Professional Documents
Culture Documents
net/publication/302590943
CITATIONS READS
47 2,098
2 authors:
All content following this page was uploaded by Edward J O'Brien on 27 March 2018.
This chapter was originally published in the book The Psychology of Learning and
Motivation, Volume 65. The copy attached is provided by Elsevier for the author's
benefit and for the benefit of the author's institution, for non-commercial research, and
educational use. This includes without limitation use in instruction at your institution,
distribution to specific colleagues, and providing a copy to your institution's
administrator.
CHAPTER SEVEN
Contents
1. Introduction 249
2. The RI-Val Model of Comprehension 252
2.1 General World Knowledge Versus Contextual Influences on Comprehension 256
2.2 Mediating Influences on Validation 259
2.3 When Validation “Fails” 263
2.4 Manipulating the Coherence Threshold 268
3. Implications of the RI-Val Model 270
4. Conclusions 271
References 272
Abstract
The goal of this chapter is to build on previous models of discourse comprehension
that break reading into two components: activation and integration. We describe a
new model of comprehension, the RI-Val model, which separates out the mechanisms
involved in activation and integration, and adds an additional validation stage to pro-
cessing. The assumptions of this model have implications for the types of information
reactivated from memory, how this information can mediate subsequent processing,
and how the timing of different influences is affected. We provide evidence from
our own work to support the model and discuss broad implications of the model for
future research and theory in discourse comprehension.
1. INTRODUCTION
All models of discourse comprehension share the assumption that
readers must link incoming content with previously encoded and reactivated
informationdwhether from previously read text or from general world
Psychology of Learning and Motivation, Volume 65
© 2016 Elsevier Inc.
j
ISSN 0079-7421
http://dx.doi.org/10.1016/bs.plm.2016.03.004 All rights reserved. 249
(eg, Myers & O’Brien, 1998; O’Brien & Myers, 1999; Sanford & Garrod,
1998, 2005). All models also share the assumption that construction and
integration (ie, activation þ integration) are continuous and cyclical, such
that encoding of new information automatically initiates new construction
and integration cycles.
Long and Lea (2005; see also Long, Oppy, & Seely, 1997; Long, Seely, &
Oppy, 1996) have noted that the issue on which researchers tend to disagree
is the role of top-down, or evaluative, processes during the integration
phase. Researchers who espouse the memory-based view have traditionally
downplayed the role of strategic processing (van den Broek et al., 1996;
McKoon & Ratcliff, 1992, 1998; Myers & O’Brien, 1998; O’Brien &
Myers, 1999), whereas others have argued that integration is more strategi-
cally evaluative or analytical in nature (eg, Graesser et al., 1994; Singer,
Graesser, & Trabasso, 1994; Zwaan & Radvansky, 1998). Long and Lea
presented a compelling argument for a modified view of postactivation pro-
cessing that allowed for both passive convergence-like integration processes
as well as processes that are more strategically evaluative in nature (see also
Ferretti, Singer, & Harwood, 2013; Isberner & Richter, 2014; Singer,
2006, 2013).
Evidence from our own work (eg, Cook et al., 2014; Cook & O’Brien,
2014) has largely supported the argument that following initial integration,
there must be an additional evaluative (but not necessarily strategic) compo-
nent involved in comprehension. Cook and Myers (2004) argued that
incoming information results in the activation of related information from
memory, and that activated content is then linked to the incoming informa-
tion. However, they argued that the link is then verified, or evaluated,
against additional information in a subsequent stage. This differed from
traditional two-stage models because of the assumption that processing con-
tinues even after initial integration appears to be complete. Similarly, Cook
et al. (2014) provided evidence that the initial inference activation process is
passive and dumb; readers activated predictive inferences that were related to
outdated (ie, no longer relevant to the current discourse model) content in
the text. That is, the initial inferences activated were irrelevant to the current
status of the discourse model. In a subsequent experiment, however, Cook
et al. demonstrated that only relevant inferences (ie, those related to the cur-
rent discourse model) were integrated into the long-term memory represen-
tation of the text, whereas irrelevant inferences were “pruned” out. They
argued that there must be some process by which activated information is
evaluated against both its relatedness and relevance to the ongoing discourse
representation (see also Rizzella & O’Brien, 1996; Singer Halldorson, Lear,
& Andrusiak, 1992; Singer, Harkness, & Stewart, 1997). These findings in
support of an evaluative stage during comprehension led us to propose
the RI-Val model outlined in the next section.
integration (I) stage. Drawing from the work by Kintsch and colleagues
(Kintsch, 1988, 1998; Kintsch & Van Dijk, 1978; Van Dijk & Kintsch,
1983), we assume that these linkages are made on the basis of general con-
ceptual overlapdor goodness of fit. Once these initial linkages have been
formed, we assume that they are then subject to validation (eg, Cook
et al., 2014; Cook & Myers, 2004; Garrod & Terras, 2000; Long & Lea,
2005). Thus, we added an additional (Val) stage to the RI-Val model.
