Psychonomic Bulletin &Review1999, 6 (4),668-676Aptness predicts preference formetaphors or similes, as well as recall biasDAN L. CHIAPPEUniversity ofBritish Columbia, Vancouver, British Columbia, CanadaandJOHN M. KENNEDYUniversity of Toronto, Scarborough, Ontario, CanadaWhymight we sometimes prefer a metaphor such as "genes are blueprints" to a simile such as "genesarelike blueprints"? One possibility is that metaphors arepreferred when thecomparison between atenor (e.g, genes) anda vehicle(e.g., blueprints) seems especially apt. That is, metaphors might beusedwhenthecomparison captures manysalient featuresof thetenor in question.Thepresent ex-periments examined the relation between the aptness of comparisons and people's preferences for ex-pressing those comparisons as metaphors or as similes, In Experiment 1, it was found that there is con-sensus on how toexpress particular comparisons. In Experiment 2, it was foundthat this preferencecan be predicted from theaptness of a comparison, It wasalso foundthat aptness canpredict errorsin therecall of comparisons. These findingshaveimplications for theories of metaphor.Metaphors andsimiles aredistinct butclosely relatedliterary devices, or tropes. Metaphors, such as "rumors areweeds," assert an identity between twodomains. To useI.A. Richards's (1936) terms, theyassert an identity be-tween the tenor andthevehicle.The tenor is thesubjectof thestatement ("rumors"in "rumors areweeds"),andthevehicleistheconceptthatweareusinginordertosaysomething new about thetenor ("weeds" in "rumorsareweeds"). Similes, suchas"rumorsarelikeweeds,"assert a similarity between thetenor and the vehicle, notanidentity relation,Therelation between metaphors andsimiles haslongbeen thesubject of debate. Whatissureis that, despitetheir differences, metaphors andsimiles shareimportantfunctions, Certainly, bothtropesareusedtoconveyin-formationabout their topics. Abrief vehicleoftenactsas a handy summary, as when onesays "In short, thisar-ticleis a goldmine." Furthermore,it is often argued thatmetaphors andsimiles might beusedwhenpeople wanttoconveyinformationabout atopic that cannotbeex-plicitly stated inliterallanguage(Fogelin, 1988;Gibbs,1994; Ortony,1975). For example, a person might use themetaphor "my job is a jail" orthesimile "my job islikeThis research was made possible by NSERC Grant OGP9947 to1.M.K.The authors thank Penny Chiappe and Chang Hong Liu for their adviceonstatisticalmatters. TheyalsothankHeather Bensonforherassis-tancein datacollectionandHiroshi Onoand graduate students fromYork University for their participation in Experiment I. Finally, the au-thors thankPsychonomicBulletin&Reviewreferees foruseful sug-gestions, including a helpful argument about nonsensicalcomparisons.Correspondenceconcerningthisarticleshouldbeaddressedto 1. M.Kennedy, University of Toronto, 1265 Military Trail, Scarborough, ONMICIA4, Canada (e-mail:kennedy@lake.scar.utoronto.ca).ajail" to convey that their job possesses an intricate systemof properties that is difficult to stateexplicitly (Glucks-berg& Keysar, 1993).Inthispaper, weaskaboutafactor thatcould playamajor roleindeterminingwhether acomparisonisex-pressed asa metaphor orasasimile. Forexample, whymight a person express a comparison between genes andblueprintswiththemetaphor"genesareblueprints"ratherthanwiththesimile"genesarelikeblueprints"?And, why might a person express a comparison betweentreetrunks and straws with the simile "treetrunks are likestraws"ratherthanwiththemetaphor "treetrunksarestraws"?We examinewhetherthepreferencefor oneform ofa comparison is simply idiosyncratic and whetherthechoice reflectsimportant differencesbetween meta-phors and similes.Before turning to our major hypothesis, let us acknowl-edgethatthereis a roleplayedhereby custom. Throughuse, certaincomparisonsbecomemore familiartopeo-ple as metaphors oras similes. As a result, people mightchoose to express thecomparisons in their more familiarform. For example, Robert Bums's comparison "my loveis likea red, redrose" is most familiar asasimile.As aresult, followers of Bums might strongly prefer to expressthiscomparisonasasimile rather thanasthemetaphor"my love is a red, red rose." Likewise, the metaphor "Godis love" is quitefamiliar in our culture, "God is like love"is quite unfamiliar and, to many, may seem peculiar, Hence,many mightchoose to express the comparison as a meta-phor, Inshort, oneformof expressionmaybecomeen-trenched in a