recapitulating patterns of activity in the at the top and negative words at the bottom of a computer screen nucleus accumbens that produce the interoceptive experience of (Meier & Robinson, 2004). In another experiment, participants were affection, and simulating fear recapitulating patterns of activity in faster to make lexical decisions on positive-valence words (e.g., the amygdala that produce the experience of fear (Damasio, 2001). brave, ethical, wealthy) when they were presented above nonword These primitive target domain representations may be too vague or distractors, and on negative-valence words (e.g., poverty, failure, fleeting to support higher order reasoning about emotional states hate) when presented below nonword distractors, even though neither and may be resistant to the kinds of verbal and imagistic coding the spatial position of the words nor their valence was relevant to the that can scaffold such reasoning. Mental metaphors import the task (Casasanto & Nabieva, 2007). These results are problematic for inferential structure of source domains like space into target do- a purely embodied theory of mental metaphor: What are the percep- mains, allowing us to envision, measure, and compare the height of tuomotor correlates of politeness, poverty, or ethics that could plau- people’s excitement, the depth of their sadness, or the breadth of sibly link these concepts with vertical space? their compassion (Boroditsky, 2000; Casasanto, 2008c; Pinker, Linguistic conventions associating valence with vertical space 1997). To the extent that mental representations in perceptuomotor are reinforced by other nonlinguistic cultural conventions, such as source domains constitute abstract concepts, these concepts can be the “thumbs up” and “thumbs down” gestures that indicate ap- instantiated by the same neural and mental structures that simulate proval and disapproval. Once these linguistic and nonlinguistic perception and action in the physical world. conventions are part of a culture, they can serve as the basis for metaphorical mappings in the minds of individual learners, obvi- Are Mental Metaphors Necessarily “Embodied”? ating the role of direct bodily experience. As such, data that support metaphor theory do not necessarily support embodiment Considerable evidence for the existence of mental metaphors theory. Mental metaphors could be learned from patterns in lan- has accumulated, but their experiential origins have remained guage and culture. Behavioral effects that reveal the existence of unknown. Two proposals have emerged from nearly distinct liter- mental metaphors could result from spreading activation between atures, the first positing that mental metaphors arise due to corre- nodes in an amodal conceptual network that are habitually coac- lations in bodily experience and the second due to correlations in linguistic experience. On the first proposal (e.g., Lakoff & John- tivated during language use, rather than resulting from correlated son, 1999; Piaget, 1927/1969), mental metaphors like Positive Is physical and emotional experiences.1 Up and Negative Is Down could be established as people implicitly learn associations between physical experiences and emotional 1 The proposal that correlations in linguistic experience give rise to states that typically co-occur (e.g., standing tall when we feel Good Is Up mental metaphors in the individual learner raises the question proud, slouching when we feel dejected). Linguistic metaphors of how linguistic metaphors for valence arose in the first place, and why then encode preexisting mental metaphors developed on the basis they are so common across languages. It may be that correlations in direct of relationships between different types of bodily experiences (i.e., bodily experience resulted in the construction of these linguistic conven- perceptuomotor experiences in source domains and interoceptive tions during the course of linguistic/cultural evolution. Yet, the existence of experiences in target domains). Social psychological studies have linguistic metaphors in a culture is not evidence of prelinguistic mental been interpreted as supporting this proposal (Barsalou, Niedenthal, metaphors in the individual. Metaphors in language and other linguistic Barbey, & Ruppert, 2003). Participants perform better and produce structures (e.g., words for exact numbers) can precede concepts in the individual learner and lead to the creation of new conceptual links via more positive evaluations on various tasks when they assume learning processes that do not need to be grounded directly in perceptuo- upright body postures in an experimental setting that are similar to motor experience (see Carey, 2004; Dehaene, Bossini, & Giraux, 1993; the postures people assume spontaneously when they are feeling Gentner et al., 1997). Even if direct bodily experience is necessary on the good in their everyday lives. In one experiment, participants per- timescale of biological evolution or cultural/linguistic evolution, it may not sisted longer in a puzzle-solving task after assuming an upright be necessary on the timescale of conceptual development in the individual posture as opposed to a slouching posture (Riskind & Gotay, learner (Tomasello, 2003; Vygotsky, 1934/1986).  BODY-SPECIFIC ABSTRACT CONCEPTS 353 Is it possible to determine whether mental metaphors arise from 1993; Maas & Russo, 2003; Tversky, Kugelmass, & Winter, 1991; correlations in linguistic experience versus correlations in nonlin- Zebian, 2005). Yet, despite these cross-cultural reversals in hori- guistic bodily experience, or from some combination of linguistic zontal mappings for number and time, linguistic and cultural and bodily experiences? Because patterns in language closely conventions in Arabic-speaking and English-based cultures reveal mirror patterns of bodily interactions with the environment (Clark, the same Good Is Right mapping. Valence is apparently insensitive 1973), language and bodies generally make the same neural and to writing direction. One possible explanation for this seemingly behavioral predictions (Machery, 2007; Zwaan & Yaxley, 2003a, universal directional preference is that it arises from universal 2003b). To overcome this obstacle, the present study tested for a properties of the human brain and mind, perhaps related to innate body-specific mapping between space and valence that is not hemispheric specialization for approach and avoidance motiva- encoded in any known language or cultural artifacts but that was tional systems. Once established due to innate neurobiological predicted on the basis of particulars of our bodily experience. factors, conventions in language and culture may reinforce this implicit preference for the right. When Is the Right Side the “Right” Side? An alternative possibility, however, is that left–right conven- tions in language and culture arise as a consequence of body- In addition to links between valence and vertical space, there are specific associations between space and valence. Bodies are lop- robust associations between valence and the right and left sides of sided. Most people have a dominant hand, usually the right hand the body. Actions performed with the right side of the body (e.g., (Corballis & Beale, 1976), and therefore interact with their envi- contracting the muscles of the right hand or the right side of the ronment more fluently on one side of body-centered space than the face, or viewing stimuli in the right visual hemifield) correlate with other. Greater perceptuomotor fluency has been shown to correlate positive affect, whereas the same actions performed with the left with more positive evaluations (Oppenheimer, 2008; Reber, Wink- side of the body correlate with negative affect (e.g., Davidson, ielman, & Schwarz, 1998). For example, expert typists have shown 1992; Natale, Gur, & Gur, 1983). This association of positive and a preference for pairs of letters that can be typed easily over pairs negative valence with the right and left sides of the body may be that are more difficult to type (even when typing is not relevant to linked to motivational systems controlling approach and avoidance the task), suggesting that motor experience can influence affective behaviors. Approach has been lateralized to the left frontal lobe judgments (Beilock & Holt, 2007; Van den Bergh, Vrana, & (which controls the right side of the body), and avoidance to the Eelen, 1990). In a sense, we are all “experts” at using our dominant right frontal lobe (which controls the left side of the body; e.g., hands. Perhaps over a lifetime of lopsided perceptuomotor expe- Davidson, Ekman, Saron, Senulis, & Friesen, 1990; Schiff & rience, people come to implicitly associate good things with the Bassel, 1996). side of space they can interact with more fluently and bad things The same associations of right with “positive” and left with with the side of space they interact with less fluently? On this “negative” that are revealed by neuropsychological tests are also body-specific possibility, the apparent universality of the Good enshrined in everyday linguistic expressions. English idioms like Is Right mapping suggested by linguistic and cultural conventions “the right answer” and “my right-hand man” associate good things could be a result of right-handers’ predominance in the population, with rightward space, and complementary idioms like “out in left worldwide: Linguistic and cultural conventions may develop ac- field” and “two left feet” associate bad things with leftward space. cording to the implicit body-specific mental metaphors of the The Latin words for right and left—dexter and sinister—form the majority. roots of English words meaning “skillful” and “evil,” respectively. Five experiments tested associations between valence and hor- Similar patterns are found across many languages. The words for izontal space in right- and left-handed individuals, to determine right in French (droite) and in German (Recht) are closely related whether these mappings are universal or body-specific. If they are to the words meaning “a ‘right’ or privilege accorded by the law,” universal, either due to invariant properties of the human brain and whereas the words for left in French (gauche) and German (Links) mind or to pervasive patterns in language and culture, then both are related to words meaning “distasteful” or “clumsy.” Such right- and left-handers should preferentially associate good with idioms are evident in nonlinguistic conventions in some cultures, right and bad with left. Alternatively, if they are body-specific, such as Ghanaian society, where pointing and gesturing with one’s then right- and left-handers should show opposite patterns, each left hand is prohibited (Kita & Essegbey, 2001). According to group associating good things more strongly with their dominant Islamic doctrine, the left hand should be used for dirty jobs, side and bad things with their nondominant side. whereas the right hand is used for eating. Likewise, the left foot is used for stepping into the bathroom, and the right foot for entering Experiment 1: Diagramming the Good and the Bad the mosque. Why, across cultures and modalities, is left conventionally as- Participants performed a pencil-and-paper diagram task in sociated with bad and right with good? This directional preference which they drew one animal in each of two boxes, located either for valence does not appear to vary with the direction of reading to the left and right of a cartoon figure (in the horizontal condition; and writing in a culture, which correlates with mappings from see Figure 1a) or above and below a cartoon figure (in the vertical horizontal space to other abstract concepts such as number and condition; see Figure 1b). Instructions indicated that the cartoon time. In cultures that use left-to-right writing systems, like English- figure likes certain animals and thinks they are good but dislikes speaking cultures, the mental number line increases from left to other animals and thinks they are bad. Participants were instructed right, and likewise the mental time line flows from left to right; the to draw a good animal in the box they thought best represented opposite is true in Arabic-speaking cultures, which use right-to-left good things and a bad animal in the box that best represented bad writing systems (Chatterjee, 2001; Dehaene, Bossini, & Giraux, things. 354 CASASANTO Figure 1. Top: Examples of stimuli and responses from Experiment 1. 1a: Horizontal condition. 1b: Vertical condition. Bottom: Results of Experiment 1. 1c: Proportion of left- and right-handers who drew the good animal in the left box (dark bars) and the right box (light bars). 1d: Proportion of left- and right-handers who drew the good animal in the top box (dark bars) and the bottom box (light bars). Error bars indicate standard error of the mean. The horizontal condition served as a test of the body-specificity ipant completed either the horizontal or the vertical task, which hypothesis: If the horizontal spatialization of valence is body- differed primarily in the orientation of the boxes; in addition, specific, then right- and left-handers should show opposite pref- different cartoon characters were used for the horizontal and erences for the placement of good and bad animals in the left and vertical conditions. This was necessary because in the horizontal right boxes. Alternatively, if the horizontal spatialization of va- condition it was critical that the cartoon character’s spatial per- lence is universal, then right- and left-handers should show the spective be the same as the participants’ so that their assignment of Good Is Right preference that is suggested by linguistic and animals to the left and right boxes could be interpreted unambig- cultural conventions. uously, whereas in the vertical condition it was critical that up/ The vertical condition served as a control. Regardless of hand- down space be distinguishable from front/back space in the two- edness, we all use the same linguistic and cultural conventions that dimensional diagram (see Figures 1a and 1b). associate up with good, and we all experience the same correla- Instructions indicated that when participants flipped the page tions between upright body posture and positive mood. Therefore, they would meet a cartoon character who was planning a trip to the both right- and left-handers should tend to place the good animal zoo. They were told that the character loves zebras and thinks they on top and the bad animal on the bottom. are good but hates pandas and thinks they are bad (or vice versa, depending on the version of the questionnaire a participant re- Method ceived). Their task was to draw a zebra in the box that best Participants. Students from the Stanford University and the represents good things like seeing zebras and a panda in the box University of California, Riverside, communities took part in that best represents bad things like seeing pandas (or vice versa). exchange for payment or course credit (N ! 219; 28 left-handers, The assignment of valence to the panda and zebra was counter- 191 right-handers by self-report). Participants were randomly as- balanced across participants. The order in which participants were signed to perform either the horizontal task (n ! 104; 19 left- instructed to draw the good and bad animals was also counterbal- handers, 85 right-handers) or the vertical task (n ! 115; 9 left- anced, to ensure that any associations between space and valence handers, 106 right-handers). in participants’ judgments were not confounded with associations Materials and procedure. A paper-and-pencil diagram task between the side of space and the temporal order in which they with instructions on one side of a page and responses on the other drew the animals. After completing the diagram task, participants side was distributed among unrelated questionnaires. Each partic- were asked to report their handedness (i.e., “Are you left-handed or  BODY-SPECIFIC ABSTRACT CONCEPTS 355 right-handed? Circle: L or R”) among three other filler questions right- and left-handers’ space–valence mappings, in general; (e.g., “Are you male or female? Circle: M or F”). The question- rather, it is specific to the space–valence mappings that were naire took 1–2 min to complete. predicted to vary on the basis of the body-specificity hypothesis. Results and Discussion Experiment 2: Do People Know That Handedness In the horizontal task, a majority (74%) of left-handers drew the Influences Their Judgments? good animal in the box on the left of the cartoon character (sign test on 5 right side vs. 14 left side, prep ! 0.91), whereas a majority In Experiment 1, more than two thirds of participants placed the (67%) of right-handers drew the good animal in the box on the good animal in the box on their dominant side. Were they aware right (sign test on 57 right side vs. 28 left side, prep ! 0.99; see that their responses corresponded to their handedness? Data col- Figure 1c).2 By Fisher’s exact test, there was a significant corre- lection procedures did not allow for an explicit debriefing, so it is lation between the handedness of the participant and the left–right possible that participants associated the good and bad animals with placement of the good and bad animals ( prep for the exact asso- their dominant and nondominant sides strategically. To find out ciation ! 0.99). The strength of this association was evaluated whether people were conscious of the handedness manipulation, I with a binary logistic regression. The odds ratio (OR) for the replicated the horizontal diagram task in a new group of partici- regression of left–right preference on handedness was estimated at pants but included a written debriefing and two additional changes. 5.70 (95% confidence interval [CI] ! 