ffi Available online at www'sciencedirect'com Ioutnal of MemorY and

s"rE ,t= fllDtREcl@\Y7 Language .."#..Siles

Journal of Memory and Language 50 (2004) 268¿88 www.elsevier'corr/locate/jml

ELSÐ/IER

Processing discourse roles in scripted narratlves:

w;ld knowledgtoThe influences of context '"d k

Anne E. Cooku'* and Jerome L' MYers"

Rnt 327' (rniuersitlt of utùlt'

,,r,,;"r;rl;:",nï:',jr';)i:ti'r';:r'þ!'¡-'',, Departrnent of Educationar

of M as sachusett s, U S A" i' Ll,ri,)¿,s¡¡1, "

Recerved 13 August 2003; revision teceived 13 November 2003

rn o IAbstract

maymay afiêct compreltt'rrùrv:' "i.1,.contextDiscourseDiscourse contextcontext lrav .'wvl '--^^r -' in ,*o experiments, we used:::*,. affect :?T:*:.:-:,1:i.i,lï',*i;;ä:,Liåîî.3ïî3"':î'""îlåïiï;i":J"iryjï'#ätrtext. In two expe'm't*' y",":î ^"-j:;;;;-; context had an early

,.eí,,ìã" "i ti'e tarset ":ï"1-l-î:t:*"'li.,:1iäfi." narratlves' t" llll::';g*x*ti.'i::ï",il:1il-.,å:i"13*flr'*L'r,ïïiîkru;:;*:dl,öïîT{:i:r:*rt#:ï:ytr::ïåxï.iijå:"Í::i;il"il.e.áiio" år'"1:-o'::' inin scriptedscripted narrauvcs' '.î*ir, *¿'i""ted that semantic knowledge

innuence on integra'on .r 'i;ñt1";i¡"-:"'liå i:"åîff;:.'i;":i'iiiiH:äiË;ä ;i'::':lllîl iîi:å'',ï å1fälä'åï'ffi";li:':+:i.'""å[i-l#",î,îi: i:iffi1::J"J:'ä{:,í*Xlgjnt'ïå',ïiJiÏi] on

.'eming text is rinked to the

continued to influence rn - orooe modol of processing: in th*ff:.ï"''Tl'fi3äi"Ïì.r"'^",'"*"i:.::::*'"n$:ï:ËXiå,',',Ël'l',i1"ã'îi 'i,*., -^.:^- ,o êr¡â'ìâevaluated't"î"rmation isis evaluated reading. The data upptu'-iJ."'ppol:::î".r,t3.:ttl. r,t between the new ";;;h

i"i";mation;äå"iiï: ä;ffiö;ä'*nn"" ",y"--..,ï..il"fft_i.ä".'.iïi,''*;;;;'othe secold stage' tn

contents of active memory; in äïoö;ilJui.' Inc' All rights reserved'

Many factors influence the speed of processing indi

uioïuiio.¿,."".1"-'1ïllå,iiåîîî:,::ä:;;ffi: factors are inhetent tn tne

it is well established that words that are more frequent rn

,i'ã e¡er*t """iî: iïilf ïi',ft:lî51'-'lä'Ï::

i,ï.:'J::l "üiiifr*i;f åi ll åru J'"ïïfl fr fii."*'*¿ i' ir'" i"n'""ce or

nor the information p'ou'oJå- tv the rreie!119^1e1äåtitr 'ïi",ii::å:

a passage whose protagonists are rn a

example, consider

-----------------;'po*n, of this research were completed-wliile Amre cookt"

was supported bv NIMH rì"i'*e c'"* t|-1Í1t îîïo* äîöïp'",,*"",-rr".n".";|å:Jü'ïîlli":l,ii;iï; setts. Additional sìrppoft was p'","'iî i:.-ã^^k ive would Research Committee grant awarded to Anne Cook' tÀ

like to thank Chuck Clittoni;; ¿;th Ravner for their helpful

ä*t*,t":lllï ,i',îTî iilrì iäi ii ill i,.(A E' Cook)'

E¿nail atilress: nnnt'èootg"a'utah'edu ",

sen­rock band Assume that the reader encounters

*ut pl.t¡* by thî^;:'a tence that begins "The "otn'o"g;*';'ät.:îchank & Áitrton' térr¡ ana data (e's''

;;';**,;ñ;kj*'lf ïiiÍ,?,îif îï.iî::ilJ:i:i;knowledge of the role ntl

or racilitate integration of' the sentence:iil;ä;' continuation' gttitar tst'.ifttttf"it' we would expect that

thetimespentprocessrngtheappropriate*"lt:Ì:t:""tist'o*ãr than that spent processlns an

ät;;;;,i;;idt' t:li:^t:' n"*­inappropriate toot'noution' ^o'og"' ever, that earlier rn ,n" purrlg., the manager had played

that tñe text included a sentence;hJil;. õ,

"tppo" åf the band members handled that indicated titut 'o"" oublicity and finances tn ã¿¿i'ion to performing' In the

such experiments reported nåtoït it"*tti"Í llt:ler contextual manipulation''huut un effect and' if so' the

nuirrr. un¿ time course of those effects'

In the rock Uut'o t*u*ff"' there are two of 'sourcesrn1wle'dfl of rock

information available iäìüt 'L"¿tt' band scripts links the goitu'itt to the role' of performer'

whereas the preceding t"*t1t *"']ip.Yr,tlat in thn

passage the usual t""fta-'ott' are not valid' Theories

0749-596>0$ - see front l1tlt' O 2003 Elsevier Inc All rights reserved

doi:10. 1016/j jml 2003'1 1 uul

269 A.E. Cook, J.L. l{1'sys I Jountul of Mentory and Language 50 (2004) 268,288

about discourse processing differ with respect to the priority they assign to the two sources of information, semantic knowledge, and discourse context. Many the­orists have also used the terms "Local" context effects and "global" context effects to refer to semantic and discourse influences (e.g., Hess, Foss, & Carroll, 1995). However, the terms "local" and "global" have also been used to refer to a variety of other linguistic factors (e.g., discourse coherence, distance in text, causal informa­tion, etc.). To avoid confusion, here we associate local context effects with those caused by information reacti­vated from readers' semantic knowledge, and global context effects with those caused by information reacti­vated from the episodic discourse context.

Kintsch's (1988) construction-integration model in­corporates both immediate effects due to the semantic properties of text and later effects due to discourse context. In the construction stage ofthe model, concepts and propositions in the reader's knowledge base are activated as a function of their relation to what is cur­rently in the reader's focus; the prior discourse context plays its role in the later, integration, stage. Till, Mross, and Kintsch (1988) presented data in support of this two-stage process. As in other studies, at short intervals associates of ambiguous words in sentences were primed whether or not they were related to the discourse con­text. At longer intervals, words unrelated to the overall discourse context ofthe passage were no longer primed; however, words that were inferentially related to the topic were primed although they were not lexical asso­ciates of words in the preceding sentence.

Similarly, Garrod and Sanford (1999) have argued in favor of immediacy in discourse processing, in which "language input is related to world knowledge at the earliest possible opportunity" þ.23). Sanford and Gar-rod (1989) reviewed several studies of anaphoric pro­cessing and argued that the majority of the evidence pointed to a two-stage process, with the first stage be­ginning immediately upon encountering an anaphor (also, see Garrod & Sanford, 1983; Sanford & Garrod, 1981, for earlier discussions of stages involved in ana­phoric processing). The process in the first stage was referred to as "bonding," in which potential antecedents (whether correct or incorrect) may be quickly reacti­vated on the basis of their semantic relation to an ana­phor through a lowlevel pattern-matching mechanism. The match made during the bonding stage is verified, and if correct, is instantiated, during the second stage, resolution.

Garrod and Sanford (1999) proposed more general versions ofthe bonding and resolution stages, which can apply to processing at the lexical, syntactic, or discourse Ievels. Consistent with this idea, Garrod and Terras (2000) demonstrated that bonding and resolution could be extended to processes beyond anaphor resolution. They required participants to read sentence palrs 1n

which the context provided by the ûrst sentence was either appropriate or inappropriate with respect to a phrase in the second sentence. An example of the f,rst sentence is (1a) Tlrc teaclter was busy writing a letter of complaint to

a parent.

._. ?'_.(Ib) I he teacher was bttsy writing an exercise on the blackboard.

The second sentence contained the phrase the pen dropped which forms an appropriate pairing with sen­tence la but not with sentence lb. Eye fixation times revealed no initial effect of appropriateness on process­ing the fiotrn, pen. However, times in the region of the verb, dropped, and regressions from it back to the noun indicated difficulties in the later stages of processing when sentence 1b provided the context. Garrod and Terras explained their results in terms of the two stages originally proposed by Sanford and Garrod (1989): in bonding, an early automatic associative process links a potential role filler (pen) with a verb encountered earlier (ruriting), and a later process, resolution, evaluates and resolves the instrument-verb link in terms of the dis­course context.

Using eye tracking technology, Morris (1994) and Duffy, Morris, and Rayner (1988) provided evidence that both lexical and contextual information can affect early lexical access processes. Morris presented readers with sentences such as, (2a) The gardener talked as the barber trimmed the

mustache after lunch. (2b) The gardener talked to the barber and trimmed the

mttstache after lunch. both of which contain identical content words (e.g., barber, trimmed) with strong lexical associations to mustache. But, in sentence 2b, mustaclte is not consistent with the overall message of the sentence-barbers usu­ally trim mustaches, not gardeners. In contrast to pre­vious demonstrations of serial effects of semantic and contextual influences, Morris found that both intralexi­cal associations and messageJevel context facilitated early processing (i.e., gaze durations) for mustaclrc. Similarly, Duffy et al. (1988) provided evidence for both lexical (e.g., meaning dominance) and contextual influ­ences (e.9., prior disambiguating context) on early pro­cessing of lexically ambiguous words, and proposed that these factors may contribute to a reordered access of lexical information during reading.

In contrast to models that assume early stages may be mediated by lexical associations, Foss and his col­leagues (Foss, 1982; Foss & Speer, 1991; Hess et a1.,

1995; Traxler & Foss, 2000) have proposed that initial access is facilitated by the discourse context. Perhaps the strongest pieces of evidence for their position are the results obtained by Hess et al. Using a cross-modal priming procedure, they presented participants with

210 A.E. Cook, J.L. M¡tst'5 | Journal of Mentory and Language 50 (2004) 268-288

sentences that either had an association between the last word and the preceding text, such as,

(3a) Tlte Englßh major wrote the poem or sentences that lacked such an association, such as

(3b) The computer science major wrote the poem. When sentences such as 3a were presented alone, the

final word, poem, was pronounced aloud more quickly than when it ended a neutral sentence such as,

(3c) The boy read tlte poem. However, when sentence 3a ended a passage in which

the discourse context was not relevant e.9., a passage

about an English major in a computer science course-no facilitation was observed. On the other hand, if the passage placed the computet science major in an English course, poem in sentence 3b was named more quickly than when it ended a control passage. In discussing their' results in relation to models that assume two stages,

based on semantic knowledge and discourse context ef­fects, Hess et al. stated that these models ". . . make two incorrect predictions: They predict a local effect where none is observed, and they predict a slow-acting global effect where in fact a fast-acting one occurs" (p. 78). The authors proposed a lexical reinterpretation model (also,

see Foss & Speer, 1991) in which discourse context modifies the representation of a concept. For example, an English major taking a computer science course may be viewed as a "student-in-a-technical-coutse-on-com­puting (not the poetic related information)" (p. 80).

Accordingly, the talget word, poern, is not readily inte­grated into this representation and therefore receives no facilitation.

One limitation of the Hess et al. (1995) study, and of many other studies of contextual effects, is the use of probe techniques. Although useful, probe response times provide only one sample in the time course of processing

and may result in an incomplete or possibly even dis­torted view of the effects of context on processing of words in discourse (see Gordon, Hendrick, & Foster, 2000, for additional issues concerning strategic pro­cessing when probe response methodologies are used).

For example, if Garrod and Terras (2000) had only probed with the target word pen, they might have con­cluded that context had no eflect. By monitoring eye

movements, they found that the context had no initial effect but that an inappropriate context intelfered with subsequent plocessing of the talgeted word.

The goal of the present set of experiments was to use

several eye tracking measures in different regions of a

text to further our understanding of context effects on the processing of words in connected discourse. In two experiments, we analyzed eye movement data for tar­geted regions of passages that evoked familiar scripts (e.g., rock band). The regions of interest were the one-word target role flllers (e.g., guitarist, manager) that were either appropriate or inappropliate with respect to the speciûc action being performed in the current sen­

tence and either appropriate or inappropriate with re­spect to the previous discourse context. We also

examined processing times on the post-target regions, the few words that immediately followed the target in the text. If incoming text is linked via lowlevel asso­

ciative processes with general world knowledge before it is integrated with the discourse context, as suggested by Garrod and Terras (2000) and Kintsch (1988), we

should obselve effects of appropriateness with respect to the action described in the current sentence in early stages of processing (e.g., ûrst pass on target regions), and discourse context effects should be observed in later stages of processing (e.g., first pass on post-talget re­

gions, second pass, regressions). In contrast, ifdiscourse context influences early stages of word plocessing, as

suggested by Foss and colleagues, discoulse context ef­fects should be observed in early stages of processing.

Experiment 1

The passage in Table 1 illustrates the two factors manipulated in Experiment 1. Note that there are two encounters between the protagonists and an individual filling the role of a performer. Our primary interest in this experiment was in participants' eye movements while they processed the second encounter. The action described there (e.g., playing a song) is carried out either by an individual appropliate to the scripted role (the

audience listened Io the guitarist, Ã2) or by someone

inappropriate to the role (the audience listened to the manager,I2). Also, the earlierfrst encounter contained either an appropriate (the grdtarist played the song, Al) or an inappropriate (the manager played the song, I1) role filler. In summary, two factors were manipulated in the second encounter: the appropriateness of the role tller in the earlier first encounter, and the appropriate­ness of the role ûller in the second encounter.

When the second encounter in lhe A2 condition is

consistent with both context and semantic knowledge-that is, when both first and second encounters are ap­propriate to scriptal knowledge-integration of the sec­

ond-encounter role flller should proceed easily, and therefore flrst pass reading times should be faster in the A1/A2 condition than in the Al/I2 condition. What diflerentiates models are their predictions when the first encounter is inappropriate. According to the Hess et al. (1995) lexical reinterpretation model, when the manager plays the song, the reader reinterprets the role of per­folmer; the manager is incorporated into the discourse as someone who also fills the guitarist's role.If this is so,

on the second encounter the inappropriate lole filler (I1/ 12 condition) should be plocessed as fast or faster than the appropriate role flller (I1/42 condition). Having re­interpreted the manager's role, readers should now find it as easy, or easier, to integrate tlte manager than the

A.E. Coolc, J.L. Myers I Journal of Mentory and Language 50 (2004) 268-288

Table I Versions of a sarnple passage for Experiment I

httroduction The fans at the rock concelt were going wild. They quieted down when the next song

Fit'st encounter Appropriate (AI): was played by the band's guilarist. Inappropriate (Il ): was played by the band's m(tnager.

Background The concert was being held in an outdoor stadium, and there ivere about ten thousand screaming fans in attendance. They were mostly all hysterical teenage girls. Some even cried as they

Seconcl encounter Appropriate ( A2 ): lisfened þ the guitaríst Inappropri.ate (12); hsTened lo The ntanager

Conclusion Play a slow and heartfelt love song. The song was about a love affair that ended badly.

Note l. 'lhe target region in this example is guit(trist or manuger; the post-target region (second encounter) is plq, a slow. Note 2. ln Experiment 1, half of the participants read a similar passage in which the manager was appropriate and the guitarist was

iñennrnnricte

gltitarist with the action of playing the song. Although Hess et al. would not necessarily predict the role rein­terpretation assumed here, such a result would be con­sistent with their model. It would also be consistent with the assumption that situation models are updated on the basis of textual information (Zwaan & Radvansky, 1998). However, models that assume an immediate effect

of world knowledge (e.g., Garrod & Sanford, 1999; Kintsch, 1988) would predict that an inappropriate second encounter will be reflected in longel reading t1mes.

Because none of the models are specific about the time course of processing, it is not clear what their pledictions would be for processing after the eyes move past the role filler. Integration may be completed while readers are fixated on the role ûller, or it may continue into subsequent fixations on the words following the role filler or on the role flller itself. In the present experiment, eye movement measures provide a means of assessing

context effects after the initial processing of the role fil­ler. Continued effects ofeither flrst-encounter or second-encounter appropriateness should be reflected in time spent in the region following the target word, ln re­gressions back to the target word, and in the time spent reprocessing the target after regressing back to it.

In summary, Experiment t had two purposes. The first was to examine the effects ofthe appropriateness of the first encounter on processing of the second encoun­ter. If such effects were obtained in early measures of processing (e.g., flrst pass), they would suggest a limi­tation of models that assume only semantic priming by the immediately preceding context and would further extend the applicability of Hess et al.'s (1995) assump­tion of lexical reinterpretation. However, a disruption of processing when the second encounter was inappropri­ate and the flrst encounter was also inappropriate would

be inconsistent with the lexical reinterpretation hy­pothesis. The second purpose was to use the measures

derived from records of participants' eye movements to examine the time course of anv effects that were ob­served.

Method

Participants Thirty-eight members of the University of Massa­

chusetts community participated in exchange for either money or course credit. The data for six participants were not analyzed due to large calibration errors during portions of the experiment.

App0ratus Eye movements were recorded by a Fourward

Technologies Dual Purkinje Eye tracker that has a res­

olution of l0min of arc. The eye tracker was interfaced with a 486 microcomputer, which ran the experiment. Viewing was binocular, with eye location recorded from the right eye. The position of the participant's eye was sampled every millisecond by the computel and aver­aged over four consecutive samples. The avelaged hor­izontal and vertical positions of the eye were compared with those of the previous sample to determine whether the eye was fixated or moving.

Materials The materials used were 24 passages similar to the

example plesented in Table 1 (see also Appendix A). Each passage began with a sentence that introduced a

script (e.g., rock band). In the next (flrst encounter) sentence, a scripted action was associated with an entity or an object that was either appropriate or inappropriate for that role. Readine times were measured for the role

272 A.E. Cook, J.L. M¡,s¡5 I Journal o.f Memory atzd Language 50 (2004) 268-288

filler (e.g., guitarisî) in the f,rst encounter. The role filler was always a single word, with a range of 4-1 1 charac­ters (mean:7.06 characters). A background section of two to three sentences followed. The next (second en­counter) sentence contained a scripted action that was associated with either an appropriate or inappropriate role filler, and which either matched or did not match the role f,ller in the first encounter. Thus, the appro­priateness of the role filler in the second encounter was crossed with the appropriateness of the role filler in the flrst encounter to create four conditions: A1/42 (both encounters appropriate), A1/I2 (flrst encounter appro­priate, second encounter inappropriate), 11/42 (ûrst encounter inappropriate, second encounter appropri­ate), and I1/I2 (both encounters inappropriate). In the second encounter, reading times were measured for the target word (the role filler; e.g., guitarist) and the post-target region, which consisted of the two to four words (mean:14.17 chancters; e.9., play a slow) that imme­diately followed the target. A short closing sentence ended the paragraph.

