Why do metaphors seem deeper than similes?...conservative in using metaphors than in using similes....

6
Zharikov,S .,&Gentner,D .(2002) .Whydometaphorsseemdeeperthansimiles?In W .D .Gray&C .D . Schunn (Eds .), ProceedingsoftheTwenty-FourthAnnualConferenceoftheCognitiveScienceSociety (pp .976-981) .Fairfax,VA :GeorgeMasonUniversity . Whydometaphorsseemdeeperthansimiles? Sergey S. Zharikov(s-zharikovt®n orthwestern .edu) NorthwesternUniversity,DepartmentofPsychology,2029SheridanRoad Evanston,IL60208USA DedreGentner(gentner® northwestern.edu) NorthwesternUniversity,DepartmentofPsychology,2029SheridanRoad Evanston,IL60208USA Abstract Figurativeexpressionsinmetaphorform (e.g., Marriage isajourney) seemstrongeranddeeperthanexpressions insimilefoam (e .g., Marriageislikeajourney) . Weran astudytoexaminethenatureofthesejudgments. Participantsreadshortparagraphsdescribing either objectattributesorrelational structure andthenmadea forcedchoiceofthegrammaticalformofafigurative expressionmentioningthetargetconceptreferredtoin thepassages.Theresultsshowedthatthemetaphorform waschosenmoreoften(I)forexpressions with conventionalbases,and(2)whenfigurativestatements followedcontextscontainingrelationalinformation .We speculate about a possible linkage between conventionalizationandrelationality . Introduction Nominalfigurativestatementscanbeexpressedintwo ways - insimileform (Xislike1) andinmetaphor form(Xis1) .Althoughthetwogrammaticalforms . largelyservethesamepurpose(showingthatoneentity isfigurativelysimilar-toanother),peoplereportthat expressionsinmetaphorformfeelmoreprofoundand expressstrongerclaimsthanexpressionsinsimileform . Forexample,saying Her heartis a stone feelsdeeper than Her heartislikeastone . Further,severalstudies (Gibb&Wales,1990 ;Bowdle, 1998 ;Aisenman,1999) havefoundthatifpeopleareaskedtomakeachoice betweenanexpressioninmetaphorformorthesame expressioninsimileform,thesimileformischosen more often . Itappears that people are more conservativeinusingmetaphorsthaninusingsimiles . Thegreaterforceofthemetaphoricformwasnoted byGlucksbergandKeysar(1990),whoproposedthat themetaphorformisthebasicformoffigurative statementsandthatsimilesareunderstoodasvariantsof metaphors .Notingthat thegrammaticalformof metaphors matches that ofcategory inclusion statements,theysuggested thatmetaphorsinfact functionascategoryinclusionstatements,wherethe categoryisanabstractionthatcanbeaccessed or createdfromthemetaphor'sbaseconcept .(Wewilluse theterms target and base, respectively,fortheXandY 976 terms,ofafigurativeexpression Xis[Tike]Y .) There hasbeendebateconcerningtheprocessingimplications ofthistheory,butforourpurposesthekeypointis GlucksbergandKeysar'sinsightthatthegrammatical formoffigurativestatementshaspsychologicalforce, withmetaphorbeingthestronger,morecategorical form.Thispaperexamines thereasonsforthis phenomenon . Tworecent theories haveproposed different explanationsforthesimile-metaphordifference . One accountsinglesouttheconventionalityofthe base termtheother,therelationality/attributionalityofthe metaphor'sinterpretation .Thefirstaccount,theCareer ofMetaphorhypothesis(Bowdle&Gentner, 1999; Gentner&Bowdle,2001)suggeststhatthedifference liesintheconventionalityofthebaseterm :figuratives withconventionalbasesareexpressedasmetaphors, andthosewithnovelbasesareexpressedassimiles . Thesecondaccount,Aisenman's(1999)Relational Precedencehypothesis,suggeststhatthedifferenceis duetothekindofinterpretationtheexpressionreceives: relationalinterpretationsarestatedasmetaphors,and attributionalinterpretationsarestatedassimiles . . Intheirresearchonmetaphorprocessing,Gentner andWolff(1999)proposedanimportantdistinction betweennewlymintedfigurativesandconventionalized figuratives .AccordingtotheCareerofMetaphor hypothesis,figurativeswithnovelbases,suchas An encyclopediais(like)auraniummine, areprocessedby comparisonbetweenthetargetandtheliteralmeaning ofthebase .Incontrast,figurativeswithconventional bases,suchas Anencyclopediaisagoldpae, canbe processedbyalignmentwithaconventionalabstraction (e.g., asourceofsomethingvaluable) associatedwith thebaseterm.Thekeydifferencebetweennoveland conventionalbasesis thattherepresentations of conventionalbasesincludeasecondarymetaphoric meaningalongwiththeoriginalliteralmeaning .They havebecomepolysemous .Incontrast,representations ofnovelbasescontainonlyaliteralmeaning . Gentner and Wolff (1997) proposed that conventionalmetaphoricmeaningsarecreated over timeasaresultofrepeatedcomparisonsofdifferent

