Brain & Language 123 (2012) 11–21

Contents lists available at SciVerse ScienceDirect

Brain & Language
journal homepage: www.elsevier.com/locate/b&l

Not all ambiguous words are created equal: An EEG investigation of homonymy and polysemy
Ekaterini Klepousniotou a,b,⇑,1, G. Bruce Pike a,b, Karsten Steinhauer b,c, Vincent Gracco b,c a

McConnell Brain Imaging Centre, Montreal Neurological Institute, McGill University, 3801 University Street, Montreal, QC, Canada H3A 2B4 Centre for Research on Language, Mind, and Brain, McGill University, 3640 de la Montagne, Montreal, QC, Canada H3G 2A8 c

School of Communication Sciences and Disorders, McGill University, 1266 Pine Avenue West, Montreal, QC, Canada H3G 1A8 b

a r t i c l e

i n f o

Article history:
Accepted 24 June 2012
Available online 21 July 2012
Keywords:
ERP
N400
Lexical ambiguity
Lateralization
Homonymy
Polysemy
Metaphor
Metonymy

a b s t r a c t
Event-related potentials (ERPs) were used to investigate the time-course of meaning activation of different types of ambiguous words. Unbalanced homonymous (‘‘pen’’), balanced homonymous (‘‘panel’’), metaphorically polysemous (‘‘lip’’), and metonymically polysemous words (‘‘rabbit’’) were used in a visual single-word priming delayed lexical decision task. The theoretical distinction between homonymy and polysemy was reflected in the N400 component. Homonymous words (balanced and unbalanced) showed effects of dominance/frequency with reduced N400 effects predominantly observed for dominant meanings. Polysemous words (metaphors and metonymies) showed effects of core meaning representation with both dominant and subordinate meanings showing reduced N400 effects. Furthermore, the division within polysemy, into metaphor and metonymy, was supported. Differences emerged in meaning activation patterns with the subordinate meanings of metaphor inducing differentially reduced N400 effects moving from left hemisphere electrode sites to right hemisphere electrode sites, potentially suggesting increased involvement of the right hemisphere in the processing of figurative meaning. Ó 2012 Elsevier Inc. All rights reserved.

1. Introduction
In everyday life, successful communication occurs even when
we must attribute correctly a speaker’s intended meaning to words that convey a wide array of possible interpretations, that is, lexical ambiguity. Theoretical linguistics distinguishes between two types of lexical ambiguity, homonymy and polysemy. The first type of lexical ambiguity, homonymy, is observed in lexical items that ‘‘accidentally’’ carry two distinct and unrelated meanings2 (Weinreich, 1964). For example, in the sentences ‘‘John lay down on the bank of the river’’ and ‘‘The Royal Bank is the largest bank in Montreal’’, the word ‘‘bank’’ has the meanings ‘‘river side’’ and ‘‘financial institution’’. Homonymy is assumed to have contrastive meanings which are contradictory in nature. The context and the discourse set-

⇑ Corresponding author at: McConnell Brain Imaging Centre, Room WB325, Montreal Neurological Institute, McGill University, 3801 University Street, Montreal, QC, Canada H3A 2B4. Fax: 1 514 398 2975.

E-mail addresses: [email protected], [email protected] leeds.ac.uk (E. Klepousniotou).
1
Present address: Institute of Psychological Sciences, University of Leeds, Leeds LS2 9JT, UK. Fax: +44 113 3435749.
2
It should be noted here that the term ‘‘meaning’’ is used to refer to the multiple distinct and unrelated interpretations of a word form such as in the case of homonymy. In contrast, the term ‘‘sense’’ is used to refer to the multiple related interpretations of a word form such as in the case of polysemy. 0093-934X/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.bandl.2012.06.007

ting help in their disambiguation selecting the appropriate meaning each time (Weinreich, 1964).
The other type of ambiguity, polysemy, involves lexical senses which relate to the same basic meaning of the word as it occurs in different contexts (Weinreich, 1964). For example, in the sentences ‘‘Mary painted the door’’ and ‘‘Mary walked through the door’’, the word ‘‘door’’ in the first sentence refers to the ‘‘physical object’’, whereas in the second sentence it refers to the ‘‘aperture’’. Yet, the basic meaning of the word is the same in both sentences. Weinreich (1964) referred to these sense distinctions as complementary polysemies (i.e., polysemy) which, unlike senses in contrastive ambiguity, are not contradictory in nature. Rather, one sense seems more appropriate or ‘‘focused’’ for the interpretation of the word in the particular context.