language and, thereby, affect how a compar-isonis subsequently expressed,Familiarity, however, cannot account forhow all com-parisonsareexpressed, becausemanycomparisonsareCopyright1999 Psychonomic Society, Inc. 668created denovo. Somefactor orfactors other thanfa-miliaritymayhavedeterminedhowacomparisonwasexpressedin thefirst place. But what?In the present experiments, aptness was examined. Wefollowed muchof the recent literatureonmetaphor anddefined aptness as the extent to which a comparison cap-tures salientfeatures ofthe topic in question (Katz, 1989,1992; Malgady&Johnson, 1976; Tourangeau &Stern-berg,1981, 1982; Trick &Katz,1986). Consider, for ex-ample,"evolutionis a lottery" and "evolution is a blindwatchmaker." In both of these comparisons, the tenor andthe vehicle share relevant features. Evolution involves ran-domnessand lack of foresight. Also,evolution involvesthe emergence of admirablestructures. But one of thecomparisons capturesmorefeaturesthantheother, andfor some purposes, it may be necessary to draw attentionto many featuresof the tenor. The comparison that cap-turesthe appropriatenumberof relevant featurescouldbe considered especially apt., It may be helpful if we expand on our example to makeourpoint clear. Formanyreaders of recentargumentsabout evolution, "evolution is a lottery" might not appearvery apt, because the comparison fails to capture the con-structiveaspect of evolution(Dawkins, 1986). In con-trast, "evolution is a blind watchmaker" capturesboth theprocess andtheproduct. Therefore, itcould appeartomany to be a moreapt comparison.The difference between"lotteries" and"blindwatch-makers" is largely one of content. Would it make any dif-ference if the formof expression of the comparison wasvaried?Is "evolution is like a lottery" to be preferred to"evolutionis a lottery"? lf so, is the preference related tothe apparentaptness of the comparison? For example,ifa comparison is deemed veryappropriate, are people morelikelyto judgethat itis best expressedasa metaphor?And if a comparison is deemednot very appropriate, arepeople morelikely to judge that it is bestexpressed as asimile?In thetraditional theoryof metaphorthat hascomedown to us from Aristotle (1926), we findlittle guidanceonthepreferenceforoneformof expressionoveran-other. In particular, Aristotle'stheoryoffersnopredic-tion about the relation between the aptness of a compar-ison and any tendency to express it as a metaphor or as asimile. If anything, theAristoteliantheoryclaims thatthere is no such relation. For Aristotle, metaphors are sim-iles with the term ofcomparison-the like-left out (mod-ern proponents ofthis view include Fogelin, 1988; Gent-ner, 1983, 1989; Gentner &Clement, 1988; Gentner &Markman, 1997; Miller, 1979;Ortony, 1979a, 1979b;and Tversky,1977). In this view, the only differencebe-tween a metaphor and a simileis thatthemetaphor is amore compact statement. As Aristotlesaid, "[The]sim-ile .. , is a metaphor differingonlyby theadditionof aword" (p. 397). As a result, one's preference for express-ing a comparison as a metaphoror as a simileshould bedeterminedby whether or notonewantstobe concise.Thereis noexplicitbasisin theAristoteliantheoryforMETAPHORSANDSIMILES 669arguing that the preference should be guidedby the apt-ness of the comparison.Incontrast, recent theoristshavecontendedthat theform of the comparison is morethana matter of brevity(Black, 1979; Morgan, 1979).For example, Glucksbergand Keysar (1990, 1993) claim that there is a relation be-tweenaperson'sbeliefsandtheforminwhicha com-parisonis expressed. Asanexample, theycomparethesimile"my jobis like a jail" to themetaphor"my job isajail." They say, "When the metaphorical form is chosenover the simileform, the very choice is communicative.If onechoosesthesimile, thenthehearerpresumablyrecognizes that choice and would therefore infer that thespeaker was not attributing all the properties ofthe class'jail'to 'my job,'but only some ofthem"(1993, p. 413).Alternatively, if a person chooses the metaphor, "thein-tention is for the hearer to understand that the job in ques-tion has all the properties of the attributivecategory thatis called 'jail'" (1993, p. 413). Hence, accordingto Glucks-bergandKeysar (1990, 1993),similesareusedwhenacomparison is not deemed to be what we have called veryapt; when a comparison is very apt, the metaphorical formis preferred. As Roberts and Kreuz (1994) say, followingGlucksberg and Keysar, "a simile