Eight materials sets were constructed. These were counterbalanced such that the appropriate and inap­propriate actions for the role filler in the first four sets (e.9., gr.titarist) were the inappropriate and appropriate actions, respectively, for the role filler in the second four sets (e.g., manager). In each materials set, six passages

appeared in each of the four conditions. Across each group of four materials sets, each passage appeared once in each of the four conditions. -fhe 24 experimental passages always appeared within a larger set of 72 pas­sages. The 48 "flller" passages were stimuli for two un­related experiments; one set of flller passages contained anaphoric references to objects previously mentioned in the text, while the second set of filler passages involved bridging inferences. The filler passages were approxi­mately the same length as the experimental passages but did not focus on scripted settings.

Rating study la. It order to ensure that the scripted actions selected were appropriate or inappropriate with respect to the role fillers, a rating study was conducted. Forty University of Massachusetts community members who did not participate in either of the experiments re­ported in this paper were asked to rate the likelihood of an action being performed by a speciflc role filler. Par­ticipants raled 32 items such as 3a-3d below on a 7­point scale (where 1 : "Extremely Unlikely" and 7 : "Extremely Likely"): (4a) At a rock concert, teenage fans cry as a sad song is

played by the band's guitarist. (4b) At a rock concert, teenage fans cry as a sad song is

played by the band's manager. (4c) For a rock band, all ofthe legal issues are handled

by the band's manager. (4d) For a rock band, all ofthe legal issues are handled

by the band's guitarist.

Appropriate role fillers (e.9., 3a and 3c; mean: 6.13) were rated as signiflcantly more likely to perform a

scripted action than inappropriate role flllers (e.g., 3b and 3d; mean:2.57), ¡'(1,31) --349.42, MSE:1.158. The items from this norming study were used to create the 24 passages used in Experiment 1.

Ratíng stLtdy lb. An additional rating study was conducted in order to ensure that a switch in role fillers between the first and second encounters was less plau­sible than the same role filler being maintained in the first and second encounters. Twenty-four University of Utah undergraduate students enrolled in Introductory Educational Psychology courses participated in ex­change for course credit. Participants were presented with one of four counterbalanced stimulus sets, in which six passages appeared in each of the four conditions, up to the target sentence. Participants were then asked, gi­ven the information in the passage, to rate the likelihood of the target action being performed by the second-en­counter role filler on a 5-point scale (where I : "Ex­tremely Unlikely" and 5: "Extremely Likely"). If maintaining the same role filler across both encounters is more likely than switching role frllers across encounters, ratings in the AlA2 and I1I2 conditions should be higher than in either the AlI2 and llÃ2 conditions, re­spectively. The results supported this hypothesis. Rat­ings in the A1A2 condition were higher than in the AlI2 condition (3.75 vs. 1.93; F1(1,23) :91.96, MSE: .4332; F2(1,23) :124.32, MSE:.3072).In addition, ratings in the I1I2 condition were higher than in the AlI2 condi­tion (3.17 vs. 1.93; Ft(l,22) :39.22, MSE:.4704; F2(1,23):30.05, MSE:.5914). Mean ratings for all passages are presented in Appendix A.

Procedure Each participant took part individually in a sesslon

that lasted approximately 60min. For each participant, a bite bar was prepared to eliminate head movements, and the eye tracker was calibrated. The initial calibra­tion procedure took approximately 5 min. Prior to reading each passage, calibration of the eye tracking system was checked to ensure that accurate records were obtained. Each participant read three practice passages followed by the set of 24 experimental and 48 flller passages. Participants were told that they would be reading a series of paragraphs displayed on a CRT screen. They were told to read for comprehension so that they would be able to answer an occasional oral comprehension question (e.9., Were the fans enjoyíng the concert?). Questions were asked on approximately one-sixth of the trials, and there were an equal number of "yes" and "no" comprehension questions. Participants had little difficulty answering the questions correctly; all participants answered at least 85Y" of the questions correctly. Because of the high accuracy rates, the fixa­tion data for all trials were analyzed, regardless of

273 A.E. Cook, J.L. Myers I Journal oJ Mentory und Language 50 (2004) 268-28g

v/hether a question for a speciûc trial had been answered & Clifton, 1989). In Experiment 2, we remedied this correctly. by including several words after the target prior to the

At the beginning of each trial, five boxes appeared end of the sentence. across the top of the screen, one box appeared in the middle, and flve boxes appeared at the bottom of the Second encounter screen. Each participant was instructed to look at the middle box until the experimenter said, "Ready," and First pass times in the target region then to look at the upper left-most box. Once the ex- 'lable 2 presents the first pass times in the target and perimenter had determined that the participant was post-target regions as a function of the appropriateness fixating the box, the entire passage was presented on the of the earlier ûrst encounter and of the current second screen. When the participant was finished reading the encounter. First consider the appropriateness effect on passage, he or she was instructed to press a button that second-encounter target times when the first encounter would end the trial. Participants were given a brief break had been appropriate; this is the contrast between the approximately halfway through the experiment. AllÃ2 and A1/I2 cells in the target region of Tab\e 2. In

these conditions, first pass times were shorter when the Results and discussion second encounter was appropriate than when it was

inappropriate. Conversely, in the right-hand column First pass reading time, the sum of all flxations in a (ûrst encounter was inappropriate, I1A2 vs. I1I2), sec­

region before the eyes first leave that region, is reported ond-encounter times were shorter when the second for two regions: the target and the post-target region. encounter was inappropriate than when it was appro-Trials in which a region was initially skipped were not priate. Simply put, times were shorter whenever there included in first pass analyses. In addition, percentage of was a match of role flllers between the two encounters. regressions made into the target region and second pass This interaction was significant; Fr (1, 30) : 5.31 ,reading times, any rereading of the target region after MSE : 3030, F20, 46) : 9.83, MSE : 5822. the eyes have regressed back to that region, are reported. The fact that times were faster when first- and sec-In all analyses reported, Fl always refers to tests against ond-encounter role flllers were the same suggests facili­error terms based on participants variability, and F2 tation due to word repetition. Effects of roughly this always refers to tests against an error term based on magnitude have been found in text processing experi­items variability. All analyses were signìficant at the ments (Rayner, Raney, & Pollatsek, 1995). It is some-standard alpha level of.05, unless otherwise indicated. what surprising, however, that the absolute differences in

the A1 and Il columns of Table 2 were almost identical First encounter (296 - 27 5 : 21; 3Q4 - 281 : 23). Although this could be

a random occurrence, the lack ofa difference implies no When the role frtler (e.g., gttitarÌst of manager) was first effect due to the inappropriate role flller on the second

encountered in the text, first pass times were significantly encounter although such an eflect was obtained on the shorter in the appropriate (mean:318ms) than in the first encounter. Hess et al.'s (1995) lexical reinterpreta­inappropriate (mean:346) condition: Fr (1,30) -- 5.52, tion hypothesis suggests an explanation that can en­MSE:4747.72, F2Q,46) : 4.85, MSE:8783.24. Sec- compass effects due to both repetition (i.e., the discourse ond pass times, post-target region reading times, or re- context) and appropriateness of the role flller. Under gressions into the target region are not reported for that hypothesis, for example, if the managet played the the ûrst encounter because the target was followed im- song at the ûrst encounter, then that individual maymediately by a period and any post-target measures may have been reinterpreted as fllling the role of guitarist and have also reflected sentence-wrapup processes (e.g., Just consequently was not truly inappropriate on the second & Carpenter, 1980; Rayner, Sereno, Morris, Schmauder, encounter; that is, although the manager was inappro­

-lable 2

Experiment l: first pass times (in ms) in the second-encou.nter regions

Second encounter Target region Post-target region

First encounter First encounter'

AI II Mean AI I1 Mean

2',7s 304 290 417 461 439 12 296 281 289 507 456 482 Mean 286 293 462 459

Note. A and I refer to appropriate and inappropliate; I and 2 to first and second encounter.

a1Á A.E. Cook, J.L. M),ers I Journal of Mentorlt and Language 50 (2004) 268-288

priate to the scripted action being performed in the second encounter, that individual was appropriate with respect to the previous discourse context. In addition, if such a reinterpretation took place, the guitarist, al­though appropriate to the scripted action, might be

viewed as inappropriate on second encounter in light of the previous discourse context. This analysis suggests

that the discourse context and the appropriateness of the role filler interacted to influence frrst past reading times.

First pass times in tlxe post-target region The mean duration of first pass in the post-target re­

gion was significantly longer when the second encounter was inappropriate (I2) than when it was appropriate (42); Fr(1,30) :4.21, MSE:13,780, F2(1,46) :4.19, MSE :22,452. This effect is clearly due to the slower times in the Al/I2 condition than in the AllA2 condition. ¡'r(1,31) : 5.55, MSE:23,415, F2(1,47) :8.15, MSE :23,629; there was almost no difference when the ûrst encounter was inappropriate (lllA2 vs. Il/I2 conditions, both .Fs < 1). The interaction was significant;.F¡ (1, 30) : 4.17, M S E : 17,396, F2(1, 46) : s.96, M SE : 16,ss9. lt is clear from the interaction that the flrst encounter ln­fluenced later processing. This effect of context extends the Hess et al. (1995) results. They used two sentences

to establish a discourse context whereas we changed a

single word (i.e, the first encounter), and Hess et al.'s relevant discourse context immediately preceded their target sentence whereas the flrst encounter was separated from the target region by several lines of unrelated text in the present experiment.

The A1/I2-All42 difference is consistent with the assumption that the process of integrating the role filler with the preceding text continued into the post-target region. It is not clear whether this integration process

has been facilitated in the A2 condition, disrupted in the 12 condition, or is the result of a combination of both facilitative and inhibitory effects. The idea that pro­cessing was disrupted in the 12 condition seems most likely; the reader had to incorporate an inappropriate role filler into the discourse model. When the role f,ller in the first encounter was inappropriate, this may no longer be a problem because the reader has previously encountered the manager in the guitarist's role. In this view, the post-target means are consistent with Hess

Table 3

et al.'s (1995) assumptions that context can lead to a

reinterpretation of a character's role, and that context affects integration processes, as weli as with the updating assumption proposed for situation models (Zwaan & Radvansky, 1998). However, as the measures we con­sider next indicate, the nature of context effects may be

somewhat more complicated.

Regressions into, and second pass tùnes in, the target regton

In calculating these statistics, the target region was

redeflned to include the entire prepositional phrase; the target referred to in Table 3 is, for example, to tlrc gui­tc¿rist, rathet fhan guitarist This was done because ap­proximately 20Y" of regressions landed on to the rather than on guitarist. Considering the regression percentages

in Table 3 flr'st, we found that readers regressed more often to the target region when the second encounter was inappropriate; fl (1,30) : 22.48, MSE:100.95, ,Pr(1,30) :10.64, MSE:301 .02. No other main effects

were significant. Note that, unlike the results for first pass times in both the target and post-target regions, these data indicate that even when the first encounter was inappropriate, the inappropriate second encounter involved more processing difficulty than did the appro­priate second encounter; that is, there were more re­gressions to the target region in rheIllI2 condition than in the 11/42 condition. The contrast of these two con­ditions was significant when tested against participant variability and nearly so when tested against item vari­ability; Fr ( 1, 30) : 4.62, M S E : 19 5.39, Fr(l, 46) : 3.52, MSE:s91.20, P: .067.

The second pass results in Table 3 also indicate in­tegration difficulty after initial processing of the target region. Readers spent more time on second passes in the target region when the flrst encounter was appropriate than when it was inappropriate; Fr ( I , 30) : 5.98,

M SE : 5184, F20, 46) : 6.2a, M SE : 5903. Second-en­counter second pass times were also longel when the second encounter was inappropriate than when it was

appropriate; ¡'r (1,30) : 16.64, MSE:3968, F2(1,46) : 11.96, MSE:6889. Although the second-encounter ef­fect is larger when the first encounter was appropriate, largely due to the very long duration in the Al/I2 cell, the interaction only approached signiûcance when tested

Experiment 1: percent of regressions into, and second pass times (in ms) in, the target regiolì

Second encounter Regressions Second pass times

First encounter Fìrst encounter

A1 I1 AI II Mean

t\¿ T2

Mean

9.47

21.00 t5.24

7.41

12.7 5

10.10

8.46

I 6.88

36

r00 68

L)

5l 37

30

75

275 A.E. Coolc, J.L. M),ers lJottrnal oI Metìrcr)'úttd Løtguilge 50 (2004) 268-2BB

against participant variability; fi (1,30) :2.90, MSE-­3592, p: .10. Tests of the simple eflects of the second-encounter appropriateness factor were also conducted. The difference between appropriate and inappropriate second encounters was signiflcant when the flrst encoun­ter was appropriate (¡'r (1,30) : 10.92, MSE-- 11,806, F2(1,46) :6.27, MSE:24,478) and when the first en­counter was inappropriate (Fr (1,30) :7.24, MSE: 3314, F2(1,46) :4.38, MSE: 7575).

These results, particularly the findings that when the second encounter was inappropriate, readers regressed more often to the target region, and spent more time there after returning to it, seem to conflict with the ûrst pass results. Those data indicated that when the initial encounter had been inappropriate. an inappropriate second encounter proved no more difficult to process

than an applopriate second encounter. The results in Table 3 cannot be readily explained by theories of fa­cilitation due to priming (e.g., Duffy et al., 1988; Duffy, Henderson, & Morris, 1989; Morris, 1994; Schwanen­flugel & Shoben, 1985; Schwanenflugel & White, 1991); expectancies generated for a specific role filler may fa­cilitate flrst pass times on the target, but regressions and second pass times indicate difficulty in integrating inap­propriate targets. These results also seem inconsistent with Hess et al.'s (1995) lexical reinterpretation hypoth­esis, or withZwaan and Radvansky's (1998) assumption of situation model updating. If the manager was rein­terpreted as ûlling the guitarist's role, why is Íhere any evidence of difficulty in processing a later mention of the manager carrying out a guitarist's duties? Before further

Table 4 Versions of a sample passage for Experiurent 2

Appropriate (A) Versions Introduction

discussion of the lesults of Experiment l, we first con­sider the results of a different manioulation of context.

Experiment 2

One purpose of Experiment 2 was to replicate the ef­fects offlrst-encounter appropriateness on first pass times on the target, in addition to examining the appropriate­ness effect on other measures. In Experiment 1, the ûrst encounter with the target word occurred at the end of a

sentence. Because ofthis, we could not be certain that any measures of post-target processing (e.g., post-target first pass times, regressions to the target, second pass times on the target) wele not confounded with sentence \¡/rap-up plocesses. As was seen in the data of Experiment 1, such post-target processing measures may reflect different ef­fects than those on the first pass target times and can provide a more complete pictrue than just first pass times or probe response times alone. Therefore, the target sentences of Experiment 2 always included several words after the target word and prior to the end ofthe sentence, to allow for examination of post-target processing.

The primary purpose of Experimenl2 was to exam­ine further the effects ofthe conflict between scriptal and contextual knowledge. There are circumstances in which a role in a script may be filled by someone ol something other than the standard role frller. For example, in the inappropriate/justification (IJ) condition of Table 4, the manager (instead of the guitarist) played the song but the action is justifled by the preceding statement that

The fans at the rock concert were going wild. The band was still new on the lock scene,

Justification Neun'al (N). and they weren't used to the adoring fans and all of the attention from the media. Justified (-/./: so some of the band members had to handle publicity and finances in addition to performing.

Conclusion The crowd quieted down when the next song was played by the band's guitarist who was standing alone on the stage. The concert was being held in an outdoor stadium, and there were about ten thousand scleaming fans in attendance.

Inappropriate (I) Versions Inn'oduction The rock band was becoming more and mole popular. The band was still new on the rock scene,

JustiJication Neun'al. ( N ): and they weren't used to the adoring fans and all of the attention from the media. Justified (-/): so some of the band members had to handle publicity and finances in addition to performing.

Conclusion A contract was signed by the band's guitarist who had arranged a six-record deal. The band had been getting lots of calls recently, and they were having a very hard time figuring out how to handle theil new success.

Note. The target region in this exarnple ts guitarist and the post-target region is wlø was standínglwln lrud arranged

276 A.E. Cook, J.L. M)¡ers I Jountal of Mentory and Language 50 (2004) 268-288

"some of the band members had to handle publicity and flnances in addition to performing." The results of the effects of appropriateness on the ûrst encounter in Ex­periment 1 indicate that in the absence of any prior justification-a departure from the reader's knowledge of scriptal roles increases first pass times on the role filler. A theory such as Kintsch's (1988) construction-integration theory, that assumes that the reader's world knowledge is consulted before any attempt to integrate any new input with prior text, would predict that this increase in ûrst pass times would be observed even when the ìnappropriate role filler had received prior justifica­tion. On the other hand, it is possible that the justifica­tion creates an expectancy, either for the alternative role flller, or at least that the scriptal role will not be fllled in the usual way. Such an expectancy seems consistent with models which suggest that the discourse context may constrain or influence semantic priming effects (e.g.,

Duffy et al., 1988, 1989; Morris, 1994; Schwanenflugel & Shoben, 1985; Schwanenflugel & White, 1991; see also MacDonald, Pearlmutter, & Seidenberg, 1994); accord­ing to this view, in the present study, the appropriateness effect would be smaller in the justifled than in the neutral condition. The facilitated integration model proposed by Foss and his colleagues (Foss & Speer, 1991; Hess et al., 1995) would make a similar prediction. On the basis of their results, Hess et al. hypothesized that integration with the context occurs very rapidly, and that contextual effects override world knowledee.

Method

Participants Thirty-five members of the University of Massachu­

setts community participated in exchange for money or course credit. The data for three participants were not analyzed, due to large calibration errors during portions of the experiment.

Apparattts tn. .nutn-ent used was the same as in Experiment

,.

Materials The materials used were altered versions of the pas­

sages from Experiment 1. The first sentence introduced the script (e.g., rock band) and was followed by a context sentence. The context sentence either justiûed a scripted action's association with an inappropriate role filler, or it was neutral with tespect to actions and role flllers but continued the stoly. The next sentence contained a scripted action that was associated with an appropriate or inappropriate role flller', followed by several words prior to the end ofthe sentence. (See Appendix B for all passages in Experiment 2.) Note that, in contrast to Experiment 1, this was the first and only encounter with

the role filler in the text. Reading times were measured for the target word (the role filler; e.g., guitarist),which ranged from 4 to I I characters (mean:6.79 characters), and the post-target region, which consisted of the three words that immediately followed the target (mean: 14.27 characters; e.9., who was standinglwho lrud atanged). Thus, context type was crossed with the ap­propriateness of the role filler to create four conditions: AN (appropriate neutral), IN (inappropriate neutral), AJ (appropriate justified), and IJ (inappropriate justi­fled). One to two closing sentences ended the paragraph.

The between-subjects manipulation in Experiment I produced no significant effects, so it was dropped in Experiment 2. Four materials sets were constructed, such that each materials set contained six passages that appeared in each ofthe four conditions. Across the four materials sets, each passage appeared once in each of the four conditions. The 24 experimental passages appeared within a larger set of 54 passages. The 30 "flller" pas­sages were stimuli for an unrelated experiment; they were similar in length to the experimental passages but did not focus on scripted settings. The number of flller passages used in this experiment was reduced from the number used in Experiment 1 in order to shorten the overall amount of time required for the experiment-this helped reduce participant fatigue and allowed for better precision on the eye tracker.

