Turcan, Alexandra and Filik, Ruth

28 Pages·2017·0.9 MB·English

by  Christine Middleton

JournalofExperimentalPsychology: ©2016TheAuthor(s)
Learning,Memory,andCognition 0278-7393/16/$12.00 http://dx.doi.org/10.1037/xlm0000285
2016,Vol.42,No.12,1867–1893
An Eye-Tracking Investigation of Written Sarcasm Comprehension:
The Roles of Familiarity and Context
Alexandra T‚urcan and Ruth Filik
UniversityofNottingham
This article addresses a current theoretical debate between the standard pragmatic model, the graded
saliencehypothesis,andtheimplicitdisplaytheory,byinvestigatingtherolesofthecontextandofthe
propertiesofthesarcasticutteranceitselfinthecomprehensionofasarcasticremark.Twoeye-tracking
experiments were conducted where we manipulated the speaker’s expectation in the context and the
familiarity of the sarcastic remark. The results of the first eye-tracking study showed that literal
comments were read faster than unfamiliar sarcastic comments, regardless of whether an explicit
expectationwaspresentinthecontext.Theresultsofthesecondeye-trackingstudyindicatedanearly
processingdifficultyforunfamiliarsarcasticcomments,butnotforfamiliarsarcasticcomments.Later
reading time measures indicated a general difficulty for sarcastic comments. Overall, results seem to
suggestthatthe familiarityof theutterancedoesindeedaffectthetimecourseofsarcasmprocessing
(supporting the graded salience hypothesis), although there is no evidence that making the speaker’s
expectationexplicitinthecontextaffectsitaswell(thusfailingtosupporttheimplicitdisplaytheory).
Keywords:sarcasm,irony,languagecomprehension,figurativelanguage,eye-tracking
Verbalironyandsarcasmareformsofnonliterallanguagethat An example of an ironic comment is No. 1 below, where the
arecommonlyusedinoureverydayinteractions.Gibbs(2000)and expressedevaluationisthattheweatheris“great”buttheimplied
Hancock (2004) both reported similar rates of ironic language evaluation is that the weather is “terrible.” Sarcasm is a specific
use—about8%ofconversationalturnsincludeanironiccomment, formofirony,whichisusedwhenthetargetobjectofthecomment
beitbetweenfriends,ortotalstrangers.However,psycholinguists isaperson(Kreuz&Glucksberg,1989).Anexampleofasarcastic
have found it difficult to define these two forms of figurative commentisNo.2below,wheretheexpressedevaluationisthatthe
languageandconceptualisethemechanismsthroughwhichpeople colleague is “early” but the implied evaluation is that they are
managetounderstandandmakeuseofthemintheireverydaylife “late.”
(Bryant,2012).
1. Nonsarcastic irony—Uttering while standing outside in
thepouringrain:“Theweatherisgreattoday!”
Operational Definitions
Ironyisdefinedasaformofindirectlanguage,usedwhenthe 2. Sarcasticirony—Utteringtoacolleaguewhoarrivedata
speakerexpressesoneevaluativeutterancebutimpliesadifferent meetinghalfanhourlate:“You’reearly!”
evaluative appraisal (Burgers, van Mulken, & Schellens, 2011).
This article is concerned with the comprehension of sarcastic
irony(oneofthemostcommonlyusedformsofirony),wherethe
expressed (positive) evaluation is the direct opposite of the in-
ThisarticlewaspublishedOnlineFirstAugust8,2016. tended(negative)evaluation,asinexampleNo.2above.
AlexandraT‚urcanandRuthFilik,SchoolofPsychology,Universityof
Nottingham. Current Theoretical Debates
This research was supported by grants from the Economic and Social
Research Council (ESRC ES/L000121/1) and the British Academy Existing theories of sarcasm processing can be classified into
(PM140296) awarded to Ruth Filik. The authors would like to thank modularaccountsandinteractiveaccounts,andtheydifferinterms
RachelGiora,andthreeanonymousreviewersfortheirhelpfulfeedback. oftheirpredictionsforthetimecourseofsarcasmprocessing,and
This article has been published under the terms of the Creative Com- the roles played by the properties of the utterance itself and by
mons Attribution License (http://creativecommons.org/licenses/by/3.0/), contextualfactors.Inwhatfollows,wegiveabriefoverviewofthe
whichpermitsunrestricteduse,distribution,andreproductioninanyme- twotheoreticalcategories,andthereportedexperimentsfocuson
dium,providedtheoriginalauthorandsourcearecredited.Copyrightfor
testingthepredictionsofspecifictheoriesfromeachcategory.
this article is retained by the author(s). Author(s) grant(s) the American
Modular accounts claim that the literal meaning of a sarcastic
Psychological Association the exclusive right to publish the article and
utterance is usually accessed first and the sarcastic meaning is
identifyitselfastheoriginalpublisher.
accessed afterward. One example of a well-known modular ac-
CorrespondenceconcerningthisarticleshouldbeaddressedtoAlexan-
draT‚urcan,SchoolofPsychology,TheUniversityofNottingham,Univer- count is the standard pragmatic model (Grice, 1975), which pre-
sity Park, Nottingham NG7 2RD, United Kingdom. E-mail: ale.turcan@ dicts that sarcastic utterances will always take longer to process
gmail.com than the same utterances used literally, because they will always
1867
1868 T‚URCANANDFILIK
require the extra step of processing and then rejecting the literal sarcasm comprehension is governed by the concept of prototypi-
meaningofthesarcasticutterance,irrespectiveofhowsupportive cality. A prototypical sarcastic utterance is one that satisfies all
thecontextis(Gibbs,1999).Amorerecentmodularaccountisthe threeconditionsforimplicitdisplay.Theclaimisthatprototypical
gradedsaliencehypothesis(Giora,1997,2003),whichintroduces sarcastic utterances that fully satisfy the three requirements of
theconceptofsalience.Asalientmeaningisonethatisstoredin implicit display will have the highest degree of ironicalness (that
the mental lexicon due to its familiarity, conventionality, fre- is, they will be perceived as most ironic). Sarcasm that fails to
quency,orprototypicality(Peleg,Giora,&Fein,2001).According satisfyoneormoreoftherequirementswillhavealowerscoreof
tothegradedsaliencehypothesis,salientmeaningsareprocessed ironicalness(thatis,theywillbeperceivedaslessironic).
first,regardlessofstrengthofcontext(Giora,1997). Utsumi (2000) gives a mathematical formula (see Equation 1
Afamiliarsarcasticremark(like,“That’sgreat!”)isassumedto below)fordegreeofironicalnessthatcontainsdegreeofmanifest-
havetwosalientmeanings:thesarcasticandthenonsarcastic,and ness as a variable (defined as the explicitness of the speaker’s
they will both be activated in parallel. Therefore, the graded expectationinthecontext).
