DrscouRSE PROCESSES, 39(2&3), 265-27 8 Copyright @ 2005, Lawrence Erlbaum Associates, Inc.

What Have We Been Missing? The Role of General World Knowledge

in Discourse Processing

Anne E. Cook University of Utah

Sabine Guéraud University of Lyon

In recent years, memory-based and explanation-based theories have dominated the discourse processing literature. Numerous studies have been conducted to show sup­port for each of the two views. Most of these studies have manipulated factors in the episodic memory trace of texts, without a great deal of focus on how general world knowledge impacts processing. We describe several studies from the reading com­prehension literature that show strong effects of general world knowledge. We also present a framework that can account for both episodic and general world knowledge effects, and that may be used to reconcile the memory-based and explanaúon-based views.

Currently, the two most prominent positions within the research literature on text processing are the memory-based processing view (e.g., Gerrig & McKoon, 1998; Myers & O'Brien, 1998; O'Brien & Myers, 1999) and the explanation-based view (e.g., Graesser, Singer, & Trabasso, 1994; Singer, Graesser, & Trabasso, 1994). Although the two views have frequently been portrayed as conflicting, the com­bined results from both perspectives have led to a considerable advancement in our understanding of the comprehension process. Both views readily acknowledge the role that general world knowledge plays in the comprehension process, although explanation-based theories (e.g., Graesser et al., 1994;Zwaan & Radvansþ, 1998) have more readily appealed to the influence that general world knowledge has on

Correspondence and requests for reprints should be sent to Anne E. Cook, Department of Educa­tional Psychology, 1705 E. Campus Center Drive, Room 327, University of Utah, Salt Lake City, UT 841 12. E-mail: anne.cook@ed.utah.edu

266 COOK AND GUERAUD

comprehension. The goal of this article is to provide a framework that can account for general world knowledge effects in reading that are due to both memory-based and explanation-based processes. In this article, we show that the influence ofgen­eral world knowledge is often greater than acknowledged, independent of theoreti­cal perspective.

Within the memory-based view of text processing, the goal has been to explain as much of comprehension as possible, relying only on low-level automatic mem­ory retrieval processes such as resonance (Myers & O'Brien, 1998; O'Brien, 1995;

O'Brien & Myers, 1999). According to the resonance model, when new concepts are encoded into working memory, a signal is sent from these concepts to all of long-term memory in parallel (both the episodic memory trace and general world knowledge). Concepts in long-term memory that share features in common with the contents of working memory will "resonate" in response (cf. Ratcliff, 1978),

and those concepts that resonate the most are also most likely to be incorporated into working memory. Factors that have been found to influence this passive reacti­vation process include elaboration (e.g., O'Brien, 1987; O'Brien, Albrecht, Hakala, & P.izzella, 1995; O'Brien, Plewes, & Albrecht, 1990), distance (e.g.,

O'Brien, 1987), featural overlap (e.g., Albrecht & Myers, 1998; Albrecht & O'Brien, 1993; O'Brien & Albrecht,l99l), and causal connections (e.g., O'Brien & Myers, t987:Prizzella & O'Brien, 1996).

Therole ofhigherorderprocesses inreading has beenmore directly addressed by proponents ofthe explanation-based approach. According to this view, both auto­matic and controlled strategic processes interact to influence online processing.

Vy'hen new information is encoded, readers search long-term memory for any mean­ingful information that is relevant to the concepts in working memory to develop a

rich and complete representation of the situation described by the text. This search

may involve bottom-up retrieval processes, such as those described by mem­ory-based researchers, as well as more active higher order processing on the part of the reader. For example, explanation-based researchers argued that readers ofnarra­tives track the causes of events (e.g., Trabasso & Sperry, 1995; Trabasso & van den

Broek, 1985), attend to goal information (e.g.,Lntz & Radvansþ,1997; Singer & Halldorson, 1996; Suh & Trabasso, 1993; van den Broek, 1990), and may engage in different strategies or hold different standards of coherence, depending on the con­textand/orpurposeof thereading situation (e.g., vandenBroek,Lorch, Linderholm, & Gustafson,200l; van den Broek, Risden, & Husebye-Hartman, 1995).

A FRAMEWORK FOR RECONCILING MEMORY-BASED AND EXPLANATION-BASED VIEWS

The memory-based and explanation-based views are often presented as conflicting with one another, in part because memory-based theories focus primarily on the

GENERAL WORLD KNOWLEDGE EFFECTS 267

bottom-up components involved in reading and explanation-based theories take

into account top-down components. Cook and Myers (2004) recently proposed a

framework derived from research used to support memory-based processing that may also be used to explain the more strategic processes described by explana­

tion-based researchers. The framework, which operates via a bottom-up reactiva­

tion process (e.g., resonance), is a modified version ofthe bonding and resolution model proposed by Garrod and Sanford (1999; see also Garrod & Tenas, 2000;