In order to demonstrate that the justifying context was strong enough to override the inappropriateness of the role flller, a rating study was conducted. Twenty-six University of Utah undergraduates enrolled in Intro­ductory Educational Psychology courses participated in exchange for course credit. Participants were presented with one of two counterbalanced stimulus sets, in which 12 passages appeared in each of the IN and IJ condi­tions, up to the target sentence. They were then asked, given the information in the passage, to rate the likeli­hood of the target action being perfolmed by the role filler on a 5-point scale, where (1 :Extremely Unlikely and 5: Extremely Likely). If the justifying context was sufficient to override the inappropriateness of the role filler, ratings should be higher in the IJ condition than in the IN condition. This was supported in the results: ratings in the IJ condition were signiûcantly higher than in the IN condition (3.29 vs. 2.45; Fy(1,25):a5.12, MSE:.185; ,4(1,23) :71.846, MSE: .116). Mean ratings for all passages are presented in Appendix B.

Procedure The procedure used was the same as in Experiment 1.

Results and discussion

First pass times in the target region Table 5 presents the first pass times in the target and

post-target regions as a function of the appropriateness

277 A.E. Cook, J.L. M)ters I Journal of Memory and Language 50 (2004) 268-288

Table 5

Experiment 2: first pass times (in ms)

Appropriate Inappropliate Mean(A) (r)

Target regíon Neutral (N) 280 304 292 Justiûed (J) 289 283 286 Mean 285 293

Post-target region Neutral (ìr) 416 525 501

Justified (J) 462 500 481

Mean 469 513

of the role flller and of whether the target action had been previously justified. First pass times are longer on the target (e.9., gtitarist) in the inappropriate than in the appropriate condition only in the neutral (i.e., unjusti­fied) condition. This resulted in a signiûcant interaction of appropriateness and justification; Fr (1,3l) :4.84, MSE:1561, F20,23) : 6.52, MSE:1546. Neither main effect approached signiflcance; both F's < l. However, the contrast of the neutral condition means (i.e., IN-All) was signiflcant; Fr (1 , 3l) : 4.31, MSE: 4243, F2(1,23) :5.12, MSE:4815. As in Experiment 1,

in the absence of any prior justification of the inappro­priate lole tller, flrst pass times were longer when the role filler was inappropriate than when it was appro­priate. When there was prior justiûcation, however, flrst pass times on the appropriate role filler were actually (nonsignificantly) slightly slower than in the inappro­priate condition. Thus, flrst pass times on the target suggest that any difficulties due to the inappropriate role flller were quickly resolved, either because the justiflca­tion facilitated lexical access to the target, or facilitated its integration with the discourse model. However, subsequent measures contradict this conclusion.

First pass times in the post-target region The pattern in the post-target region differed from that

in the target region in that times were slower in the inap­propriate than in the appropriate condition whether or not the inappropriate role flller had been justified in the pleceding text. Only the appropriateness main effect was significant; F'r(1,31) : 8.19, MSE :7311, F2(1,23) : 5.42, MSE:9004. Furthermore, the interaction ,E's were iess than 1. When there was no preceding justification, the il;rqer times in the inappropriate than in the appropriate .-r:irdition were a continuation of the eflect found in flrst pass times on the target; apparently, the eyes moved be­fore integration was complete. The simple effect contrast of the AN and IN conditions was more variable than the rnain effect ofappropriateness, resulting in a less powerful test; the contrast was marginal when based on partici­¡rants variability and nonsignificant when based on iterns valiability, F'r(1,31) : 3.85, MSE:9825, p : .059;

F2(1, 23) : 2.68, M S E : 9189, p : . I 1. In contrast, when the inappropriate role tller was justifled, the pattern in the post-target region differed from that in the target region; although there was no effect in the target region, times for the inappropriate role filler were (nonsignificantly) slower than for the appropriate role filler in the post-tar­get region: 4 (1, 31) :2.29, MSE:9997, p : .14; Fz(|, 23) : 2.94, MSE : 1r,840, p : .16.

Regressions into, and second pass times in, the target regton

Table 6 presents the mean percentage of regressions into the target region from the post-target region, and the mean second pass times on the target. As in Exper­iment 1, for these analyses the target region was ex­panded to include the two to three words immediately preceding the role f,ller. Although there were more re­gressions in the inappropriate than in the appropriate condition, with a larger difference in the justified con­dition, no ,E's were signiûcant; all p values were greater than .12. However, there was a more convincing effect of appropriateness in the second pass times on the target than in the post-target reading times or the regression percentages; second pass reading times were significantly longer in the inappropriate than in the appropriate condition; ¡(1,31) :4.68, MSE :3206, F2(1,23): 7.60, MSE:1617. All other -F's were less than 1.25. Thus the pattern of regressions and second pass times were consistent with the pattern of times in the post-target region in exhibiting more difficulty in the in­appropriate than in the appropriate condition. Most notably, despite the absence of an appropriateness efect on ûrst pass times on the target when it had been justifled, measures of subsequent processing indicated difficulty integrating the inappropriate role filler. One explanation for these results is that the inappropriate role flller was tentatively accepted due to the preceding justification, but a further evaluation of the inappro­priate role filler was slowed when it was checked against the reader's scriptal knowledge.

Table 6

Expeliment 2: percent of regressions into, and second pass

times (in ms) in. the target region

Appropriate Inappropliate Mean(A) (r)

Reg"essions

Neutral (N) Justified (J)

Mean

8.91

7.15

8.33

9.44 12.12

10.78

9.18

9.94

Second pass tùnes

Neutral (N) Justifled (J)

Mean

34

36

35

48

65

4t 5l

A.E. Cook, J.L. Myers I Journal qf Memory and Language 50 (2004) 268-288

General discussion

In Experiment l, participants read script-based sto­ries in which some individual or object fllled a scriptal role in an inappropriate or unexpected way. When the inappropriate role f,ller had not been encountered pre­viously in the story, trst pass times in the target and post-target regions, and second pass times on the target, were longer for the inappropriate than for an appro­priate role filler, and there were more regressions to the target region in the inappropriate condition. The same evidence of greater processing difficulty in the inappro­priate condition was obtained in Experiment 2 when there was no prior justification of an inappropriate scriptal role filler. These effects of appropriateness are

consistent both with models in which context can have an early influence on word processing by either influ­encing lexical access or integration (e.g., Duffy et al., 1988, 1989; Hess et al., 1995; Morris, 1994; Schwanen­flugel & Shoben, 1985; Schwanenflugel & White, 1991;

Traxler & Foss, 2000), and with models which assume

that incoming information is first linked with general world knowledge (e.g., Garrod & Terras, 2000; Kintsch, 1988; Sanford & Garrod, 1989).

When the inappropriate role filler had been previ­ously encountered (Experiment 1) or previously justified (Experiment 2), the pattern of effects for the different measures are more difficult to explain. First pass times rn both the target (Experiments 1 and 2) and post-target regions (Experiment 1) showed an advantage of the in­appropriate role f,ller when it had been previously en­countered or justiûed. These results cannot be explained by models which assume that general world knowledge influences processing before contextual information (e.g., Garrod & Terras, 2000; Kintsch, 1988; Sanford & Garrod, 1989). However, they can be explained by Hess et al.'s (1995) lexical reinterpretation hypothesis. For example, when readers encountered the first mention of the manager playing the song, the manager may have been reinterpreted as an individual who carried out a

guitarist's duties. This is essentially an update of the situation model (Zwaan & Radvansky, 1998). These first pass results can also be accounted for by models in which context can constrain or influence what infor­mation is reactivated during the lexical access process (e.g., Duffy et al., 1988, 1989; Morris, 1994; Schwanen­flugel & Shoben, 1985; Schwanenflugel & White, 1991).

However, if processing of "the manager" was facilitated due to either a reinterpretation of the original role, an update of the situation model, or some expectancy generated as a result of the context, there should have been no subsequent difficulty in processing the inap­propriate role filler. But, there were more regressions

back to the target, and longer second pass times after regressing, when the previously encountered or justified role ûller was inappropriate. Similarly, without addi­

tional assumptions, Kintsch's (1988, 1998) constluction­integration model appears to have no basis for ac­counting for difficulties in processing after integration appears to be complete.

As described earlier, Garrod and Terras (2000) also found a discrepancy between patterns of flrst pass times and measures of subsequent processing. They proposed a two-stage "bonding and resolution" process to ac­count for their results. In the ûrst stage, a dominant role filler, such as pen for writing is automatically integrated with the text, and this is "impervious to the influence of the context in which the role was introduced" (p. 540). In the second stage, the link formed between pen and writing rs evaluated, and when the context was inap­propriate (e.g., the writing had been on a blackboard), processing difficulties were revealed in such measures as

regressions and second pass times. These ideas are con­sistent with other studies of discourse processing, in which early measures were dominated by lowJevel pro­cesses and evidence of higher order processes only ap­peared in later measures of processing (e.9., Cook, 2000; Duffy & Rayner, 1990; Ehrlich & Rayner, 1983). Similar arguments have also been made in more general de­scriptions of eye movements during reading (e.g., the E­Z Reader model, Reichle, Pollatsek, Fisher, & Rayner, 1998; see also Rayner, 1998; Rayner & Pollatsek, 1989).

Although the Kintsch (1988) and Garrod and Terras (2000; Sanford & Garrod, 1989) models cannot account for the patterns ofresults in the present experiments, the genelal idea of two stages can provide an explanation for our effects. As in the previous models, the first stage involves a tentative linkage initiated between informa­tion curlently being processed and information reacti­vated by a low-level automatic retrieval process, and the second stage involves a veriflcation of the link formed in the first stage. We assume that the stages are overlapping and continuous; the processor may verify a linkage just established as the eye fixates on the next word and new linkages are made. As each new concept is encountered, a lowlevel automatic retrieval process is initiated; some possible mechanisms for this process include pattern-matching (Sanford & Garrod, 1989), spreading activa­tion (e.g. Anderson, 1916, 1983), or resonance (e.g.,

Myers & O'Brien, 1998; O'Brien & Myers, 1999). Propositions and concepts highly related to the focal concept are reactivated and are available for integration.

In contrast to previous models, we assume that the link in the first stage of processing can occur between the focal concept and semantic or contextual traces, de­pending on which is reactivated more quickly. In the present experiments, when prior context supported an inappropriate role ûller, the appropriateness effect was no longer signiflcant, which suggests that the contextual trace (i.e., first encounter in Experiment 1, justification in Experiment 2) might have influenced, if not overrid­den, integration of the semantic information (i.e., ap­

ffi Available on I i ne at www.sciencedirect.com Journal ofnI.sclENcE lt-tfDlREcr@ Memory andv Language

ELSH/IER Journal of Memory and Language 50 (2004) 268-288

www.elsevier. com/locate/jml

Processins discourse roles m scnptedl' narratlves: The influences of context and world knowledgt*

Anne E. Cooku'* and Jerome L. Myersb

" Department of Eclucational Psychology, 1705 East Cantpus Center Driue, Rnt. 327, Uníuersity of Utalt,

Salt Lake City, UT 84112-9255, USA o (Jniuersity of Massachusetts, USA

Received 13 Aueust 2003: revision received i3 November 2003

Abstract

Discourse context may affect comprehension of a word in a text by facilitating lexical access or by facilitating in­tegration of the target concept with the preceding text. In two experiments, we used eye tracking measures to examine

contextual influence on the integration of role fillers in scripted narratives. In both experiments, context had an early

influence on integration of role fillers. However, measures of post-target processing indicated that semantic knowledge

continued to influence integration. The results are discussed in terms of several theories of contextual influence on

reading. The data appear to support a two-stage model of processing: in the first stage, incoming text is linked to the

contents of active memory; in the second stage, the link between the new and old information is evaluated.

O 2003 Elsevier Inc. All rishts reserved.

Many factors influence the speed of processing indi- rock band. Assume that the reader encounters a sen­

vidual words in a connected discourse. Some of these tence that begins "The song was played by the. . ." Both factors are inherent in the individual word; for example, theory (e.g., Schank & Abelson, 1977) and data (e.g.,

it is well established that words that are more frequent in Bower, Black, & Turner, 1979) suggest that the reader's

the English language are read more quickly than less knowledge of the role fillers in rock band scripts should

frequently encountered words, and shorter words are either prime, or facilitate integration of, the sentence

processed more quickly than longer words (Inhoff & continuation, guitarist. Therefore, we would expect that Rayner, 1986; Rayner, 1998). At least as important in the time spent processing the appropriate continuation, determining how words are processed is the influence of gtùtarist, would be shorter than that spent processing an

the information provided by the preceding text. For inappropriate continuation, manager. Suppose, how-example, consider a passage whose protagonists are in a ever, that earlier in the passage, the manager had played

the song. Or suppose that the text included a sentence

that indicated that some of the band members handled publicity and finances in addition to performing. In the

-ì were completed while Anne Cook Portions of this research experiments reported here, we investigated whether such

was supported by NIMH Training Grant l|ll.}116745 awarded to contextual manipulations have an effect and, if so, the the Department of Psychology at the University of Massachu­ nature and time course of those effects. setts. Additional support was provided by a University of Utah In the rock band example, there are two sources of Research Committee grant awarded to Anne Cook. We would

information available to the reader. Knowledge of rocklike to thank Chuck Clifton and Keith Rayner for their helpful band scripts links the guitarist to the role of performer,comments on earlier drafts of this manuscript.

''Corresponding author. Fax: l-801-581-5566. whereas the preceding context may imply that ìn this

E-tnail a d dres s ; Anne.Cook@ed.utah.edu (,A..E. Cook). pâssage the usual scriptal roles are not valid. Theories

0749-596xlg - see fiont matter' @ 2003 Elsevier Inc. All lights reserved.

doi: 10. i016/j.jml.2003. 1 1.003

A.E. Cook, J.L. M)¡ers I Journal of Mentory ard Lattguage 50 (2004) 268-288

about discourse processing differ with respect to the priority they assign to the two sources of information, semantic knowledge, and discourse context. Many the­orists have also used the terms "local" context effects and "global" context effects to refer to semantic and discourse influences (e.g., Hess, Foss, & Carroll, 1995). However, the terms "local" and "global" have also been used to refer to a variety of other linguistic factors (e.g.,

discourse coherence, distance in text, causal informa­tion, etc.). To avoid confusion, here we associate local context effects with those caused by information reacti­vated from readers' semantic knowledge, and global context effects with those caused by information reacti­vated from the episodic discourse context.

Kintsch's (1988) construction-integration model in­corporates both immediate effects due to the semantic properties of text and later effects due to discourse context. In the construction stage ofthe model, concepts and propositions in the reader's knowledge base are activated as a function of their relation to what is cur­rently in the reader's focus; the prior discourse context plays its role in the later, integration, stage. Till, Mross, and Kintsch (1988) presented data in support of this two-stage process. As in other studies, at short intervals associates of ambiguous words in sentences were primed whether or not they were related to the discourse con­text. At longer intervals, words unrelated to the overall discourse context of the passage were no longer primed; however, words that were inferentially related to the topic were primed although they were not lexical asso­ciates of words in the preceding sentence.

Similarly, Garrod and Sanford (1999) have argued in favor of immediacy in discourse processing, in which "language input is related to world knowledge at the earliest possible opportunity" þ.23). Sanford and Gar-rod (1989) reviewed several studies of anaphoric pro­cessing and argued that the majority of the evidence pointed to a two-stage process, with the first stage be­ginning immediately upon encountering an anaphor (also, see Garrod & Sanford, 1983; Sanford & Garrod, 1981, for earlier discussions of stages involved in ana­phoric processing). The process in the first stage was referred to as "bonding," in which potential antecedents (whether correct or incorrect) may be quickly reacti­vated on the basis of their semantic relation to an ana­phor through a lowlevel pattern-matching mechanism. The match made during the bonding stage is veriûed, and if correct, is instantiated, during the second stage, resolution.

Garrod and Sanford (1999) proposed more general versions ofthe bonding and resolution stages, which can apply to processing at the lexical, syntactic, or discourse levels. Consistent with this idea, Garrod and Terras (2000) demonstrated that bonding and resolution could be extended to processes beyond anaphor resolution. They required participants to read sentence pairs in

which the context provided by the first sentence was either appropriate or inappropriate with respect to a

phrase in the second sentence. An example of the ûrst sentence is (1a) The teacher was bttsy writing a letter of complaint to

a parent. or

(lb) The teaclter was busy writing an exercise on tlrc blackboard.

The second sentence contained the phrase the pen dropped which forms an appropriate pairing with sen­tence la but not with sentence lb. Eye f,xation times revealed no initial effect of appropriateness on process­

ing the îour¡ pen. However, times in the region of the verb, dropped, and regressions from it back to the noun indicated difficulties in the later stages of processing when sentence lb provided the context. Garrod and Terras explained their results in terms of the two stages

originally proposed by Sanford and Garrod (1989): in bonding, an early automatic associative process links a potential role flller (pen) with a verb encountered earlier (writing), and a later process, resolution, evaluates and resolves the instrument-verb link in terms of the dis­course context.

Using eye tracking technology, Morris (1994) and Duffy, Morris, and Rayner (1988) provided evidence that both lexical and contextual information can affect early lexical access processes. Morris presented readers with sentences such as,

(2a) The gardener talked as the barber trimmed the mustache after lunch.

(2b) The gardener talked to the barber and trimmed the mustache after lunch.

both of which contain identical content words (e.g.,

barber, trímmed) with strong lexical associations to tnustache. But, in sentence 2b, mustache is not consistent with the overall message of the sentence-barbers usu­ally trim mustaches, not gardeners. In contrast to pre­vious demonstrations of serial effects of semantic and contextual influences, Morris found that both intralexi­cal associations and messageJevel context facilitated early processing (i.e., gaze durations) for mustache. Similarly, Duffy et al. (1988) provided evidence for both lexical (e.g., meaning dominance) and contextual influ­ences (e.g., prior disambiguating context) on early pro­cessing of lexically ambiguous words, and proposed that these factors may contribute to a reordered access of lexical information during reading.

In contrast to models that assume early stages may be mediated by lexical associations, Foss and his col­leagues (Foss, 1982; Foss & Speer, 1991; Hess et a1.,

1995; Traxler & Foss, 2000) have proposed that initial access is facilitated by the discourse context. Perhaps the strongest pieces of evidence for their position are the results obtained by Hess et al. Using a cross-modal priming procedure, they presented participants with

I

j A.E. Cook, J.L. M¡,srs I Journal oJ Memory and Language 50 (2004) 268-288

sentences that either had an association between the last word and the preceding text, such as,

(3a) Tlrc Englßh major wrote the poem or sentences that lacked such an association. such as

(3b) The computer science major wrote the poem. When sentences such as 3a were presented alone, the

final word, poem, wal pronounced aloud more quickly than when it ended a neutral sentence such as,

(3c) Tlrc boy read tlte poem. However, when sentence 3a ended a passage in which

the discourse context was not relevant e.g., a passage

about an English major in a computer science course

no facilitation was observed. On the other hand, if the passage placed the computer science major in an English course, poem in sentence 3b was named more quickly than when it ended a control passage. In discussing their lesults in relation to models that assume two stages,

based on semantic knowledge and discourse context ef­fects, Hess et al. stated that these models ". . . make two incorrect predictions: They predict a local eflect where none is observed, and they predict a slow-acting global effect where in fact a fast-acting one occurs" (p. 78). The

authors proposed a lexical reinterpretation model (also,

see Foss & Speer, 1991) in which discourse context modiûes the representation of a concept. For example, an English major taking a computer science course may be viewed as a "student-in-a-technical-course-on-com­puting (not the poetic related information)" (p. 80).