saliencehypothesispredictsthatfamiliarsarcasticremarksshould
d(U)(cid:2)d *d (cid:3)(1(cid:4)d )*d (cid:3)d (cid:3)d (1)
nothavelongerprocessingtimesthantheirliteralcounterparts.An m a m d i e
unfamiliar sarcastic remark however, has only one salient mean- This is Equation 1: The mathematical formula for degree of
ing, which is usually the literal one. In the case of unfamiliar ironicalness according to the Implicit Display Theory (Utsumi,
sarcastic remarks, the graded salience hypothesis predicts a very 2000).Theabbreviationsareasfollows:d(U)(cid:2)degreeofirony;
similarcomprehensionprocesstothestandardpragmaticmodel— d (cid:2)degreeofmanifestness;d (cid:2)degreeofallusion;d (cid:2)degree
m a d
unfamiliar sarcastic comments will take longer to process com- of polarity; d (cid:2) degree of pragmatic insincerity; d (cid:2) degree of
i e
paredwithliteralcounterparts,becausethesalientliteralmeaning indirectexpressionofnegativeattitude.
willbeactivatedfirst,followedbythenonsalientsarcasticmean- In the series of experiments presented in this article, d (the
m
ing (Giora, 1997, 2003). Diverging from the standard pragmatic degreeofmanifestness)wastheonlyfactorfromtheformulathat
model,however,thesalience-basedinterpretationwillnotbedis- wasmanipulated.Allotherfactorshavebeenkeptconstantandat
carded,becauseitcontributestotheinterpretationprocess. theirmaximalvalues(d :allsarcasticcommentssaidtheopposite
a
Interactiveaccountsclaimthatthesarcasticmeaningisaccessed of what the speaker meant, d : polarity of the comments was
d
directly in supportive contexts. A classic example is the direct always positive, that is, only sarcastic criticisms were employed,
accessview(Gibbs,1986),whichpredictsthatsarcasticutterances d:themaximofqualitywastheonlymaximviolated,andd :the
i e
shouldbeprocessedinequaltimetotheirliteralcounterpartswhen samesarcasticcueswereusedacrosscomments,thatis,anexcla-
embeddedinsupportivecontexts(thatis,contextswherethereisa mation mark at the end), so that the ironic environment and
discrepancybetweenexpectationandreality),becausereadersdo implicitdisplaywereprototypicalandcouldonlyvarywithdegree
nothavetoperformacompleteanalysisoftheliteralmeaningfirst ofmanifestness.
(Gibbs&Colston,2012).Amorerecentinteractiveaccountisthe One prediction of the implicit display theory is that more pro-
parallel constraint satisfaction model (Pexman, 2008). This is a totypical sarcastic utterances, that is, those that are made in con-
more general model, that allows for many different and unspeci- texts in which the speaker’s expectation is made explicit, will be
fied contextual factors to act as cues for sarcasm and therefore processed faster than or as fast as their literal counterparts (see
facilitatesarcasmprocessing.Inthisrespect,theparallelconstraint Utsumi, 2000). Less prototypical sarcastic remarks, that is, those
satisfactionmodelcouldbeconsideredtobeaframeworktheory, that are uttered in contexts in which the speaker’s expectation is
inthatitdoesnothaveaspecificsetoffactorsforwhichitmakes implicit(andhencehardertoinfer)willbeprocessedmoreslowly
testablepredictions. thanliteralequivalents.
Atestableinteractivetheoryofsarcasmprocessingspecifically
(and not figurative language in general), is the implicit display
Empirical Evidence
theory(Utsumi,2000)whichexpandsonthedirectaccessview’s
claimthatcontextcanaidsarcasmcomprehension,butdissociates The first question that researchers have typically addressed is,
itself from the idea that only one factor can influence sarcasm Dosarcasticutterancestakelongertoprocessthanliteralones?In
comprehension. The implicit display theory postulates that sar- atypicalexperiment,participantswouldbepresentedwithscenar-
casm requires an ironic environment, which is a property of the ios that would end in an utterance that could be interpreted as
context.Anironicenvironmentincludesthreecomponents:(a)the either literal or sarcastic. On the one hand, evidence from self-
speakerhastohaveanexpectation(knowntobothinterlocutors), pacedreadingstudies(e.g.,Giora,1995;Giora,Fein,&Schwartz,
(b)theexpectationhastobeunmetbythecurrentsituation,and(c) 1998; Spotorno & Noveck, 2014), and eye-tracking studies (e.g.,
the speaker has to have a negative emotional attitude toward the Filik & Moxey, 2010; Kaakinen, Olkoniemi, Kinnari, & Hyönä,
incongruity between expectation and reality (Utsumi, 2000). Ac- 2014) showing that sarcasm comprehension takes longer than
cordingtotheimplicitdisplaytheory,sarcasticremarksimplicitly literallanguagecomprehension,hasbeentakentosupportmodular
display this ironic environment, and they can do so to different accounts. Other evidence showing that sarcasm can be compre-
degrees. hended as fast as literal language, again from self-paced reading
The implicit display is a property of the ironic utterance; to (e.g., Gibbs, 1986), and additionally from visual-world paradigm
achieve implicit display, this utterance should (a) allude to the studies(e.g.,Kowatch,Whalen,&Pexman,2013),hasbeentaken
speaker’sexpectation,(b)violateatleastoneofGrice’spragmatic assupportformoreinteractiveaccounts.
principles, and (c) indirectly express the speaker’s negative atti- Torefinethedebate,researchersstartedtoaddressthequestion
tude (Utsumi, 2000). According to the implicit display theory, ofwhetherpropertiesoftheutterance(e.g.,itssalience)affectthe
WRITTENSARCASMCOMPREHENSION 1869
time course of sarcasm comprehension. Although “salience” is a The present studies set out to contribute to the debate outlined
concept loosely defined within the graded salience hypothesis, above, specifically, investigating the roles of familiarity of the
researchershavegenerallyequateditwith“familiarity,”thatis,if commentitselfandexplicitnessofexpectationinthecontextusing
a sarcastic utterance is deemed “salient,” that means that that tightlycontrolledliteralandsarcasticstimuliandasensitivemeth-
utteranceisfamiliartoreadersinitssarcasticinterpretation.Over- odology. The eye-tracking method has been previously used to
all, research seems to support the graded salience hypothesis investigatetheeffectoffamiliarity(e.g.,Filiketal.,2014),butnot
predictionthatfamiliarsarcasticutterancesareprocessedinequal ofcontextualfactors.Thismethodisamoresensitiveandprecise
timetoliteralones,butunfamiliarsarcasticutterancestakelonger measure of comprehension than simple self-paced reading times,
toprocessthantheirliteralcounterparts(e.g.,lexicaldecisiontask, anditalsoallowsustoinvestigatebothearlyandlateeffectsofour
Giora&Fein,1999,eye-tracking,Filik,Leuthold,Wallington,& manipulationsonsarcasmprocessing.Thetargetcommentsinour
Page,2014,Experiment1). experiments will be disambiguated by a single word, which will
In the literature presented above, contextual factors have not also allow us to distinguish early and late stages of processing
been manipulated at all, other than ensuring that sarcastic utter- whichmighthavebeenconfoundedinstudiesthathaddisambig-
ances were embedded in contexts where there was a mismatch uatingregionsmadeupofseveralwords.