Sanford & Garrod, this issue). The first stage involves linking the contents of ac­

tive memory with the first information reactivated from long-term memory via an

automatic retrieval mechanism-semantic or episodic information. Which type of information, if either, dominates this initial processing stage depends on the degree

of featural overlap between the concepts in active memory and long-term memory and the types of episodic manipulations used. According to this view, both epi­

sodic and semantic information may be reactivated and may interact to influence early stages of processing (e.g., Myers, Cook, Kambe, Mason, & O'Brien, 2000;

Rizzella & O'Brien, 2002; Sereno, Brewer, & O'Donnell ,2003; Wiley & Rayner,

2000). The second stage of the Cook and Myers (2004) framework describes pro­cessing as information continues to be reactivated and integrated with the contents

of active memory even after the reader has moved on in the text. That is, "slower"

information may not show its influences until later downstream (e.g., spíllover sentences, posttarget regions. regressions, second pass reading times, etc.).

What this framework leaves out is any description of when more active pro­cesses on the part of the reader come into play. O'Brien and Myers (1999) sug­

gested that when resonance fails to reactivate sufficient information for compre­

hension to proceed, readers may refocus on the contents of active memory and thus

"reboot" the resonance process or they may engage in active problem-solving pro­cesses such as those described by explanation-based theorists (Graesser et al.,

1994; Singer et a1.,1994; van den Broek, 1990; van den Broek et al., 1995). Con­

sidering that automatic processes have been shown to influence priming at an ear­

lier stage than controlled processes (e.g., Kintsch, Welsch, Schmalhofer, &Zimny, 1990; Neely, 1977), it may be that the flrst stage in the framework described by Cook and Myers (2004) is dominated by memory-based processes, whereas expla­

nation-based processes do not show their influences until the second, slower stage.

Therefore, this framework can explain both memory-based and explanation-based processes, as well as effects due to episodic and semantic variables.

Much of the evidence for the memory-based and explanation-based views has

come from studies in which variables in the episodic memory trace of a text have

been manipulated. This has often involved investigating how the relation between

information presented early in a passage influenced processing time on text pre­

sented later (e.g., a target sentence) or affected the availability ofa specific concept (e.9., a probe word). Studies of this nature often attempted to minimize semantic

influences on processing by holding general world knowledge constant. This is es­

268 COOK AND GUERAUD

pecially true for studies conducted within the memory-based framework (although

see Rizzella & O'Brien, 2002, for an exception). For example, following an ana­

phoric phrase, the availability of an antecedent is influenced by contextual factors

such as distance (e.g., O'Brien,1987; O'Brien & Myers, 1987), elaboration (e.g.,

O'Bríen et al., 1995; O'Brien et al., 1990), causal relations (e.g., O'Brien & Myers, 1981;Rizzella & O'Brien,1996), and distractors (e.g., O'Brien et al.,

199s). However, because an anaphor and its antecedent are not always lexically identi­

cal, the process of reactivating or accessing an antecedent in response to an

anaphor must be mediated by their pre-existing relation to one another in general

world knowledge; that is, antecedent retrieval processes must involve both general

world knowledge and episodic information (Cook, 2000; Duffy & Rayner, 1990;

Garrod & Sanford, 1977; O'Brten & Albrecht, 1991).

Even in those studies that assume that general world knowledge is activated, se­

mantic influences on processing are assumed to be held constant. This is especially

true in examinations of predictive inferencing (e.g., Cook, Limber, & O'Brien, 2001; Keefe & McDaniel, 1993; Klin, Guzman, & Levine, 1999; McKoon & Ratcliff, 1986; Murray, Klin, & Myers, 1993). For example, Cook et al. presented

readers with passage contexts (e.g., a person is standing on a scaffolding at the 15th

story of a building and then loses his balance and falls) and tested for the activation

of a predictive inference (e.g., dead). Although they held semantic factors (e.g.,

lexical associations) constant, the context was only supportive ofthe predictive in­ference because readers' general world knowledge contained the information that

something terrible usually happens to a person when he falls from great heights,

and that information was available to the reader and influenced comprehension. In the next section, we describe several studies that show the influence of semantic

variables in reading.

GENERAL WORLD KNOWLEDGE INFLUENCES IN READING

Given that general world knowledge can have a strong influence on the accessibil­

ity of information contained in the episodic memory trace, it is important to under­

stand how these two sources of information interact in determining what informa­tion ultimately becomes available to the reader (see O'Brien, Cook, & Derepentigny ,2001, for a discussion of the importance of understanding the inter­action between general world knowledge and information contained in the epi­

sodic memory trace). O'Brien et al. provided considerable evidence that informa­

tion from earlier portions of a text are reactivated when the reader encounters

related information; this occurs independent of whether that reactivated informa­

tion facilitates or hinders the processing of current information (e.g., Albrecht &

GENERAL WORLD KNOVTLEDGE EFFECTS 269

O'Brien, 1993; Myers & O'Brien, Albrecht, & Mason, 1994; O'Brien, Rizzella, Albrecht, & Halleran, 1998). Because the same activation mechanism is assumed

to operate on the episodic memory trace and general world knowledge in parallel, similar effects ought to occur as the result of the reactivation of information in gen­

eral world knowledge; that is, there should be conditions in which activated gen­

eral world knowledge facilitates comprehension and conditions in which it hinders comprehension.