Accordingly, the talget word, poem, is not readily inte­grated into this representation and therefore receives no facilitation.

One limitation of the Hess et al. (1995) study, and of many othel studies of contextual effects, is the use of plobe techniques. Although useful, probe response times provide only one sample in the time course of processing

and may result in an incomplete or possibly even dis­torted view of the effects of context on processing of words in discourse (see Gordon, Hendrick, & Foster, 2000, for additional issues concerning strategic pro­cessing when plobe response methodologies are used).

For example, if Garrod and Telras (2000) had only probed with the target word pen, they might have con­cluded that context had no effect. By monitoring eye

movements, they found that the context had no initial effect but that an inappropriate context interfered with subsequent processing of the targeted word.

The goal of the plesent set of experiments was to use

several eye trackirg measures in different regions of a

text to further oul understanding of context effects on the processing of words in connected discourse. In two experiments, we analyzed eye movement data for tar­geted regions of passages that evoked familìar scripts (e.g., rock band). The regions of interest were the one-word target role flllers (e.g., guitarist, manager) that were either appropriate or inappropriate with respect to the speciflc action being performed in the current sen­

tence and either appropriate or inappropriate with re­spect to the previous discourse context. We also

examined processing times on the post-target regions, the few words that immediately followed the target in the text. If incoming text is linked via lowlevel asso­

ciative processes with general world knowledge before it is integrated with the discourse context, as suggested by Garlod and Terras (2000) and Kintsch (1988), we should observe effects of appropriateness with respect to the action described in the current sentence in early stages of processing (e.g., first pass ol'r target regions), and discourse context effects should be observed in later stages of processing (e.g., fir'st pass on post-target re­gions, second pass, regressions). In contrast, ifdiscourse context influences eally stages of wold processing, as

suggested by Foss and colleagues, discourse context ef­fects should be observed in early stages of processing.

Experiment I

The passage in Table 1 illustrates the two factols manipulated in Experiment 1. Note that there are two encounters between the protagonists and an individual tlling the role of a performer. Our primary interest in this experiment was in participants' eye movements while they processed the second encounter. The action described there (e.g., playing a song) is carried out either by an individual applopriate to the scripted lole (the

audience listened to the gnitarist, A2) or by someone inappropriate to the role (the audience listened to the manager, I2). Also, the earlierfrst encounter contained either an appropriate (the guitarist played the song, Al) or an inappropriate (the manager played the song, I1) role flller. In summary, two factors were manipulated in the second encounter: the appropriateness of the role filler in the earlier first encounter, and the appropriate­ness of the role filler in the second encounter.

'When the second encounter in the A2 condition is

consistent with both context and semantic knowledge that is, when both flrst and second encounters are ap­propriate to scriptal knowledge integration of the sec­

ond-encounter role filler should proceed easily, and thelefore first pass reading times should be faster in,the A1/A2 condition than in the A1/I2 condition. What differentiates models are their predictions when the frst encounter is inappropriate. According to the Hess et al. (1995) lexical reinterpretation model, when the manager plays the song, the reader reinterprets the role of per­former; the manager is incorporated into the discoulse as someone who also fills the gtdtarist's role.If this is so,

on the second encounter the inappropriate role filler (I1l 12 condition) should be processed as fast or faster than the appropriate role filler (I1l42 condition). Having re­interpreted the manager's role, readers should now find it as easy, or easier, to integrate the manager than tlrc

í,:!t/î

t

:

211 A.E. Cook, J.L. M1,svs I Jountal of Memor¡, and Ltutguage 50 (2004) 268-288

Table I

Velsions of a sample passage for Experiment I

Intt'oduction The fans at the rock concert were going wild. They quieted down when the next sonq

Fitst encounter Appropriate (AI ): was played by the band's guitarist. Inappropriate (Il ): was played by the band's n1(n1ager.

Background The concert was being held in an outdoor stadium, and there were about ten thousand scleaming fans in attendance. They were mostly all hysterical teenage gills. Some even cried as they

Seconcl encounter Appropriate (A2): listened To rhe guitarist hlctppropriate ( 12 ): listened to Íhe ntanager

Conclusion Play a slow and heartfelt love song. The song was about a love affair that ended badly.

Note L The target legion in this example is guitarist or tllanager; the post-target region (second encounter) is pkry a slow. Note 2. ln Experiment 1, half of the participants read a similar passage in which the manager was appropriate and the guitarist was

inqnnrnnriata

guitarist \/ith the action of playing the song. Although Hess et al. would not necessarily predict the role rein­terpretation assumed here, such a result would be con­sistent with their model. It would also be consistent with the assumption that situation models are updated on the basis of textual information (Zwaan & Radvansky, 1998). However, models that assume an immediate effect of world knowledge (e.g., Garrod & Sanford, 1999;

Kintsch, 1988) would predict that an inappropriate second encounter will be reflected in longer reading times.

Because none of the models are specific about the time course of processing, it is not clear what their predictions would be fol processing after the eyes move past the role filler. Integration may be completed while readers are flxated on the role flller, or it may continue into subsequent flxations on the words following the role filler or on the role filler itself. In the present experiment, eye movement measures provide a means of assessing

context effects after the initial processing of the role fil­ler'. Continued effects ofeither first-encounter or second-encounter appropriateness should be reflected in time spent in the region following the target word, in re­gressions back to the target word, and in the time spent reprocessing the target after regressing back to it.

In summary, Expeliment I had two purposes. The first was to examine the effects of the appropriateness of the first encounter on processing of the second encoun­ter. If such effects were obtained in early measures of processing (e.g., flrst pass), they would suggest a limi­tation of models that assume only semantic priming by the immediately preceding context and would further extend the applicability of Hess et al.'s (1995) assump­tion of lexical reinterpretation. However, a disruption of processillg when the second encounter was inappropri­ate and the flrst encounter vr'as also inappropriate would

be inconsistent with the lexical reinterpretation hy­pothesis. The second purpose was to use the measures derived from records of participants' eye movements to examine the time course of any effects that were ob­served.

Method

Participants Thirty-eight members of the University of Massa­

chusetts community participated in exchange for either money or course credit. The data for sìx participants were not analyzed due to large calibration errors during portions of the experiment.

Apparahrs Eye movements were lecorded by a Fourward

Technologies Dual Purkinje Eye tracker that has a res­olution of l0min of arc. The eye tracker was interfaced with a 486 microcomputer, which ran the experiment. Viewing was binocular, with eye location recorded from the right eye. The position of the participant's eye rvas

sampled every millisecond by the computer and aver'­aged over four consecutive samples. The averaged hor­izontal and vertical positions of the eye were compated with those of the previous sample to determine whether the eye was fixated or moving.

Materials The materials used were 24 passages similar to the

example presented in Table I (see also Appendix A). Each passage began with a sentence that introduced a

script (e.g., roclc band). In the next (first encounter) sentence, a scripted action was associated with an entity or an object that was either appropriate or inappropriate for that role. Reading times were measured for the role

A.E. Cook, J.L. M)¡ers I Journal of Mentory and Lengu(tge 50 (2004) 268-288

f,ller (e.g., gltitarist) in the first encounter. The role filler was al',vays a single word, vvith a range of +l 1 chaÍac­ters (mean:7.06 characters). A background section of two to three sentences followed. The next (second en­counter) sentence contained a scripted action that was associated with either an appropriate or inappropriate role filler, and which either matched or did not match the role filler in the flrst encounter. Thus, the appro­priateness of the role filler in the second encounter was

crossed with the appropriateness of the role flller in the flrst encounter to create four conditions: A1/A2 (both encounters applopriate), Al/I2 (first encounter appro­priate, second encounter inappropriate), I1l42 (first encounter inappropriate, second encounter appropri­ate), and I1/I2 (both encounters inappropriate). In the second encounter, reading times were measured for the target word (the role filler; e.g., gtùtarist) and the post-target region, which consisted of the two to four words (mean:74.77 characters; e.9., play a slow) that imme­diately followed the target. A short closing sentence ended the paragraph.

Eight materials sets were constructed. These were counterbalanced such that the appropriate and inap­propriate actions for the role f,ller in the first four sets

(e.g., guitarist) were the inappropriate and appropriate actions, respectively, for the role flller in the second four sets (e.g., manager). In each materials set, six passages

appeared in each of the four conditions. Across each group of four materials sets, each passage appeared once in each of the four conditions. 'fhe 24 experimental passages always appeared within alarger set of 72 pas­sages. The 48 "f,ller" passages were stimuli for two un­related experiments; one set of filler passages contained anaphoric references to objects previously mentioned in the text, while the second set of flller passages involved bridging inferences. The flller passages were approxi­mately the same length as the experimental passages but did not focus on scripted settings.

Rating study 1a. In order to ensure that the scripted actions selected were appropriate or inappropriate with respect to the role fillers, a rating study was conducted. Forty University of Massachusetts community members who did not participate in either of the experiments re­ported in this paper were asked to rate the likelihood of an action being performed by a specifrc role f,ller. Par­ticipants rafed 32 items such as 3a-3d below on a 7­point scale (where 1 : "Extremely Unlikely" and 7 : "Extremely Likely"): (4a) At a rock concert, teenage fans cry as a sad song is

played by the band's guitarist. (4b) At a rock concert, teenage fans cry as a sad song is

played by the band's manager. (4c) For a rock band, all ofthe legal issues are handled

by the band's manager. (4d) For a rock band, all of the legal issues are handled

by the band's guitarist.

Appropriate role f,llers (e.g., 3a and 3c; mean:6.13) were rated as signiflcantly more likely to perform a

scripted action than inappropriate role tllers (e.g., 3b and 3d; mean:2.57), ,q(1,31) :349.42, MSE:1.158. The items from this norming study were used to create the 24 passages used in Experiment 1.

Rating stLtdy Ib. Ãn additional rating study was conducted in order to ensure that a switch in role fillers between the first and second encounters was less plau­sible than the same role flller being maintained in the flrst and second encounters. Twenty-four University of Utah undergraduate students enrolled in Introductory Educational Psychology courses participated in ex­change for course credit. Participants were presented with one of four counterbalanced stimulus sets, in which six passages appeared in each of the four conditions, up to the target sentence. Participants were then asked, gi­ven the infolmation in the passage, to rate the likelihood of the target action being performed by the second-en­counter role frller on a 5-point scale (where I : "Ex­tremely Unlikely" and 5: "Extremely Likely"). If maintaining the same role filler across both encounters is more likely than switching role fillers across encounters, ratings in the AlA2 and Ill2 conditions should be

higher than in either the A112 and ll{2 conditions, re­spectively. The results supported this hypothesis. Rat­ings in the AlA2 condition were higher than in the Al12 condition (3.75 vs. 1.93; Ft0,23) :91.96, MSE: .4332; Fz(l, 23) : 124.32, M S E : .307 2). ln addition, ratings in the I1I2 condition were higher than in the A1I2 condi­tion (3.17 vs. 1.93; Ft(1,22) :39.22, MSE:.4704; F2(1,23):30.05, MSE:.5914). Mean ratings for all passages are presented in Appendix A.

Procedure Each participant took part individually in a session

that lasted approximately 60min. For each participant, a bite bar was prepared to eliminate head movements, and the eye tracker was calibrated. The initial calibra­tion procedure took approximately 5min. Prior to reading each passage, calibration of the eye tracking system was checked to ensure that accurate records were obtained. Each participant read three practice passages

followed by the set of 24 experimental and 48 filler passages. Participants were told that they would be reading a series of paragraphs displayed on a CRT screen. They were told to read for comprehension so

that they would be able to answer an occasional oral comprehension question (e.g., Were the fans enjoying the concert?). Questions were asked on approximately one-sixth of the trials, and there were an equal number of "yes" and "no" comprehension questions. Participants had little difficulty answering the questions correctly; all participants answered at least 85% of the questions correctly. Because of the high accuracy rates, the fixa­tion data for all trials were analyzed, regardless of

A.E. Cook, J.L. M),ers I Journal oJ Metnory and Language 50 (2004) 268-288 273

whether a question for a specific trial had been answered & Clifton, 1989). In Experiment 2, we remedied this correctly. by including several words after the target prior to the

At the beginning of each trial, five boxes appeared end of the sentence. across the top of the screen, one box appeared in the middle, and five boxes appeared at the bottom of the Second encounter screen. Each participant was instructed to look at the middle box until the experimenter said, "Ready," and First pass times in the target region then to look at the upper left-most box. Once the ex- -lable 2 presents the first pass times in the target and perimenter had determined that the participant was post-target regions as a function of the appropriateness ûxating the box, the entire passage was presented on the of the earlier first encounter and of the current second screen. When the participant was finished reading the encounter. First consider the appropriateness effect on passage, he or she was instructed to press a button that second-encounter target times when the first encountel would end the trial. Participants were given a brief break had been appropriate; this is the contrast between the approximately halfway through the experiment. AllA2 and A1/I2 cells in the target region of Table2.In

these conditions, flrst pass times were shorter when the Restilts and discussion second encounter was appropriate than when it was

inappropriate. Conversely, in the right-hand column First pass reading time, the sum of all fixations in a (first encounter was inappropriate, 11A2 vs. I1I2), sec­

region before the eyes flrst leave that region, is reported ond-encounter times were shorter when the second for two regions: the target and the post-target region. encounter was inappropriate than when it was appro- \ Trials in which a region was initially skipped were not priate. Simply put, times were shorter whenever there included in flrst pass analyses. In addition, percentage of was a match of role fillers between the two encounters. regressions made into the talget region and second pass This interaction was signiflcant; Fr ( 1 30) : 5.37, ,

reading times, any rereading of the target region after MSE : 3030, F2(1, 46) : 9.83, MSE : 5822. the eyes have regressed back to that region, are reported. The fact that times were faster when ûrst- and sec-In all analyses reported, fi always refers to tests against ond-encounter role fillers were the same suggests facili­error terms based on participants variability, and F2 tation due to word repetition. Effects of roughly this always refers to tests against an error term based on magnitude have been found in text processing experi­items variability. All analyses were significant at the ments (Rayner, Raney, & Pollatsek, 1995). It is some-standard alpha level of .05, unless otherwise indicated. what surprising, however, that the absolute differences in

the A1 and I1 columns of Table 2 were almost identical First encounter (296 - 27 5 : 21 304 - 281 : 23¡. Although this could be

a random occurrence, the lack of a difference implies no When the role frller (e.g., guitarist or manager) was first effect due to the inappropriate role filler on the second

encountered in the text, flrst pass times were signiflcantly encounter although such an effect was obtained on the shorter in the appropriate (mean:318ms) than in the first encounter. Hess et al.'s (1995) lexical reinterpreta­inappropriate (mean : 346) condition: Fr (1, 30) : 5.52, tion hypothesis suggests an explanation that can en-M SE : 47 47 .7 2, F2(1, 46) : 4.85, M SE : 8783.24. Sec- compass effects due to both repetition (i.e., the discourse ond pass times, post-target region reading times, or re- context) and appropriateness of the role flller. Under gressions into the target region are not reported for that hypothesis, for example, if the manager played the the flrst encounter because the target was followed im- song at the flrst encounter, then that individual may mediately by a period and any post-target measures may have been reinterpreted as fllling the role of guitarist and have also reflected sentence-wrapup processes (e.g., Just consequently was not truly inappropriate on the second & Carpenter, 1980; Rayner, Sereno, Morris, Schmauder, encounter; that is, although the manager was inappro-

Table 2

Experiment l: first pass times (in ms) in the second-encountel regions

Second encounter Target region Post-target region

First encounter First encounter

AI II AI II Mean

A2 275 304 290 4t7 46r 439 T2 296 281 289 507 456 482 Mean 286 293 462 4s9

Note. A and I refer to appropriate and inappropriate; I and 2 to fust and second encountet

214 A.E. Cook, J.L. M)¡ets I Joutnal of Memory ntcl Langtruge 50 (2004) 268-288

priate to the scripted action being performed in the second encounter, that individual was appropriate with respect to the previous discourse context. In addition, if such a reinterpretation took place, the guitarist, al­though appropriate to the scripted action, might be

viewed as inappropriate on second encountel in light of the previous discourse context. This analysis suggests

that the discourse context and the appropriateness of the role filler interacted to influence ûrst past reading times.

First pass times in tlrc post-target region The mean duration of first pass in the post-target re­

gion was signif,cantly longer when the second encounter was inappropriate (I2) than when it was appropriate (42); 4 (1,30) :4.21, MSE:13,780, F2(1,46) :4.19, MSE :22,452. This eflect is clearly due to the slower times in the A1/I2 condition than in fhe Allé^2 condition, Fr(1,31) : 5.55, MSE:23,415, F2(1,47) : 8.15, MSE :23,629; there was almost no difference when the ûrst encounter was inappropriate(I1lA2 vs. I1lI2 conditions, both.Fs < 1). The interaction was significant; F¡ (1, 30) : 4.17, MSE:17,396, Fr(1,46) :5.96, MSE: 16,559. It is clear from the interaction that the first encounter in­fluenced later processing. This effect of context extends

the Hess et al. (1995) results. They used two sentences

to establish a discourse context whereas we changed a

single word (i.e, the flrst encounter), and Hess et al.'s relevant discourse context immediately preceded their target sentence whereas the first encounter was separated

from the target region by several lines of unrelated text in the present experiment.

The A1/I2-41142 difference is consistent with the assumption that the process of integrating the role filler with the preceding text continued into the post-target region. It is not clear whether this integration process

has been facilitated in the A2 condition, disrupted in the 12 condition, or is the result of a combination of both facilitative and inhibitory effects. The idea that pro­cessing was disrupted in the 12 condition seems most likely; the reader had to incorporate an inappropriate role flller into the discourse model. When the role filler in the first encounter was inappropriate, this may no longer be a problem because the reader has previously encountered the manager in the guitarist's role. In this view, the post-target means are consistent with Hess

Table 3

et al.'s (1995) assumptions that context can lead to a

reinterpretation of a character's role, and that context affects integration processes, as well as with the updating assumption proposed for situation models (Zwaan & Radvansky, 1998). However, as the measures we con­sider next indicate, the nature of context effects may be

somewhat more complicated.

Regressions into, and second pass times in, the target region

In calculating these statistics, the target region was

redefined to include the entire prepositional phrase; the

target referred to in Table 3 is, for example, to the gui­tarist, rather than guítarist. This was done because ap­proximately 20% of regressions landed on to tlxe ralher than on gttitarist. Considering the regression percentages

in Table 3 first, we found that readers regressed more often to the target region when the second encounter was inappropriate; .F¡ (1, 30) : 22.48, MSE:100.95,

(1, 30) : 10.64, MSE :307 .02. No other main efects were significant. Note that, unlike the results for flrst pass times in both the target and post-target regions, these data indicate that even when the first encounter was inappropriate, the inappropriate second encounter involved more processing difficulty than did the appro­priate second encounter; that is, there were more [e­gressions to the target region in theIllI2 condition than in the 11/42 condition. The contrast of these two con­ditions was signiûcant when tested against participant variability and nearly so when tested against item varr­ability; .Fr ( 1, 30) : 4.62, MS E : 19 5.39, F2(1, 46) : 3.52, MSE:591.20, p: .067.