Transcript of Why do metaphors seem deeper than similes?...conservative in using metaphors than in using similes....

Page 1: Why do metaphors seem deeper than similes?...conservative in using metaphors than in using similes. The greater force of the metaphoric form was noted by Glucksberg and Keysar (1990),

Zharikov, S ., & Gentner, D . (2002). Why do metaphors seem deeper than similes? In W. D. Gray & C. D .Schunn (Eds .), Proceedings of the Twenty-Fourth Annual Conference of the Cognitive Science Society(pp. 976-981) . Fairfax, VA : George Mason University.

Why do metaphors seem deeper than similes?

Sergey S. Zharikov (s-zharikov t®northwestern .edu)Northwestern University, Department of Psychology, 2029 Sheridan Road

Evanston, IL 60208 USA

Dedre Gentner (gentner®northwestern.edu)Northwestern University, Department of Psychology, 2029 Sheridan Road

Evanston, IL 60208 USA

Abstract

Figurative expressions in metaphor form (e .g., Marriageis a journey) seem stronger and deeper than expressionsin simile foam (e .g., Marriage is like a journey) . We rana study to examine the nature of these judgments.Participants read short paragraphs describing eitherobject attributes or relational structure and then made aforced choice of the grammatical form of a figurativeexpression mentioning the target concept referred to inthe passages. The results showed that the metaphor formwas chosen more often (I) for expressions withconventional bases, and (2) when figurative statementsfollowed contexts containing relational information . Wespeculate about a possible linkage betweenconventionalization and relationality .

IntroductionNominal figurative statements can be expressed in twoways - in simile form (X is like 1) and in metaphorform (X is 1). Although the two grammatical forms .largely serve the same purpose (showing that one entityis figuratively similar- to another), people report thatexpressions in metaphor form feel more profound andexpress stronger claims than expressions in simile form .For example, saying Her heart is a stone feels deeperthan Her heart is like a stone. Further, several studies(Gibb & Wales, 1990; Bowdle, 1998; Aisenman, 1999)have found that if people are asked to make a choicebetween an expression in metaphor form or the sameexpression in simile form, the simile form is chosenmore often . It appears that people are moreconservative in using metaphors than in using similes .

The greater force of the metaphoric form was notedby Glucksberg and Keysar (1990), who proposed thatthe metaphor form is the basic form of figurativestatements and that similes are understood as variants ofmetaphors. Noting that the grammatical form ofmetaphors matches that of category inclusionstatements, they suggested that metaphors in factfunction as category inclusion statements, where thecategory is an abstraction that can be accessed orcreated from the metaphor's base concept . (We will usethe terms target and base, respectively, for the X and Y

976

terms, of a figurative expression X is [Tike] Y.) Therehas been debate concerning the processing implicationsof this theory, but for our purposes the key point isGlucksberg and Keysar's insight that the grammaticalform of figurative statements has psychological force,with metaphor being the stronger, more categoricalform. This paper examines the reasons for thisphenomenon .