between expectation and reality, whereas literal utterances were
embedded in contexts where there was no such mismatch. How-
Experiment 1
ever,itseemsintuitivetoassumethatcontextualfactorsmustplay
a role in sarcasm comprehension along with properties of the Theaimofthefirstexperimentwastoinvestigatetheroleofthe
utterance itself, because an utterance can only be interpreted as speaker’s expectation on sarcasm processing. Therefore, the ex-
sarcasticincontext(Calmus&Caillies,2014;however,seeGiora plicitness of the speaker’s expectation in the context and the
et al., 2013, for a different view). Thus, researchers wanted to literalityofthetargetutteranceweremanipulated,whilealltarget
addressathirdquestionof,specifically,whethercontextualfactors utterances were unfamiliar (because the predictions made by the
affectthetimecourseofsarcasmcomprehension.Evidenceforthe two modular accounts and those made by the implicit display
roleofcontextisalsomixed.Therearestudiesthatshowedthatthe theoryaremostclearlydistinctincaseswherethesarcasticutter-
degree of negativity of the event described in the context does ances are unfamiliar). Specifically, both the standard pragmatic
indeed affect the time course of sarcasm comprehension (e.g., modelandthegradedsaliencehypothesiswouldpredictaprocess-
readingtask,Ivanko&Pexman,2003),andthatthisnegativitycan ing difficulty for unfamiliar sarcastic utterances in both explicit
alsointeractwiththeexplicitnessoftheexpectationinthecontext andimplicitcontexts.Inotherwords,inordertosupportthetwo
to influence how sarcastic a comment is perceived to be (e.g., modular theories, we would expect to find a main effect of liter-
Utsumi,2005).Furthermore,thetimecourseofsarcasmcompre- ality, but no interaction between literality and explicitness. In
hensioncanalsobeaffectedbyfactorsliketheoccupationofthe contrast, the implicit display theory would predict that when the
speaker in the context (Pexman, Ferretti, & Katz, 2000), context speaker’s expectation is made explicit in the context, sarcastic
incongruity, and relationship between characters (Pexman, utteranceswouldbereadasfastasliteralones,however,whenthe
Whalen, & Green, 2010). These studies support interactive ac- expectation is implicit, sarcastic utterances would take longer to
counts of sarcasm comprehension by showing that context can read than literal ones. In other words, in order to support the
indeed affect the time course of sarcasm processing (see also implicit display theory, we would expect to find an interaction
Campbell&Katz,2012). between literality and explicitness, with longer reading times for
Ontheotherhand,therearestudiesshowingthatthetimecourse sarcasticthanliteralcommentsonlyintheimplicitcondition.
ofsarcasmcomprehensionisnotaffectedbyhavinganexpectation
made explicit in the context (e.g., Giora, Fein, Kaufman, Eisen-
Method
berg,&Erez,2009),orbyhavingacharacterthatisknowntobe
sarcastic in the context (e.g., Giora et al., 2007). These studies Participants. Thirty-twostudentsfromtheUniversityofNot-
supportthemodularaccounts. tingham (M (cid:2) 18 years and 4 months, SD (cid:2) 6 months, 31
age
Inconclusion,evidenceintheliteraturewithregardstothetime females and 1 male) participated in the experiment. All partici-
course of sarcasm processing is mixed and conflicting (as also pantswerenativeEnglishspeakers,notdiagnosedwithanyread-
shown by Gibbs & Colston, 2012). Possible reasons for this are ingdisorders,andhadnormalorcorrected-to-normalvision.They
that some of the studies described above either lacked a good receivedcoursecreditinreturnfortheirparticipation.
controloverconfoundingfactors(e.g.,saliency),theydidnothave Materials and design. Twenty-four experimental materials
thedesignrequiredinordertodistinguishbetweenthetwogroups wereconstructed(seeTable1foranexampleandAppendixAfor
of theories, or the methodology they employed was not sensitive aselection;thefullsetofitemsisavailablefromthefirstauthor).
enough to reveal reading time differences between sarcastic and Each scenario was made up of five sentences, describing an
literalcomments,oreffectsofthecontext. interaction between two characters, and ending with a comment
Thisarticlefocusesontestingthepredictionsofthreetheories: thatonecharactermadetowardtheotherone.Thefirstsentenceof
two modular (the standard pragmatic model and the graded sa- thecontextsimplyintroducedthetwocharactersandthesituation
liencehypothesis),andoneinteractive(theimplicitdisplaytheory, theywerein(e.g.,DeanandChloewereonholidayinValenciafor
because it makes testable predictions about the way in which aweek.).
specificcontextualfactorsshouldaffectsarcasmcomprehension). Thesecondsentencehadtwoversions,whichdifferedbetween
Therefore,fromhereonward,thefocusofthearticleshiftstothese the explicit and implicit conditions. In the explicit condition, the
threespecifictheories. second sentence contained an explicit expectation of the speaker
1870 T‚URCANANDFILIK
Table1
ExampleScenario(Experiment1)
Literality Explicitness Examplescenario
Literal Explicit DeanandChloewereonholidayinValenciaforaweek.TheendofthetripwasapproachingsoDeanaskedChloetothink
ofsomethingthrillingtodoontheirlastday.ChloesuggestedtheygoandwatchtheFormula1race,whichwasDean’s
favouritesport.“Your/suggestionisprecriticalregion/stirring!”criticalregion/Deansaidtoher.Postcriticalregion/
Theywentout.
Implicit DeanandChloewereonholidayinValenciaforaweek.Theirtripwasquicklycomingtoanend,andtheyweren’tsure
whattodoontheirfinalday.ChloesuggestedtheygoandwatchtheFormula1race,whichwasDean’sfavouritesport.
“Your/suggestionisprecriticalregion/stirring!”criticalregion/Deansaidtoher.Postcriticalregion/Theywentout.
Sarcastic Explicit DeanandChloewereonholidayinValenciaforaweek.TheendofthetripwasapproachingsoDeanaskedChloetothink
ofsomethingthrillingtodoontheirlastday.ChloesuggestedtheystayinthehotelandwatchTV,whichwasquite
boring.“Your/suggestionisprecriticalregion/stirring!”criticalregion/Deansaidtoher.Postcriticalregion/Theywent
out.
Implicit DeanandChloewereonholidayinValenciaforaweek.Theirtripwasquicklycomingtoanend,andtheyweren’tsure
whattodoontheirfinalday.ChloesuggestedtheystayinthehotelandwatchTV,whichwasquiteboring.“Your/
suggestionisprecriticalregion/stirring!”criticalregion/Deansaidtoher.Postcriticalregion/Theywentout.
regarding how the other character should behave, which was theory,thatis,theyfulfilledtheofflinepredictionsofthetheory:
known to both characters, as required by the implicit display thereader’sexpectationforsarcasmwasincreasedintheexplicit
theory (e.g., The end of the trip was approaching so Dean asked conditioncomparedwiththeimplicitcondition(seeAppendixB).