However, it is difficult to draw direct comparisons between the influences of the

two types of information. Reactivated episodic informatíon appears to have its greatest influence at the level of the text base or situation model, whereas general

world knowledge appears to influence processing at a much wider variety of lev­els. For example, general world knowledge has been found to influence processing

at the lexical level (e.g., Binder & Rayner, 1998; Duffy, Morris, & Rayner, 1988;

Kambe, Rayner, & Duffy, 2001; Rayner &.Frazier, 1989; Rayner, Pacht, & Duffy, 199Ð, rhe featural level (e.g., Albrecht & Myers, 1995, 1998; Cook, 2000;

Guéraud, 2003; Guéraud & Tapiero, 2003; O'Brien & Albrecht, 1991), and the

scenario or script level (e.g., Bower, Black, & Turner, 19.19; Cook &}llyers,2004; Garrod & Terras, 2000; Rizzella& O'Brien, 2002). This is certainly not an exhaus­

tive list, but it highlights some of the major ways that general world knowledge im­pacts comprehension. In what follows, we discuss a select set of frndings that dem­

onstratÞ the impact of general world knowledge at each of these levels, how it interacts with information contained in the episodic memory trace, and how the

framework proposed earlier can account for these effects.

Lexical Knowledge

A considerable number of studies have been conducted on lexically ambiguous words (e.g., Binder & Rayner, 1998; Duffy et al., 1988; Kambe et al., 2001; Rayner &.Frazier, 1989; Rayner etal.,1994). One major finding from this research is the subordinate bias effect, which is clearly influenced by both general world knowl­edge and the episodic memory trace. This effect refers to the finding that, for bi­ased ambiguous words (i.e., ambiguous words for which one meaning is more dominant than the other meaninç, e.g., "bank"), when the prior episodic context supports the dominant meaning of the word (e.g., money), fixation times on that word a¡e shorter than in a control condition. But, when the episodic context sup­ports the subordinate meaníng of that word (e.g., river), fixation times on that word are longer than in a control condition. The subordinate bias effect is good evidence for the interaction of information from general world knowledge with information in the episodic memory trace. However, it also shows that although disambiguat­

ing context in the episodic memory trace may speed access of the subordinate

meaning of a word, it does not completely override general world knowledge (see

also Binder & Rayner, 1998; Kambe et a1.,2001).

270 cooKANDcuÉReuo

Wiley and Rayner (2000) investigated the subordinate bias effect by manipulat­ing the episodic memory trace in a way that was different from the manipulations used in previous studies. They compared reading times on lexically ambiguous words containing more than two meanings (e.g., fly) and matched control words, which were embedded in texts that were preceded by titles (e.g., "Worries of a

Baseball Manager") that disambiguated the words toward a particular meaning. Contrary to studies on the subordinate bias effect, Wiley and Rayner found that ti­tles that disambiguated words toward their more frequent subordinate meanings were suffrcient to eliminate the effect. Consistent with previous lexical access the­ories (e.g., Duffy et al., 1988), they concluded that context can interact with lexical information about a word 1o influence the speed of access of a particular meaning. The impact of context in this interaction may depend on such factors as the diffr­culty level of the texts (Rayner & Pollatsek, 1989), the strength (i.e., frequency) of the subordinate meanings, and the type of context used. This frnding demonstrates that when the information coming from the episodic memory trace is sufficient, it can override a strong general world knowledge effect very early in the reading pro­cess. Thus, according to the framework that we have proposed, at the lexical level, meaning activation is determined by an interaction of information coming from both general world knowledge and the episodic memory trace (see also Sereno et aI.,2003).

Featural Knowledge

In addition to lexical information, featural informatíon is also stored in general

world knowledge. Although "features" have never been specihcally defined in the

discourse processing literature, many theorists agree that they are an important ele­ment in reading (e.g., construction-integration [CI] model, Kintsch, 1988; reso­nance model, Myers & O'Brien, 1998; O'Brien & Myers, 1999). Features involve more basic units of meaning than can be captured by propositions (e.g., Kintsch, 1998; Kintsch & van Dijk, 1978) or lexical items. That is, concepts are made up of clusters of features and features can overlap among concepts. For example, the concept, "cat" contains the feature "fur," which is also a feature of the concept "dog"; however, it is important to note that each feature may also be composed of additional features.