Two recent theories have proposed differentexplanations for the simile-metaphor difference. Oneaccount singles out the conventionality of the baseterm the other, the relationality/attributionality of themetaphor's interpretation . The first account, the Careerof Metaphor hypothesis (Bowdle & Gentner, 1999;Gentner & Bowdle, 2001) suggests that the differencelies in the conventionality of the base term : figurativeswith conventional bases are expressed as metaphors,and those with novel bases are expressed as similes .The second account, Aisenman's (1999) RelationalPrecedence hypothesis, suggests that the difference isdue to the kind of interpretation the expression receives:relational interpretations are stated as metaphors, andattributional interpretations are stated as similes . .In their research on metaphor processing, Gentner

and Wolff (1999) proposed an important distinctionbetween newly minted figuratives and conventionalizedfiguratives . According to the Career of Metaphorhypothesis, figuratives with novel bases, such as Anencyclopedia is (like) a uranium mine, are processed bycomparison between the target and the literal meaningof the base. In contrast, figuratives with conventionalbases, such as An encyclopedia is a gold pae, can beprocessed by alignment with a conventional abstraction(e.g., a source of something valuable) associated withthe base term. The key difference between novel andconventional bases is that the representations ofconventional bases include a secondary metaphoricmeaning along with the original literal meaning. Theyhave become polysemous . In contrast, representationsof novel bases contain only a literal meaning .

Gentner and Wolff (1997) proposed thatconventional metaphoric meanings are created overtime as a result of repeated comparisons of different

Page 2: Why do metaphors seem deeper than similes?...conservative in using metaphors than in using similes. The greater force of the metaphoric form was noted by Glucksberg and Keysar (1990),

target terms with the same base . The idea is thatthrough progressive alignments of the base, a set ofproperties or a relational schema belonging to the baseemerges as a separable abstraction. This can become anadditional word sense - a kind of metaphoric categoryassociated with the base.Bowdle's Grammatical Concordance principle links

the Career of Metaphor hypothesis with the simile-metaphor distinction. It states that metaphoricexpressions are interpreted by the process of structuralalignment (Gentner & Markman, 1997), but the natureof the invited alignment differs for metaphors andsimiles . The simile form invites directly aligning theliteral base and target concepts (e.g., encyclopedia andgold mine in the above example), whereas the metaphorform suggests that the listener should first access theabstraction associated with the base - e.g ., source ofsomething valuable-- and then align it with the targetrepresentation. Consistent with this explanation,Gentner and Bowdle (2001) found that novel metaphorsare slow to process. This follows from the claim thatsuch statements lead to a false start in processing . Forexample, hearing That encyclopedia is a uranium mineis infelicitous, because there is no conventionalabstraction associated with uranium mines.

Thus, the claim is that (1) repeated alignments canlead to the formation of an abstraction, and (2)figurative statements can occur in metaphor form only,when there is existing abstraction (or metaphoricalcategory) associated with the base. Perhaps the moststriking evidence for this claim is Bowdle's (1998)study showing 'in vitro' conventionalization. Afterseeing novel bases in parallel comparisons with threetarget terms in simile form, subjects preferred toexpress further statements involving that base inmetaphor form. They also (mis) recalled the statementsthey had seen as having been in metaphor form .Gentner and Bowdle (2001) found that as figurativestatements became increasingly conventional, there is ashift in people's preference from the simile form to themetaphor form.

A second explanation for the subjective differences inperception of similes and metaphors was recentlyoffered by Aisenman (1999) . She extended Gentner's(1988; Gentner & Clement, 1988) distinction betweenattributional and relational comparisons and suggestedthat people primarily use the metaphor form tohighlight common relations between the base andtarget, and the simile form to highlight commonattributes (Aisenman, 1999) . Thus, the metaphor formis likely to convey a deep common system of relations .This theory fits well with the intuition that metaphorsoften seem more profound than similes. In her study,Aisenman presented subjects with base and target termsand asked whether they would be more likely to putsentences with those terms in simile or metaphor form.