Chloetothinkofsomethingthrillingtodoontheirlastday.).Inthe Theothertestinvestigatedwhethertheconditionsdifferedinterms
implicit condition, the second sentence of the context did not ofhownaturaltheysoundtothereader,andtheresultsindicated
containanymentionofanexpectation(e.g.,Theirtripwasquickly the literal materials sounded more natural than the sarcastic ones
comingtoanend,andtheyweren’tsurewhattodoontheirfinal (see Appendix B). This is perhaps to be expected, given that
day.). Results from a stimulus norming test revealed that as in- sarcastic comments are employed significantly less in every day
tended,participantsthoughtthatthematerialsintheexplicitcon- speechthanliteralones.
dition created an expectation for how the other character should Thirty-six filler materials accompanied the 24 experimental
behave significantly more than the materials in the implicit con- materials.Athirdofthefilleritemsalsocontainedtwocharacters
dition(seeAppendixB).
butendedinaliteralnegativeutterance,anotherthirddidnothave
The third sentence contained the outcome of the second char-
anycharactersandwereinformativetexts,whereasthefinalthird
acter’s behavior and it had two versions, which differed between
contained two characters and ended in a literal positive utterance
the literal and sarcastic conditions. In the literal conditions, the
(seeaselectionoffilleritemsinAppendixC).
outcome fulfilled the expectation mentioned in the previous sen-
The software used to display the texts (Eye Track; http://blogs
tence (e.g., Chloe suggested they go and watch the Formula 1
.umass.edu/eyelab/software/)ensuredtherandomizationandcoun-
race, which was Dean’s favorite sport.). In the sarcastic condi-
terbalancing of the scenarios. For each scenario, there were four
tions, the outcome frustrated the expectation mentioned in the
stimulus presentation files, each containing only one version of
previoussentence(e.g.,Chloesuggestedtheystayinthehoteland
each scenario, and a total of six experimental items for each
watch TV, which was quite boring.). Results from a stimulus
condition. Each participant was presented with one stimulus file,
norming test verified that indeed the materials in the sarcastic
so that in the end data were collected from eight participants for
conditionwereperceivedassignificantlymoresarcasticthanthose
eachstimulusfile.Theorderinwhichthescenarioswerepresented
intheliteralcondition(seeAppendixB).
withineachstimulusfilewasrandomizedforeachparticipant.
The final comment was contained in the fourth sentence (e.g.,
Procedure. Eye movements were recorded via an SR Re-
“Your suggestion is stirring!” Dean said to her.). In the literal
search Eyelink 1000 eye tracker that sampled eye position every
conditions,thespeakermeantwhattheyliterallysaidthroughthe
millisecond. Viewing was binocular, but only one eye was re-
final comment, which had a positive meaning, whereas in the
corded for each participant. Materials were displayed on a com-
sarcastic conditions, the speaker said the opposite of what they
puterscreenapproximately56cmfromparticipants’eyes.Before
meant, that is, they said something positive in order to convey a
negativemeaning.Allfinalsarcasticcommentswerenonconven- the start of the experiment, the procedure was explained to the
tional,meaningtheywerenotfamiliartothereaders(asshownby participants.Theywereinstructedtoreadastheywouldnormally,
a familiarity stimulus norming test; see Appendix B). The fifth taking as much time as they needed in order to understand the
sentencewasawrap-upsentencethatconcludedthescenario(e.g., texts.Participantswerethenseatedattheeyetrackerandplaced
They went out.). Thus the experiment consisted of a 2 literality on a chin- and forehead-rest to minimize head movements.
(literal vs. sarcastic) (cid:3) 2 speaker’s expectation (explicit vs. im- Theythencompletedacalibrationprocedure.Beforeeachtrial,
plicit) design, with both factors being within-subject and within- a fixation box appeared in the top left quadrant of the screen.
item. Once the participant fixated this box, the texts would be pre-
Besides the literality, explicitness, and familiarity stimulus sented. If the participants’ apparent point of fixation did not
normingtests,twomorenormingtestswereconducted.Onever- match with the fixation box, the experimenter recalibrated the
ifiedthatthematerialsweresuitablefortestingtheimplicitdisplay eye tracker. Each trial consisted of one scenario, presented as
WRITTENSARCASMCOMPREHENSION 1871
fourlinesoftext,withtwoblanklinesbetweeneachlineoftext. Table2
Once the participants finished reading it, they looked away Summaryof0-msReadingTimeRemoval(Experiment1)
from the text and toward a post-it note affixed to the bottom
right hand edge of the monitor, and then pressed the right- Analysisregion Readingmeasure %ofmissingdata
shoulder button on the console to progress to the next trial.
Precritical fp 17.4
After 25% of the trials, a yes/no comprehension question ap- rp 17.4
peared to ensure that the participant actually read and compre- tt 7.2
hended the text. The comprehension question (e.g., “Were Dean Critical fp 11.6
rp 11.6
andChloeonholidayinValencia?”)relatedsolelytothecontext
tt 8.4
of the scenario, and it was not a test of sarcasm comprehension. Postcritical fp 0.5
Theaveragecorrectresponserateof94.7%indicatesthatpartici- rp 0.5
pantswereindeedreadingforcomprehension. tt 0.1
Note. fp(cid:2)first-pass;rp(cid:2)regressionpath;tt(cid:2)totalreadingtime.
Results and Discussion
Eachscenariohadthreeanalysisregions.Thecriticalregionwas
the word that disambiguated the target utterance as being either and within-item, respectively. However, because the maximal
sarcastic or literal. For example, in the scenario in Table 1, the modelfailedtoconverge,therandomeffectsstructurehadtobe
criticalwordinthefinalcomment“Yoursuggestionisstirring!” simplified in order to obtain convergence. This was done by
was“stirring!”Theprecriticalregionconsistedofthetwowords progressively removing one random component at a time—the
precedingthecriticalregion(e.g.,“suggestionis”).Thepostcriti- onethatexplainedtheleastamountofvarianceintheprevious
cal region was the remainder of the target utterance (e.g., Dean nonconverging model.
said to her.) Three measures of reading behavior are reported: The best way of establishing the appropriate random effects
first-pass reading time (the sum of all fixations in a region from structure is currently a debatable issue. Barr et al. (2013)
firstenteringituntilleavingiteitherviaitsleftorrightboundary, recommend always fitting the maximal model, with random
also known as gaze duration when the region comprises a single slopesforallfixedeffectsofinterest.However,thissuggestion
word), regression path (or go-past) reading time (the sum of all is often not practical—fitting the maximal model might often
fixationsfromthetimethataregionisfirstentereduntiltheregion fail to converge because the model is overparameterized—the
is left via its right region boundary), and total reading time (the researcher is overfitting the data. The reason why Barr et al.
sum of all fixations in a region, including fixations made when (2013) argue for fitting the maximal model is because they
rereadingtheregion). claimthatexcludingarandomslopeforoneofthefixedeffects
Priortothestatisticalanalysis,thedatawerepreprocessedusing of interest increases the likelihood of making a Type I error.
theEyeDoctorsoftware(http://www.psych.umass.edu/eyelab).For However, Matuschek and his colleagues have recently coun-
each participant, the blinks were removed, and also the fixations tered this point. They conducted a simulation study and con-
werealignedontheverticalplane.TheEyeDrysoftwarewasthen cluded that the maximal model is not in fact the best choice,
used to create the files needed for data analysis. Trials that had because although it reduces the likelihood of making a Type I
zero first-pass reading times for two consecutive regions (where error, it also significantly reduces power (Matuschek, Kliegl,
regions were defined as a whole sentence in the context, the Vasishth,Baayen,&Bates,2015).Theirstudyindicatedthatfor
precritical,critical,andpostcriticalregions)wereeliminated(dis- factorial studies such as those reported in this article, it is
cardedtrialsaccountedfor2.6%ofthedata). recommendedthattherandomeffectsstructureofnonconverg-
Data analysis was performed in R (R Core Team, 2013) using ing models be reduced until a significant decrease in goodness
linearmixedeffectsmodeling(lme4package)andpotentialinter- of fit is observed. The model that is supported by the data (the
actionsweredecomposedinRusingthefunctiontestInteractions one prior to a significant decrease in goodness of fit) provides
fromthephiapackage(wherethechi-squareisthedefaulttest;all the best balance between Type I error rates and power.