One example from the memory literature in which featural information from general world knowledge can influence processing is the Moses lllusion. Erickson and Mattson (1981) asked participants questions such as, "How many animals of each kind did Moses take on the Ark?" and found that over 807o answered, "Two," even though they knew that it was actually Noah who sailed the biblical ark. How­ever, when the question referred to Nixon instead of Moses, error rates decreased

dramatically. Erickson and Mattson argued that although Moses and Noah repre­

sent two distinct individuals, the figures share several features in common (e.g.,

GENERAL WORLD KNOWLEDGE EFFECTS 27 1

Old Testament hgures, received messages from God, stories involved water), which may lead them to be easily mistaken for one another. These features in com­mon do not exist in the episodic memory trace, but instead are a parf of general

world knowledge; and that these two figures share features in common can lead to interference in the comprehension of a question about one of them (see also van

Oostendorp & de Mul, 1990; van Oostendorp & Kok, 1990). Although the Moses

Illusion can be mediated by manipulating the amount of information contained in the episodic memory trace (and by strategic factors), the effect has never been

completely eliminated (e.g., Bredart & Docquier, 1989; Bredart & Modolo, 1988;

Hannon & Daneman,2001; Kamas, Reder, & Ayers, 1996; Reder & Cleeremans, 1990).

O'Brien and his colleagues have provided evidence for the role offeatural in­formation during reading with their inconsistency paradígm (Albrecht & O'Brien, 1993; Cook, Halleran, & O'Brien, 1998; Hakala & O'Brien, 1995; Myers et al., 1994; O'Brien & Albrecht, 1992; O'Brien et al., 1998). In several studies, they demonstrated that readers exhibit longer reading times on a target sentence (e.g., "Mary ordered a cheeseburger and fries") that is inconsistent with previous text (e.g., "Mary was a health nut and a strict vegetarian"), compared to a control condition. It is important to point out that the target sentence and the

previous text do not create a direct contradiction (e.g., Mary eats meat. Mary does not eat meat). The two portions of text are inconsistent because they con­flict with information already stored in general world knowledge (e.g., vegetari­ans do not eat meat). That is, when readers encode the information that Mary is a vegetarian, the knowledge that vegetarians don't eat meat is likely activated and may or may not be integrated as part of the episodic representation of the text in memory. When the target sentence is encountered, it too would activate the information about meat from either the episodic memory trace or from gen­

eral world knowledge. In either case, it is the activation of this information from general world knowledge that creates the conflict between the target sentence

and previous text (see also, Cook et al., 1998; Long & Chong,2001; O'Brien et a1., 1998; O'Brien, Cook, & Peracchi, 2004).

Additional evidence for the role of featural information in reading comes

from studies investigating anaphoric resolution processes. Although general

world knowledge effects have been found with both nominal and pronominal anaphors (see Garnham,200l, for a review), this discussion is limited to investi­gations of nominal anaphoric references. Duffy and Rayner (1990; see also Garrod & Sanford, 1977) manipulated the typicality of an antecedent (e.g.,

robin/goose) and distance (close versus distant). Using eye tracking methodol­ogy to measure f,rxation times on the categorical anaphor (e.g., bird), they found interacting effects of typicality and distance, with the shortest fixation times on anaphors in the close, high-typical condition. In addition, typicality (but not dis­tance) continued to influence processing times after the eye moved away from

272 COOK AND GUERAUD

the anaphor. These studies demonstrate that the underlying semantic relation in general world knowledge between an anaphor and an antecedent is an important factor in anaphor resolution and that this relation may have stronger and longer lasting effects than some episodic relations (e.g., distance between concepts in the discourse).

O'Brien and Albrecht (i991) found that concepts that were highly related to an anaphor in general world knowledge, but not present in a text, can also im­pact anaphor resolution processes. They used passages containing statements

that shared several features in common with an unmentioned concept, "skunk" (e.g., terrifrc odor, small black animal, had a white stripe down its back, etc.). They used naming probe methodology to demonstrate that "skunk" was acti­vated in response to an anaphoric phrase, even though it had never been men­tioned and the correct antecedent, "cat," had been explicitly mentioned in the

text. They also found that if the passage contained highly supportive context for the unstated concept, and if the unstated concept had sufficíent featural overlap with the anaphoric phrase, the unstated concept could even be instantiated in place of the correct antecedent.

Cook (2000) also showed that when there are strong pre-existing associations between anaphors and potential antecedents, they can have strong effects on the resolution process. She presented readers with passages containing an object (e.g., "cello") that was referenced later in the text with an anaphor. The anaphor was either a correct reference for the antecedent (e.g., "cello"), an incorrect but highly related anaphor (e.g., "violin"), or incorrect and not highly related anaphor (e.g., "oboe"). Reading times for the sentence containing the anaphor decreased as a function of the pre-existing relation in general world knowledge between the anaphor and the antecedent. The correct anaphor condition (e.g., "cello") yielded faster reading times than the incorrect-high related condition (e.g., "violin"), which yielded faster reading times than the incorrect-low related condition (e.g., "oboe"). This effect persisted despite several attempts to override it with manipulations that were designed to increase the availability of the epi­sodic memory lrace of the antecedent (e.g., describing and elaborating on dis­tinctive characteristics of the antecedent, syntactic focus, and distance). It was

only reduced when all featural information about the antecedent was removed from the text. The results of this study, in combination with those described ear­

lier, provide support for the proposed framework by demonstrating that anaphor resolution processes are affected by semantic factors (i.e., general world knowl­edge) to the same degree, if not more than, episodic factors (e.g., elaboration or distance). This is especially important, given that most of the focus in research

on anaphoric resolution has focused on factors governing the episodic memory trace (e.g., O'Brien, 1987; O'Brien & Myers, 1987; O'Brien et al., 1990;

O'Brien et al., 1995).