1.87–17.41), indicating that First, for the debriefing, two of the filler questions that followed right-handers were nearly six times more likely than left-handers the diagram task were replaced with free-response questions ask- to place the good animal on the right and the bad animal on the left. ing participants to guess the purpose of the task and to explain their Since horizontal space has been associated with spatial and answer. Second, an objective measure of handedness was included temporal sequences (Clark, 1973; Dehaene, Bossini, & Giraux, to corroborate participants’ self-report. Finally, the task was ad- 1993) as well as with valence, an analysis was conducted to ministered to native speakers of Dutch in the Netherlands, to evaluate the possibility that the order in which participants drew generalize the findings of Experiment 1 to a new population. As in the animals influenced their placement. Due to counterbalancing, English, conventions in the Dutch language and culture associate approximately equal numbers of participants were instructed to good with right but not with left. draw the good animal first (45 right-handers, 10 left-handers) and the bad animal first (40 right-handers, 9 left-handers). When the good animal came first, 69% of participants placed the good Method animal on their dominant side. Likewise, when the bad animal came first, 67% of participants placed the good animal on their Participants. Students from the Radboud University and Uni- dominant side. The proportions of body-specific responses did not versity of Amsterdam communities participated (N ! 100; 14 differ significantly as a function of the temporal order in which left-handers, 86 right-handers by self-report and by observed writ- participants drew the good and bad animals ( prep ! 0.64), sug- ing/drawing hand). gesting that the temporal order of responses was not responsible Materials and procedure. The horizontal diagram question- for the observed effect. There was also no association between naire from Experiment 1 was used. Instructions were translated left–right preference and the animal (zebra, panda) assigned to be into Dutch by a native-Dutch-speaking linguist who is fluent in good ( prep ! 0.76). English. After completing the questionnaire, participants gave In the vertical task, by contrast, a majority of both left-handers written responses (in Dutch) to three questions: one filler (“What (89%) and right-handers (83%) drew the good animal in the top is your major in school?”), and two free-response debriefing ques- box and the bad animal in the bottom box (left-handers: sign test tions. First, they answered “What do you think this experiment was on 8 top vs. 1 bottom, prep ! 0.93; right-handers: sign test on 88 about? What do you think it was testing?” and then “Why do you top vs. 18 bottom, prep ! 0.99; see Figure 1d). By Fisher’s exact think you placed the good animal in the box that you did?” There test and binary logistic regression, there was no reliable association was no mention of handedness among the printed questions. between the handedness of the participant and the top– bottom Handedness was assessed in two ways. Participants were tested placement of the good and bad animals ( prep for the exact asso- individually or in small groups. Unbeknownst to the participants, ciation ! 0.74; OR ! 0.61, 95% CI! 0.07–5.19). the experimenter observed and recorded the hand they used to In short, the association between horizontal space and valence is draw the animals and write their responses. After the participants body-specific: The number of participants who placed the good had completed the diagram task and the written posttest questions, animal in the box on their dominant side outnumbered those who the experimenter asked them to write the word right or left at the placed it on their nondominant side by a ratio of more than 2 to 1. Right-handers’ responses were consistent with the mental meta- bottom of the page to indicate whether they were right- or left- phor Good Is Right, and left-handers’ with the mental metaphor handed. Good Is Left. The association between vertical space and valence is different. People reliably apply the mental metaphor Good Is Up 2 prep indicates the probability of replicating an observed effect, given an even when the task does not require them to activate any linguistic equipotent replication (Killeen, 2005). A prep value of.92 corresponds to a metaphors for valence or to use language at all in their response. p value of.05 and can be interpreted as estimating a 92% probability of a This is true regardless of handedness. Thus, the effect of handed- replication producing a difference with the sign in the same direction as the ness in the horizontal task cannot be due to differences between observed difference. 356 CASASANTO Results and Discussion self-reported handedness. Surreptitiously observing participants’ handwriting is an efficient, objective, and reliable means of estab- As in the previous experiment, a majority (71%) of left-handers lishing handedness, given that writing hand accounts for the ma- drew the good animal in the box on the left of the cartoon character jority of the variance in the results of more extended measures (sign test on 4 right side vs. 10 left side, prep ! 0.83), whereas a such as the Edinburgh Handedness Inventory (Oldfield, 1971; majority (66 %) of right-handers drew the good animal in the box on the right (sign test on 57 right side vs. 29 left side, prep ! 0.99; S. M. Williams, 1986). The present result corroborates previous see Figure 2a). There was a significant correlation between the research validating simple binary self-report as a measure of handedness of the participant and the left–right placement of the handedness and showing strong agreement between single-item good and bad animals by Fisher’s exact test ( prep for the exact self-report and other measures (Coren, 1993). association ! 0.99) and binary logistic regression of left–right Only 1 person out of the 100 participants guessed that the preference on handedness (OR ! 4.91; 95% CI ! 1.42–17.03), purpose of the experiment had anything to do with handedness. indicating that right-handers were nearly five times more likely The most common response to the first debriefing question was to than left-handers to place the good animal on the right and the bad interpret the task at face value as investigating whether good and animal on the left. bad are more strongly associated with one side or the other (n ! Due to counterbalancing, approximately equal numbers of par- 31). The second most common response was “no idea” (n ! 21), ticipants were instructed to draw the good animal first (43 right- followed by several speculations about hemispheric dominance for handers, 9 left-handers) and the bad animal first (43 right-handers, memory or emotions, effects of writing direction, political prefer- 5 left-handers). When the good animal came first, 69% of partic- ence, or preferences for one animal or the other. In total, partici- ipants placed the good animal on their dominant side. Likewise, pants offered 29 distinct explanations of the task that had nothing when the bad animal came first, 65% of participants placed the to do with handedness. good animal on their dominant side. The proportions of body- When asked to speculate further about why they placed the good specific responses did not differ significantly as a function of the animal in the box that they did, 8 participants offered handedness temporal order in which participants drew the good and bad as the explanation, and 6 more mentioned handedness among other animals ( prep ! 0.75). There was also no association between alternative explanations or after indicating that they were not left–right preference and the animal (zebra, panda) assigned to be certain about this conjecture— or that they were guessing wildly. good ( prep ! 0.79). Among these 14 participants (14% of the total sample), there were For 100 out of 100 participants, the hand that the experimenter equal proportions of right-handers (12/86; 14%) and left-handers recorded as the participant’s writing/drawing hand matched their (2/14; 14%). The fact that 13 of the 14 did not mention handedness when asked to explain what the experiment was testing suggests that handedness came to mind later, when they were asked to rationalize their response. It seems unlikely that they were thinking about handedness when making their responses in the diagram task. Importantly, the body-specific association between handed- ness and left–right preference remained when these 14 participants were removed from the analysis ( prep for the exact association ! 0.92; OR ! 3.10, 95% CI ! 0.86 –11.25). In summary, the results of Experiment 2 provide nearly an exact replication of Experiment 1 in a new population of participants. The objective handedness measure validates the self-report mea- sure used in Experiment 1 (and in subsequent experiments reported here). The debriefing indicated that 99% of the participants were unaware of the purpose of the experiment, suggesting that the observed pattern of body-specific associations between space and valence does not depend on conscious or strategic awareness of handedness. Experiment 3: Does Handedness Affect Judgments When People Don’t Use Their Hands? Why did right- and left-handers associate good with their dom- inant side in Experiments 1 and 2? On a skeptical interpretation, Figure 2. Results of Experiments 2 and 3. 2a: Proportion of left- and these results could reflect a task-related preference for the side of right-handers in Experiment 2 who drew the good animal in the left box (dark bars) and the right box (light bars). 