reportedpvaluesareBonferronicorrected).Thefirststepwasto The analyses reported in this article reduce the random ef-
discard0-msreadingtimesfromtheanalysis(seeTable2forthe fects structure in a similar way to the procedure suggested in
percentage of data removed due to the reading time being 0 Matuschek et al. (2015), except that the simplification proce-
ms—typically due to participants skipping over the respective dureisonlycontinueduntilconvergenceisachievedratherthan
region). This was in addition to the 2.6% of the trials already until a significant decrease in goodness of fit is observed (i.e.,
removed due to having zero first-pass reading times for two it is stopped earlier). That is because continuing until a signif-
consecutiveregions. icant decrease in goodness of fit is obtained leads to an even
The second step was to establish the appropriate random simplerrandomeffectsstructurethanthestructureobtainedby
effects structure for each analysis. We started by fitting the stoppingthesimplificationprocesswhenamodelconverges.It
maximal model to the data, as recommended by Barr, Levy, is important to note that for the key findings reported in this
Scheepers,andTily(2013).Therandomeffectsstructureofthe article,bothproceduresleadtothesamefixed-effectsstructures
maximal model was: (1 (cid:4) literality (cid:3) explicitness|subject) (cid:4) in the final models.
(1 (cid:4) literality (cid:3) explicitness|item). The reason why literality Oncetherandomeffectsstructurehadbeenestablished,thethird
and explicitness were introduced as random slopes for both stepwastoperformaseriesoflikelihoodratiotestscomparingthe
subjects and items is because both factors were within-subject, fit of models with different fixed-effects structures in order to
1872 T‚URCANANDFILIK
Table3
BestFittingModelsandFixed-EffectsParameters(Experiment1)
Analysis Reading
region measure Model Fixedeffects Coefficient SE t
Precritical fp (cid:6)1(cid:4)(1(cid:4)explicitness|subject)(cid:4)(1(cid:4)explicitness|item) (Intercept) 252.5 10 25
rp (cid:6)explicitness(cid:4)(1|subject)(cid:4)(1|item) (Intercept) 307.9 21 14.6
Explicitness 54.4 22.3 2.4
tt (cid:6)1(cid:4)(1(cid:4)literality(cid:3)explicitness|subject)(cid:4)(1(cid:4)literality(cid:3) (Intercept) 383.4 23.4 16.4
explicitness|item)
Critical fp (cid:6)1(cid:4)(1(cid:4)literality(cid:3)explicitness|subject)(cid:4)(1(cid:4)literality(cid:3) (Intercept) 276.2 16.6 16.6
explicitness|item)
rp (cid:6)literality(cid:4)(1|subject)(cid:4)(1(cid:4)explicitness|item) (Intercept) 424 45.4 9.3
Literality 92 25.7 3.6
tt (cid:6)literality(cid:4)(1(cid:4)literality|subject)(cid:4)(1(cid:4)literality(cid:3)explicitness|item) (Intercept) 343.5 32.2 10.7
Literality 87.9 23 3.8
Postcritical fp (cid:6)literality(cid:3)explicitness(cid:4)(1(cid:4)literality(cid:3)explicitness|subject)(cid:4)(1(cid:4) (Intercept) 394.3 21.5 18.4
literality(cid:3)explicitness|item) Literality 66.1 26.7 2.5
Explicitness 78.5 27.8 2.8
Literality(cid:3) (cid:7)87.3 35.5 (cid:7)2.5
explicitness
rp (cid:6)literality(cid:4)(1(cid:4)literality(cid:3)explicitness|subject)(cid:4)(1(cid:4)literality(cid:3) (Intercept) 530.2 35.3 15.02
explicitness|item) Literality 80.5 30 2.7
tt (cid:6)literality(cid:4)(1(cid:4)literality(cid:3)explicitness|subject)(cid:4)(1(cid:4)literality(cid:3) (Intercept) 518.3 33 15.7
explicitness|item) Literality 66.1 23.3 2.8
Note. fp(cid:2)first-pass;rp(cid:2)regressionpath;tt(cid:2)totalreadingtime.
reachthebestmodelfitforourdata.1Theprocedureusedwasto sarcasticutterancesincontextscontaininganexplicitexpectation
compare the model with the two factors in interaction with pro- will be read as fast as literal utterances. In other words, it seems
gressivelysimplerfixed-effectsstructures(thatis,twomaineffects thatwefailedtosupportthepredictionthatincreasingthedegree
but no interaction, or only one main effect). See Table 3 for the ofmanifestnessofthespeaker’sexpectationinthecontextoffers
models that had the best fit for our data and the values of their an initial processing advantage for sarcastic utterances. These
fixed-effects parameters. Furthermore, see Appendix D for the t resultsareinlinewiththoseofpreviousstudiesofironyprocess-
values associated with the fixed factors that did not have signifi- ing that report a literality effect (e.g., Filik et al., 2014 for unfa-
cant effects (i.e., were not included in the best models), and the miliar ironies; Filik & Moxey, 2010; Giora, 1995; Giora et al.,
seriesoflikelihoodratiotestsperformedinordertoreachthebest
1998,2007;Kaakinenetal.,2014).
models.
Thepostcriticalregion. Aninteractionbetweenliteralityand
Theprecriticalregion. Noeffectswereobservedinfirst-pass
explicitness was observed in first-pass reading time—see Figure
ortotalreadingtimes—seeFigures1aand1c.However,regression
3a. Post hoc comparisons showed that (a) the region of text
pathreadingtimewasshorterfollowingexplicitcontexts(M (cid:2)
rp-explicit followingaliteralcommentwasreadfasterwhenthecontextwas
309ms,SEM(cid:2)11ms)thanimplicitones(M (cid:2)362ms,
rp-implicit explicit (M (cid:2) 396 ms, SEM (cid:2) 15 ms) than when it
SEM (cid:2) 20 ms)—see Figure 1b. This suggests that even before fp-literal-explicit
was implicit (M (cid:2) 471 ms, SEM (cid:2) 18 ms): (cid:5)2(1,
readingthedisambiguatingword,participantsrereadthecontextin fp-literal-implicit
N(cid:2)32)(cid:2)8,p(cid:2).009,and(b)theregionfollowingacomment
the implicit condition more than in the explicit one. Importantly,
presentedinanexplicitcontextwasreadfasterwhenthecomment
thenulleffectobservedinfirst-passreadingtimewasreassuringin
wasliteral(M (cid:2)396ms,SEM(cid:2)15ms)thanwhen
thatitsuggestedthattherewasnobaselinereadingtimedifference fp-literal-explicit
itwassarcastic(M (cid:2)456ms,SEM(cid:2)18ms):(cid:5)2(1,
betweentheexperimentalconditions. fp-sarcastic-explicit
N(cid:2)32)(cid:2)6.1,p(cid:2).027.Interestingly,thispatternofresultswas
Thecriticalregion. Therewasamaineffectofliteralityonall
readingmeasures.Literalutteranceswerereadfaster(M (cid:2) notduetosarcasticutterancesbecomingmoredifficultinimplicit
fp-literal
275 ms, SEM (cid:2) 11 ms; M (cid:2) 441 ms, SEM (cid:2) 22 ms; contexts, but due to literal utterances becoming more difficult in
rp-literal
M (cid:2)355ms,SEM(cid:2)15ms)thansarcasticones(M (cid:2) implicitcontexts.Wecanconcludethatthecontextualmanipula-
tt-literal fp-sarcastic
299 ms, SEM (cid:2) 10 ms; M (cid:2) 527 ms, SEM (cid:2) 23 ms; tionseemstohaveaneffectonthelaterstagesofliterallanguage
rp-sarcastic
M (cid:2)446ms,SEM(cid:2)15ms)—seeFigure2a,2b,and2c processing,butnotonthelaterstagesofsarcasmprocessing.