GENERAL WORLD KNOWLEDGE EFFECTS 273

ScripVScenario Knowledge

A final level of general world knowledge that we discuss refers to knowledge about

scripts and scenarios. For example, when reading a narrative about a common ac­

tivity or situation (e.g., writing a letter, going to a rock concert, going to a restau­

rant, etc.), readers quickly reactivate information in general world knowledge re­

lated to the situation being described (e.g., Bower et al., 1979;Ptizzella & O'Brien, 2002). Several theorists (e.g., Garrod & Sanford, 1999; Kintsch, 1988) argued that

this semantic information may even be reactivated more quickly than episodic in­formation (i.e., information explicitly stated in the text).

Garrod and Terras (2000) investigated whether effects of general world knowl­edge on comprehension processes are immediate or delayed and how general

world knowledge effects are influenced by episodic factors. They tracked readers' eye movements for texts in which the first sentence contained common verbs (e.g.,

writing) in specihc scenarios (e.g., writing a letter, writing on a blackboard). The next sentence referenced a role filler for the verb that was: either explicitly stated

or implied in the first sentence; either dominant (e.g., pen) or nondominant (e.g.,

chalk) with respect to the verb; and either appropriate (e.g., writing a letter with a

pen) or inappropriate (e.g., writing on the blackboard with a pen) with respect to

the scenario described. They tested for effects of general world knowledge by comparing explicit conditions to implicit conditions. If "pen" is a dominant role filler for "write," it should be easy to process regardless of whether it had been ex­plicitly stated or not. However, if "chalk" is a nondominant role filler for "write," it should be easier to process when it was explicitly stated than when it was only im­plied. Context effects were examined by comparing the appropriate conditions (e.g., write a letter with a pen, write on the blackboard with chalk) to the inappro­priate conditions (e.g., write a letter with chalk, write on the blackboard with a

pen). They found that initial processing times (i.e., frst pass reading times) on the

role hller were influenced by general world knowledge and that information in the

episodic memory trace did not influence processing until later downstream (e.g.,

second pass reading times). Garrod and Terras argued that their results supported a

two-stage process.In the first stage, bonding, a role filler is linked with its verb via a bottom-up memory retrieval process that is dominated by general world knowl­edge. Once this link is formed, it is resolved with respect to the discourse (i.e., the episodic memory trace) in a second, resolution stage (see Sanford & Ganod, 1999,

this issue).

Cook and Myers (2004) investigated whether bonding and resolution would extend beyond verbs to more general scripted actíons. Their passages were script-based stories (e.g., about a rock concert) in which individuals (e.g., guitar­ist) performed actions that were either appropriate or inappropriate with respect to readers'general world knowledge about scripts (e.g., played a slow and heart­

274 cooKANDGUÉRAUD

felt love song vs. handled publicity and finances). In addition, in half of their passages, the individuals'inappropriate actions were supported by previous pas­

sage context. Like Garrod and Terras (2000), Cook and Myers tracked readers'

eye movements and found that when a character's action was not previously sup­ported by information in the episodic memory trace, general world knowledge

influenced first pass reading times on the role f,rller (e.g., guitarist)-fixation du­

rations were longer when the action was inappropriate than when it was appro­priate. However, when the action was previously supported, information in the

episodic memory trace appeared to dominate initial processing times on the role filler-there was no disadvantage for the inappropriate condition in first pass

reading times, but inappropriateness effects did appear later downstream (i.e., in second pass reading times). Cook and Myers argued that although general world knowledge will often become available more quickly than episodic information to influence the initial linkage stage, these results show that in some situations, strong episodic information may be reactívated quickly enough to influence early processing. As information (semantic or episodic) continues to be reacti­

vated and integrated with the incoming text, those effects appear in the second,

slower verification stage. To explain these results, Cook and Myers presented a

modified version of bonding and resolution-the framework presented earlier in this article.