977

When the base and target shared mostly surfaceattributes (e .g., The sun is (like) an orange - both areround and orange), participants preferred to statesentences in simile form. When the base and targetshared common relational structure (e.g., Television is(like) a magnet - both attract), participants were morelikely to use the metaphor form. Aisenman's resultssuggest that the metaphor form is preferred forrelational commonalities .

There are thus two accounts for form differences infigurative language : metaphors tend to be preferredover similes (a) when the base is conventional or (b)when the interpretation is relational. To compare theseaccounts, we varied both factors - conventionality ofthe base and the type of commonalities between thebase and target - and obtained people's preferences forstating figurative expressions' in simile or metaphorform.

Experiment 1. Context PrimingWe selected 20 metaphors from prior metaphor studies(Ortony, 1979; Gentner & Clement, 1988 ; Aisenman,1999). The metaphors used were classified as doublemetaphors (Gennner & Clement, 1988) in that theypermitted both attributional and relationalinterpretations. We presented subjects with shortparagraphs describing the target, focusing either on itsattributes or on its relational structure. Examples ofrelational and attributional contexts are listed in Table1. Then, participants were asked to choose which of thetwo figurative sentences they preferred . Both sentencesfeatured the target coupled with the same base anddiffered only in that one of them was a simile and onewas a metaphor. Half the bases were novel, and halfwere conventional. Conventionality of the base wasoperationalized as having the metaphoric meaninglisted in the Merriam-Webster Collegiate dictionary.The base terms. never appeared in the contextspreceding the simile and metaphor statements . Table 1shows a sample stimulus with a conventional base .

The Career of Metaphor account predicts that peoplewould be more likely to prefer the metaphor form forstatements with conventional rather than novel bases .Aisenman's Relational Precedence hypothesis predictsthat people would be more likely to prefer the metaphorform when given the paragraph priming the relationalinterpretation .

MethodSixty-four Northwestern University undergraduateswere presented with 20 short paragraphs . Eachparagraph supported either an attributional or relationalinterpretation of a figurative expression . After readingthe paragraph, participants chose between simile andmetaphor forms as shown in Table 1 and were asked tochoose the sentence they preferred by circling it . Four

Page 3: Why do metaphors seem deeper than similes?...conservative in using metaphors than in using similes. The greater force of the metaphoric form was noted by Glucksberg and Keysar (1990),

random orders were used across participants. Whetherthe sentence in simile or metaphor form was presentedon the left side of the page was counterbalanced .

Table 1 : Example of attributional andrelational contexts

Conventional base

Attributional interoretatton ;

Mr. White, a sociologist, is writing an article aboutpoverty in urban America. He considers poverty a horribleblight on our society and argues that the government mustintervene with a welfare reform. He thinks that

Poverty is a disease .

Poverty is like a disease .

Relational interpretation.

Mr. White, a sociologist, is writing an article aboutpoverty in urban America He considers poverty to beincreasing and argues that, unless the governmentintervenes with a welfare reform, poverty will spreadfurther. He thinks that

Poverty is like a disease.

Poverty is a disease.

Results

We computed the number of metaphor choices bycoding preference for simile form as 0 and preferencefor metaphor form as 1 . Analysis of variance performedwith base conventionality and context type as between-subjects factors showed a significant effect of baseconventionality (F,, p = 7.50, MSE = 0.31, p < 0.01).The proportion of , metaphor form choices wassignificantly higher for statements with conventionalbases (Mc = 0.39) than for statements with novel bases(MN = 0.22). The number of metaphor preferences wassignificantly lower than chance for both novel andconventional bases (p < 0.05) .

We also obtained a marginally significant effect ofpreceding context type (Fi,'39 = 3.66, ACE = 0.15, p =0.06). Statements following relational contexts werepreferred in metaphor form more often than 'statementsfollowing attributional contexts (MR= 0.37, MA = 0.24).The number of metaphor preferences was significantlylower than chance for both relational and attributionalcontexts (p < 0.05). The results are summarized inFigure 1 .