The idea that information in text is validated against a reader’s broader
knowledge base is well supported in the research and writings of Singer
and colleagues (eg, Singer, 2006, 2013; Singer & Doering, 2014; see also
Isberner & Richter, 2014; Richter, 2015). However, considerably less
research has been devoted to understanding the specific mechanisms
involved in validation; and unpacking and exploration of the validation pro-
cess is the most important contribution of the RI-Val model.
Within the RI-Val model, the validation process is based on simple, pas-
sive, pattern-matching processes (similar pattern matching mechanisms have
been used to explain several components of comprehension; see Kintsch,
1988, 1998; Sanford, 1990). Specifically, we have adopted a mechanism
similar to that described by Reder in the partial matching assumption of
her Featural Familiarity Hypothesis (eg, Kamas & Reder, 1995; see also
Kamas, Reder, & Ayers, 1996; Reder & Kusbit, 1991). Although Reder
proposed this account to describe access of information from memory, we
have adapted her assumptions to the process through which linkages from
the I (Integration) stage are validated against the activated contents from
long-term memory. That is, initial linkages are matched against activated in-
formation at the featural level; linkages “match” activated content if they
share many clusters of features, even if the relations between the clusters
of features or the lexical items representing them are not identical. As noted
by Kamas and Reder, this processing heuristic is “quick, easy, and relatively
accurate” (p. 184), and that “a rigid comprehension system would have a
difficult time indeed” (p. 191) of handling the kind of complex linguistic
input readers and speakers are faced with on a regular basis.
When combined, the activation and integration stages of the RI-Val
model are comparable to the Construction stage in Kintsch’s (1988) CeI
model. We chose to divide this initial phase of processing into two distinct
stages to clarify that: information is activated via a purely associative reso-
nance process; and that the activated content is then linked in an integration
(I) stage to existing information in working memory on the basis of “good-
ness of fit” (primarily semantic overlap). The primary difference between
our RI-Val model and Kintsch’s CeI model is between our Validation stage
and his Integration stage. Kintsch assumed that activation of individual link-
ages between propositions are stabilized via a connectionist convergence
mechanism. Based on previous research, we assume that these linkages are
validated, albeit passively, against activated information from memory via
a more flexible, passive pattern matching process, and that this process plays
out over time.
Prior to an in-depth discussion of the predictions and consequences of a
passive validation process, it is important to outline in more detail the critical
assumptions of the RI-Val model, which is illustrated in Fig. 1. The first crit-
ical assumption is that activation, integration, and validation are all assumed
to be passive in nature; as a consequence, once each process starts, it runs to
completion. By “completion” we assume that each process arrives at a stable
state, similar to the stabilization of activation networks and vectors assumed
in Kintsch’s (1988) CeI model. The three curves in the figure depict the
activation, integration, and validation processes assumed in the RI-Val
model. We assume that upon encoding new information, that information
sends a signal to all of memory, reactivating related information from inac-
tive portions of memory via resonance (R). As soon as activation levels rise
above a minimum threshold (horizontal dotted line), the reactivated infor-
mation has the potential to influence comprehension. Information at higher
Figure 1 The RI-Val Model of Comprehension. This figure represents the parallel asyn-
chronous nature of the Resonance, Integration and Validation stages assumed by the
model. Reprinted from O’Brien, E.J., & Cook, A.E. (2016). Coherence threshold and the
continuity of processing: the RI-Val model of comprehension. Discourse Processes,
Accepted for publication by Taylor & Francis LLC (http://www.tandfonline.com).
to subsequent text) (O’Brien & Cook, 2016). This flexible account of the
timing of validation effects is a consequence of the assumption that the vali-
dation process is continuous. This differs from other instantiations of valida-
tion in which the process is assumed to be discrete (eg, Richter, 2015;
Singer, 2006, 2013). We assume that validation begins immediately after in-
formation is reactivated above threshold and integrated with active memory;
but more importantly, we assume that validation processes continue to oper-
ate even after a reader has accrued sufficient activation, integration, and vali-
dation (ie, reached their coherence threshold) and moved on in a text.