2b: Proportion of left- and space that is easiest to interact with in the moment, while writing right-handers in Experiment 3 who indicated that the good animal should and drawing with one’s dominant hand. Experiment 3 investigated be placed in the left box (dark bars) and the right box (light bars). Error whether a similar pattern of body-specific preferences is found bars indicate standard error of the mean. when the participants respond without using their hands.  BODY-SPECIFIC ABSTRACT CONCEPTS 357 Method of body-specific space–valence associations across experiments. This could indicate that using one’s dominant hand during the task Participants. Students from the Radboud University commu- increases the strength of the effect. However, this comparison must nity participated (N ! 87; 7 left-handers, 80 right-handers by be interpreted cautiously, in part because of the small number of self-report). left-handers in the Experiment 3 sample. Moreover, whereas right- Materials and procedure. Participants performed a version of handers showed a stronger body-specificity effect in the task the horizontal diagram task used in Experiments 1 and 2. Rather where they used their dominant hand to respond, left-handers than responding by drawing animals in boxes, they responded showed a stronger effect in the task in which they responded orally orally to indicate where they would place the good and bad (albeit neither of these differences was significant). Further exper- animals. The picture of the cartoon character and empty boxes iments are needed to determine conclusively whether using one’s used in the previous tasks was printed in the center of a page. At hands in the moment affects left–right valence judgments. the top of the page sentences were printed in Dutch, indicating What is clear from Experiment 3 is that handedness influences which animal the character liked and which he disliked (order of people’s judgments even when they’re not using their hands to mention and assignment of good and bad animals were counter- perform the task. This suggests that the results of Experiments 1 balanced across participants). A native-Dutch-speaking experi- and 2 were not an artifact of the response modality and cannot be menter approached participants one at a time and explained the explained in terms of performance factors related to which side of task while showing them the diagram, which was on one side of a the page is more convenient to interact with in the moment. Rather, notebook cover. At the moment that the experimenter was asking they reflect long-term associations between goodness and one’s the participant to respond, she folded the notebook cover, hiding dominant side of space. Attesting to the robustness of these asso- the diagram— ostensibly so that she could record the responses ciations, the body-specific Good Is Left mapping was activated in on the data sheet. This forced the participant to indicate which left-handers even when the task required them to use words for boxes the good and bad animals should be placed in orally, rather “right” and “left” in the response, which could have activated the than by pointing, since the diagram was no longer visible. If the Good Is Right mapping found in language and culture, instead. response was not clear, the experimenter asked for clarification until the participant used the Dutch words for “right” or “left” to designate at least one of the boxes. After completing the diagram Experiment 4: Body-Specific Judgments of Aliens’ task, the experimenter asked participants to report whether they Attributes were right- or left-handed, among three filler questions. All par- Do mental metaphors from space to valence influence how we ticipants were tested in one extended data collection session, in one evaluate things we encounter in different spatial locations? In location on one day, to minimize the chance that participants tested Experiments 1–3, right- and left-handers demonstrated body- later would have heard anything about the experiment from par- specific space–valence mappings when they were required to as- ticipants tested earlier. sign good and bad stimuli to spatial locations, explicitly. Experi- ment 4 tested for implicit influences of the left–right position of Results and Discussion stimuli on right- and left-handers’ judgments of positive or nega- tive traits in alien creatures. Consistent with the results of the previous experiments, a ma- jority (86%) of left-handers indicated that the good animal should Method go in the box to the left of the cartoon character (sign test on 1 right side vs. 6 left side; Fisher’s exact prep ! 0.83), whereas a majority Participants. Students from the Stanford University and the (58%) of right-handers indicated that it should go in the box on the University of California, Riverside, communities participated in right (sign test on 46 right side vs. 34 left side, prep ! 0.81; see exchange for payment or course credit (N ! 286; 40 left-handers, Figure 2b). There was a significant association between the hand- 246 right-handers by self-report). edness of the participant and the left–right placement of the good Materials and procedure. Participants were given a one-page and bad animals ( prep for the exact association ! 0.94; OR ! 8.11; paper-and-pencil questionnaire in a packet of unrelated question- 95% CI ! 0.93–70.53). naires. Instructions welcomed participants to the planet Fribbalia Due to counterbalancing, approximately equal numbers of par- and asked them to make judgments about some of its inhabitants. ticipants were instructed to place the good animal first (42 right- Below the instructions were 12 pairs of Fribbles (alien creatures handers, 3 left-handers) and the bad animal first (38 right-handers, created by Michael J. Tarr, Brown University, www.tarrlab.org). 4 left-handers). The proportions of body-specific responses did not Members of each Fribble pair were visually similar but clearly differ significantly as a function of the temporal order in which distinguishable by several salient features. Fribbles were arranged participants responded about the good and bad animals ( prep ! in two columns, one on each side of a list of questions printed in 0.75), or as a function of which animal was assigned to be good a center column. This arrangement placed the members of each ( prep ! 0.77). Fribble pair on opposite sides of the page without calling attention The pattern of body-specific responses was compared across to their spatial arrangement. Experiments 2 and 3. Overall, 67% of participants placed the good Participants were asked to circle one member of each Fribble animal on their dominant side in Experiment 2, compared with pair, located on either the left or right side of the question, to 60% in Experiment 3 ( prep ! 0.90, corresponding to a p value of indicate a judgment about one of four personal characteristics (i.e.,.07). According to a traditional significance threshold of p ".05, intelligence, attractiveness, honesty, happiness), for a total of 12 this result shows a marginally significant difference in the strength judgments per participant. Participants were randomly given either 358 CASASANTO the positive-wording version of the questionnaire (i.e., “Circle the Fribble who looks more intelligent”/“more attractive”/“more hon- est”/“happier”) or the negative-wording version (i.e., “Circle the Fribble who looks less intelligent”/“less attractive”/“less honest”/ “sadder”). Both positive- and negative-polarity questions were used for two reasons. First, it is known that question polarity can influence scalar judgments in general (Clark, 1969), so using both positive and negative questions increases the generality of the results. Second, and more importantly, using both positive and negative wording allowed the valence of the response to be or- thogonal to its location: For example, it was possible to make a happiness judgment consistent with the Good Is Right mapping either by circling the right Fribble in the positive wording condi- tion (indicating that the right Fribble was happier) or by circling the left Fribble in the negative wording condition (also indicating that the right Fribble was happier). As such, it was possible to determine whether observed effects of stimulus location were due to participants’ evaluation of the stimuli per se or simply to their preference to make circles on one side of the page versus the other, regardless of their assessment of the Fribbles’ relative virtues. The assignment of questions to Fribble pairs was counterbal- anced across participants, and crucially, assignment of the mem- bers of each pair to the right or left side of the page was also counterbalanced to ensure that interitem differences could not result in spurious preferences for Fribbles located on the left or Figure 3. Results of Experiments 4 and 5. 3a: Mean number of responses right side of the page (even if one Fribble of a pair was inherently (out of 12) made by left- and right-handers in Experiment 4 (Fribbles task) more appealing than the other). After completing the Fribble task, indicating a preference for the Fribble on the left (dark bars) and the participants were asked to report their handedness, among other Fribble on the right (light bars). 