The preference for metaphoric form for relationalinformation was only marginally significant . However,an item analysis indicated a disparity in the quality ofthe items used. Some items were strongly preferred insimile form (e .g., only one out of 64 participants choseto put Titanium chips are (like) diamonds in metaphorform). It thus seemed possible that not all the itemswere suitable as metaphors. To ensure that theRelational Precedence view was fairly tested, we

978

removed items that were put in metaphor form by lessthan seven participants (2 with novel bases and 2 withconventional bases). An ANOVA performed on theremaining items yielded a significant effect of contexttype (F,, 3, = 5.10, MSE = 0. 16, p < 0 .05) in addition tothe significant effect of base conventionality (F,, 3, =10.15, ACE = 0.32, p < 0.01). The interaction betweenbase conventionality and context type was notsignificant .

IAttributional NRelational

novel conventional

bass conventionality

Figure 1 : Proportion of metaphor choices fornovel and conventional bases. Error bars show

standard error.

Experiment 2

The results of the first study offered support for boththe Career of Metaphor account - in which -themetaphor-simile distinction is one of conventionality -and the Relational Precedence hypothesis. However,one concern here is to what extent the results simplyreflect the nature of the materials. First, although thefigurative expressions used in Experiment I allow bothrelational and attributional interpretations, it is possiblethat people may prefer one kind of interpretation overthe other. Previous studies have demonstrated thatpeople find relational interpretations of figurativeexpressions more interesting and apt (Clement &Gentner, 1988). Second, and more importantly, it ispossible that the conventional metaphors we used were

biased in favor of either relational 'or attributionalinterpretations, relative to the novel metaphors .

To calibrate the materials, we gave the figurativeexpressions used in Experiment I to a new group ofparticipants, either in simile or metaphor form, and

asked the participants to rate how much they agreedwith the relational and attributional interpretations .

Both interpretations were shown together for eachfigurative statement, but participants rated each

Page 4: Why do metaphors seem deeper than similes?...conservative in using metaphors than in using similes. The greater force of the metaphoric form was noted by Glucksberg and Keysar (1990),

separately. Thus they were free to assign high or lowratings to both the relational and attributionalinterpretations if they .chose .

MethodThirty-two Northwestern University undergraduateswere presented with 28 statements: 20 figurativestatements taken from Experiment 1, and 8 fillers . Thestatements were blocked so that each participant saweither all statements in simile form or all statements inmetaphor form. Two random orders were used . Aftereach statement, a relational and attributionalinterpretations of the statement appeared . The order ofthe interpretations on the page was counterbalanced .The participants were asked to rate how much theyagree with each of the interpretations on a 1 to 7 scale .

ResultsWe computed the scores for the relational andattributional interpretations for each item . Table 2shows mean ratings for each of the four item categories,along with the number of relational and attributionalinterpretations that received ratings of 4 or greater (outof 7). Consistent with prior research, relationalinterpretations are preferred over attributionalinterpretations overall (Gentner, 1988 ; GentnerClement, 1988). ,The key question for our purposes is whether the

materials were skewed such that conventionalmetaphors had more or better relational interpretationsthan the other categories . This does not appear to be-thecase. Relational interpretations received high ratings (4or above) for 8 out of 10 items in each of the four itemcategories -- conventional metaphors, novel metaphors,conventional similes, and novel similes . (Attributionalinterpretations were rated lower overall, as shown inTable 2.) It appears that the intended relationalinterpretations were highly apt for both metaphor andsimile forms. These data offer some reassurance thatthe shift towards relationality in metaphor preferencewas not simply determined by disproportionateavailability of relational interpretations for metaphorsover similes .

We also created a relational preference score (Rpref),which was the difference between the relational ratingand the attributional rating. An analysis of variancewith base conventionality (novel or conventional) andgrammatical form (simile or metaphor) as between-subjects factors revealed no significant differences inrelational preference scores (F3, 636 = 1.25, MSE _13.07, p < 0.3) .