The second pass results in Table 3 also indicate in­tegration difficulty after initial processing of the target region. Readers spent more time on second passes in the target region when the first encounter was appropriate than when it was inappropriate; Fr (1 , 30) : 5.93,

MSE : 5784, F20,46) : 6.24, MSE :5903. Second-en­counter second pass times were also longer when the second encounter was inappropriate than when it was

appropriate;,Fr ( 1, 30) : 16.64, M S E : 3968, F2(1, 46) : 11.96, MSE:6889. Although the second-encounter ef­fect is larger when the frrst encounter was appropriate, largely due to the very long duration in the Al/I2 cell, the interaction only approached signiflcance when tested

Expeliment 1: percent of regressions into, and second pass times (in ms) in, the target region

Second encounter Regressions Second pass times

First encounter First encounter

AÌ II Mean AI I1 Mean

12

Mean

9.47 21.00 15.24

7.41

t2.75 t0.10

8.46

r6.88 36

i00 68

¿J

5i 37

30

75

A.E. Cook, J.L. Ml,ers I Journal of Mentory utd Language 50 (2004) 268-288 2',75

against participant variability; Fr (1,30) :2.90, MSE: 3592, p: .10. Tests of the simple effects of the second-encounter appropriateness factor were also conducted. The difference between appropriate and inappropriate second encounters was significant when the f,rst encoun­ter was appropriate (Fr (1,30) :10.92, MSE:11,806, Fr(|,46):6-27, MSE:24,478) and when rhe ûrsr en­counter was inappropriate (d(1,30) :7.24, MSE: 3314, F2(1,46) : 4.38, MSE: 7s7s).

These results, particularly the flndings that when the second encounter was inappropriate, readers regressed more often to the target region, and spent more time there after returning to it, seem to conflict with the fi¡st pass results. Those data indicated that when the initial encounter had been inappropriate, an inappropriate second encounter proved no more difficult to process than an appropriate second encounter. The results in Table 3 cannot be readily explained by theories of fa­cilitation due to priming (e.g., Duffy et al., 1988; Duffy, Henderson, & Morris, 1989; Morris, 1994; Schwanen­flugel & Shoben, 1985; Schwanenflugel & White, 1991); expectancies generated for a specific role flller may fa­cilitate first pass times on the target, but regressions and second pass times indicate difficulty in integrating inap­propriate targets. These results also seem inconsistent with Hess et al.'s (1995) lexical reinterpretation hypoth­esis, or withZwaarr and Radvansky's (1998) assumption of situation model updating. If the manager was rein­terpreted as f,lling the guitarist's role, why is there any evidence of difficulty in processing a later mention of the mallager carrying out a guitarist's duties? Before further

Table 4

Velsions of a sarnple passage for Expeliment 2

Appropriate (A) Versions Introductiotl

discussion of the results of Experiment 1, we first con­sider the results of a different manipulation of context.

Experiment 2

One purpose of Experiment 2 was to replicate the ef­fects offlrst-encounter appropriateness on first pass times on the target, in addition to examinirg the appropriate­ness effect on other measures. In Experiment 1, the first encounter with the target word occurred at the end of a sentence. Because ofthis, we could not be certain that any measures of post-target processing (e.g., post-target first pass times, regressions to the target, second pass times on the target) were not confounded with sentence wrap-up processes. As was seen in the data of Experiment 1, such post-talget processing measures may reflect different ef­fects than those on the flrst pass target times and can provide a more complete picture than just first pass times or probe response times alone. Therefore, the target sentences of Experimenf 2 always included several words after the target word and prior to the end ofthe sentence, to allow for examination of post-target processing.

The primary purpose of Experiment 2 was to exam­ine further the effects ofthe conflict between sctiptal and contextual knowledge. There are circumstances in which a role in a script may be filled by someone or something other than the standard role flller. For example, in the inappropriate/justification (IJ) condition of Table 4, the manager (instead of the guitarist) played the song but the action is justified by the preceding statement that

The fans at the rock concert were going wild. The band was still new on the rock scene,

Justification Neutral (N); and they weren't used to the adoring fans and all of the attention from the media. Justified ("r). so some of the band members had to handle publicity and finances in addition to performing.

Conclusion The crowd quieted down when the next song was played by the band's guitarist who was standing alone on the stage. The concert was being held in an outdoor stadium, and there were about ten thousand screaming fans in attendance.

Inappropriate (I) Versions httroduction The rock band was becoming mo.re and more populal. The band was still new on the rock scene,

Justification Neutral (N); and they weren't used to the adoring fans and all ofthe attention from the media. Justified /"I). so some of the band members had to handle publicity and finances in addition to performing.

Cottclusion A contract was signed by the band's guitarist who had arranged a six-record deal. The band had been getting lots ofcalls recently, and they were having a very hard time figuring out how to handle their new success.

Note. The target region in this example is guirarist and the post-target region is wlto was standinglwlrc lnd arranged.

¿ lÍt A.E. Cook, J.L. Myers I Journal oJ Mentory and Language 50 (2004) 268-288

"some of the band members had to handle publicity and finances in addition to performing." The results of the effects of appropriateness on the first encounter in Ex­periment 1 indicate that-in the absence of any prior justification-a departure from the reader's knowledge of scriptal roles increases first pass times on the role flller. A theory such as Kintsch's (1988) construction-integration theory, that assumes that the reader's world knowledge is consulted before any attempt to integrate any new input with prior text, would predict that this increase in first pass times would be obselved even when the inappropriate role filler had received prior justifica­tion. On the other hand, it is possible that the justifica­tion creates an expectancy, either for the alternative role filler, or at least that the scriptal role will not be filled in the usual way. Such an expectancy seems consistent with models which suggest that the discourse context may constrain or influence semantic priming effects (e.g.,

Duffy et al., 1988, 1989; Morris, 1994; Schwanenflugel & Shoben, 1985; Schwanenflugel & White, 1991; see also MacDonald, Pearlmutter, & Seidenberg, 1994); accord­ing to this view, in the present study, the appropriateness effect would be smaller in the justifled than in the neutral condition. The facilitated integration model proposed by Foss and his colleagues (Foss & Speer, 1991; Hess et al., 1995) would make a similar prediction. On the basis of their results, Hess et al. hypothesized that integration with the context occurs very rapidly, and that contextual effects override world knowledse.

Method

Participants Thirty-ûve members of the University of Massachu­

setts community participated in exchange for money or course credit. The data for three participants were not analyzed, due to large calibration errors during portions of the experiment.

Apparatus tn. equipment used was the same as in Experiment

,.

Materials The materials used were altered versions of the pas­

sages from Experiment 1. The first sentence introduced the script (e.g., rock band) and was followed by a context sentence. The context sentence either justifled a scripted action's association with an inappropriate role filler, or it was neutral with respect to actions and role fillers but continued the story. The next sentence contained a

scripted action that was associated with an appropriate or inappropriate role filler, followed by several words prior to the end ofthe sentence. (See Appendix B for all passages in Experiment 2.) Note that, in contrast to Experiment 1, this was the first and only encounter with

the role flller in the text. Reading times were measured for the target word (the role filler; e.g., guitarist), which ranged from 4 to 1 I characters (mean :6.79 characters), and the post-target region, which consisted of the three words that immediately followed the target (mean: 74.27 characters; e.g., wlzo r)as standinglwho had arranged). Thus, context type was crossed with the ap­propriateness of the role flller to create four conditions: AN (appropriate neutral), IN (inappropriate neutral), AJ (appropriate justified), and IJ (inappropriate justi­ûed). One to two closing sentences ended the paragraph.

The between-subjects manipulation in Experiment 1

produced no signiflcant effects, so it was dropped in Experiment 2. Four materials sets lvere constnrcted, such that each materials set contained six passages that appeared in each ofthe four conditions. Acloss the four materials sets, each passage appeared once in each ofthe four conditions. -lhe 24 experimental passages appeared within a larger set of 54 passages. The 30 "flller" pas­sages ',vere stimuli for an unrelated experiment; they were similar in length to the experimental passages but did not focus on scripted settings. The number of filler passages used in this experiment was reduced from the number used in Experiment I in order to shorten the overall amount of time required for the experiment-this helped reduce participant fatigue and allowed for better precision on the eye tracker.

In order to demonstrate that the justifying context was strong enough to override the inappropriateness of the role filler, a rating study was conducted. Twenty-six University of Utah undergraduates enrolled in Intro­ductory Educational Psychology courses participated in exchange for course credit. Participants were presented with one of two counterbalanced stimulus sets, in which 12 passages appeared in each of the IN and IJ condi­tions, up to the talget sentence. They were then asked, given the information in the passage, to rate the likeli­hood of the target action being pelformed by the role filler on a 5-point scale, where (l:Extremely Unlikely and 5 : Extremely Likely). If the justifying context was sufficient to override the inappropriateness of the role ûller, ratings should be higher in the IJ condition than in the IN condition. This was supported in the results: ratings in the IJ condition were significantly higher than in the IN condition (3.29 vs. 2.45; F(1,25) :45.12, MSE:.185; .4(1,23) :71.846, MSE:.116). Mean Iatings for all passages are presented in Appendix B.

Procedtte The procedure used was the same as in Experiment 1.

Restilts and discussion

First pass times in the target region Table 5 presents the flrst pass times in the target and

post-target regions as a function of the appropriateness

277 A.E. Cook, J.L. Myers I Journal of Mentory and Language 50 (2004) 268-288

Table 5 F2(1,23) :2.68, MSE--9189,p: .11. In conrrasr, when Experiment 2: first pass times (in ms) the inappropriate role filler was justifled, the pattern in the

Appropriate Inappropriate(A) (r)

Target region Neutral (N) 280 304 292 Justified (J) 289 283 286 Mean 285 293

Posl-larget region Neurral (N) 476 s25 501

Justiûed (J) 462 500 481

Mean 469 513

of the role flller and of whether the target action had been previously justifled. First pass times are longer on the target (e.9., guitarist) in the inappropriate than in the appropriate condition only in the neutral (i.e., unjusti­fied) condition. This resulted in a signiflcant interaction of appropriateness and justification; 4 (1,31) : 4.84, MSE:1561, F2(1,23) :6.52, MSE:1546. Neither main effect approached signiflcance; both F's < 1.

However, the contrast of the neutral condition means (i.e., IN-Alfi was signiflcant; Fr ( I , 3l) : 4.31, MSE : 4243, F2(1,23) : 5.12, MSE:4815. As in Experiment 1,

in the absence of any prior justiflcation of the inappro­priate role filler, flrst pass times were longer when the role ûller was inappropriate than when it was appro­priate. When there was prior justiflcation, however, flrst pass times on the appropriate role filler were actually (nonsignificantly) slightly slower than in the inappro­priate condition. Thus, flrst pass times on the target suggest that any difficulties due to the inappropriate role filler were quickly resolved, either because the justifica­tion facilitated lexical access to the target, or facilitated its integration with the discourse model. However, subsequent measures contradict this conclusion.

First pass times in the post-target region The pattern in the post-target region differed from that

in the target region in that times were slower in the inap­propriate than in the appropriate condition whether or not the inappropriate role tller had beenjustifled in the preceding text. Only the appropriateness main effect was signiflcant; ¡'r(1,31) : 8.19, MSE :7311, F\(1,23) : 5.42, MSE:9004. Furthermore, the interaction.F/s were icss than 1. When there was no preceding justiflcation, the ir:rÊer times in the inappropriate than in the appropriate ,r¡ìldition were a continuation of the effect found in ûrst pass times on the target; apparently, the eyes moved be­fote integration was complete. The simple effect contrast ol'the AN and IN conditions was more variable than the maìn effect of appropriateness, resulting in a less powerful tcst; the contrast was marginal when based on partici­rr¿nts variability and nonsigniflcant when based on items r.ariability, 4 (1,31) : 3.85, MSE:9825, p: .059;

post-target region differed from that in the target region; although there was no effect in the target region, times for the inappropriate role flller were (nonsigniflcantly) slower than for the appropriate role filler in the post-tar­get region: ¡'r (1,31) :2.29, MSE:9997, p: 14; Fz(|, 23) : 2.94, MSE : 11,840, p : .16.

Regressions into, and second pass times in, the target regton

Table 6 presents the mean percentage of regressions into the target region from the post-target region, and the mean second pass times on the target. As in Exper­iment l, for these analyses the target region was ex­panded to include the two to three words immediately preceding the role ûller. Although there were more re­gressions in the inappropriate than in the appropriate condition, with a larger difference in the justified con­dition, no F's were signiflcant; all p values were greater than .72. However, there was a more convincing effect of appropriateness in the second pass times on the target than in the post-target reading times or the regression percentages; second pass reading times were signiflcantly longer in the inappropriate than in the appropriate condition; rr(1,31) : 4-68, MSE :3206, F2(1,23) : 7.60, MSE:1617. All other F,s were less than 1.25.

Thus the pattern of regressions and second pass times were consistent with the pattern of times in the post-target region in exhibiting more difficulty in the in­appropriate than in the appropriate condition. Most notably, despite the absence of an appropriateness eflect on flrst pass times on the target when it had been justified, measures of subsequent processing indicated difficulty integrating the inappropriate role filler. One explanation for these results is that the inappropriate role flller was tentatively accepted due to the preceding justification, but a further evaluation of the inappro­priate role flller was slowed when it was checked against the reader's scriptal knowledge.

Table 6

Expeliment 2: percent of regressions into, and second pass

times lin ms) in. the target region

Appropriate Inappropriate Mean(A) (r)

Regressions

Neutral (N) Justified (J) Mean

8.91

7 .7 5

8.33

9.44 t2.t2 10.78

9. l8 9.94

Second pass tùnes

Neutral (N) Justifled (J)

34

36

48

65

41

5l Mean 35 57

278 A.E. Coolc, J.L. M),ers I Joutnal of Memory and Language 50 (2004) 268-288

General discussion

In Experiment 1, participants read script-based sto­ries in which some individual or object filled a scriptal role in an inappropriate or unexpected way. When the inappropriate role f,ller had not been encountered pre­viously in the story, flrst pass times in the target and post-target regions, and second pass times on the target, were ionger for the inappropriate than for an appro­priate role filler, and there were more reglessions to the target region in the inappropriate condition. The same

evidence of greater processing difficulty in the inappro­priate condition was obtained in Experiment 2 when

there was no prior justification of an inappropriate scriptal role ûller. These effects of appropriateness are

consistent both with models in which context can have an early influence on word processing by either influ­encing lexical access or integration (e.g., Duffy et al., 1988, 1989; Hess et al., 1995; Morris, 1994; Schwanen­flugel & Shoben, 1985; Schwanenflugel & White, 1991;

Traxler & Foss, 2000), and with models which assume

that incoming information is frrst linked with general

world knowledge (e.g., Garrod & Terras, 2000; Kintsch, 1988; Sanford & Garrod, 1989).

When the inappropriate role f,ller had been previ­ously encountered (Experiment 1) or previously justified (Experiment 2), the pattern of effects for the different measures are more difficult to explain. First pass times in both the target (Experiments 1 and 2) and post-target regions (Experiment 1) showed an advantage of the in­appropriate role filler when it had been previously en­countered orjustif,ed. These results cannot be explained by models which assume that general world knowledge influences processing before contextual information (e.g., Garrod & Terras, 2000; Kintsch, 1988; Sanford & Garrod, 1989). However, they can be explained by Hess

et al.'s (1995) lexical reinterpretation hypothesis. For example, when readers encountered the flrst mention of the manager playing the song, the manager may have been reinterpreted as an individual who carried out a

guitarist's duties. This is essentially an update of the situation model (Zwaan & Radvansky, 1998). These first pass results can also be accounted for by models in which context can constrain ol influence what infor­mation is reactivated during the lexical access process

(e.g., Duffy et al., 1988, 1989; Morris, 1994; Schwanen­flugel & Shoben, 1985; Schwanenflugel & White, 1991).

However, if processing of "the manager" was facilitated due to either a reinterpretation of the original role, an update of the situation model, oI some expectancy generated as a result of the context, there should have been no subsequent difficulty in processing the inap­propriate role f,ller. But, there \ryere more regressions back to the target, and longel second pass times after regressing, when the previously encountered or justif,ed

role filler was inappropriate. Similarly, without addi

tional assumptions, Kintsch's (1988, 1998) construction-integration model appears to have no basis for ac­

counting for difficulties in processing after integration appears to be complete.

As described earlier, Garrod and Terras (2000) also

found a discrepancy between patterns of flrst pass times and measures of subsequent processing. They proposed a two-stage "bonding and resolution" process to ac­

count for their results. In the first stage, a dominant lole filler, such as pen for writing is automatically integrated with the text, and this is "impervious to the influence of the context in which the lole was introduced" (p. 540). In the second stage, the link formed between pen and writing is evaluated, and when the context was inap­propriate (e.g., the writing had been on a blackboard), processing difficulties were revealed in such measures as

regressions and second pass times. These ideas are con­sistent with other studies of discourse processing, in which early measures were dominated by low-level pro­cesses and evidence of higher order processes only ap­peared in later measures of processing (e.g., Cook, 2000;

Duffy & Rayner, 1990; Ehrlich & Rayner, 1983). Similar arguments have also been made in more general de­

scriptions of eye movements during reading (e.g., the E­Z Reader model, Reichle, Pollatsek, Fisher, & Rayner, 1998; see also Rayner, 1998; Rayner & Pollatsek, 1989).

Although the Kintsch (1988) and Garrod and Terras (2000; Sanford & Garrod, 1989) models cannot account for the patterns ofresults in the present experiments, the general idea of two stages can provide an explanation for our effects. As in the previous models, the first stage

involves a tentative linkage initiated between informa­tion currently being processed and information reacti­vated by a low-level automatic retrieval process, and the second stage involves a verification of the link formed in the first stage. We assume that the stages are overlapping and continuous; the processor may verify a linkage just established as the eye fixates on the next word and new linkages are made. As each ne'w concept is encountered, a lowlevel automatic retrieval process is initiated; some possible mechanisms for this process include pattern-matching (Sanford & Garrod, 1989), spreading activa­tion (e.g. Anderson, 1916, 1983), or resonance (e.g.,

Myers & O'Brien, 1998; O'Brien & Myers, 1999).

Propositions and concepts highly related to the focal concept are reactivated and are available for integration.

In contrast to previous models, 'ñ/e assume that the link in the f,rst stage of processing can occur between the focal concept and semantic or contextual traces, de­pending on which is reactivated more quickly. In the present experiments, when prior context supported an

inappropriate role filler, the appropriateness effect was

no longer significant, which suggests that the contextual trace (i.e., first encounter in Experiment 1, justification in Experiment 2) might have influenced, if not overrid­den, integration of the semantic information (i.e., ap­

A.E. Coolc, J.L. M),ers I Jountal of Mentory and Language 50 (2004) 268-288 279

propriateness of the scriptal lole f,ller). There are at least two explanations for why this may have occurred. First, in Experiment 1, the inappropriate ûrst encounter in­volved extensive processing, as evidenced by lengthy first pass times, which may have led to a stronger encoding and therefore faster reactivation and integration of this information upon encountering the inappropriate sec­

ond encounter. Second, in Experiment 2, Íhe justiûca­tion may have generated an expectancy that there would be a violation of scriptal norms, and this expectancy may have been quickly reactivated when the target role flller was encountered. Several recent results suggest that context can affect an early stage of processing. Wiley and Rayner (2000) demonstrated that first pass reading times on lexically ambiguous words (e.g., fly) were facilitated compared to a control condition when they were em­bedded in a text that was preceded by a disambiguating title (e.9., "Worries of a Baseball Manager"). This context effect was strong enough to "allow for the se­

lective access of all but the least frequent meanings of ambiguous words" (p. 1018). Using ERP methodology, Sereno, Brewer, and O'Donnell (2003) provided evi­dence that biasing sentence contexts facilitated process­ing of lexically ambiguous words very early in the reading process. Context effects appeared in the Nl component (the window of time ranging from 132 to 192ms post-stimulus) of the ERP lecord, a time window that has been associated with very early reading pro­cesses, such as lexical access. Finally, Myers, Cook, Kambe, Mason, and O'Brien (2000) manipulated both the semantic and episodic availability of information needed to comprehend a sentence and showed that or­dinalily strong semantic effects disappealed when in­formation that could be used for comprehension had been highly elaborated several sentences earlier in the text.