3b: Mean number of responses (out of 12) demographic and filler questions. The questionnaire took approx- made by left- and right-handers in Experiment 5 (shopping and jobs tasks) imately 1 min to complete. indicating a preference for the description on the left (dark bars) and the description on the right (light bars). Error bars indicate standard error of the mean. Results and Discussion For each of the 286 participants, the number of responses attributing positive characteristics to Fribbles on the right versus Was this body-specific effect simply due to left-handers prefer- the left side of the page was tabulated. A total of 210 participants ring to write responses on the left of the page and right-handers on (74%; 31 left-handers, 179 right-handers) showed either a right- the right? The positive and negative wordings required participants ward or a leftward preference on average, and 76 participants to make circles on different sides of the page to indicate which (26%; 9 left-handers, 67 right-handers) showed no preference, Fribbles had more positive characteristics. In a chi-square analysis, selecting six Fribbles from the left column and six from the right the pattern of body-specific responses was compared across the column. The proportion of participants who showed no preference positive-wording and negative-wording versions of the question- did not differ as a function of handedness (27% of left-handers, naire, in right- and left-handers combined. There was no reliable 23% of right-handers), #2(1) ! 0.24, Fisher’s exact prep ! 0.63, so association between the wording of the questionnaires and the these participants’ data were excluded from further analysis. frequency with which participants’ responses followed the Dom- Of the participants who showed a directional preference, a majority inant Side Is Good mental metaphor, #2(1) ! 0.44, prep ! 0.68. of left-handers (65%) attributed positive characteristics more often to Thus, the two versions of the questionnaire showed similar results. Fribbles on the left of the page, whereas a small majority of This rules out the possibility that participants simply circled right-handers (54%) attributed positive characteristics more often Fribbles on the side of the page that was most convenient for them to Fribbles on the right of the page (left-handers: sign test on 11 to reach. Rather, in judging intelligence, attractiveness, honesty, right-side preference vs. 20 left-side preference, prep ! 0.85; and happiness, participants were influenced by the left versus right right-handers: sign test on 96 right-side preference vs. 83 left-side location of the Fribbles, per se. preference, prep ! 0.74). As in the previous experiments, there was In summary, the right–left location of Fribbles influenced par- a reliable association between participants’ handedness and their ticipants judgments of their personal characteristics differently in preference for Fribbles on the left versus the right of the page right- and left-handers, even though spatial location was irrelevant ( prep ! 0.94; see Figure 3a). Binary logistic regression of left– to the judgments. Among the majority of participants who showed right preference on handedness showed that the right-handers were a directional preference, left-handers were more likely than right- about twice as likely as the left-handers to attribute more positive handers to show a Good Is Left bias, and right-handers were more characteristics to Fribbles on the right side of the page (OR likely than left-handers to show a Good Is Right bias. These results estimate ! 2.10; 95% CI ! 0.95– 4.64). support the body-specificity hypothesis.  BODY-SPECIFIC ABSTRACT CONCEPTS 359 Experiment 5: Body-Specific Decisions About the Office was orthogonal to the temporal primacy of responses (first item and Marketplace circled or crossed out, second item circled or crossed out). The order of the job and shopping tasks was also counterbalanced Experiment 5 was designed to determine whether the body- across subjects, as was the assignment of the members of each pair specific influences of spatial location apply not only to alien of descriptions to the right or left side of the page, to ensure that creatures but also to the kinds of everyday decisions that we make interitem differences could not result in spurious preferences for here on Earth. Rather than making decisions on novel pictures with items located on the left or right side of the page. Each participant no preexperimental semantic content, participants judged brief made a total of 12 responses. After completing both tasks, partic- verbal descriptions of job applicants (in the job task) and familiar ipants were asked to report their handedness, among other demo- products that might be advertised in a newspaper or website (in the graphic and filler questions. The questionnaire took 1–2 min to shopping task). complete. Method Results and Discussion Participants. Students from the Stanford University and the Responses to the job and shopping questions were combined for University of California, Riverside, communities participated in all analyses. For each of the 371 participants, the number of people exchange for payment or course credit (N ! 371; 50 left-handers, and products selected on the right side versus the left side of the page 321 right-handers by self-report). was tabulated. A total of 272 participants (73%; 35 left-handers, Materials and procedure. The participants were given a paper- 237 right-handers) showed either a rightward or a leftward pref- and-pencil questionnaire on two sides of a page, among unrelated erence on average, and 99 participants (27%; 15 left-handers, 84 questionnaires. On one side of the page was the job task. Partici- right-handers) showed no preference, selecting six items from the pants were asked to imagine that they were in charge of hiring new left column and six from the right column. The proportion of personnel for their company and that they had narrowed the participants who showed no preference did not differ as a function choices to two candidates for each of six jobs (programmer, of handedness (30% of left-handers, 26% of right-handers), graphic designer, security guard, advertising copywriter, adminis- #2(1) ! 0.18, Fisher’s exact prep ! 0.62, so these participants’ data trative assistant, and child daycare manager). Their job was to were excluded from further analysis. decide whom to hire on the basis of brief descriptions of the Of the participants who showed a directional preference, a candidates arranged in columns on the left and right sides of the majority of left-handers (74%) attributed positive characteristics page (as in the Fribbles task). The names of the jobs were listed in more often to people or products on the left side of the page, the center of the page, separating the pairs of descriptions to the whereas a small majority of right-handers (52%) attributed posi- right and left without calling attention to their spatial location. For tive characteristics more often to items on the right side of the page each job, participants were instructed to circle the description of (left-handers: sign test on 10 right-side preference vs. 29 left-side the candidate they would be more likely to hire and to cross out preference, prep ! 0.99; right-handers: sign test on 123 right-side (with an X) the description of the candidate they would be less preference vs. 114 left-side preference, prep ! 0.64). There was a likely to hire. Descriptions were intended to make candidates reliable association between handedness and preference for de- comparable yet distinctive. For example, one applicant for the scriptions on the left versus the right of the page by Fisher’s exact programmer position was an engineering major from Virginia test ( prep for the exact association ! 0.99; see Figure 3b). Binary Tech who programs in Perl, and the other was a math major from logistic regression showed that right-handers were more than twice Georgia Tech who programs in Python. as likely as left-handers to attribute more positive characteristics to On the other side of the questionnaire was the shopping task, people or products described on the right side of the page (OR identical in design to the job task. Participants were asked to estimate ! 2.69; 95% CI ! 1.25–5.76). These results demonstrate imagine that they were shopping for six products (dish soap, implicit body-specific preferences, corroborating the results of mattress, carpet, desk chair, family car, inflatable kiddie pool) and Experiment 4. that they had searched the Internet and narrowed their choices to A final set of analyses was conducted with the pooled data from two competing brands for each product. Now their task was to Experiments 1–5 to assess the association of handedness and decide which product to buy on the basis of brief product descrip- left–right preference overall. When all participants in the horizon- tions, which were listed in columns on the right and left of the tal condition of Experiment 1, all in Experiments 2–3, and all who product names. For example, one brand of dish soap boasted that showed a left–right preference in Experiments 4 –5 were combined it killed germs, was fresh-scented, and moisturized hands, and the (n ! 