979

Table 2. Mean interpretation ratings andnumber of interpretations that received high

ratings (in parentheses)

Discussion

As predicted by the Career of Metaphor hypothesis,participants in Experiment 1 were likely to choose themetaphor form for figurative statements withconventional bases, and the simile form for those withnovel bases. Aisenman's , Relational Precedencehypothesis also received support: the metaphor formwas chosen more often for relational meanings i .,following a relational context) than for attributionalmeanings (following an attributional context) .Might both claims be true? Some intriguing,

possibilities arise if we consider the implications ofthese two patterns taken together . Suppose that, as inthe Career of Metaphor hypothesis, nominal figurativeexpressions are initially phrased as similes. As theseexpressions become conventionalized, the metaphorform becomes more felicitous . Suppose further thatrelational meanings of novel bases have more potentialto get conventionalized . Then we would find apreponderance of relational meanings amongconventional bases . An informal survey of the literatureusing conventional metaphors suggests that most ofthem do convey relational meanings . For example, themetaphors used by Ortony (1979) and by Glucksbergand Keysar (1990) are primarily relational (e .g .,Cigarettes are time bombs ; Some jobs are jails ;Sermons are sleeping pills). Assuming that thesestimuli are roughly typical of conventional metaphors,we might speculate that there is a preponderance ofrelational figuratives within the class ofconventionalized metaphors . How might such a linkbetween relationality and conventionality have comeabout?

One possibility is that different forms are used forconventionalized relational and attributional figurativestatements. English has a special form for conventionalbases that is often used for property attribution - "as Xas, Y," where X is the shared attribute, and Y is the baseterm - for example, as white as snow; as strong as anox (Ortony, 1979) . Perhaps conventional attributionalmeanings are siphoned off by this dedicated form .

Attributlonal Relational

ConventionalMetaphor 3.81(5) 4.95(8)Simile 4.35(6) 4.74(8)

NovelMetaphor 3 .68 (4) 4.51(8)Simile 3 .81(4) 4.78(8)

Page 5: Why do metaphors seem deeper than similes?...conservative in using metaphors than in using similes. The greater force of the metaphoric form was noted by Glucksberg and Keysar (1990),

However, relational adjectives can enter the as X as Yframe as well (e.g., as delicious as an apple; asfierceas a tiger) . The only requirement for the descriptor Xseems to be that it be orderable on some dimension.Thus a possible special form for attributive figurativesdoes not seem like a viable explanation for thepreponderance of conventional relational metaphors .

Another possibility is preemption by existing terms .Over the course of development, languages havedeveloped names for attributes, which preempt thecreation of new ones (Clark, 1992) . On this account,creation of attributional metaphoric meanings might beless likely simply because we already have names forattributes. However, this explanation carries the hiddenassumption that the number of attributes we want toexpress is smaller than the number of relations .

This brings us to the third and most speculativepossibility . There is evidence that (1) people findshared relational structure more interesting or importantthan shared attributes ; and that (2) relational meaningsare relatively slow to emerge in cognitive development(Gentner & Rattermann, 1991 ; Halford, 1993) andarguably in the history of science. Applying this to theevolution of metaphor suggests that new relationalabstractions are more likely to become entrenched thanattribute meanings. Coherent relational systems arelikely to be preserved in comparison processing, andthis may carry over into the conventionalization ofmeanings and the formation of new categories(Gentner & Bowdle, 2001 ; Ramscar & Pain, 1996;Shen, 1992). On this account, a simile that expressesshared relational structure is more likely to give rise toparallels than one that expresses an attributionallikeness . This would lead to differential likelihoods ofconventionalization for relational and attributionalfiguratives.

Some evidence for this account can be obtained fromstudies of word meaning extension over time . One ofthe ways one can extend the meaning of a word is byanalogy. For example;, words like bridge and sanctuaryinitially had only concrete meanings, but now candenote metaphoric categories such as somethingconnecting two points and a safe place, respectively.Table 3 shows the timeline of the fast occurrences ofthe literal and figurative meanings of sanctuary, aslisted in the Oxford English Dictionary, as well as othersample occurrences. (All senses are written exactly asin the OED.)For sanctuary, the literal meaning of a holy building

appears in 1340 . Extensions to the church or the bodyof believers also appear in the 14`s century. The firstfigurative usage appears two centuries later, in 1568 .Interestingly, the first figurative use is signaled by anexplicit comparison phrase "counted as a sanctuary" .The first `metaphorical' occurrence, unmarked by acomparison phrase, occurs considerably later, in 1685 .