The first pass data for the target regions suggest that the link in the ûr'st stage is made with the first infor­mation that becomes available in working memory this information may be either semantic or contextual in nature. However, it is clear from the second pass data that even after the eyes have moved away from the target, additional information continues to be reacti­vated and integrated. Although the inappropriate role frller appeared to be easily integrated when it was sup­ported by previous context, measures of subsequent processing indicated that the violation of the scriptal norm did disrupt the reading process. In both experi­ments, regression percentages and second pass times were elevated when the role flller was inappropriate, regardless of the prior context condition. Thus, even after integration of a role filler appears to be complete, the role flller continues to be evaluated and verifled ac­cording to any new information that becomes available. This is similar to the "resolution" process described by Garrod and Terras (2000; also Sanford & Garrod, 1989).

Although many models of reading assume that se­

mantic knowledge has an earlier influence than contex­tual information, the results of the present study are consistent with a more interactive process. The view that contextual knowledge can interact with semantic knowledge to influence early processing is consistent with assumptions made by models used to describe processing on the word level (e.g., reordered access

model; Duffy et al., 1988), the syntactic level (e.g., con­straint satisfaction model; MacDonald et a1.,7994), and the discourse level (e.g., memory-based text processing view; Cook, 2000; Gerrig & McKoon, 1998; Myers & O'Brien, 1998; O'Brien & Myers, 1999; O'Brien, Cook, & Derepentigny, 2001; Rizzella & O'Brien, 2002). The present work goes beyond earlier work to demonstrate that context can influence both early integration pro­cesses (the linkage stage), and later processes in which the links made in the frrst stage are evaluated (the veri-f,cation stage).

1\

Appendix A

Stimuli and Rating Study lb results (in parentheses) for Experiment 1. (Note that for the 5-point rating scale,

I : "Extremely Unlikely" and 5 : "Extremely Likely.") L Appropriate second-encounter conditions (AlA2: 3.67,

I1A2:2.15) The fans at the rock concert were going wild. They quieted

down when the next song was played by the band's guitaristl mctnager- The concert was being held in an outdoor stadium, and thele wele about ten thousand screaming fans in atten­dance. They were mostly all hysterical teenage gills. Some

even cried as they listened to The guitarist play a slow and heartfelt love song. The song was about a love affail that ended badly.

Inappropr iate second- encounter conditions (ÃII2: l. 56, IlI2 4.00)

The lock band was becoming more and more popular'. A record company producer wanted to sign them onto a six-re­cord contract, and he was told to talk to the band's ntanagerl guitarist. The band had been getting lots of calls recently, and they were having a very hard time frguring out how to handle their new success. All of the legal issues weLe handled by the guitarist so the musicians could just focus on playing. They signed a six million dollar contract.

2. Appropriate seconcl-encounter conditions (4142: 3.89,

llA2:2.89) The movie was being tlmed on locatìon in the Sahara

Desert. In preparation for the first day of fllming, the script was rehearsed by the actressldireclor. The movie was going to be an epic about the nomads that lived in the desert and the horrible conditions they had to withstand. The first line of the ûlm was

spoken by The acn'ess while crawling on hands and knees across

the desert. The movie was sure to be a big hit at the box office. Inappropriate second-encounter condítíons (All2: 2.33, Ill2:

3.1 3)

The movie was being filmed on location in the Sahara Desert. Everyone cleared the set, waiting to hear, "Action!"

280 A.E. Cook, J.L. Myers I Jounml of Mentory and Language 50 (2004) 268-288

called by the directorlactress. They were behind schedule, so

they were trying to film as much as possible every day. Every­one was hot and tired and wanted to go home. Suddenly, "Cut!" was yelled by The actress when a cameraman acciden­tally walked across the set. They were never going to finish at this rate.

3. Appropriate second-encounter conditions (AlA2: 5.00,

I1A2: 3.38)

It was time to do the yard work at the manor. First thing in the morning, the rose beds wele weeded by The gardenerl

cltauffeur. The roses were the pride and joy of the owner of the manor. Some of the roses had even taken first place at presti­gious garden shows. The roses were pruned by lhe gardener

with a pair of shears fi'om the garden shed. The rose beds were picture perfect by the end of the day.

Inappropriate second-encounter conditions (All2: 2.22, IlI2: 3.83)

It was time for the day to start at the manor. First thing in the morning, the owne¡ was driven to his office in the city by the clmuffeurlgardener. It was only about a thirty minute ride through the country into the business district of the city. When the car arrived at its destination, the owner was dropped off by the gardener right in front of his office building. He was picked up in the same spot at the end of the day.

4. Appropriate second-encounter conditions (4142: 4.00, IlA2:2.67)

A terrible crime had been committed ìn the palace. Unfor­tunately, the crown jewels had been stolen by a palace tltiefl guard. -[he queen was beside herself with grief, because some of the jewels that had been taken had been in her family for over four hundred years. One of the pieces taken by the tlúefhad been her wedding present lrom the king. The queen ordered a

full-scale investigation into the matter. Inappr opr iate second- encontter condit ions (All2: 2.7 5, lll2:

3.i1) There had been many break-ins at the palace lately. To

ensure the queen's safety, the king ordered that she be under constant protection by a palace guardlthief. They couldn't take any unnecessary risks that would result in her being hurt. To­day, the queen was giving a speech from one of the palace

balconies. She felt very secure under the protection of the tlùef the entire time she was giving her speech. The king felt very relieved as well.

5. Appropriate second-encounter conditions (4142: 3.83,

IlA2:3.61) Lucy lived in a huge apartment complex. On the first day of

every month, the ¡ent was collected by tlte landlordla tenant. Lucy always had her rent check ready on time, but it seemed

like her neighbors were always behind on their rent. They were always begging for more time. Late payments were frowned upon by the landlord and eviction was always a threat. Lucy was careful, so she knew she wouldn't have to worry about that.

Inappropr iate second-encounter conditions (Ã112: 2.7 5, I1I2: 3.11)

Lucy lived in a huge apartment complex. On the first day of every month, the rent had to be paid by each tenantlthe land­lord. Lttcy always had her rent check ready on time, but it seemed like her neighbors were always behind on their rent. They were always begging for more time. The rent check sent in by fhe landlord had to be received by the end of the day. The management didn't accept any excuses.

6. Appropriate second-encounter condítions (AlA2: 3.00,

ll{2:2.00) The kidnapper was brought to the state prison. After he was

photographed and fingerprinted, he was led to his cell by tlrc warrlenla prisoner. As he walked slowly down the hallway to­ward his cell, the kidnapper looked at some of the other men sitting on the floor in their cells. Then he was pushed into his cell by the warden and was told that dinner was in two hours. The kidnapper knew that prison was going to be miserable.

Inappropriate second-encounter condi tions (A112: 2.38, lIl2: 3.33)

The kidnapper was brought to the state prison. After he was photographed and frngerprinted, he was led to his cell, which he

had to share wtth anotlter prisonerltlrc warden. The kidnapper looked around the tiny room. There was a sink, a toilet, a table and chair. and one small window that had bals on it. The bed that had been taken by the warden was obviously the most comfortable one. The kidnapper knew that prison was going to be miserable.

7. Appropriate second-encontter conditions (4142: 3.00,

IlA2: 1.88)

The Johnsons were visiting the local state forest. They were excited, because a nature tour was being given by a rangerl camper. The Joh¡sons were big nature enthusiasts, and they spent a great deal of thei¡ free time doing things outside, such as

hiking, bird watching, and gardening. They were told about the local wildlife by the ranger who seemed to know a great deal

about the topic. The Johnsons took lots of photographs along the way.

Inappropriate second-encounter conditions (AlI2: 1.67 , I1I2: 2.83)

The Johnsons were visiting the local state forest. As they were driving thlough the campground, they saw a tent being put up by a c(tmper I tlrc ranger. The Johnsons were glad they lived close enough that they could enjoy the beauty ofthe forest and then go to their warm comfortable house at night. Sud­denly, the tent fell down on top of the ranger making him yell and curse loudly. The Johnsons all thought that was very funny.

8. Appropriate second-encounter conditions (AlÃ2: 3.75, ll{2:2.33)

Sheila was spending her vacation at the seashore. While she

was on the beach, she saw a drowning man being saved by a

lifeguardlcltild. The man had apparently swum out too far and then was swept out by the undertow. He seemed alright after he

was brought to the sho¡e. The man was very grateful to the lifeguard for helping him and saving his life. Sheila couldn't even make herself go into the water after watching that.

Inappropriate second-encounter condi.tíons (All2:1.67, Ill2: 4.00)

Sheila was spending her vacation at the seashore. While she

was sunning herself on the beach, she watched a childllifeguard playing in the waves. Sheila thought about going into the water, but she was so warm on the sand that she decided to stay there for a while longer. She saw a wave knock down the lifeguard and get him completely soaked. She shivered and was glad she

was dry on the beach. 9. Appropriate second-encounter conditions (4142: 3.33,

I1A2:2.89) The nursing home was engulfed in flames and smoke.

Luckily, the frre was put out in the nick of time by a firemanl uictitn.The fire was extinguished before it could spread to other

281 A.E. Cook, J.L. Mlts¡s I Journal of Mentory and Languuge 50 (2004) 268-288

buildings on the block. There was a big story about it in the newspaper the following day. Twelve people had been saved by thefirentan after the building started to collapse. The building was ruined, but all the ¡esidents were safe.

Inappropriate second-encounter condit îons (A1I2: 2.00, IlI2: 3.s0)

The nursing home was engulfed in flames and smoke. Luckily, a man who lived in the house next door was able to run in and give help to a trapped aictintlfirentan. The tre was ex­tinguished before it spread to other buildings on the block. There was a story about it in the newspapet the following day. It was all about the rescue of they'r.eman by the man from the house next door. The story called the man a helo.

10. Appropriate second-encounter conditions (AlA2: 3.56, IlA2:1.38)

Bill and Joe decided to go to the movies. After getting his ticket, Bill walked over to the concession stand and asked for some popcontla donut.Then he found his seat in the theater and watched the movie. After the movie, Joe asked Bill if he wanted to grab a pizza. Blll said he was still full from the popcom bttt said he'd go and have a beer with Joe. They decided to go to a local cafè on the corner.

Inappropriate second-encounter conditions (All2: 1.56, IlI2: 3.s0)

Bill and Joe decided to go out for breakfast. After sipping from his cup of coffee, Bill decided to start with a donutlsome popcorn- While he and Joe were eating, they tried to figure out what they would do for the rest ofthe day. Joe suggested that they go hiking. Bill quickly finished the ¡est ofhìs popcorn so tha|They could get on the trail. The day was perfect for a nice hike.

11. Appropriate second-encounter conditions (A1A2: 3.22, IlA2:2.33)

It was Sunday, so Joan got up, got dressed, and went to church. Today, a sermon about charity was given by the preacherlorganis¡. Joan decided that she should volunteer her time to the needy more often. After church, Joan saw a f¡iend of he¡s and went over to say hello. Joan was inspired by the ser­mon given by Lhe preacher and told her friend all about it. Joan and her friend both decided to volunteer at the soup kitchen.

Inappropri.ate second-encounter conditions (All2: 2.17, lll2: 2.88)

It was Sunday, so Joan got up, got dressed, and went to church. Today, a beautiful hymn was played by the org(tnistl preacher. After church, Jill went to visit her mother, who was in the hospital after having heart surgery. Jill told her.mother about the church seruice that morning. She sang the hymn played by the preaclrcr while her mother listened quietly. She thought Jill had a beautiful singing voice.

12. Appropriate second-encounter conditions (AlA2: 3.67, llA2:2.78)

The city council had decided to build a new school. They were shown the plans for the building by the architectlplwnber. The building would have forty classrooms, two teachers'lounges, and a beautiful two-story library. It would be a model for. futu¡e schools in the area. The council loved the plans drawn up by the arclxitect and voted for building to begin right away. They knew the taxpayers would stand behind their decision.

htappropriate second-encounter conditions (AlI2: 1.88, Il12: 2.56)

The city council was touring the new school in their district. They were very impressed by the bathroom fixtures installed by

The pluntber I cu'clùtect.The building was costing the taxpayers a lot of money, so the council members were extra careful to make sure that everything was in top shape. They banged on the pipes put in by the u'chitect and said he had done a great job. The school was going to be a model for others in the area.

13. Appropriate second-encoatter conclitions (AlA2: 4.25, I1A2:2.83)

The army unit was being defeated by a larger-, better armed enemy. Finally, the order to retreat was given by a generall priuate- Almost all of the men quickly ran back to their tren­ches, but one malt foolishly stayed behind and kept shooting at the approaching enemy folces. He disobeyed the order given by The general and almost lost his life because of it. Luckily, he made it back to camp unharmed.

htappropriate second-encounter conditions (AlI2: 1.7 B, I1l2: 3.89)

The army unit was being defeated by a larger-, better armed enemy. Finally, a sergeant gave an or.der to a priuateltlæ gen­eral. They welen't going to give up this battle without a tough fight. Their coLlntry's ìndependence was at stake in this war, and they were determined to win. The order that was given to the general was to pull out the heavy artillery. The cannons were their last defense.

\4. Appropriate second-encomtter conditions (AlA2: 4.38, IlA2: 2.50)

The first half of the football game came to an end. The players ran off the field and the crowd was entertained during the halftime show by a bandlclotvn. The home team was ahead of their rival by twenty-one points, so the crowd's spirits were high. People were yelling and cheer-ing in the stands. The half­time show put on by the bandkept them happy until the second half of the game stalted. The home team won the game by forty polnts.

hmppropriate second-encowtter conditions (AIl2: 1.89, lll2: 3.22)

The bilthday party \/as just getting underway. The parents of the birthday girl had arranged for all of the active fivc-year­olds to be entertained by a clotwtlband There were loads of presents, party favors, and cake, and everyone was having a great time. The parents sat back and watched all of the children having a great time. The kids all laughed at the band and ran around playing games. The birthday party vvas a huge success.

15. Appropriate second-encowtter concfitions (AlA2: 3.56, IlA2:2.22)

Paul had been deeply deplessed lately. After weeks of pe­culiar behavior, he flnally agreed to be treated by a therapistl mínister. Paul thought that maybe his problems were related to stress. He had just recently gotten manied and then was promptly fired frorn his job two weeks later.. He was hypnotized by the tlterapisl and woke up feeling much calmer and more balanced. He just hoped that he wor,rld stay that way.

Inappropriate second-etxcoutxter conditions (Ãll2: 1.33, I1I2: 3.2s)

Paul had fallen in love. After several months of dating his girlfriend Judy, they were finally married by Paul's tninisterl llterapist. The cerernony was somewhat small, with only a few friends and family rnembers present. They exchanged their vows and then gave each other the wedding rings. They were pro­nounced mar¡ied by the therapis¡ and Paul was told he could kiss his bride. Paul couldn't imagine being any happier than he was at that moment.

282 A.E. Cook, J.L. M),ers I Journol of Mentory and Latzgnage 50 (2004) 268-288

16. Appropriate second-encoatter conditions (4142: 4.50,

I1Ã2:2.22) Sam was having a barbecne at his house. While his guests

wele playing a game of badminton, Sam fired up the grill and put on ror?¿ burgers I u turldqt. He called all of his guests to cor¡e and eat, and everyone sat down at the picnic table. After selving everyone else, Sam finally sat down to eat. He put some

ketchnp on his burger and took a big bite of it. Everyone thought the meal was great.

Inappropriate second-encotnúer contlitions (AIl2: 1.7 5, III2: 2.00)

Sam was having Thanksgiving dinner at his house. While his guests watched football, Sam stayed in the kitchen and tended

1o the turkq,lburgers. Afíer a while, he called his guests to come and eat, and everyone sat down irr the dining room. After selvirrg everyone else, Sam frnally sat down to eat. He poured some gravy over his burger and took a big bite of it. Everyone thought the meal was gleat.

I7. Appropriate second-encounter conditions (4142: 3.00,

I1A2: 3.00) It was deadlile day at the big publishing cornpany. Just in

time to meet the deadline, the last chapter of the novel was written by tbe uutlørleditor. The president of the company decided to read the chaptel while he was having his lunch. He liked the novel, but he had some concerns about it. He thought that the ending written by Lhe antlnr was longer that it leally should be. He decided to bring it up at the next neetlng.

Inappropriate seconcl-encototter conditions (AlI2: 2.44, I1I2: 2.67)

It was deadline day at the big publishing company. Finally, the last chaptel of the novel was proofi'ead, corrected, and

critiqned by the editorlautltor. The president of the company decided to read the chapter while he was having his lunch. He liked the chapter, but he had a few concerns about it. He considered the criticisms made by the uutlnr too vague and unnecessarily harsh. He decided to bring it up at the next

meetlng. 18. Appropriate second-etzcottn.ter concfiti.ons (AlA2: 2.83,

IlA2:2.44) A young woman had been found in the park with three

bullet wounds in her head. Several witnesses on the scene were questioned by the detectiuelcoroner. He was careful not to reveal

any details for fear they might be released to the press. Every­thing was to be kept strictly top-secret. The clues that were

found by fhe detectiue led the police to the killer. The suspect

was latel tlied and found guilty of the murder. Inappropriate second-en.counter conclitiotts (1'112: 2.13, III2:

3.s6) A young woman had been found in the park with three

bullet wounds in her head. An autopsy on the body was per­

formed by the coroner I detectiue. After careful consideration of the evidence, the time and cause of death were recorded. The details ofthe woman's death were surprising to evelyone on the case. The autopsy results frorn rhe detectiue indicated that she

irad actually been hung. The bnllet wounds must have been an attenpt to cover up the hanging.

19. Appropriate second-encounter conclitions (4142: 3.33,

IIA2:2.44) Mr. Jones was remodeling his living room. He had bought

a new sofa and chairs, refinished a table, and had a set of

shelves made by a carpenterlnteclrunic. He was doing all of this as a birthday surprise to his wife, who had been out of town visiting her mother. She was so happy with the shelves

made by lhe carpenter that she wanted to hire him to redo her kitchen cabinets. Mr. Jones was ecstatic with his wife's reactton.

htappropriate second-encounter conditions (Al12: 2.00, I112: 2-7s)

Mr. Jones needed to have some work done on his car. He went to the local service station and had his fan belt and ra­diator replaced by a nteclnniclcarpenter. The charges were much more than he could really affold right now, but he knew that his car would run much better as a result. He hoped that the repairs done by lhe curpetlîer would extend the life of his car for a few more years. He had had the cal since he was in college and didn't want to give it up.