Apart from the distinction between homonymy and polysemy,
according to theoretical linguistics, there is a further distinction within polysemy into two types, which are motivated by two distinct figures of speech, namely metaphor and metonymy (Apresjan, 1974). In metaphorical polysemy, in which a relation of analogy is assumed to hold between the senses, the basic sense is literal, whereas the secondary sense was originally figurative when this use of the word emerged. For example, the ambiguous word

‘‘eye’’ has the literal primary sense ‘‘organ of the body’’, and the secondary sense ‘‘hole in a needle’’. Metaphorically motivated polysemy seems to be quite unconstrained in that the relatedness

12

E. Klepousniotou et al. / Brain & Language 123 (2012) 11–21

in meaning between the primary and the derivative meanings is not always so obvious (Apresjan, 1974).
The other type of polysemy is motivated by metonymy and a
relation of contiguity or connectedness is assumed to hold between the senses. Metonymically motivated polysemy respects the usual notion of lexical polysemy, namely the ability of a word to have several distinct but related senses (Apresjan, 1974). The changes of meaning in metonymic polysemy are not accidental, as in homonymy, but systematic or ‘‘regular’’ (Apresjan, 1974), and both the primary and the secondary senses are quite literal. For example, the ambiguous word ‘‘rabbit’’ has the literal primary sense ‘‘the animal’’, and the literal secondary sense ‘‘the meat of that animal’’.

Drawing on the observation that homonymy and polysemy are
relative concepts, it seems that some types of metaphorically motivated polysemy are closer to homonymy. On the other hand, metonymically motivated polysemy is a step further away from homonymy and possibly represents ‘‘pure’’ polysemy (Apresjan, 1974). Several types of metonymic changes of meaning (or shifts of meaning), which seem to hold cross-linguistically and are systematic in nature (thus, in a way, predictable and productive), have been identified such as count/mass, container/containee, producer/ product, product/institution, figure/ground, and place/people alternations to name just a few (Pustejovsky, 1995).

Although polysemy is much more frequent in language, most
psycholinguistic and neurolinguistic studies to date have focused mainly on homonymy (see Simpson, 1994 for a review). With respect to homonymy, most models agree that the multiple, unrelated meanings are represented separately in the mental lexicon; however, the representation of polysemy in general, and of its two subtypes – metaphor and metonymy – in particular, has been very controversial. The question we address here, using electrophysiological correlates, is how the multiple, closely related senses of polysemous words, both metaphors and metonymies, are activated and represented in the mental lexicon. Are the multiple related senses of metaphorical and metonymic polysemous words processed just like the multiple unrelated meanings of homonymous words or do they employ a qualitatively different process? 1.1. Behavioural evidence for the distinction between homonymy and polysemy

There is currently an ongoing debate in the literature, based on behavioural studies, on the representation of polysemy. On the one hand, there is evidence that the senses of polysemous words, unlike the meanings of homonymous words, are stored together in the mental lexicon. For example, words with multiple meanings associated with a single derivation (i.e., all the meanings have the same etymology) are accessed faster than words with an equal number of meanings that are associated with multiple derivations (i.e., the meanings are associated with different etymologies; Jastrzembski, 1981). In addition, several other studies suggest that homonymous and polysemous words are represented and processed differently (Frazier & Rayner, 1990; Klepousniotou, 2002; Klepousniotou & Baum, 2007; Klepousniotou, Titone, & Romero, 2008; Pickering & Frisson, 2001; Williams, 1992). For example, Frazier and Rayner (1990) found that participants’ eye movement patterns differed for polysemous and homonymous words in that polysemous words required shorter fixation times. They argued that because the meanings of homonymous words are mutually

exclusive, selection of the appropriate meaning must occur before processing can proceed. In contrast, because the different senses of polysemous words are not mutually exclusive and may share a
core representation, all possible meanings can remain activated so that selection and disambiguation, if necessary, is delayed. Further evidence for the facilitatory effects of the interrelatedness of

multiple senses on the processing of polysemous words is observed in the processing advantage for lexical decisions for words with many senses, a trend for a disadvantage for words with many meanings (Rodd, Gaskell, & Marslen-Wilson, 2002) and the inability to suppress priming effects of the contextually irrelevant central meaning of polysemous words even over long prime-target delays (Williams, 1992).