980

Table 3. Timeline of occurrences of literal andfigurative meanings for sanctuary.

[Initial literal meaning)I . a holy place - a building or place set apart for theworship of God or of one or more divinities: applied, e .g.,to a Christian church, the Jewish temple and the Mosaictabernacle, a heathen temple or site of local worship, andthe like; alsofig. To the church or the body of believers

1340 . . . in that sanctuary ours lordsail be kynge . . .1382 And thel shulen make to me a seyntuarye, and Yshaldwelle in the myddil ofhem.1530. Sanctwarle, aplace hallowed and dedicate unto god.

[La - a church or other sacred place in which, by the lawof the medieval church, a fugitive from justice, or adebtor, was entitled to immunity from arrest. Hence, in awider sense, applied to any place in which by law orestablished custom a similar immunity is secured tofugitives.

1374 To whiche Jugement they nolden net obeye butdefendedyn hem by the sikernesse of holy homes, that isto seynfedden in to sentuarye.1463-4 Eny persone. .that shall dwelle or inhabit withinthe Sayntwarle and Procyncte ofthe same Chapell.

[First figurative meaning)1568 Vstng aiwaise soch discrete moderation, as thescholehouse should be counted a sanctuarie againstfeare.

[First unmarked figurative meaning[1685 My house Is your Sanctuary, and here to offer youviolence, wou'dpreJudice myself.1770 The reformation was preceded by the discovery ofAmerica, as if the Almighty graciously meant to open asanctuary to the persecuted in future years . . .

Table 4. Timeline of occurrences of literal andfigurative meanings for bridge.

[Initial literal meaning]1. A structure forming or carrying a road over a river, a 4ravine, etc., or affording passage between two points at aheight above the ground .

c1000 theos brycg1131 Men weorth on adrencte and brigges to brokene.c1449 The brigge ofLondoun.1660 This was so severe a bill upon the Women, that, if abridge was made from Dover to Calais, the women wouldall leave this kingdom.

[Figurative[1225 The beoth ouer thisse worldes see, uppen the bruggeof heouene.1742 Faith builds a bridge from this world to the next.1863 The bridgefor thought to pass from one particular tothe other.1874 Gestures . . . forming the bridge by which we may passover into spoken language.

The pattern for bridge, shown in Table. 4, is similar.The first literal meaning of bridge as a structureaffording passage between two points above the groundgoes back to the 11`h century. However, the figurative

Page 6: Why do metaphors seem deeper than similes?...conservative in using metaphors than in using similes. The greater force of the metaphoric form was noted by Glucksberg and Keysar (1990),

uses are not listed until the middle of the 18` s century,except for a single reference to the bridge of heaven(which may have been meant literally) in 1225 .

These patterns suggest that, at least in some cases, themore abstract, figurative meanings appear later inwritten language. In both cases, these figurativemeanings are relational in nature . Interestingly, at leastfor sanctuary, the derived category no longer seemsmetaphoric; it has become a literal sense .We suggest that the Relational Precedence account

and the Career of Metaphor account may both beoperative in the evolution of metaphor, and that theyinteract. Beginning with a pool of novel figuratives, theCareer of Metaphor hypothesis states that for some ofthese the base teen is repeatedly used in parallelcomparisons, so that a conventional abstractionbecomes associated with the base. What we suggest isthat figurative expressions that yield coherent relationalsystems are most likely to be found novel and useful .Their bases are thus most likely to be reused andthereby conventionalized. For example, the simile Thecloud is like a marshmallow elicits common attributes0 the target and base, such as fluffy and white. But thepotential abstraction `white and fluffy' is unlikely tobecome a conventionalized word sense, both because oflexical preemption (we already have words for whiteand fluffy) and because the category it suggests issimply not very interesting. (Indeed, the conventionaluse of marshmallow as a metaphor is relational, as inThat boxer turned out to be a marshmallow.)Metaphors are a source of polysemy in language -