tt-sarcastic
below. Regression path and total reading times only reflected a main
It seems that when the disambiguating word is encountered in effectofliterality—seeFigure3band3c.Theregionfollowinga
thetext,readerstakelongertoreaditifitpointstowardasarcastic literal utterance was read faster (M (cid:2) 522 ms, SEM (cid:2) 18
rp-literal
interpretation of the comment, than if the comment’s intended
meaning is literal. These results clearly support the predictions
1Factorslikethefrequencyorlengthoftheanalysesregionswerenot
made by the modular accounts of sarcasm interpretation (the
includedaseffectsinthelinearmixedmodelsbecausetheseexperiments
standardpragmaticmodelandthegradedsaliencehypothesis),but weredesignedsothatweonlycomparedbetweenthesamecriticalwords,
offer no support for the implicit display theory’s prediction that precritical,andpostcriticalregions.
WRITTENSARCASMCOMPREHENSION 1873
model and the graded salience hypothesis), however, it fails to
supporttheimplicitdisplaytheory.
Inconclusion,theresultsfromExperiment1didnotprovide
any support for the implicit display theory’s predictions that
explicitness of the speaker’s expectation in the context would
affectreadingtimesforsarcasticutterances,bymakingthemas
easy to read as literal utterances when the expectation is ex-
plicit. However, it did provide support for both modular ac-
counts’ predictions (the standard pragmatic model and the
gradedsaliencehypothesis),byshowingthatunfamiliarsarcas-
ticutterancestooklongertoreadthanliteralcounterparts.Inthe
next experiment we wanted to replicate the current results and
Figure1. Meanreadingtimesontheprecriticalregion(Experiment1).
Figure 1a: first-pass reading time. Figure 1b: regression path reading
time.Figure1c:totalreadingtime.Errorbarsrepresent(cid:8)1SEM.
ms;M (cid:2)531ms,SEM(cid:2)15ms)thanfollowingasarcastic
tt-literal
one(M (cid:2)607ms,SEM(cid:2)23ms;M (cid:2)605ms,
rp-sarcastic tt-sarcastic
SEM(cid:2)17ms).ThispatternofresultswasalsoobservedinFilik
and Moxey’s (2010) study, and was taken to reflect difficulty in
integrating the comment with the context when the comment is
sarcastic. This difficulty in contextual integration seems to be
independentoftheexplicitnessofthespeaker’sexpectationinthe
context. Rather, as suggested by Filik and Moxey (2010), these
results provide some evidence that after a sarcastic utterance is
encountered, more reinspection of the text is required before the
reader can comprehend the material, as compared with when a Figure 2. Mean reading times on the critical region (Experiment 1).
literalutteranceisencountered,whichisinlinewiththemodular Figure 2a: first-pass reading time. Figure 2b: regression path reading
accountsofsarcasmcomprehension(boththestandardpragmatic time.Figure2c:totalreadingtime.Errorbarsrepresent(cid:8)1SEM.
1874 T‚URCANANDFILIK
familiaritywasaddedtothepreviousdesign.Underthesecircum-
stances, the two modular accounts would make different predic-
tionsforthetimecourseofsarcasmcomprehension.Thestandard
pragmatic model would predict that sarcastic utterances would
take longer to read under all circumstances, irrespective of com-
ment familiarity or speaker’s expectation in the context. In other
words,inordertosupportthestandardpragmaticmodel,wewould
expecttofindamaineffectofliterality,andnointeractionswith
familiarity or explicitness. The graded salience hypothesis would
predict that sarcastic utterances would take longer to read than
literal counterparts if they were unfamiliar; however, if sarcastic
utterances were familiar, they would be read as fast as literal
counterparts. In other words, in order to support the graded sa-
liencehypothesis,wewouldexpecttofindaninteractionbetween
literalityandfamiliarity,butnointeractionwithexplicitness.The
implicit display theory would predict that sarcastic utterances
would take longer to read than literal ones if they are uttered in
contextsinwhichthespeaker’sexpectationisimplicit;however,if
thespeaker’sexpectationisexplicit,sarcasticutteranceswouldbe
read as fast as literal ones. Thus, as in Experiment 1, we would
expecttofindaninteractionbetweenliteralityandexplicitness,but
nointeractionwithfamiliarity.
Method
Participants. Sixty-fourstudentsfromtheUniversityofNot-
tingham participated (M (cid:2) 22 years and 6 months, SD (cid:2) 7
age
months,42femalesand22males).Noneofthemhadtakenpartin
Experiment 1. All participants were native English speakers, not
diagnosedwithanyreadingdisorders,andhadnormalorcorrected-
to-normalvision.Theyeitherreceiveda£4inconvenienceallowance
fortakingpart,orcoursecredit.
Materials and design. The experimental materials consisted
of 48 short texts, with the same structure as the materials of the
previousexperiment(seeAppendixEforaselection;thefullsetof
itemsisavailablefromthefirstauthor).Theonlydifferenceinthis
experiment was that in the familiar condition the final utterance
wouldbeforexample“Soexcited!”insteadof“Yoursuggestion
isstirring!”Thustheexperimentconsistedofa2literality(literal
vs.sarcastic)(cid:3)2speaker’sexpectation(explicitvs.implicit)(cid:3)2
familiarity (familiar vs. unfamiliar) design, with literality and
Figure 3. Mean reading times on the postcritical region (Experiment 1).
expectation as within-subject and within-item factors, and famil-
Figure 3a: first-pass reading time. Figure 3b: regression path reading
iarityasawithin-subjectandbetween-itemsfactor.Thesameset
time.Figure3c:totalreadingtime.Errorbarsrepresent(cid:8)1SEM.
of stimulus norming tests were conducted as in Experiment 1,
showing that (a) familiar comments were rated as significantly
morefamiliartothereaderthantheunfamiliarones,(b)sarcastic
additionally address the question of what role the properties of
commentswereratedassignificantlymoresarcasticthattheliteral
the utterance play in sarcasm comprehension. To this end, in
ones,(c)theexpectationforhowtheothercharactershouldbehave
Experiment 2 we investigated the online reading patterns of
wassignificantlyclearerintheexplicitthaninimplicitconditions,
both familiar and unfamiliar sarcastic utterances presented in
(d)literalcommentssoundedmorenaturalthansarcasticones,and
explicit and implicit contexts.