777; 110 left-handers, 667 right-handers), the results showed other that it was antibacterial, was clean smelling, and conditioned that 59% of participants made judgments consistent with the skin. For each item, participants were instructed to circle the predictions of the body-specificity hypothesis (sign test on 458 description of the product they would be more likely to buy and to with the hypothesis vs. 319 against the hypothesis, prep ! 0.99). Of cross out (with an X) the description of the product they would be the left-handers, 72% showed a Good Is Left preference (sign test less likely to buy. on 31 right side vs. 79 left side, prep ! 0.99), and of the right- For both the job and shopping tasks, participants made an equal handers, 57% showed a Good Is Right preference (sign test on 379 number of responses on the right and left of the page, one circle right side vs. 288 left side, prep ! 0.99). By Fisher’s exact test, and one cross for each pair of descriptions. The order in which they there was a highly significant association between handedness and were instructed to make circles and crosses was counterbalanced left–right preference ( prep for the exact association ! 0.99). Bi- across subjects, so preference (preferred item, dispreferred item) nary logistic regression showed that, overall, right-handers were 360 CASASANTO more than three times as likely as left-handers to associate positive periments 1–5 cannot be explained by conventions in language or responses with the right side of the page (OR estimate ! 3.35; culture. In English-speaking cultures (and many others), cultural 95% CI ! 2.15–5.21). and linguistic conventions associate the right with ideas and ac- tions that are good or allowable and the left with ideas and actions General Discussion that are bad or prohibited. Conversely, there appear to be no linguistic or cultural conventions that link the left with things that Five experiments demonstrated associations between horizontal are good.3 Both enculturation and body-specificity could poten- space and the mental representation of abstract concepts with tially account for the Good Is Right mapping, which was stronger positive and negative emotional valence. These associations dif- in right-handers, but only body-specificity can explain (a) the fered between right- and left-handers. Right-handers were more likely Good Is Left mapping found in the majority of left-handers and (b) than left-handers to associate right with positive ideas and left with the difference between right- and left-handers’ judgments. negative ideas. Left-handers were more likely than right-handers to associate left with positive ideas and right with negative ideas. Experience and Mental Simulation of Experience Right- and left-handers tended to link good things like intelligence, attractiveness, honesty, and happiness with opposite sides of left– The question of whether concepts and word meanings are em- right space, each group associating them more strongly with their bodied must be addressed on at least two time scales: How did they dominant side. By contrast, both left- and right-handers showed develop in the individual learner, and how are they instantiated the same preference to associate good things with up and bad online during thinking and language use? An answer to the former things with down. This pattern of results was predicted on the basis question concerning concepts’ experiential origins seems a prereq- of the body-specificity hypothesis and demonstrates that people uisite for the latter concerning the representational format in which with different bodies (in this case, right- and left-handers) form experiences are recapitulated online. In order to support the claim correspondingly different mental representations, even in highly that instantiating concepts requires people to create modality- abstract conceptual domains. specific mental simulations of physical actions or bodily states (see Barsalou, 1999; Feldman, 2006; Gallese & Lakoff, 2005; Gold- stone & Barsalou, 1998; Prinz, 2002) one must specify the actions Distinguishing Influences of Language, Culture, and Body or states that could serve as their basis. Previous experiments have been unable to determine whether The present data provide evidence that abstract concepts with mental metaphors from space to valence arise in the individual positive and negative valence are embodied insomuch as their learner due to correlations in bodily experience or to correlations mental representation depends, in part, on particulars of an indi- in linguistic experience. The vertical condition of Experiment 1 vidual’s body: possibly on the mere fact of having the bodily trait illustrates this difficulty. The data from this nonlinguistic task of being right- or left-handed but presumably also on correlations validate the mental metaphor Good Is Up, which is evident in between emotional states and lateralized physical actions. Thus, at many linguistic metaphors. Yet, like previous results that have least some abstract concepts have an embodied origin. It remains revealed implicit mappings from vertical space to valence (e.g., to be determined whether these concepts also have an embodied Casasanto & Dijkstra, 2009; Casasanto & Nabieva, 2007; Meier & instantiation online, in modality-specific regions of the brain. Robinson, 2004; Stepper & Strack, 1993), these data are equivocal regarding the origins of this mental metaphor. On an embodied Formation of Body-Specific Mental Metaphors interpretation, participants put the good animal on top because they have generalized associative links between bodily action and emo- What is the mechanism by which bodily differences cause tion (e.g., standing tall when we feel proud). But on an alternative body-specific mappings to emerge? It is proposed that the Good Is interpretation, they put the good animal on top because they Right and Good Is Left mental metaphors are created in right- and habitually use linguistic expressions that conflate space and va- left-handers, respectively, via correlational learning (Hebb, 1949) lence (e.g., high on life), and this linguistic experience has led to over a lifetime of lopsided perceptuomotor experience. People the creation of links between nodes representing height and hap- come to implicitly associate good things more strongly with the piness in an amodal conceptual network. Because embodied the- side of space they can interact with more fluently (their dominant ories and amodal alternatives make the same behavioral predic- side) and bad things with the side of space they interact with less tions (in this and many other cases), it is not clear to what extent fluently (their nondominant side). the mental metaphor Good Is Up arises in individuals (a) from This proposal is consonant with other experimental data linking correlations between their perceptuomotor and emotional experi- perceptuomotor fluency with positive affective judgments ences or (b) from correlations between vertical space and valence (Beilock & Holt, 2007; Oppenheimer, 2008; Reber et al., 1998; in linguistic and cultural conventions. Everyone is exposed to both Van den Bergh et al., 1990), and it makes a number of testable kinds of correlations during development, so it is likely that both predictions. For example, if asymmetries in perceptuomotor flu- direct bodily experience and linguistic/cultural experience contrib- ute to the formation of Good Is Up. But the relative contributions 3 One possible exception to the general rule that “left is bad” in linguistic of these sources of experience remain unknown. and cultural conventions is that in the United States and some European The body-specificity hypothesis allows us to distinguish the countries, liberal political views are said to be on the left of the political influences of bodily and linguistic/cultural experience unequivo- spectrum. However, the valence of this spatial–political mapping is not cally for mappings from left–right space to valence. The contrast- fixed with respect to positive or negative valence; rather, whether “polit- ing mental metaphors exhibited by left- and right-handers in Ex- ically left” is good or bad varies according to the politics of the individual.  