20. Appropriate second-encounter conditions (AlA2: 3. i 3,

IlA2: 4.00) Lori was a student at the local art school. Today she

was observing an advanced ceramics course, where she saw a beantifnl sculpture cleated by an artistla model. Lori hoped that she would be able to deveiop her talents to create such beautiful works of art. She was really only an arnateur right now. She quickly sketched the piece made by the artist so that she would be able to remember it later. Then she packed away her things alrd went to her next class.

hruppropriate second-encontîer cotzditions (AlI2: 2.00, IlI2: 3.33)

Loli was a student at the local art school. Today she was

sitting in on an advanced drawing class, where she was at­temptìng to sketch a dificult pose held by a ntodellan artist. Lori had to leave to go to another class right now, so she would have to work on the sketch later that night. Later, she imagined the expression on the face of the artist and tried to pick up sketching where she had left off. She was very happy with the sketch when it was finished.

2L Appropriate second-encounter conditions (1^11^2: 3.67, IlA2:2.75)

Jeannie and her daughter were having dinner at a very fancy French restaurant. After they were seated by the maitre'd, they looked over the menus, and then their order was taken by the waiterlbusboy. While they waited for theil food, Jeannie asked her daughter about her day at school. Jeannie's daughter was

still talking when their food arrived. They found that the wrong order was blought by the waíter so they sent it back to the

kitchen. Fortunately, their correct orders were brought out right away.

Inappropriate second-encounter conditions (AIl2: I.56, III2: 4.00)

Jeannie and her daughter were having dinner at a very fancy French restaurant. After dinnel was over, Jeannie paid the bill. As they were leaving the table, the table was cleared by the busboylwaiter. Jeannie and her daughter got up and started walking toward the front door of the restaurant, and then Jeannie realized that she was missing something. Their doggie

bag had been picked up by the waiter by mistake, so they tried to find him. Luckily, he was still holding the bag when they found him.

22. Appropriate second-encounter conditions (AlA2: 4.00, I1A2:3.67)

283 A.E. Cook, J.L. M),ers I Journal oJ Memory and Lntguage 50 (2004) 268-288

The criminal shufied into the courtroom. Aftel' all of the charges against him were [ead, he was given a sentence of five years in prison with no parole by fhe judgelbaíLiff. The criminal looked down in shame. He knew that he deselved whatever punishment he received, but at the same time he had hoped for a lighter sentence. He was also given a stern lecture by thejudge about hoping that he had learned his lesson. The criminal honestiy thought he could reform.

Inappropríate seconcl-encounter condítíons (AlI2: 2.33, Ill2: 1.13)

The criminal shuffed into the colÌrtl'oom. After he was led to the wìtness stand, he was sworn in by the bailiffljudge. The criminal agreed and proceeded to tell his side ofthe stoly. After hours of grueling testimony, the court adjoumed for the day. The criminal was led back to the jail by the judge and locked up again in his cell. He had nothing to do but wait and hope.

23. Appropriate second-encounter conditions (AlA2: 3.83,

l1A2:3.44) The fifth grade students were nervously studying for a big

social studies test. After attendance was taken, they were given their exam by their teaclterlpríncipal. The examwas much harder and longel than any of the students had expected, so they all quickly got to work. The roorn soon fell silent. They were later given their gradesby The teaclter who patiently explained all ofthe questions. Most of the students had done quite well.

Inappropriate second-encounter condilions (All2: 2.50, IlI2: 2.tt)

The fifth glade class was quietly taking a social studies test in the classroom. One student had been caught cheating and was sent down the hall to be punished by the principal.lteaclter. The student had been in trouble many times before at this school, so he was very nervous. He hoped that his father would not be told about this. The stÌrdent was suspended by the rea¿lzer and then his father was called. The student knew he was in trouble now.

24. Appropriate second-encoaúer corxdiîiot1s (AIA2: 4.22, I1A2: 3.38)

About three weeks before the construction on the new r¡all began, all of the workers wele hired by the contracîorlelecu'i­cian. Unfottunately, some of the men that wete on the ploject were incompetent and did not take ploper safety precautions. Some of them were also repeatedly late for work. One guy was immediately fired by the contractor andwas told that he was not welcome back on the site. Unfortunately, this put the con­struction even more behind schedule.

Inappropriate second=encounter conditions (AII2: 2.56, lll2: 4.00)

When the constrllction ploject on the mall was almost fin­ished, some vandals ¡uined all the wiring that was put in by the electricianlcontr'(tctor. UrfortLtnately, this was discovered by a

safety inspector who was writing up a repott on the site. The inspector checked everything carefully and wrote his report. He looked over the aleas wiled by fhe contractor and asked who had damaged them. He told the construction company to file a

leport with the local police.

Àppendix B

Stimuli and Likelihood Ratìngs (in parentheses) fol Ex­periment 2. (Note that for the 5-point rating scale, 1 : "Extremely Unlikely" and 5: "Extremely Likely.")

l. Appropriate neutral ancl justified conclitions The fans at the rock concert were going wild. Tlte butd was

still new on tlte rock scene, cutd tltelt \:¿¡s¡1'¡ used to tlrc adoring

fans and all oJ the (tttention fi'oÌ11 tlte ntedia.lTlrc bcutd wcts still netv on \|rc rock scene, so some oJ tlte bad ntentbers hud to htutdle publicitlt ancl finances in addition to perfornting. The crowd quieted down when the next song was played by the band's guilarist who was standing alone ou tlìe stage. The concert was being held in an outdoor stadium, and there were about ten thousand screaming fans in attendance.

htappropriate neutt'al and jr.tstified conditiotts (IN: 3.58, IJ: 4.00)

The rock band was becorning more and mole popular. //ia bancl was still netv on tlrc rock scene, and tltey weren't used to tlte adoring Jans and all of tlte attentíon Jron¡ tlæ ntediq.lTlrc bcutd tras still netu- on tlte roclc scene, so some of tlte band ntentbers ltad to hartdle publicit¡t cutd finances in addition to petfornling. A contract was signed by the band's guitqrisÍ who had arranged a

six-record deal. The band had been getting lots ofcalls recently, and they were having a very hard time figuring out how to handle theil rlew success.

2. Appropri.ate neutral and justified conditÌons The movie was being filmed on location in the Sahala

Desert. ã tvcts a ynall indepen.dent flnl with e lotv budget und sntall stctff, so lltelt v¿v¿ careJful ttot îo t1)aste too mucll time or ntoney.llt was a sntall independent flnt tvitlt a lov, budget and sma.ll staff, so eùerytone inuolued ltad to take on extra jobs cutd responsíbilities. On the first day of filming, the script was le­hearsed by The actress for sevelal hours with no bleaks. The movie was going to be an epic about the nomads that lived in the desert and the horrible conditions they had to withstand.

[nappropriate neutral and justified conditions (IN: 2.58, IJ: 2.58)

The movie was being fihned on locatiou in the Sahara Desert. If tuqs a sn¡all independent Jtlm y,itlt a lotv budget and smc¿ll stalÏ, so tlrcy were careful not 10 ttaste îoo nrucll tinle or money.llt was a sntall independent flm witlt o low budget mtd unall staff, so eüery)otle inuolued lnd to îalce on extra jobs and responsibiliti.es. On the fir'st day of filming, "Action!" was called by |he actress so that shooting could begin. They were behind schedule, so they were trying to film as much as possible every day. Everyone was hot and tired and wanted to go home.

3. Appropriate neutral. and justified conditions It was time to do the yald work at the manor. Altlnugh the

owner oJ tlrc ntmtor was Der)t riclt, lrc was ulso ntecut, tutd lrc dentanded tltat ltis seruants workfi'ont dqtvrl to sunset.lAltlnugll tlrc otvner of tlze ntanor was uery) riclt, lrc was also clteap, culd lte lùred only a Jew seruan.ts to do all oJ the worlc. Each mom ing, the rose beds were weeded by fhe garclener who took great pt'ide in his wol'k. The roses were the pride and joy of the owner of the manor. Some of the roses had even taken first place at presti­gious garden shows.

Inappropriate neutral and justiJied cottditions (IN: 1.42, IJ: 3.00)

It was time for the day to stalt at the manor. Although the

owner of tlrc ntanor ntas Dery riclt, læ was also ntean, ancl lte dentanded lltat ltis seruants work from dawn to sunsef.lAltlrcugh tlte owner of the ntanor wûs uerv riclt, Irc was also clrcap, and he lúred ottb, a lew serùatxts to do all oJ the work. Each morning, the owner was driven to his office by The gardenet'who took great pride in his driving. He was very careful thror"rghout the thirty

284 A.E. Coolc, J.L. M),ers I Journal of Mentory and Language 50 (2004) 268-288

minute drive into the city and was relieved when they arrived safely.

4. Appropriate neutral and justifed conditions

There had been many break-ins at the palace lately. Under great pressure fr-ont tlte king and queen, tlrc in\estigators wete

doing euer)ttltinT tlrcy çev¡¿ to solue tlte crirnes.lAfter seueral

unsuccessful attempts 1o solue tlte crimes, llrc in\estigators were

at a loss and so tltey toolc an unusual step.They set a trap for the

thief who they believed would go after the crown jewels. They left the jewels apparently unguarded but had gualds in place.

After several days, the plan worked and the thiefwas captuled. Inappropriate neutral and justifed conditions (IN: 3.08, IJ:

3.58)

There had been many break-ins at the palace laiely. Under great pressure J|ont tlrc king and clueen, \|rc inuestigators were

doing euery¡tlting llte1t ¿6sl¿ to solue tlte crimes.lAfter seueral unsuccessful ettenlpts to solue tlrc crimes, tlte in)estigators were

at a loss and so tlæy took an unusual step. They asked for help

from a thief who they believed knew where the criminal was.

They finally tracked down the criminal and brought him to justice. The king and queen were very grateful, and they re­warded the investigators handsomely.

5. Appropriate neutral and justffied conditions

Lncy lived in a huge apartment complex. Tlrc contplex was

úctualb) a large co-op, and no one wlrc liued tltere could intngitte a

better place to liue.lTlrc contplex v,as actually a large co-op, so

eqerytone wlto liued tlrcre ltad to pitch in and do tlrcù' share. OnThe first day of every month, the rent had to be paid by each tenant

who lived in the building. Lucy always had her rent check ready

on time, but it seemed like her neighbots were always behind on their rent. They were always begging for more time.

hruppropriate neutral and justified conditions (IN: 3.08, IJ: 3.67)

Lucy lived in a huge apartment complex. Tlte contplex v,as

actually a large co-op, tutd no one vtlto lit¡ed tlrcre could imagine a

better place to liue.lTlrc contplex was actually a large co-op, so

eüery)one wln liued tlrcre had to pitch in and do their slwre. Onthe first day of every month, the rent was collected by a tenant who lived in the building. Lucy always had her rent check ready on time, but it seemed like her neighbors wele always behind on their rent. They were always begging for more time.

6. Appropriate neutrel and justified conditions

The kidnapper was brought to the state prison. f1e ruas

placed in a new program, and was told about all of tlrc details of the progrant wlten lte arriued.lHe was placed fut a netu progrû11,

vlære the inmates and tlte ernployees all liued and worked 1o­

gether. After his admission, he was led to his ceil and was visited by fhe warden who told him the rules of the prison. As he

walked slowly down the hallway toward his cell, the kidnapper'

looked at some of the other men sitting on the flool in their cells.

Itxappropriate neutral and justified conditions (IN: 1.08, IJ: 2.s0)

The kidnapper was blonght to the state prison. I1e wns

placed in a new- prograril, and was told about all of the details of tlrc program wlten lte arriued.lHe was placed in a new progrant, wlrcre tlte inntates and the entployees all liued and worked to­gether. After his admission, he was led to his cell, which he

shaled with The v'arden who ran the prison. The kidnapper looked around the tiny loom. There was a sink, a toilet, a table and chair, and one small window that had bars on it.

7. Appropriate neutral and justffied conditions

The Johnsons wele visiting the local state park. The park was not only a beautiful natural forest, but tlte entry fee was also uery reasonable.lTlte park was not only a beautiful natural forest, but tlrc park attendants were also ahvays uery lrclpful. On This

day, the Johnsonsjoined a nature tour that was being given by a" ranger who was very familiar with the park. The Johnsons thought it was a terrific tour, and they started making plans to come back again the following weekend.

Inappropriate tteutt'al and justifed condítíons (IN: 2.42, IJ: 3.2s)

The Johnsons were visiting the local state park. Tlrc park was not only a beautiful natural forest, but tlrc entry fee was also

uery reasonable.lTlte park was not only a beautiful naturalforest, but tlte park atrendants were also alwal,s ¿¿vy ltelpful. On this day, the Johnsons saw a tent being put up by a ranger who was helping some teenagers. The Johnsons thought this was a nrce

thing to do, and they hoped that the teenagers would be grateful for his assistance.

8. Appropriate neutral and justified conditions

Sheila was spending her vacation at the seashore. It was a sntall local beaclt, and Slteila thought all of tlle people tltere were

uery f iendly and generally pleasant to be around.llt was a small Iocal beaclt, and all of tlrc people wlrc worked and liued there were

uery experienced and sîrottg swinuners. While she was sunning herself on the beach, she watched a child in the water near the

beach. Sheila thought about going into the \¡/ater, but she was

so warm on the sand that she decided to stay there for a while longer.

Inappropriüte neutral and justified conditions (IN: 1.67, IJ: 2.s0)

Sheila was spending hel vacation at the seashore. It was a

sntall local beaclt, and Slrcila thought all of tlte people there were

uery fiendly and generally pleasunt to be around.llt was a small local beaclt, and all of tlte people who worked and liued tlrcre were

uery experiettced and sn'ong swinunerc. While she was on the beach, she saw a drowning man being saved by a child in rhe

water near the beach. The man had apparently swum out too far and then was swept out by the undertow. He seemed alright after he was brought to the shore.

9. Appropriate neutral and justified conditions

The nursing home was engulfed in flames and smoke. Tltere would not be ntuclt time before the bui.lding collapsed, so

people fu nearby bvil¡l¡1gs took precautions and tried to help

otlters.lTlrcre w'ould not be ntuclt tinte before the building col­

lapsed, so eDery)one had to help put out tlrc fl.ames and get eu­

eryone out safely. Lrl.ckily, a man who lived in the house

nearby was able to save a trapped uictint who was in the burning building. The fire was extinguished before it spread to other buildings on the block. There was a story about it in the

newspaper the following day.

hmppropriate neutral and justffied conditions (IN: 1.67, IJ: 2.67)

The nursing home was engulfed in flames and smoke.

Tltere would not be ntuclt time before tlrc buílding collapsed, so

people in nearby buildings took precautions and tt'ied to ltelp

otlrcrs.lTltere would not be ntuclt tinte before tlze building col­lapsed, so eDetyone lud to lrclp put out tlrc flames and gel eu­

eryone out safely. Ltrcksly, the fire was put out in the nick of time by a uictint who was using a garden hose. The fire was extinguished befo¡e it could splead to other buildings on the

A.E. Cook, J.L. M)¡ets I Journal of Mentory and Language 50 (2004) 268-288 28s

block. There was a big story about it in the newspaper the followirg day.

10. Appropriate neuîral and justified conditions Bill and Joe had gone out for breakfast. The place they went

to was nevì and uery n'end1,, and Bill ltoped tlwt ít would be good.l Tlæ place tlzey went to was new and uery n'endy, and it sold all kinds offoods. After entering, Bill went to the counter and or­dered a donut and a large coffee. While he and Joe were eating, they tried to frgure out what they would do for the rest of the day. Joe suggested that they go hiking.

Inappropriate neuÍral and justifi.ed conditions (IN: 3.17, IJ: 4.08)

Bill and Joe had gone to the movies. Tlß pl(tce tlrcy went to wqs new and uery tretzdy, and Bill hoped tlzat it n¡ould be good.l Tlte place they wetxt to ttas nev) and uery tendy, and it sold all kinds offoods. After entering, Bill went to the concession stand and asked for a donut and a large coffee. Then he found his seat

in the theater and watched the movie. After the movie, Joe

asked Bill if he wanted to grab a pizza.

11. Appropriate neutral and justified conditions It was Sunday, so Joan got up, got dressed, and went to

church. 1¡ was a small qnd nontraditional clturclt, and Joan lød been going to seruices tlrcre for as long as slte could rentember.llt was a small and nontraditional clturclt, in whiclt eaclt part of tlte seruice was led by a different person euery week. -|oday, a

beautiful hymn was played by the organist who was a very fine musician. After church, Jill went to visit her mother, who was in the hospital after having heart surgery. Jill told her mother about the church service that morning.

Inappropriate neutral and justified conditíons (IN: 2.75, IJ: 4.00)

It was Sunday, so Joan got up, got dressed, and went to church. ¿ was a small and nontraditional clrurcl4 and Joan had been going to serzices tlrcre for as long qs slte could rentember.llt was a stnall and nontraditional church, in whiclt eaclt part of tlrc seruice was led by a diferent person euery week. Today, a sermon about charity was given by the organist who was a very in­spiring speaker. Joan decided that she should volunteer her time to the needy more often. After church, Joan saw a friend of he¡s and went over to say hello.

12. Appropriate neutral and justffied condítions The city council had decided to build a new school. Tltey

Itad debated about tlrc school a lot, but eueryone was ltappy when

tlrcy saw lzow well the design ltad turned out.lTltey hqd debated about tlæ school a lo1, and tltey finally lùred one person to design

and build tlrc new scltool. They were shown the plans for the building by the arclùtecl that they had hired. The building would have forty classrooms, two teachers' lounges, and a

beautiful two-story library. It would be a modei for future schools in the area.

Inappropriate neutral and justified conditions (IN: 2.17, IJ: 2.92)

The city council was touring the new school in their district. They lrud debated about tlrc scltool a lot, but eueryone was happy when tltey saw lzow well tÌrc desigtz had turned out.lThey ltad debated about tlæ scltool a lot, and tltey finally ltired one person to design and build tlrc new scltool. They liked the bathroom fix­tures installed by the architect that they had hired. The building was costing the taxpayers a lot of money, so the council members were extra careful to make sure that evelything was rn too shaoe.

13. Appropriate neutrql and justified conditions The army unit was being defeated by a larger, better armed

enemy. Tension was extt'emely ltíglt antong tlle troops, and tlrcy weren't sure whether they would get out aliue.lTensiotx was ex-n'emely high antong llze troops, and nobody seented uery sure of wlrc was in conmtand. Finally, the order to retreat was given by a general who was very well-respected. Almost all of the men quickly ran back to their trenches, but one man foolishly stayed behind and kept shooting at the approaching enemy forces.

hmppropriate neutral and justified condítíons (IN: 2.42, lJ: 2.67)

The army unit was being defeated by a larger, better armed enemy. Tensíon was extretnely ltiglz antong the troops, and tlrcy weren't sure wlætlrcr they would get out aliue.lTensíot1 was ex-u'emely ltiglt among tlle troops, and nobody seented uery sure of *'lzo was in comtnand. Finally, a sergeant gave an order to the general who was still in his tent. They weren't going to give up this battle without a tough fight. Their country's independence was at stake in this war, and they wele determined to win.

14. Appropríate neutral and justified conditions The fust half of the football game came to an end. Tlze honte

te(Jn't h)(JS winning by seueral poínts so far and tlrc crowd was

excited about the game.lTlrc lrcme team was losing tlrc game so

far and the crowd was in an ugly ntood. As haiftime started, they heard music from the band which was entering the ûeld. Then the crowd settled down and many of them went to the con­cession stand To geT a snack befo¡e the second half started.