(e) the materials met the implicit display theory’s offline predic-
tion:anexpectationforsarcasmwassignificantlyhigherwhenthe
Experiment 2
context was explicit compared with implicit (see Appendix F for
TheaimofExperiment2wastoreplicatetheresultsofExper- thefullsetofresults).
iment1,andextendthembyinvestigatingtheroleoftheproperties Therewerealso48filleritems,followingasimilarstructureas
oftheutteranceinsarcasmcomprehension,thusfurtherdiscrimi- inExperiment1:halfofthematerialscontainedtwocharactersbut
nating between the predictions of the standard pragmatic model endedinaliteralnegativeutterance,andtheotherhalfdidnothave
andthegradedsaliencehypothesis.Hence,thefactorofcomment anycharactersandwereinformativetexts.
WRITTENSARCASMCOMPREHENSION 1875
Procedure. TheprocedurewasexactlythesameasinExper- precriticalregionofunfamiliarcommentswerereadinequaltimes
iment 1. In terms of the comprehension questions, the average in explicit and implicit contexts (M (cid:2) 258 ms,
fp-unfamiliar-explicit
correctresponseratewas93.9%,indicatingagainthatparticipants SEM(cid:2)6ms,M (cid:2)248ms,SEM(cid:2)6ms,(cid:5)2(1,
fp-unfamiliar-implicit
readandcorrectlycomprehendedthescenarios. N(cid:2)64)(cid:2)1.4,p(cid:2).5).Thisindicatesthatevenbeforethereaders
knewwhetherthecommentwasgoingtobeliteralorsarcastic,the
Results and Discussion contexthadanimpactonthereadingtimesoffamiliarcomments,
butnotontheunfamiliarones.
The scenarios in this experiment had the same three analysis Inregressionpathreadingtime,thefamiliarityeffectindicated
regions as in Experiment 1 (i.e., critical region: disambiguating that the precritical region of familiar utterances was read slower
word, precritical region: two words prior to the disambiguating than that of unfamiliar ones—see Figure 4b. However, because
word, and postcritical region: the remainder of the target utter- this specific comparison is between reading times on different
ance). The data were preprocessed using the same software and words,anysimplemaineffectsoffamiliarityareverydifficultto
proceduresasbefore.Trialsthathadzerofirst-passreadingtimes interpretmeaningfully.
for two consecutive regions were eliminated (removed trials ac- Intotalreadingtimes,theliteralitymaineffectindicatedthatthe
countedfor3.5%ofthedata). precriticalregionofliteralcommentswasreadfaster(M (cid:2)
tt-literal
DataanalysiswasperformedinthesamewayasinExperiment 343ms,SEM(cid:2)6ms)thanthatofsarcasticcomments(M (cid:2)
tt-sarcastic
1. The first step was to discard 0-ms reading times from the 378ms,SEM(cid:2)7ms)–seeFigure4c.Themostlikelyinterpre-
analysis(seeTable4forthepercentageofdataremovedduetothe
tationoftheliteralitymaineffectisthattheprecriticalregionhas
reading time being 0 ms—typically due to participants skipping
beenrereadmoreinsarcasticscenariosthaninliteralones,which
overtherespectiveregion).Thiswasinadditiontothe3.5%ofthe
might suggest a difficulty in the interpretation of the sarcastic
trialsalreadyremovedduetohavingzerofirst-passreadingtimes
materials as predicted by the standard pragmatic model. The fa-
fortwoconsecutiveregions.Thesecondstepwastoestablishthe
miliaritymaineffectindicatedthattheprecriticalregionwasread
appropriate random effects structure, which was done following
faster in familiar than unfamiliar utterances, but as explained
thesameprocedureasinExperiment1.Oncetherandomeffects
above, the familiarity main effect alone cannot be interpreted
structure had been established, the third step was to perform a
meaningfully.
series of likelihood ratio tests comparing the fit of models with Thecriticalregion. Infirst-passreadingtime,thereweretwo
differentfixed-effectsstructuresinordertoreachthebestmodelfit
main effects—see Figure 5a. The critical word of familiar utter-
forourdata.2Theprocedureusedwastocomparethemodelwith anceswasreadfaster(M (cid:2)223ms,SEM(cid:2)3ms)thanthe
fp-familiar
the three factors in interaction with progressively simpler fixed- critical word of unfamiliar utterances (M (cid:2) 260 ms,
fp-unfamiliar
effectsstructures(thatis,threemodelswithtwo-wayinteractions SEM (cid:2) 4 ms). Again, although this result is in the direction that
andonemaineffect,followedbyamodelwiththreemaineffects,
one might expect, it should be interpreted with caution, because
thenthreemodelswithtwomaineffectsandfinallythreemodels
this specific comparison is between reading times on different
withjustonemaineffect).SeeTable5belowforthemodelsthat words (e.g., excited in the familiar condition vs. stirring in the
had the best fit for our data and the values of their fixed-effects
unfamiliarcondition).Theliteralitymaineffectindicatedthatthe
parameters.Furthermore,seeAppendixGforthetvaluesassoci- criticalwordofaliteralcommentwasreadfaster(M (cid:2)237
fp-literal
atedwiththefixedfactorsthatdidnothavesignificanteffects(i.e., ms,SEM(cid:2)3ms)thanthatofasarcasticcomment(M (cid:2)
fp-sarcastic
werenotincludedinthebestmodels),andtheseriesoflikelihood 248 ms, SEM (cid:2) 4 ms). This pattern of results supports the
ratiotestsperformedinordertoreachthebestmodels.
predictionsofthestandardpragmaticmodel,indicatingthatinthe
Theprecriticalregion. Infirst-passreadingtimetherewasa
earlyprocessingstages,sarcasmseemstoindeedbeoverallmore
familiarity-explicitness interaction—see Figure 4a. Post hoc tests
difficulttoprocessthanliterallanguage,irrespectiveoffamiliarity
indicatedthat(a)theprecriticalregionoffamiliarcommentswas
orcontextualinformation.
read faster if the context was explicit rather than implicit
In regression path reading time, an interaction was observed
(Mfp-familiar-explicit (cid:2) 262 ms, SEM (cid:2) 6 ms, Mfp-familiar-implicit (cid:2) between literality and familiarity - see Figure 5b. Post hoc com-
283ms,SEM(cid:2)6ms,(cid:5)2(1,N(cid:2)64)(cid:2)6.6,p(cid:2).02),but(b)the parisonsshowedthat(a)unfamiliarutteranceswereslowertoread
inthesarcasticconditionthanintheliteralcondition(M (cid:2)
rp-unfamiliar-literal
402ms,SEM(cid:2)14ms,M (cid:2)477ms,SEM(cid:2)17
rp-unfamiliar-sarcastic
Table4 ms):(cid:5)2(1,N(cid:2)64)(cid:2)15.7,p(cid:9).001,and(b)familiarutterances
Summaryof0-msReadingTimeRemoval(Experiment2) werereadequallyfastirrespectiveofwhethertheyweresarcastic
or literal (M (cid:2) 311 ms, SEM (cid:2) 12 ms,
Analysisregion Readingmeasure %ofmissingdata M rp-fa(cid:2)mil3ia3r3-litmersal,SEM(cid:2)12ms):(cid:5)2(1,N(cid:2)64)(cid:2)0.8,
rp-familiar-sarcastic
Precritical fp 17.1 p (cid:2) .8. This pattern of results fully supports the graded salience
rp 17.1 hypothesis,butoffersnosupportforthestandardpragmaticmodel
tt 8.2 or the implicit display theory. Sarcastic utterances do not always
Critical fp 17
take longer to read than literal ones (as the standard pragmatic
rp 17
tt 14 modelwouldpredict),andthereiscurrentlynoevidencethatthey
Postcritical fp 1.9 are influenced by the strength of contextual information (as the
rp 1.9 implicit display theory would predict). However, when they are
tt 1
familiar,sarcasticutterancesarereadasfastasliteralutterances,as
Note. fp(cid:2)first-pass;rp(cid:2)regressionpath;tt(cid:2)totalreadingtime. predictedbythegradedsaliencehypothesis.