CONCLUSIONS

Although much of the research in discourse processing has focused on the role of episodic variables in reading, it is clear from the research described here that gen­

eral world knowledge also has a strong influence on reading processes. The frame­

work presented in the introduction provides a mechanism that can explain both the

passive and strategic reactivation of episodic and semantic information. This framework allows for the possibility that early stages of reading are dominated by memory-based processes, whereas later stages of reading are controlled by expla­

nation-based processes. Because many of the studies that support either the mem­

ory-based view or the explanation-based view have used methodologies that do not provide a window into the time course of processing (e.g., Iine-by-line reading,

single-SOA probe methodologies, talk aloud protocols, etc.), this question remains

open. Eye tracking methodology, such as that used by Garrod and Terras (2000)

and Cook and Myers (2004), is one possibility for investigating this issue because

it allows for a more fine-grained analysis of on-line reading processes (see Rayner

& Pollatsek, 1989). Future research should make use of eye tracking and other

on-line methodologies to examine how and when memory-based and explana­tion-based processes influence reading.

GENERAL WORLD KNOWLEDGE EFFECTS 275

ACKNOWLEDGMENTS

We thank Ed O'Brien forhis helpful comments on an earlier version of this article.

REFERENCES

Albrecht, J. E., & Myers, J. L. (1995). The role ofcontext in accessing distant information during read­ing. Journal of Experimental Psychology: Learning, Mentory, and Cognition,2I, 1459-1468.

Albrecht, J. E., & Myers, J. L. (1998). Accessing dista¡t text info¡mation during reading: Eftècts of contextual cues. D i s cours e P ro c e s s e s, 26, 87 -107 .

AJbrecht, J. E., & O'Brien, E. J. (1993). Updating a mental model: Maintaining both local and global

coherence. Journal ofExperimental Psychology: Learning, Mentory, and Cognition, 19, 106l-1070­Binder, K. S., & Rayner, K. (1998). Contextual strength does not modulate the subordinate bias effect:

Evidence from eye fixations ald self-paced readrng. Psychonomic Bul.letin & Review 5,271-276. Bowe¡ G. H., Black, J. B., & Ti:mer, T. J. (1979). Scripts in memory for text. Cognitive Psychology, 11,

1'77-220. B¡edart, S., & Docquier, M. (1989). The Moses illusion: A follow-up on the focalization effect. Euro­

pean Bulletin of Cognitive Psychology, 9,357-362­Bredart, S., & Modolo, K. (1988). Moses strikes again: Focalization effect on a semantic illusion. Actd

Psychologíca, 67, 135-144. Cook, A. E. (2000). The Moses ilLusion in reading comprehension. Unpubìished doctoral dissertation,

University of New Hampshire. Cook, A. E., Halleran, J. G., & O'Brien, E. J. (1998). What is readily available during reading?: Amem­

ory-based view of text processing. Discourse Processes,26,109-129.

Cook, A. E., Limber, J. E., & O'Brien, E- J. (2001). Situation-based context and the availability ofpre­dictive inferences. loumal of Memory and Language, 44,220-234.

Cook, A. 8., & Myers, J. L. (2004). Processing discourse roles in scripted nanatives: The influences of context and world knowledge. Journal of Memary and Itnguage, 50,268-288.

Duffy, S. 4., Morris, R. K., & Rayner, K. (1988). Lexical arnbiguity and ñxation times in readrng. Jour­nal of Memory and lanzguage,27,429-446.

Duffy, S. 4., & Rayner, K. (1990). Eye movements and anaphor resoìution: Effects of antecedent typi­cality and distance. Language and Speech,33(2),103-1 19.

Erickson, T. D., & Matlson, M. E. (1981). Fromwords to meaning: A semantìc illusion. Journal oJVer­bal l-eaming and Verbal Behavior 20, 54U551.

Garnham, A. (2001). Mental models and the interpretation of anaphora. Philadelphia: Psychology hess.

Gar¡od, S., & Sanford, A. (1977). Interpreting anaphoric relations: The integration of semantic infor­mation while reading. Jounnl of Verbal Leaming and Verbal Behavior, 16,77-90.

Garrod, S., & Sanford, A. (1999). Incrementality in díscourse understanding. In H. van Oostendorp & S. R. Goldman (Eds.), The constructio,t of mental representations during reading. (pp. 3-27). Mahwah, NJ: Lawrence Erìbaum Associates, Inc.

Garrod, S., & Terras, M. (2000). The contribution oflexical and situational knowledge to resolving dis­

course roles: Bonding and resolution. Jountal of Memory and Langunge,42, 526-544. Gerrig, R. J., & McKoon, G. (1998). The readiness is alì: The functionality of memory-based text pro­

cessing. Drcoørse Processes, 26, 67-86. Graesse¡ A. C.. Singe¡ M., & Trabasso,'1. (1994). Constructing inferences dwing narrative text com­

prehension. P sy cho lo gi c al R ev ietv, I 0 I. 37 1-395.

276 cooKANDcuÉRAUD

Guéraud, S. (2003). Updating a situational model during narrative text reading: Strategíc or automatic process? Contributions and limits of the resonance model. Unpublished doctoral dissertation, Uni­versity ofLyon.

Guéraud, S., & Tapiero, I., (2003). Distinguishing befileen reactivation and integration mechanisms in tart comprehension. Poster presented at the 44th annuaì meeting of the Psychonomic Society, Van­

couvet Canada.