may aid a listener in de-termining that a comparison is notintendedquite as se-riously" (p. 162).Inshort, theliteratureisdivided. Proponentsof theAristoteliantheory(including, amongothers, Fogelin,1988, and Miller,1979) do not offer groundsfor predict-inga relationbetweenaptness andtheformin whichacomparisonisexpressed, but GlucksbergandKeysar(1990, 1993) andRobertsandKreuz(1994)do.Thepresent experimentstestedtherelationbetweenaptness andthepreferredform of comparisons. InEx-periment 1, we examined whether there is consensus be-tweenpeopleon how a particular comparison shouldbeexpressed. This was donebypresentingsubjectswithpairs of metaphorsand similes and askingthem to indi-catewhichwas thebest way of expressingeachof thecomparisons.In Experiment 2, we explored two issues. First, we ex-aminedwhether thepreferenceforthemetaphor orthesimile form of comparisons found in Experiment I couldbepredicted fromthe aptness of thosecomparisons, asratedby a separategroup of subjects. Second,we exan.-inedwhethertheaptnessof comparisonscouldpredictperformanceonacued-recall task. A cued-recall taskwas usedas an indirectmeasure of people's preferencesfor expressing comparisons asmetaphorsor as similes.Thisisbecause, asSchacter(1995) says, "informationavailablein the present retrieval environment combineswithstoredinformation to yieldan emergent patternofactivitythat we experience as'memory" (p. 24). Thus,recall isusuallyinfluencedbyconstructiveprocesses(Bartlett, 1932; Neisser, 1967). Any preference foronetype of comparison could therefore play a role during re-call and couldbe evident in the form of a bias to recall acomparisonasa metaphor orasasimile, regardlessof670 CHIAPPEANDKENNEDYthe formin which oneinitiallysaw the comparison. Forexample, if a comparisonbetweensmilesandrazors isnot veryapt, peoplemight bemorelikelytomisrecall"smilesarerazors" as"smiles arelike razors"thanthereverse. However, if acomparisonbetweengenesandblueprints is especially apt,people might be morelikelytomisrecall "genesarelikeblueprints"as"genesareblueprints" than the reverse.EXPERIMENT 1In Experiment 1, we examined whether there is a con-sensusbetweenindividuals on whetherparticular com-parisons are best expressed as metaphors or as similes. Ispreference merelyidiosyncratic? Ifnot, what is thepre-ferred form? The subjects were shown the metaphor andthe simile forms of severalcomparisons and were askedto indicate which formbestexpressed the comparison.MethodSubjects. Thirtysubjects(24femaleand6 male), withameanageof 23years, participatedinthisexperiment. Therewere twogroups,to provide a check on the replicability of the results.Twenty-onesubjectswerevolunteers froma third-year courseinpsychol-ogy at the University ofTorontoat Scarborough. Nine subjects weregraduatestudentvolunteersfromYork University. The subjectswere testedeitherindividually or in pairs.Stimuli and Apparatus.Thesubjects wereshown30compar-isonsin boththeirmetaphor andtheirsimileforms. Thisyielded atotalof 60statements. Thestatements-listedinTable I-weredrawn from varioussources in the psychological literature on figu-rative representation. Itemswerepresented in booklets thathadin-structionsonthefront page andonepagewith30pairsof meta-phorsandsimiles. Thus, eachmetaphorwas presentedwithitscorresponding simile. Theorder of themetaphors andsimiles wasrandomized, so thathalf of thetimethemetaphors appearedfirst,and half ofthe time the similes appeared first. The number" I"waswrittenbeside the comparisons appearing first, and the number "2"beside the ones appearing second. The order of the 30 comparisonswas also randomized, withonerandomorder for eachsubject.Procedure. The subjects weregiven the booklets containing the30pairsof metaphorsandsimiles. Thesubjects weretoldtoreadeachpairof statementsandthendecidewhether themetaphororthe simile best expressed the comparison in question. For example,theywereaskedwhether acomparison betweenpriestsandshep-herdsis bestexpressed by themetaphor "priests are shepherds" orby thesimile"priests arelikeshepherds." Thesubjects respondedby circling the number written beside the statement of their choice.Results and DiscussionThis experiment used two groups: undergraduates andgraduate students,with eachgroup drawn from a differ-ent institution. Did thetwogroupsrespondin a similarway? Weexamined whether the preferences of one groupcould predict the preferences ofthe other group. The anal-ysisrevealeda correlation of. 78 (p