they allow words with specific meanings to take onadditional, related meanings (e .g., Glucksberg &Keysar, 1990; Lakoff, 1987; Lehrer, 1990 ; Miller,1979; Murphy, 1996) . We suggest that mappings thatfocus on relational structures are more likely togenerate stable abstractions than mappings that focuson object attributes. In sum, conventionalization ofrelational meanings may fulfill an important cognitivefunction in creating new abstractions .

References

Aisenman, R A . (1999). Structure-mapping and thesimile-metaphor preference . Metaphor and Symbol,14,45-51 .

Bowdle, B . F . (1998). Conventionality, polysemy, andmetaphor comprehension . Doctoral dissertation,Department of Psychology, Northwestern University .

Bowdle, B. F., & Gentner, D. (1999). Metaphorcomprehension: From comparison to categorization.In Proceedings of the Twenty-first Annual Meetingof the Cognitive Science Society (pp . 90-95),Vancouver, B.C .

Clark, E. V. (1992) . Conventionality and contrast :Pragmatic principles with lexical consequences . In A .

981

Lehrer & E . F . Kittay (Eds .), Frames, fields andcontrasts. Hillsdale, NJ : Erlbaum.

Gentner, D. (1988). Metaphor as structure mapping:The relational shift . ChildDevelopment, 59,47-59.

Gentner, D., & Bowdle, B . F. (2001). Convention,form, and figurative language processing. Metaphorand Symbol, 16,223-247.

Gentner, D., & Clement, C. (1988). Evidence forrelational selectivity in the interpretation of analogyand metaphor. In G . H. Bower (Ed .), The psychologyof learning and motivation, advances in research andtheory . New York: Academic Press.

Gentner, D., & Markman, A. B. (1997). Structure-mapping in analogy and similarity . AmericanPsychologist, 52, 45-56 .

Gentner, D., & Rattermann, M. J. (1991a). Languageand the career of similarity. In S. A. Gelman & J . P.Brynes (Eds.), Perspectives on thought andlanguage: Interrelations in development (pp . 225-277). London: Cambridge,University Press.

Gentner, D., & Wolff, P. (1997) . Alignment in theprocessing of metaphor . Journal of Memory andLanguage, 37, 331-355 .

Gibbs, H., & Wales, R (1990). Metaphor or simile :Psychological determinants of the differential use ofeach sentence form. Metaphor and Symbolic Activity,5,199-213 .

Glucksberg, S., & Keysar, B. (1990). Understandingmetaphorical comparisons : Beyond similarity .Psychological Review, 97,3-18 .

Halford, G . S . (1993) . Children's understanding: Thedevelopment of mental models. Hillsdale, NJ :Erlbaum.

Lakoff, G. (1987) . Women, fire, and dangerous things .Chicago: University of Chicago .

Lehrer, A. (1990). Polysemy, conventionality, and thestructure of the lexicon. Cognitive Linguistics, 1,207-246.

Miller, G. A. (1979). Images and models, similes andmetaphors . In A. Ortony (Ed .), Metaphor andthought (1st ed.) . Cambridge: CambridgeUniversity.

Murphy, G. L . (1996). On metaphoric representation.Cognition, 60,173-204,

Ortony, A. (1979). Beyond literal similarity .Psychological Review, 86, 161-180.

Oxford English Dictionary, online editionhttp://www.oed.com/.

Ramscar, M., & Pain, H. (1996). Can a real distinctionbe made between cognitive theories of analogy andcategorization? In Proceedings of the EighteenthAnnual Conference of the Cognitive Science Society .Hillsdale, NJ : Erlbaum.

Shen, Y. (1992) . Metaphors and categories. PoeticsToday, 13, 771-794 .