In contrast with the notion that homonymy and polysemy are
represented and processed differently, a number of experiments support the opposite view, namely that polysemy functions just like homonymy. In particular, an influential study by Klein and Murphy (2001) found no evidence that polysemous words embedded in phrasal contexts (e.g., daily paper vs. shredded paper) function differently from homonymous words (but cf. Klepousniotou et al., 2008). Their results showed that contextual consistency facilitated comprehension while contextual inconsistency inhibited comprehension. Similar findings were obtained from a series of off-line experiments as well (Klein & Murphy, 2002). Critically then, based on the findings of the studies undertaken by Klein and Murphy (Klein & Murphy, 2001, 2002), the separate representations view is supported for both homonymy and polysemy; namely, the multiple senses of polysemous words have separate representations just like the multiple meanings of homonymous words.

1.2. Behavioural evidence for the distinction within polysemy into metaphor and metonymy
As described above, polysemy is not a uniform phenomenon;
rather, according to theoretical linguistics (Apresjan, 1974), polysemy is divided into metaphor and metonymy. Importantly, studies to date (Klepousniotou 2002; Klepousniotou & Baum, 2007; Klepousniotou et al., 2008) that exploited the distinction within polysemy into metaphor and metonymy, both in context and isolation, yielded further information about lexical ambiguity processing. In particular, in a study directly comparing homonymous, metaphorically polysemous and metonymically polysemous

words, Klepousniotou (2002) found that metonymically polysemous words demonstrated stronger facilitation effects and were processed significantly faster than homonymous words, while metaphors fell somewhere in the middle between homonymy

and metonymy. The distinction of metaphor and metonymy within polysemy was further supported in a set of lexical decision experiments focused on the so-called ‘‘ambiguity advantage’’ effect (i.e., an assumed processing advantage for ambiguous than unambiguous words) by comparing the processing of balanced homonymy, unbalanced homonymy, metaphorical polysemy, and metonymic

polysemy to unambiguous frequency-matched control words
(Klepousniotou & Baum, 2007). Although no processing advantage was found for homonymy (both balanced and unbalanced), a processing advantage was evident for all polysemous words – metaphorical and metonymic – (i.e., they were processed faster than unambiguous words), suggesting a ‘‘sense-relatedness advantage’’ effect. In addition, even in the presence of a processing advantage, metaphorically polysemous words took longer to process than

metonymically polysemous words, providing further support to the theoretical linguistics division of polysemy into metaphor and metonymy.
Finally, in another study (Klepousniotou et al., 2008), participants judged whether ambiguous words embedded in word pairs (e.g., tasty chicken) made sense as a function of a cooperating, conflicting, or neutral context using the paradigm of Klein and Murphy (2001). The ambiguous words were independently rated as having low, moderate, and highly overlapping senses/meanings to map onto a homonymy to metonymy continuum. The effects of sense/ meaning dominance were also examined. The results indicated

E. Klepousniotou et al. / Brain & Language 123 (2012) 11–21

that words with highly overlapping senses (metonymy) showed
reduced effects of context and dominance compared to words with moderately or low overlapping meanings (metaphorical polysemy and homonymy), suggesting that the comprehension of ambiguous words is mediated by the semantic overlap of alternative senses/ meanings.

Based on these results, it was suggested that processing differences probably indicate representational differences depending on the type of ambiguity that the words exhibit (Klepousniotou, 2002; Klepousniotou & Baum, 2007; Klepousniotou et al., 2008). Homonymous words show longer reaction times both in sentential contexts and isolation possibly because their multiple unrelated meanings are competing, thus slowing down the activation process. Homonymous words, then, could be seen as having several distinct mental representations in the mental lexicon and the appropriate meaning is chosen from a pre-existing, exhaustive list of senses (i.e., sense selection). Polysemous words, on the other hand, and in particular metonymous words...