1876 T‚URCANANDFILIK
Table5
BestFittingModelsandFixed-EffectsParameters(Experiment2)
Analysis Reading
region measure Model Fixedeffects Coefficient SE t
Precritical fp (cid:6)familiarity(cid:3)explicitness(cid:4)(1(cid:4)literality|subject)(cid:4)(1|item) (Intercept) 257.9 9.6 26.9
Familiarity (cid:7)3.3 8 (cid:7)0.4
Explicitness 20.3 7.9 2.6
Familiarity(cid:3) (cid:7)29.6 11.2 (cid:7)2.6
explicitness
rp (cid:6)familiarity(cid:4)(1(cid:4)literality|subject)(cid:4)(1|item) (Intercept) 375.9 17 22.2
Familiarity (cid:7)42.1 13.3 (cid:7)3.2
tt (cid:6)literality(cid:4)familiarity(cid:4)(1(cid:4)literality(cid:3)explicitness|subject)(cid:4)(1|item) (Intercept) 328.5 15.9 20.7
Literality 37.9 9 4.2
Familiarity 23 8 2.9
Critical fp (cid:6)literality(cid:4)familiarity(cid:4)(1(cid:4)explicitness|subject)(cid:4)(1(cid:4)literality|item) (Intercept) 212.2 8 26.5
Literality 10.3 4.9 2.1
Familiarity 38.9 4.6 8.5
rp (cid:6)literality(cid:3)familiarity(cid:4)(1(cid:4)explicitness|subject)(cid:4)(1(cid:4)literality|item) (Intercept) 307.3 20.8 14.7
Literality 17.6 20.2 0.9
Familiarity 92.4 19 4.9
Literality(cid:3) 57.9 26.8 2.2
Familiarity
tt (cid:6)literality(cid:3)familiarity(cid:4)(1(cid:4)literality(cid:3)familiarity|subject)(cid:4)(1|item) (Intercept) 258.7 13.2 19.6
Literality 26.5 10.9 2.4
Familiarity 59.3 10.6 5.6
Literality(cid:3) 29.1 14.6 2
familiarity
Postcritical fp (cid:6)literality(cid:4)(1(cid:4)literality(cid:3)explicitness|subject)(cid:4)(1|item) (Intercept) 408.9 17.2 23.8
Literality 42.5 7.8 5.5
rp (cid:6)literality(cid:4)familiarity(cid:4)(1(cid:4)literality(cid:3)familiarity |subject)(cid:4)(1|item) (Intercept) 470.7 23.9 19.7
Literality 90.4 22.5 4
Familiarity 46.7 17.7 2.6
tt (cid:6)literality(cid:4)familiarity(cid:4)(1(cid:4)literality(cid:3)explicitness|subject)(cid:4) (Intercept) 490.6 25.4 19.3
(1(cid:4)literality|item) Literality 74.2 18.6 4
Familiarity 31.6 10.5 3
Note. fp(cid:2)first-pass;rp(cid:2)regressionpath;tt(cid:2)totalreadingtime.
Finallyintotalreadingtime,aninteractionbetweenliteralityand (M (cid:2) 590 ms, SEM (cid:2) 14 ms; M (cid:2) 584 ms,
rp-sarcastic tt-sarcastic
familiaritywasobservedagain—seeFigure5c.However,thistime, SEM(cid:2)10ms).Alsotheregionfollowingafamiliarutterancewas
literalcommentswerereadfasterthansarcasticonesinbothfamiliar readfaster(M (cid:2)524ms,SEM(cid:2)12ms;M (cid:2)533
rp-familiar tt-familiar
(M (cid:2) 264 ms, SEM (cid:2) 6 ms, M (cid:2) 293 ms, SEM (cid:2) 9 ms) than the region following an unfamiliar utter-
tt-familiar-literal tt-familiar-sarcastic
ms,SEM(cid:2)8ms,(cid:5)2(1,N(cid:2)64)(cid:2)5.9,p(cid:2).03)andunfamiliar ance(M (cid:2)565ms,SEM(cid:2)13ms;M (cid:2)561
rp-unfamiliar tt-unfamiliar
conditions (M (cid:2) 321 ms, SEM (cid:2) 8 ms, ms,SEM(cid:2)9ms).Theseresultsseemtosupportthefindingsfrom
tt-unfamiliar-literal
M (cid:2) 377 ms, SEM (cid:2) 10 ms, (cid:5)2(1, N (cid:2) 64) (cid:2) Experiment 1 and those observed in Filik and Moxey’s (2010)
tt-unfamiliar-sarcastic
22.3,p(cid:9).001).Thisresultsuggestedthattheadvantageforfamiliar study,whichshowedthattheregionoftextfollowingsarcasticutter-
sarcasticcommentsdidnotcarryoverintothelaterstagesofprocess- ancesisreadmoreslowlythanthetextfollowingliteralutterances.
ing, and instead both familiar and unfamiliar sarcastic comments Thecurrentexperimentshowedthatalthoughfamiliarityoffersan
seemed to require additional processing time compared with literal advantage for the processing of familiar sarcastic utterances when
theircounterparts.InlinewiththefindingsfromExperiment1,these theyareinitiallyencountered(asevidencedinregressionpathreading
results also fail to support the implicit display theory, because the timesonthedisambiguatingword),thisadvantageislostinthelater
explicitness of the speaker’s expectation did not facilitate sarcasm stages of processing (as illustrated by the lack of an interaction
processinginanyoftheconditions. betweenliteralityandfamiliarityonthepostcriticalregion).
The postcritical region. In first-pass reading time, a main
effectofliteralitywasobserved—seeFigure6a.Theregionoftext
General Discussion
following a literal utterance had shorter first-pass reading times
(M (cid:2) 411 ms, SEM (cid:2) 6 ms) than the region following a Two experiments were carried out to contribute to the current
fp-literal
sarcastic utterance (M (cid:2) 453 ms, SEM (cid:2) 7 ms). In theoretical debate on the factors affecting sarcasm processing,
fp-sarcastic
regression path reading times and total reading times, two main using tightly controlled materials, and a method (eye-tracking)
effects of literality and familiarity were observed—see Figure 6b sensitive enough to reveal both early and late effects of our
and 6c. The region of text following a literal utterance was read manipulations.Inbothexperiments,participantsreadshortscenar-
faster (M (cid:2) 499 ms, SEM (cid:2) 11 ms; M (cid:2) 510 ms, ioswhiletheireyemovementswererecorded.InExperiment1,the
rp-literal tt-literal
SEM (cid:2) 8 ms) than the region following a sarcastic utterance contextsofthesescenarioseitherincludedanexplicitexpectation