Inappropriate neutral and justified conditions (IN: 2.08, IJ: 2.58)

The first half of the football game came to an end. TIte ltome teanl was winning by set)eral points so far and tlte crowd was

excited about the game.lThe hotne leam was losing tlte gante so

far and tlrc crowd was in an ugly tnood. As halftime started, they started to boo the bandwhich was entering the field. Then the crowd settled down and many of them went to the concession stand to get a snack before the second half started.

15. Appropriate neutral and justified conditions Paul and Judy were planning their wedding ceremony. Al­

tltouglt Paul knew tltere were seueral options aoailable to tlrcm, Ite

had strong doubts about wlzat to do.lAltltouglt Paul knew there were seueral options aÐailable to tlæm, lrc wanted guidance frorn someone lte could tru$. Affel much discussion, they agreed to be

married by Paul's minister who he really admired a lot. The ceremony was somewhat small, with only a few lriends and family members present. They exchanged thei¡ vows and then gave each other the wedding rings.

Inappropriate neutral and justified condìtions (IN: 3.33, IJ: 4.r1)

Paul had been deeply depressed for seve¡al months. l/­though Paul knew tlrcre were seueral options auailable to ltim, læ

lød strong doubts about what to do.lLltltouglt Paul knew tlære

were seoeral options duailable to Iùm, he wanted guidance from someone lte could trust. After much thought, he decided to be

treated by a local minister who he really admired a lot. Paul thought that maybe his problems were related to st¡ess. He had just recently gotten married and then was promptly fired from his job two weeks later.

16. Appropriate neutral and justified conditions Sue was having a barbecue at her house. After greeting her

guests, slte told tltent to relax wltile slte finished preparing tlte food for tonight.lAfter greeting lter guests, slrc told tltem tltat tltey

286 A.E. Cook, J.L. M1,s¡s I Journal of Mentor¡, and Language 50 (2004) 268-288

vere in for sontething a little different tortight. Whlle they chat­ted, Sue tended to the burgers and prepared some new side

dishes. After a while, she called his guests to come and eat, and everyone sat down in the dining room. After serving everyone else, Sue frnally sat down to eat.

Inapproprìate neutral and justified conditions (IN: 1.42, IJ: 3.08)

Sue was cooking Thanksgiving dinner at her house. After greetùry lter guests, slrc told tltent to relax wltile slte finished preparùry tlrc food for tonigltt.lAfter geeting lrcr guests, she told tlænt that they were in for sontetltí.ng a little different tonigltt. While they chatted, Sue tended to the burgers and prepaled some side dishes. After a while, she called his guests to come and eat, and everyone sat down in the dining room. After selving everyone else, Sue finally sat down to eat.

17. Appropriate neutral ancl justifecl conditions

It was deadline day at the big publishing cornpany. Tlte

nouel being publislrcd was ahnost finished, and it was expected to be fu bookstores by the next sununer.lTlte ttouel being publíslrcd

was almost fníshed, but tlrc ending still needed to Inae major changes m.ade /o i/. Just in time, the last chapter was received by the editor who sent it to the publisher. The president of the

company decided to read the chapter while he was having his lunch. He liked the chapter, but he had a few concerns about it.

Inappropriate neutral and justified conditions (IN: 2.00, IJ: 2.83)

It was deadline day at the big publishing company. Tlte nouel being publislted was almost finislted, atzd it was expectecl

to be in bookstores b),tlte next summer.lTlrc nouel being pub­lislzed was ahnost finished, but tlte endùry still needed to lmue

major clrunges made to l¡. Just in time, the last chapter was written by the edítor who sent to the publisher. The president

of the company decided to read the chapter whìle he was having his lunch. He liked the novel, but he had some con­cerns about it.

18. Appropriate neutrdl and justified condi.tians

A young woman had been found in the park with three

bullet wounds in her head. It was a small town, and the identitlt of llte wontan was a big mystery to eueryone.llî was a small totvn, so tlrcre was only one man who was auailable to imestigate the

crùne. An autopsy on the body was performed by The coroner who was in charge of the case. After careful consideration of the evidence, the time and cause of death were recorded. The details of the woman's death were surprising to everyone on the case.

Inappropriate neutral and justffied conditions (IN: 2.58, IJ: 3.2s)

A young woman had been found in the palk with three bnllet wounds in her head. It was a sntall town, and tlte identity of the wontan wds a big tnystery to eaeryone.llt was a small town, so tlrcre was only olxe man who was auailable to inuestigate tlrc crime. Sevetal witnesses were questioned by the coroner who was in charge of the case. He was careful not to reveal any details for fear they might be released to the press. Everything was to be kept strìctly topsecret.

19. Appropriate neutr al and jus tified conditiotts

Mr. Jones needed to have some work done on his car. .É1e

couldn't afford to go to a real professional, but lrc still Inped tlrut lrc would be able to get work of decenî quality.l¡!¿ couldn't afford to go to a real professi.onal, so lte went to someone wlto did tlte work as a ltobby on tlle side. He had his radiato¡ replaced by a

ntecltanic who wo¡ked there on the weekends. The charses were

much more than he could really afford right now, but he knew that his car would run much better as a result.

Inuppropriate neutral ancl justified conditions (IN: 2.17, IJ: 3.08)

Mr'. Jones was remodeling his living room. He couldn't afford to go to areal professional, but he still lrcped tlrut he v,ould be able to

geî n,ork ofdecenr qualiry.lHe couldn't afford to go to a real pro­

fessional, so he ti)ent to someone tulø did tlrc work as a ltobb), on the

side. He had a set of shelves made by a nteclnnic who built them ovel the weekend. He was doing all of this as a birthday sr.rrprise

to his wife, who had been out of town visiting her mother 20. Appropríate neutral and justified conditions

Lori was a student at the local art school. At tltis sclrcol. Lori ltad noticed tllat eDeryone wqs uery talented and ltoped tlrut sonteday she migltt be tlut good.lAt tltis scltool, Lori had noti.ced

that all of tlæ sludents took turns being \|rc subjects Jor tlte otlrcr studerlts' projeus. Today she was sketching a difficult pose held by a ntodel who was also a student in the class. Lor i had to leave

to go to another class right now, so she would have to work on the sketch later that night.

hruppropriate neutral and justified conditiot'ts (IN: 3.50, IJ: 4-33)

Lori was a student at the local art school. At tlùs sclnol, Lori had notíced tlw.t euetyorxe wqs uery talented and lrcped tlrut sonteday slrc rnigltt be that good.lAt tltis scltool, Lori lrud noticed tlrut all oJ the students took turns being tlrc subjects for rlrc otlrcr students' projects. Today she saw a beautiful sculpture created by a model who was also a student in the class. Lori hoped that she would be able to develop her talents to create such beautiful works of art. She was really only an amateur right now.

21. Appropríate neutral and justified condi.ti.ons

Jeannie and her daughter were having dinner at a very fancy French ¡estaurant. After they were seated, they looked over the menus. Zftey noticed tlmt the restaur(tnt wos l)ery liuely that night, and tlttt eueryone seented to be really enjoying themselues.l

Tlæy noticed tlrut the resteurant vas üery slrcrtstaJfed tltut txight,

and tlnt eDeryone left workùzg ltad taken on some extra duties.

After a few minutes, their order was taken by the waiter who then returned to the kitchen with it. While they waited for their food, Jeannie asked her daughter about her day at school.

Inappropriate neutral and justified conditions (IN: 3.67, IJ: 4.s8)

Jeannie and her daughter were having dinner at a very fancy French restaurant. Aftel dinner was over, Jeannie paid the bìll. TIrcy noticed tlzat tlæ restaurat'tt tvús uery liaely that nigltt, and \lrut eueryone seented to be really enjoying tlrct¡tseloes.lTlzq, no­ticed that tlte restawant was uery shor|staffed that nigltt, and tløt eDeryone left worlci.ng ltad talcen on solne extra duties. As they were leaving, the table was cleared by lhe waiter who then brought all the dishes to the kitchen. Jeannie and her daughter got up and started walking toward the fiont door of the res-

IauranI. 22. Appropriate neutral and jusîifi.ed conditiotts The criminal shufled into the courtroom. It vas a sntall

rural courthouse, and it looked uery muclt like sontetlting out of an

old tnoaie.llt was a small rural courtltouse, atd all of the duties oJ

tlrc court were perfonned by just one ntan. Alter entering the witness stand, the criminal was sworn in by the bailiffwho then told him to sit down. The criminal agleed and ploceeded to tell hìs side of the story. After hours of grueling testimony, the court adjourned for the day.

\

287 A.E. Cook, J.L. Ml,st'y I Journal of Mentory and Language 50 (2004) 268-2B8

htappropriate neutral and justffied cortditions (IN: 3.75, IJ: 3.58)

The criminal shuffied into the courtroom. It was a sntall rural courtlnuse, and it looked ueryt much like sometlzing out of an olcl ntouie.llt was a st¡tall rural courtlrcuse, and all oJ tlrc duties of tlrc court were perfornted by just one ntan. After the charges were read, the climinal was sentenced by the baíliffwho then had him taken away. The criminal [eft, protesting all the way. He knew that he deserved whatever punishment he received, but he had hoped for a lightel sentence.

23. Appropriate neutral and justified conditions The fifth grade students were nervously studying for a big

social studies tes|. Because of a flu outbreak at tlrc scltool, rnany students were absent and tlte rest were ntisbeluoing. I Because of a

flu outbreak at llte sclrcol, tltere were only a handful of people leJi to superDise tlæ students. After attendance was taken, they were given their exam by their teacher who wished them good luck. The exam was much harder and longel than any of the students had expected, so they all quickly got to wolk.

Inappropriate neutral and justified conditions (IN: 2.08, IJ: 2.92)

The fifth grade class was quietly taking a social studies test in the classroom. Because of a flu outbrealc at tlrc sclzool, manlt studetxts were absent and tlte rest were ntisbelruuíng.lBecause of u

flu outbreak at r|rc scltool, tltere ytere onbt a hqlt¿fr¡.¡teople leJi

to superuise tlrc students. A student caught cheating was sent to the main offce to be punished by fhe teaclrcr who was quite angry. The student had been in trouble many times before at this school, so he was very nervous.

24. Appropriate neutral and justffied conditions The constmction project on the new mall was almost

complete. Altltouglt it was a big project, it was expected to be

completed in a sltort arnount of tinte.lAhlzouglt it was a big pro­ject, one man wqs left in charge of doing ntost of ilrc important

7bås. The foreman criticized the work by the electrician who was in charge of the wiring. Unfortunately, this rnan proved to be

ircompetent and did not take proper safety precautions. He had also been lepeatedly late for work.

Inappropriate neutral and justifed conditions (IN: 1.25, IJ: 3.08)

The construction project on the mall was about to begin. Altlrough it was a big project, it tvas expected to be contpleted in a slxort qmoutlt of tinte.lAlthough ít was a big project, one man w(ts

left in charge of doing tnost of tlte imporîant joås. All of the workels were hired by The electt'ícía,? who was in charge of the project. Unfortunately, some of the men that were on the project were incompetent and did not take proper safety ple­cautions. Some of them were also reoeatedlv late fol work.

References

Anderson, J. R. (1976). Language, memory, and tlnugltt. Hillsdale, NJ: Etlbaurn.

Anderson, J. R. (1983). The architecture oJ cognition. Canl­bridge, MA: Harvard University Press.

Bower, G. H., Black, J. 8., & Turner, T. J. (1979). Scripts in memoly for text. Cognitiue Psyclrclogy, 11,177-220.

Cook, A. E. (2000). Tlrc Moses lllusion in readíng cotrytrehen­slon. Unpublished doctoral dissertation. University of New Hampshire.

Duffy, S. 4., Henderson, J. M., & Molris, R. K. (1989).

Serrantic facjlitation of lexical access during sentence processing. Joru'nal of Experimental Psycft6l6g1t; Leaming, Metnoty, and Cogtlitiotl, I 5, 191-801.

Duffy, S. A., Morris, R. K., & Rayner', K. (1988). Lexical ambiguity and fixation times in reading. Journal of Memor¡, and Language, 27, 429446.

Duffy, S. ,A.., & Rayner, K. (1990). Eye movements and anaphor resolution: Effects of antecedent typicality and distance. Language atú Speeclt, 33(2), 103-119.

Ehrlich, S. F., & Rayner', K. (1983). Pronoun assignment and semantic integration during reading: Eye movements and immediacy of processing. Jounul of Verbal Learning and Verbal Behauiot', 22, 7 5 8'1 .

Foss, D. J. (1982). A discourse on semantic priming. Cogrtitiue Psycft6l6tr, 14, 590-60'7.

Foss, D. J., & Speer', S. R. (1991). Global and locai context effects in sentence processing. In R. R. Hoffman & D. S.

Palemo (Eðs.), Cognítion and tlrc symbolic processes:

Applied and ecological perspectiues (pp. I l5-139). Hillsdale, NJ: Erlbaum.

Garlod, S., & Sanford, A. (1983). Topic dependent effects in language processing. In G. B. Flores d'Arcais & R. J.

Jarvella (Eds.), Tlte process of language understanding (pp. 271-296). John Wiley.

Garrod, S., & Sanford, A. (1999). Incrementality in discourse understanding. In H. van Oostendolp & S. R. Goldman (Eds.), The constructiotl of mental representations duritxg reading (pp. 327). Mahwah, NJ: Lawrence Erlbaum Assoclates.

Garlod, S., & Terras, M. (2000). The contribution of lexical and situational knowledge to resolving discourse roles: Bonding and resolution. Journal of Mentory tmd Language, 42,526-544.

Gerrig, R. J., & McKoon, G. (1998). The readiness is all: The fnnctionality of memory-based text processing. Discourse Processes. 26. 67-86.

Goldon, P. C., Hendrick, R., & Foster', K. L. (2000). Language comprehension and probeJist memory. Jountal of Experi­ntental Psyclnlog)t: Learrxirxg, Mernory, and Cognítíon, 26,

766-77s. Hess, D. J., Foss, D. J., & Carroll, P. (1995). Effects of global

and local context on lexical plocessing during language comprehension. Journal of Experùnental Psychl¡ofr: 6r,r­eral. 124.62-82.

Inhoff, A. W., & Rayner, K. (1986). Parafoveal word processing dnling eye fixations in reading: Effects of word frequency. Perception & Psychoplrysics, 40, 431439.

Just, M. 4., & Carpenter, P. A. (1980). A theory of reading: From eye fixations to complehension. Psyclnlogical Reuiew,

87,329-354. Kintsch, W. (1988). The role of knowledge in discourse

complehension: A construction-integration tnodel. P s),cln­logi.cal Reuietv, 95, 163-182.

Kintsch, W. (1998). Contpreltension; A paradignt for cognitiott. New York, NY: Cambridge University Press.

MacDonald, M. C., Pearlmutter, N. J., & Seidenberg, M. S.

(1994). The lexical nature of syntactic ambiguity resolution. Psyclnlogical Reuiew, I01 ,6'76 :703.

Mo¡ris, R. K. (1994). Lexical and message-level sentence context effects on fixations times in reading. Journal of

288 A.E. Cook, J.L. trtf1'¿vs I Journal of Mentory and Langtøge 50 (2004) 268-288

Experitnental Psyclpl6g¡,; Learning, Mentory, and Cogni­tion. 20.92-103.

Myers, J. L., Cook, A. E., Kambe, G., Mason, R. 4., & O'Brien, E. J. (2000). Semantic and episodic effects on bridging inferences. Discourse Processes, 29, 179-199

Myers, J. L., & O'Brien, E. J. (1998). Accessing the discourse representation during reading. Discourse Processes, 2 6, 131­r57.

O'Brien, E. J., Cook, A. 8., & Derepentigny, K. A. (2001). The psychoiogy of knowledge activation in text comprehension and problem solving. In W. Kintsch (Ed.), The international encyclopedia of the social and behauíoral sciences (Vol. 12,

pp. 8114-8117). Oxford: Pergamon. O'Brìen, E. J., & Myers, J. L. (1999). Text comprehension: Aview

from the bottom-up. In S. R. Goldman, A. C. Graesser, & P.

van den Broek (Eds.), Narratiue compreltension, causality, and coherence: Essays ín ltonor of Tom Trabasso (pp. 35-53). Mahwah. NJ: Lawrence Er'lbaum Associates.

Rayner, K. (1998). Eye movements in reading and information processing: 20 years of research. Psychological Bulletin, 124,

372422. Rayner, K., & Pollatsek, A. (1989). The psyclzology of reading.

Hillsdale, NJ: Lawrence Erlbaum Associates. Rayner, K., Raney, G. 8., & Pollatsek, A. (1995). Eye

movements and discourse processing. In R. F. Jr., Lorch & E. J. O'Brien (Eds.), Sozrces of colrcrence in reading (pp. 9-35). Hillsdale, NJ: Lawrence Erlbaum Associates.

Rayner, K., Sereno, S. C., Morris, R. K., Schmauder, A. R., & Clifton, C. (1989). Eye movements and online langnage comprehension processes. Language and Cognitiue Pro­cesses, 4,2149.

Reichle, E. D., Pollatsek, 4., Fisher, D. L., & Rayner, K. (1998). Toward a model of eye movement control in reading. P sltclxo¡ofrro, Reuiew, I 0 5, 125-157.

Rizzella, M. L., & O'Brien, E. J. (2002). Retrieval of concepts in EJ

script-based texts and narratives: The influence of general

world knowledge. Journal of Experintental Psycholog¡,; Learning, Mentory, and Cognition,28,780 790.

Sanford, A. J., & Ganod, S. C. (1981). Understanding vritten language. Chichester: John Wiley.

Sanford, A. J., & Garrod, S. C. (1989). What, when, and how?:

Questions of immediacy in anaphoric reference resolution. Language and Cognítiue Processes, 4,235-262.

Schank, R., & Abelson, A. (1977). Scripts, plans, goals, and understanding: An enquiry into ltutnan knowledge sîructures. Hillsdale, NJ: Erlbaum.

Schwanenflugel, P. J., & Shoben, E. J. (1985). The influence of sentence constraint on the scope of facilitation for upcoming words. Journal of Mentory and Language, 24,232-252.

Schwanenflugel, P. J., & White, C. R. (1991). The influence of paragraph information on the processing of upcoming words. Reading Researclt Quarterly, 26, 160-177.

Sereno, S. C., Blewer, C. C., &. O'Donnell, P. J. (2003). Context effects in word recognition: Evidence for eariy interactive processing. Psycftl¡of¡ro, Science, 14, 328-333.

Till, R. E., Mross, E. F., & Kintsch, W. (1988). Time course of priming for associate and inference words in a discourse context. Metnory & Cognition, 16,283¿98.

Traxler, M. J., & Foss, D. J. (2000). Effects of sentence

constraint on priming in natural language comprehension. Jounml of Experimental Psyclnlogy: Learning, Mentory, and Cognition, 26, 1266-1282.

Wiley, J., & Raynel, K. (2000). Effects of titles on the processing of text and lexically ambiguous words: Evidence from eye movements. Mentory & Cognition, 28, 1011102I.

Zwaan,R.4., & Radvansky, G.A. (1998). Situation models in language comprehension and memory. Psychological Bulle­tin. 123.162-185.