+44 -934 /10.1016/j.bandl.2012.06.007 /ground /locate/b&l 0 0.00005 0.0005 0.005 0.01 0.044 0.05 0.061 0.07 0.083 0.09 0.132 0.16 0.17 0.186 0.2 0.23 0.3 0.314 0.55 0.65 0.7 0.74 0.78 0.8 0.9 0.915 0.92 0.98 0.99 0093 1 1.1 1.11 1.2 1.249 1.29 1.3 1.48 1.49 1.66 1.74 1.781 1.85 1.96 10 10.446 100 1000 103 109 11 111 112.088 113 116 119 12 12.962 120 121 123 126 1266 127 13 131.329 133 1332 1343 1366 138 1383 14 142 15 1500 1525 1534 1542 1543 16 167 17 17.3 18 18.281 181 188 19 19.3 192 193 1964 1970 1974 1975 1980 1981 1982 1983 1987 1988 1990 1992 1994 1995 1998 1999 1a 1a8 1b 2 2.1 2.1.1 2.1.2 2.1.3 2.2 2.22 2.27 2.3 20 20/20 200 2001 2002 2003 2004 2005 2005a 2005b 2006 2007 2008 2009 2010 2012 205 208 21 218 22 230 233 238 24 24.4 240 245 25 255 259 26 266 27 272 278 282 284 29 29.11 2975 2a8 2b4 3 3.1 3.19 3.2 3.2.1 3.2.2 3.2.3 3.2.4 3.4 3.6 3.65 3.67 3.69 3.78 3.91 30 300 305 308 31 32 326 33 34 343 3435749 348 35 350 36 3640 365 37 375 380 3801 382 387 398 4 4.02 4.07 4.1 4.11 4.2 4.27 4.3 4.59 4.6 4.8 40 400 405 409 41 428 429 446 45 450 46 468 47 470 478 48 484 487 494 5 5.3 5.8 5.984 50 500 512 514 53 548 55 556 566 57 570 573 59 6.478 63 64 65 68 691 7 7.31 7.792 7.90 700 8 80 81 8279 86 9 93 95 97 9jt abil academ accept access accident accord accur accuraci acknowledg acquisit across activ active-two actual ad ad-box addit address adult advantag af3 af4 af7/af8 affect afz ag agcl age agre agreement aim al alberta algebra algorithm allow almost alreadi also altern although alway alzheim ambigu american among amplitud amplifi amsterdam anaki analog analys analysi analyz anatom anim animal/human anoth anova answer anterior apart apertur appar appear appli appreci approach appropri approxim apresjan argu arm arrang array articl artifact ask assess assign associ assum asymmetri atchley attempt attest attribut author avail avenu averag away axi b balanc ball bandpass bank bank-money bank-riv bar base baselin basic baum beeman began behav behavior behaviour beretta berlin better bias bigram bilater biosemi bipolar blink bodi boston bottl box brain brain-damag brainvis brigg broader brown brownel bruce bulletin burgess c c1/c2 c3/c4 c5/c6 cage call callos cambridg canada canadian canthi care carri case categori centr central cerebr certain cf chalk chang characterist chiarello chicken choic chosen chwilla cihr clark classific clay clear clinic clock close closer cm cms coach coars code cognit cole colour combin come comfort commit common communic compar compat compet competit complementari compon comprehens compris compromis comput concept conceptu concert conclus concret condit conduct congruous connect connected connot consid consist construct consult contain container/containee contamin content context contextu contigu continu continuum contradictori contrast contribut control controversi convey confin confirm conflat conflict cooper core correct correl correspond corrobor cortex couch could count/mass counterbalanc cours cp1/cp2 cp3/cp4 cp5 cp6 cpz creat creation critic critiqu cross cross-hemispher cross-linguist cross-mod crucial cue current cz d daili damag danc data date de debat decis degre delay demonstr depend depict deriv describ detail detect dictionari diego differ differenti difficult digit dim direct disadvantag disambigu discours discuss diseas dish disord dispar display distant distinct distinguish distribut divid divis dixon doctor document domin dominance/frequency dominant-rel donchin door downward draw driven drl druri due duffi dvf dx.doi.org dx.doi.org/10.1016/j.bandl.2012.06.007 dynam e e-mail e.g e.klepousniotou earli earlier ed educ eeg eeg/erps effect eighteen either ekaterini [email protected] electrod electroencephalogram electroencephalographi electrophysiolog elicit elsevi embed emerg employ encount end energi english entri eog eog-contamin epoch equal equibias equival erlbaum erp error especi establish et etc etymolog evalu eve even event event-rel eventrel everyday evid examin exampl exclam exclud exclus exhaust exhibit exist expect experi experiment explain exploit explor extend extended/subordinate extens extent extrem eye eye-track f f1/f2 f3/f4 f5/f6 f7/f8 face facilit facilitatori fact factor familiar familiarity/frequency faster faust fax fc1/fc2 fc3/fc4 fc5 fc6 fcz fell fellowship femal fig final fiorentino first florida fmri focus follow follow-up form formal found four four-way fox fp1/fp2 fpz fragment franci frazier free freedom frequenc frequency-match frequent frisson front frontal ft7/ft8 full function furthermor fz g garfield gaskel geisser general generat gernsbach gilhooli given gmbh gracco grade grammar grammat grand grant graph gratton greater green greenhous ground group h h3a h3g hagoort half hand handbook handed handi headcap health healthi hear held help hemispher high histori hit hoc hold hole homepag homograph homonym homonymi homotop hop horizont houghton hour howev human hypothes hypothesi hz i.e ident identifi ii illustr imag imageri immedi impair imped implic import imposs inabl inc includ incongruous/unrelated inconsist incorpor increas inde