BODY-SPECIFIC ABSTRACT CONCEPTS 361 ency give rise to body-specific associations between space and approaches to conceptual structure (e.g., Evans, 2004; Lakoff & valence, then preferences for the dominant side of space should be Johnson, 1980, 1999; Talmy, 1988). Linguistic data could be strongest in individuals who are strongly “handed” and weakest in interpreted as evidence that (a) people conceptualize goodness individuals who are ambidextrous (Oldfield, 1971). Furthermore, metaphorically in terms of rightward space, (b) the Good Is Right the development of body-specific mental metaphors in children mental metaphor is universal, both within and across cultures, and should be linked to the emergence of their handedness. Finally, if (c) Good Is Left does not exist as a mental metaphor; its existence body-specific mental metaphors from space to valence result from is made doubtful by the absence of supporting linguistic metaphors asymmetries in perceptuomotor experience (and assuming that the and is ruled out by the presence of conflicting “left is bad” idioms resulting associations remain somewhat plastic throughout the across language and cultures. But the present data challenge all lifetime), then changes in the body that necessitate systematic three of these conclusions. We must be cautious when trying to changes in the way an individual interacts with the physical infer conceptual structure from patterns in language and to dis- environment should produce corresponding changes in the strength cover conceptual universals on the basis of linguistic universals. of implicit directional preferences: A right-hander whose dominant It is possible, in principle, that people have two mental meta- hand is injured or lost should, over time, exhibit a weakening or phors from right–left space to valence, one based on patterns in even a reversal of the Good Is Right mapping. Further experiments language and culture and the other on patterns of direct bodily are needed to test these predictions. experience. If so, the two would be congruent for right-handers but An alternative possibility is that body-specific mental metaphors incongruent for left-handers (for whom language and culture as- have a purely genetic origin, and their development is not mediated sociate good with right, but bodily experience associates good with by asymmetric perceptuomotor activity. Rather, that which gives left). This makes a simple prediction: Assuming the influences of rise to handedness also gives rise to contrasting directional pref- the two metaphors on an individual’s behavior are roughly addi- erences in left- and right-handers (see the Are the Neural Sub- tive, the Good Is Right bias in right-handers should be stronger strates of Affect and Motivation Body-Specific? section below). than the Good Is Left bias in left-handers. But this prediction is not The proposal that body-specific space–valence mappings are based supported by the results of Experiments 1–5. On the contrary, entirely on genes rather than physical experience seems difficult to across experiments, the body-specific pattern tended to be stronger motivate, but it is not ruled out by the present data, which can only in left-handers than in right-handers (see the How Does “Good Is establish a correlation between handedness and preference. Impor- Left” in Lefties Compare With “Good Is Right” in Righties? tantly, the data presented here would still support the main theo- section below). These data do not provide any evidence for an retical claim of body-specificity (that people with different bodies implicit Good Is Right mental metaphor based on patterns in think differently in corresponding ways) and the main empirical language and culture that are shared by both right- and left- claim of this article (that right- and left-handers have body-specific handers. mental metaphors for valence), even if body-specific preferences Does the existence of linguistic expressions like “the right were found to be activity-independent. answer” imply the existence of the mental metaphor Good Is Right in all English speakers (including left-handers), despite the ab- Relationships Between Mental Metaphors and Linguistic sence of evidence in the present data? Not necessarily. Expressions Metaphors linking right with “good” do not appear to be metaphors but rather polysemies or frozen idioms (Keysar & Bly, 1995; Murphy, 1996, Why is the right side conventionally the good side across many 1997). The fact that speakers understand “the right answer” does languages and cultures? The fact that left- and right-handers ad- not entail that they do so by activating associative mappings from here to the same linguistic and cultural conventions may appear, at the source domain of horizontal space; right in this sense may be first, to argue against the possibility that these conventions have a understood nonmetaphorically. As such, left-handers should un- bodily origin. In principle, linguistic and cultural conventions derstand these linguistic expressions as easily as do right-handers, could always reflect the body-specific mental metaphors of their despite the difference in their implicit mental metaphors. users, but this is not the case: Left-handers do not greet others with Productivity and systematicity are hallmarks of the kinds of left-handed handshakes or refer to a correct response as “the left linguistic expressions that are understood via mental metaphors answer.” Consider the confusion that could ensue if this were the (Lakoff & Johnson, 1980, 1999). These traits are evident in lin- case: When someone told you that your answer was “right,” you guistic metaphors that associate vertical space and valence. The would need to find out whether the speaker was right- or left- idea “he got happier” can be expressed with a conventional met- handed in order to determine whether your answer was correct or aphor like “his spirits rose” but also with a variety of related incorrect! A tacit agreement among members of a community to metaphors that are likely to be understood effortlessly even if they conform to linguistic and cultural conventions that reflect the are novel (e.g., ”his spirits soared,” “sailed,” “climbed,” “skyrock- implicit body-specific mental metaphors of the majority avoids eted”). Furthermore, these expressions linking up with “happy” are confusion and facilitates communication. Right-handers predomi- systematically related to expressions linking down with “unhappy” nate in the population worldwide. Thus, the finding of a body- (e.g., “his spirits sank,” “plunged,” “plummeted,” “nosedived”). specific preference for the right in right-handers provides a poten- Degrees of happiness or sadness can be expressed as locations tial explanation for the prevalence of the Good Is Right mapping along a vertically spatialized happiness continuum. across languages and cultures. By contrast, linguistic expressions that associate horizontal Good Is Left is a mental metaphor without any corresponding space and valence tend to be unproductive and unsystematic. The linguistic metaphors. This discovery, coupled with the ubiquity of idiom “I’ve got two left feet” means “I’m clumsy,” but this the Good Is Right mapping, has implications for language-based meaning is lost if the expression is varied even slightly (e.g., ”I’ve 362 CASASANTO got two left soles,” “two left ankles,” “two left toes,” “two left the leftmost (Dehaene, Bossini, & Giraux, 1993). Linguistic ex- shoes”). Furthermore, having two left feet does not indicate a pressions like “the prime example” conflate primacy with good- position along a horizontally spatialized clumsiness continuum. ness (i.e., this phrase can mean the first example, the best example, “I’ve got two right feet” does not mean “I’m graceful.” Likewise, or both). Speakers of languages like English may be predisposed to “the left answer” does not mean “the wrong answer.” “My right- consider the leftmost item to be the first and therefore the best. This hand man” means “a close associate,” but “my left-hand man” metaphorical blend of left, first, and best should result in a cultur- does not mean “a distant associate”; “out in left field” means ally constructed Good Is Left bias in all participants, which would “eccentric,” but “out in right field” does not mean “mainstream.” reinforce the body-specific Good Is Left bias in left-handers but If space is used unsystematically as in these right–left idioms, then work against the body-specific Good Is Right bias in right-handers. spatial relations cannot be used to support inferences about relative goodness or badness. As such, it is unlikely that these linguistic Body-Specific Interpretation of Previous Findings expressions are understood metaphorically via mappings from space to valence. Studies from disparate literatures have reported associations of Right–left idioms in language and culture that express ideas direction, preference, and handedness in specialized domains such with positive and negative valence may be related to body-specific as spatial navigation, music perception, and facial emotion pro- mental metaphors in right-handers. But based on the nature of cessing. Body-specificity may provide a unifying account of these these idioms in English and Dutch, and on the experimental data data. Patrons in American museums reportedly show a strong bias presented here, it appears that this relationship is largely historical. to turn right when they enter a gallery (Robinson, 1933). This is This is true, at least, for English speakers in California and Dutch true in spite of museum curators’ tendency to create exhibits that speakers in the Netherlands. In cultures with strong left-hand begin on the left and proceed rightward, consistent with the direc- taboos, where left–right valence mappings are more salient, these tion of reading and writing in English. Scharine and McBeath cultural conventions could give rise to active mental metaphors (2002) tested two explanations for the right-turn bias: handedness (Kövecses, 2005). If so, it may be possible to observe additive and driving experience. Right- and left-handed American partici- relationships between culture-specific and body-specific map- pants (who drive on the right side of the road) and English pings. For example, left-handers in Islamic cultures might show an participants (who drive on the left) performed a simple T maze. attenuation of the body-specific preference for the left (which They were asked to walk down an aisle formed by two parallel sets conflicts with the strong culture-specific Bad Is Left metaphor), of bookshelves in a university library and to find the target (a and right-handers an accentuation of the body-specific preference sticky note) hidden at the end of one of the bookshelves. There for the right (which is supported by the culturally constructed were significant effects of driving experience: The Americans Good Is Right metaphor). were more likely to turn right at the end of the bookshelves, and the English to turn left. But the strongest predictor of turning How Does “Good Is Left” in Lefties Compare With preference was handedness: Both American and British partici- “Good Is Right” in Righties? pants were biased to search for the target by turning toward their dominant side. Scharine and McBeath characterized this finding as Across all five experiments, there was a tendency for the Good a “locomotive directional bias” (2002, p. 249), which may be Is Left mapping in left-handers to be stronger than the Good Is limited to spatial navigation. Alternatively, this turning bias may Right mapping in right-handers. The strength of the body-specific be one instance of a more general body-specific preference for preference differed significantly between groups only in Experi- things located on one’s dominant side of space. ment 5 (Fisher’s exact test for the difference between the propor- Affective judgments about music also depend, in part, on loca- tions of body-specific responses in left- and right-handers, prep ! tion and handedness. McFarland and Kennison (1989) played 0.99) and in the combined data from all participants who showed right- and left-handers excerpts from instrumental musical pieces a left–right bias in Experiments 1–5 ( prep ! 0.99). This finding that evoke positive or negative emotions. Excerpts were presented was unexpected and should be interpreted with caution for a through a single headphone placed over the participant’s right or combination of reasons, foremost of which is that this unpredicted left ear. Right-handers reported experiencing more positive emo- trend was nonsignificant in four out of five experiments. Further- tions when music was presented on the right; left-handers showed more, the large difference in the number of left- and right-handed the opposite pattern, reporting more positive emotions when music participants complicates comparison of the strength of body- was presented on the left. specific preferences across groups. Studies of emotional face processing show compatible results. Two potential explanations for the observed pattern merit fur- In one experiment, right-handers tended to judge faces as express- ther investigation. First, asymmetries in perceptuomotor experi- ing a more negative emotion when they appeared briefly on the left ence may be more salient for left-handers, who are habitually side, consistent with the Good Is Right mapping (Natale, Gur, & inconvenienced by customs and devices designed for right- Gur, 1983). In another experiment, left-handers rated faces as handers. More salient perceptuomotor asymmetries could result in expressing a more positive emotion when they appeared on the left stronger associations between actions with the dominant hand and of the screen and a more negative emotion when they appeared on experiences with positive emotional valence. Second, culture- the right, consistent with the Good Is Left mappings (Everhart, specific mental metaphors may be interacting with the body- Harrison, & Crews, 1996). specific left–right spatialization of valence through a blending of It appears that no domain-general Good Is Left preference has three metaphorical mappings. According to the implicit mental been reported previously, but a Good Is Right preference consis- number line in English speakers, the first in a series (or a pair) is tent with the present results was discovered serendipitously by  BODY-SPECIFIC ABSTRACT CONCEPTS 363 Wilson and Nisbett (1978). They asked participants to evaluate the comes of actions with our nondominant hands, and (c) defensive quality of four nylon stockings that, unbeknownst to them, were actions. identical in all but one respect: Each stocking had been treated First, if we use our nondominant hand to push away things that with a different scent. The stockings were hung side-by-side on a we don’t want, this leaves the dominant hand free to grasp and rack with their left to right order counterbalanced across partici- manipulate things we that do. Imagine picking an apple or shelling pants. Scent had no effect on participants’ evaluations, but to the a nut. To separate an apple from the tree, we’re likely to hold the experimenters’ surprise, the left to right position did: The farther branch with our nondominant hand (pushing it away) while pick- an item was to the right of the row, the more likely it was to be ing the fruit with our dominant hand (pulling it toward us). selected as superior. Nisbett and Wilson did not know the cause of Likewise, after we crack a nut, we’re likely to hold the shell the rightward bias but speculated that it was a temporal order effect (which we don’t want) in our nondominant hand while picking out (since participants generally inspected the stockings serially from the meat with our dominant hand. left to right). While temporal order may have been one influence, Second, even though we prefer to use our dominant hand for the present data suggest an alternative account: Perhaps partici- manipulating objects, sometimes we end up using our nondomi- pants preferred stockings on their dominant side. Given the pre- nant hand due to functional constraints: when the object is located dominance of right-handers in the population, most of Nisbett and on our nondominant side or when our dominant hand is busy. The Wilson’s participants were probably right-handed, and so on av- outcome is often frustrating. Such experiences could discourage erage they would have preferred items on the right. approach behaviors with our nondominant hands and help to establish an association between our nondominant side and nega- tive outcomes. Are the Neural Substrates of Affect and Motivation Body- Third, if we use our nondominant hand to fend off attack (an Specific? avoidance behavior), this leaves our dominant hand free to retaliate (an approach behavior) or to perform more complicated defensive The results of Experiments 1–5 suggest a possible reinterpreta- actions. Sword fighters in previous centuries raised the shield with tion of a well-established pattern in cognitive–affective neuro- the nondominant hand while wielding the sword offensively with science. Dozens of experiments have investigated the hemispheric the dominant hand (Gould, 1908, in Scharine & McBeath, 2002). laterality of emotional valence and of approach and avoidance Consistent with this proposal, Coren (1992) demonstrated a natural behavior. Approach-related behavior (correlated with positive va- startle response in which right-handers raised their left hand higher lence) has been lateralized to the left hemisphere, and avoidance- than their right, and left-handers raised their right hand higher than related behavior (correlated with negative valence) to the right their left, both groups shielding their faces with the nondominant hemisphere, particularly in the frontal lobes (e.g., Ahern & hand in response to perceived danger. Schwartz, 1979; Cretenet & Dru, 2004; Davidson, 1992; Davidson If the hemispheric laterality of affective-motivational systems et al., 1990; Davidson & Fox, 1982; Dimond, Farrington, & correlates with manual action tendencies, it should reverse be- Johnson, 1976; Killgore & Yurgelun-Todd, 2007; Maxwell & tween right- and left-handers. This proposal is speculative, but it is Davidson, 2007; Natale, Gur, & Gur, 1983; Schiff & Bassel, consistent with the music-processing and face-perception studies 1996).4 This pattern of hemispheric specialization has achieved the reviewed above. Due to the crossing of auditory and visual path- status of scientific fact. Yet, most of the studies that demonstrate ways, stimuli were processed initially by the hemisphere contralat- the pattern have exclusively tested right-handed participants. Thus, eral to the side on which they were presented. Thus, positive the established laterality of positive/approach and negative/ judgments could have resulted from initial processing by the right avoidance behaviors may obtain only (or primarily) for right- handers. hemisphere in left-handers and by the left hemisphere in right- On a body-specific interpretation of the available data, perhaps handers, consistent with a body-specific reversal of hemispheric positive emotions and approach behaviors are not mediated by the specialization for positive and negative affect. left frontal lobe, universally; rather, perhaps they are mediated by Although suggestive, these studies do not conclusively establish the frontal lobe that controls the dominant hand. From early whether hemispheric specialization for positive and negative af- childhood, we tend