Hakala, C. M., & O'Brien, E. J. (1995). Strategies for resolving coherence breaks in reading. Discourse P rocess es, 20, 167 -185.

Hannon, 8., & Daneman, M. (2001). Susceptibility to semantic illusions: An individual-differences perspective. Memory & Cognition, 29, 449461 -

Kamas, E. N., Reder, L. M., & Ayers, M. S. (1996). Partial matching in the Moses lllusion: Response

bias not sensitivity. Memory & Cognition, 24,687-699. Kambe, Rayner, & Duffy 2001

Keefe, D. 8., & McDaniel, M. A. (1993). The time course and durability of predictive inferences. .Iaur­nal of Memory and Ia nguage, 32, 446-4$.

Kintsch, W. (1988). The role ofknowledge in discourse comprehension: A construction-integration model. Psycholo gical Review, 95, 163-182.

Kiutsch, W., Welsch, D., Schmalhofer, F., &Ztmay, S. (1990). Sentence comprehension: A theoretical analysis- Joumal of Memory and Language,29,133-159­

Klin, C. M,, Guzman, A- E., & Levine, W H. (1999). hevalence and persistence of predictive infer­ences. Joumal of Memory and Innguage,40,593-6M.

Long, D. L., & Chong, J. L. (2001). Comprehension skill and global coherence: A paradoxical picture

ofpoor comprehenders' abilities. .Iournal ofExperimental Psychology: Itamíng, Memory and Cog­nition.27. 1424-1429.

Lutz,M. F., & Radvansky, G. A. (1997). The fate of completed goal information in narrative compre­hension. Journal of Memory and ktnguage, 36,293-310.

McKoon, G., & Ratcliff, R. (1986). Inferences about predictable eveits. Journal of Experimental Psy­

chology: Leamìng, Memory, and Cognition, 12,82-91. Murray, J. D., Klin, C. M., & Myers, J. L. (1993). Forwa¡d inferences in nar¡ative text.,Ioumal of Mem­

ory and language, 32,464473. Myers, J. L., Cook, A. E., Kambe, G., Mason, R.A'., & O'Brien, E. J. (2000). Semantic andepisodic ef­

fects on bridging inferences. Discourse Processes, 29, 179-199. Myers, J. L., & O'Brien, E. J. (1998). Accessing the discourse representation during reading. Discourse

Processes, 26, 131-157 .

Myers, J. L., O'Brien, E. J., Albrecht, J. E., & Mason, R. A. (1994). Maintaining global coherence.

Journal of Experimental Psychology: Leaming, Mentory and Cognition,20,876-886. Neely, J. H. (1977). Semantic priming and ret¡ieval from lexical memory: Roles of inhibitionless

spreading activation and limited-capacity attention. Journal of Experùnental Psychology: GeneraL,

106,22Ç254. O'Brien, E. J. (1987). Antecedent search processes and the structure of lext. k¡umal of Experimental

P sycho lo gy : I-e arning, M emo ry, and C o gnit ion, I 3, 2'l 8-290. O'Brien, E. J. (1995). Automatic conponents of d.iscourse comprehension. In R. F. Lorch & E. J.

O'Brien (Eds.), Sources of coherence in reading (pp. 159-176). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.

O'Brien, E. J., & Albrecht, J. E. (1991). The role of context in accessing antecerJents in text. Joumal of Exp e rimental P sy chol o gy : Le arn in g, Memo ry, and C o gttition, 1 7 (1), 9 4-102.

O'Brien, E. J., & Albrecht. J. E. (1992). Comprehension strategies in the development of a mental model. Journal of Experimental Psychology: Learning, Mentory, and Cognition, 18,'1'7'1-184.

GENERAL WORLD KNOWLEDGE EFFECTS 277

O'Brien, E. J., Albrecht, J. E., Hakala, C. M., & Rizzella, M. L. (1995). Activation and suppression of antecedents during reinstatement. Journal of Experìmental Psychology: Itaming, Memory and

Co gnition, 2 1, 626-634. O'Brien, E. J.. Cook, A. E., & Derepentigny, K. A. (2001). Psychology of knowledge activatiolÌ in text

comprehension and problem solving. In W Kintsch (Ed.), International encyclopedia of the behav­

ioral and social sciences (Yol. 12, pp. 81 13-81 17).

O'Brien, E. J., Cook, A. E., & Peracchi, K. A. (2004). Updating situation models: Reply to Zwaan and

Madden (2O04). Journal of Experimental Psychology: Lzanùng, Memory', and Cognition, 30,

289-291. O'Brien, E. J., & Myers, J. L. (1987). The role of causal connections in the relrieval of rcxt: Memory &

Cognition, 15,419427. O'Brien, E. J., & Myers, J. L. (1999). Text comprehension: A view from the bottom up. In S. R.