independ index indic individu induc inform inhibit initi instead institut instruct instrument integr intend interact interest intern interpret interrel interrelated interv introduct inventori investig involv influenc influenti irregular irrelev isi isol item iz j jastrzembski john journal judg judgement juli june k kahan karsten keep kept keul key keyword klein klepousniotou known kolk kravetz kucera kuta kwasni kx l la languag larg larger largest last latenc later latter lawrenc lay lead learn least leed leeds.ac.uk left left-later left/right leftlater legal length lesion less letter level lexic lexical-semant lexicon lh liber lie life light like likert limit line linguist link lip list lit liter literatur lliná lock logi logic long long-stand longer low ls2 luck lv m m350 ma made magazin mahwah mail main maintain major make male malfunct mani manipul map mari marie-ev mark marslen marslen-wilson mastoid match materi matter maxim maximum maxfield may mcconnel mcevoy mcgill mean meaning-depend meaning/sense meanings/senses meanings2 measur meat mechan medial medial/lateral mediat meg memori mental metaphor method methodolog metonym metonymi microvolt middl midlin mifflin might mild militari millikan mind minim mirror mit mix modal model moder modul mold money monitor montagn month montreal moreov morri motiv mount mous mouth move movement ms mse much multipl murphi must mutual n n400 name nativ natur nebe necessari necessarili need needl negat nelson network neural neurocognit neuroimag neurolinguist neurolog neuron neurophysiolog neuropsychologia neurosci neutral never nevertheless new newman next nine nj non non-word nonrespons norm normal note notion noun number o o1/o2 object object/human observ obtain obvious occupi occur occurr ocular off-lin often older omnibus on-lin one ongo onlin onset onto opposit order organ origin other outer overal overlap oz p p-valu p1/p2 p3/p4 p5/p6 p7/p8 p9/p10 paint pair panel paper paradigm parallel parsimoni part part/object particip particular pascal patient pattern peak pen per percent perform perhap petten phase phenomenon phillip phonological/orthographic phonotact phrasal phrase physic picker pike pine pintyp place/people plot po3/po4 po7 po8 poeppel point polysem polysemi popul portner posit possibl post post-hoc post-lex postdoctor posterior potenti poz pp practic pre pre-exist preced precis predict predomin prefer presenc present press presum previous primari prime prime-target prior probabl problem procedur proceed process produc product product/institution promin pronounc propos provid profil psycholinguist psycholog psychonom pure pustejovski pylkkänen pz qc qualit qualifi question quit r rabbit rabbit-hop rabbit-stew random rang rare rate rather rayner re re-referenc reaction read real reason recent recognit record reduc reef refer referenc regard region regular reject relat related relev reli remain remov reorder repeat replac replic report repres represent requir research reserv resolut resolv respect respond respons rest result retain retriev reveal revers review reflect rh rhyme rich right right-hand right-later rivard river robust rodd romero room royal rubenstein ruin rule sake salisburi salti sampl san satur save scalar scale scalp school schreiber scienc sciencedirect scivers screen sd seat second secondari see seem seen segment select semant senior sens sense-related senses/meanings sent sentenc sententi separ seri serv session set seven seven-point sever share shed shift short shorter show shown shred side signal signific similar simpson simultan sinc singl single-word site slow slower slowli smaller so-cal softwar somewher sourc south space spatial speaker specifi specific speech speech-languag springer springer-verlag stand standard star start statist status steinhauer step stew still stimuli stimulus stop store street strength strike strong stronger strongest structur student studi sub sub-set subcondit subdivis subject subordin subordinate-rel subordinaterel subsequ subserv subset subtyp success sufficient suggest support suppress suscept swaab syllabl syntax system systemat t7/t8 tabl take taler target task tasti taylor techniqu tempor tend tentat term terminolog test thank theoret therebi therefor third thirti though thought three three-way thus tick time time-cours time-lock time-window ting titon togeth took total toward tp7/tp8 track train transit translat tremblay trend trial trigram triplet true twilley two two-way two-word type u uk ullman unambigu unbalanc unconstrain under undergo underli understand understood undertaken unequivoc uniform unilater univers unlik unrel us usag use usual v valu van vari varianc varieti various verb verbal verifi verlag version versus vertic view vincent vision visual voltag vs w walenski walk wall waveform way wb325 weaker webster weinreich well well-control west wheeler wherea wherebi whether wide william wilson window within word would wrist www.elsevier.com www.elsevier.com/locate/b&l x x-axi xaxi y y-axi year yes yet yield à100 â ó field figurat figure filler filter financial finding first fixation fixed