Goldman, A. C. Graesser, & P. van den Broek (Eds.),Narrative contprehension, causalily, and co­

herence: Essays inhonor ofTomTraå¿sso (pp. 35-53). Mahwah, NJ: Lawrence Etlbaum Associates,

Inc. O'Brien, E. J., Plewes, P. S., & Albrecht, J. E. (1 990). Antecedent retrieval processes. Joumtù ofùrper­

imental P sy c holo gy : lc arning, M emo ry and C o gnition, I 6, 24 | -249. O'Brien, E. J., Rizzella, M. L., Albrecht, J. 8., & Halleran, J. G. (1998). Updating a situation model: A

memory-based text processing víew. Journal of Experimental Psychology: Learníng, Memory, and

Co gnition, 24, 1200-1210. Ratcliff, R. (1978). A theory of memory retrieval. Psychological Rev-ievu 85, 59-108.

Rayner, K., & Frazier, L. (1989). Selection mechanisms in reading lexically ambiguouswords. Journal of Lrperimental Psychobgy: Learning, Memory, and Cognition, 15,'779-790.

Rayner, K., Pacht, J. M., & Duffu, S. A. (1994). Etrects of prior encounter and global discourse bias on

the processing of lexically ambiguous rvords: Evidence from eye ñxations. Joumal of Memory and

Language, 3 3, 52'7 -544. Rayner, K., & Pollatsek, A. (1989).The psychology of reading.Hillsdale, NJ: Law¡ence Erlbaum Asso­

ciates, Inc-

Reder, L. M., & Cleeremans, A. (1990). The role of partial matches in comprehension: The Moses lllu­sion revisited. In A. C. Graesser & G. H. Bower (Hs.), The psychology of leaming and motivation (Vol. 25, pp.233-258). San Diego, CA: Academic.

Reder, L. M., & Kusbit, G. W. (1991)- Locus of the Moses lllusion: Imperfect encoding, retrieval, or match? Joumal of Memory and Lantguage, 30,385-406.

Rizzella, M. L., & O'Brien, E. J. (1996). Accessing global causes during reading. Joumal ofExperi­menral Psychology: Leaming, Memory, and Cognition.,22, 1208-1218.

Rizzella, M. L., & O'B¡ien, E. J. (2002). Retrieval ofconcepts in script-based texts and na¡ratives: The

influence ofgeneral world knowledge. Journal ofExpertmental Psychology: Itaming, Memory, and

Cognition,28,78lu--790. Sereno, S. C., Brewer, C. C., & O'Donnell, P. J. (2003). Context effects in word recognition: Evidence

for early interactive processing. P sy cho Io g ic al S c ienc e, I 4, 328-333. Singer, M., Graesser, A. C., &Trabasso, T. (1994). Mìnimal orglobal inference during reading. Joumal

of Memory and Language, 33,427-444. Singer, M., & Halldorson, M. (1996). Constructing and validating motive bridging inferences. CogzË

tive P sycholo gy, 30, 1 -38. Suh, S., & Trabasso, T, (1993). Inferences during reading: Converting evidence from discou¡se analy­

sis, talk-aloud protocols, and recognition púming. Journal oJ'Memor¡^ andktnguage,32,2'79-3W. Trabasso, T., & Sperry, L. L. (1985). Causal relatedness and importance of story events. Journal of

Memory and Language, 24,595-61 l

278 cooKANDGUÉRAUD

T¡abasso, T., & van den Broek, P. (1985). Causal thinking and the representation of nârrative events.

Journal of Memory and Language,24,612-630. van den Broek, P. (l 990). The causal inference maker: To*,ards a process model ofinfe¡ence generation

in text comprehension. In D. A. Balota, G. B. Flores d'Arcais, & K. Rayner (Eds.), Comprehension

processes in reading (pp. 423-445). Hillsdale, NJ: Lawrence E¡lbaum Associates, Inc. van den Broek, P.. Lorch. R. F., Linderholm, T., & Gustafson. M. (2001). The effects of readers'goals

on inference generation and memory for texts. Memory & Cognitiott,29, 1081-1087 .

van den Broek, P., Risden, K., & Husebye-Hartmann, E. (1995). The role of readers' standards for co­herence in the generation ofinferences during reading. ln R. F. Lorch & E. J. O'Brien (Eds.), Sourc'es

of coherence in reading (pp. 353-373). Hillsdale, NJ: Lawrence Erìbaum Associates, Inc. van Oostendorp, H., & de Mul, S. (1990). Moses beats Adam: A semantic relatedness effect on a se­

mantic illusion. Acta Psychologica, 74, 3546. van Oostendorp, H., & Kok, I. (1990). Failing to notice erro¡s in sentences. Language and Cognitive

Processes.5. 105-1 13.

Wiley, J., & Rayner, K. (2000). Effects of titles on the processing of text and ìexically ambiguous

words: Evidence from eye movemenTs. Memory & Cognition,28,1011-1021. Zwaan, R. 4., & Radvansþ G. A. (1998). Situation models in language comprehension and memory.

P sycho lo gical Bulletin, I 23, 162-185.