In the following sections, we will review evidence from our work that
supports these assumptions. Section 2.1 will focus on the passive and unre-
stricted nature of activation, which allows for a “race” for initial reactivation,
integration, and validation between information reactivated from general
world knowledge and previously encoded context. Section 2.2 will provide
evidence that the factors that influence reactivation of information also
mediate subsequent validation processes. Section 2.3 will describe new ev-
idence regarding validation when readers are faced with semantic anomalies.
And Section 2.4 will focus on the role of the coherence threshold. After
describing this evidence, we will end with a discussion of broader implica-
tions of the RI-Val model regarding current and future research in discourse
comprehension.
Cook, Kambe, Mason, & O’Brien, 2000). This implies that strong contex-
tual information (eg, elaborated or causally connected content) has the po-
tential to influence initial processing (see, for example, Colbert-Getz &
Cook, 2013; Kendeou, Smith, & O’Brien, 2013). But because processing
is continuous, general world knowledge will still be contacted, reactivated,
and will eventually influence comprehension, albeit after a delay. This view
that either source of information can dominate initial processing differs from
other accounts in the literature that assume that general world knowledge
has a stronger influence on initial processing than contextual information
(eg, Garrod & Terras, 2000; Kintsch, 1988; Sanford & Garrod, 1989), or
vice versa (eg, Hess, Foss, & Carroll, 1995; Nieuwland & Van Berkum,
2006; Van Berkum, 2008).
Rizzella and O’Brien (2002) examined this issue in the context of script-
based texts and narratives. Script-based texts naturally rely heavily on general
world knowledge, whereas narratives are more strongly driven by contextual
information. They measured the accessibility of concepts that were either
central or peripheral to both kinds of texts. For example, in a short narrative
passage about “Memories of a Marriage,” they embedded a concept that
was central to the narrative (eg, train) and a concept that was peripheral to
the narrative (eg, ladder). Consistent with prior work (eg, Albrecht &
O’Brien, 1991), they found that in a subsequent recognition test, central
concepts were correctly recognized more quickly than peripheral concepts.
Within the context of RI-Val, this occurred because contextual information
tends to dominate narratives and accessibility was driven by context. Indeed,
because narratives tend to be unique, the amount of activation from general
world knowledge would be low relative to the activation from the episodic
memory trace. As a result, access should be dominated by activation
emanating from the episodic memory trace (ie, context), leading to easier
retrieval of central concepts than peripheral concepts.
However, Rizzella and O’Brien (2002) also used a script-based passage
about “Going to a Restaurant” in which they embedded a concept that
was central to the script (eg, waiter) and a concept that was peripheral to
the script (eg, spoon). They found the opposite pattern: peripheral concepts
were recognized more quickly than central concepts. According to RI-Val,
the strong link between the script and general world knowledge led to the
activation of concepts from both the episodic memory trace of the script (ie,
context) as well as from general world knowledge (ie, script-based knowl-
edge). Further, the activation from general world knowledge would be
greater for central concepts than peripheral concepts. The activation
Introduction
The fans at the rock concert were going wild. They quieted down when the next
song
First encounter
Appropriate (AI): was played by the band’s guitarist.
Inappropriate (Il): was played by the band’s manager.
Background
The concert was being held in an outdoor stadium, and there were about ten
thousand screaming fans in attendance. They were mostly all hysterical teenage
girls. Some even cried as they
Second encounter
Appropriate (A2): listened to the guitarist.
Inappropriate (12): listened to the manager.
Conclusion
play a slow and heartfelt love song. The song was about a love affair that ended
badly.
(ie, whether it is large or small), but that size information did not influence
processing unless it was incongruent with the surrounding context. More-
over, these incongruence effects only showed up in measures of processing
that are clearly postintegration in nature, lending support to the idea that
validation occurs after, and depends on output of the reactivation and inte-
gration stages.
The RI-Val model’s asynchrony assumption also means that validation
should be mediated by factors that influence earlier-occurring processes.
This hypothesis was tested by Cook (2014). As illustrated in Table 2, she
created passages in which an anaphor (eg, cello) was either a correct refer-
ence to a previously encountered antecedent (eg, cello), an incorrect but
highly related reference to the earlier antecedent (eg, violin), or an incorrect
and low-related reference to the antecedent (eg, oboe). Time to process the
target sentence containing the anaphor was a function of the relation be-
tween the anaphor and the antecedent; correct anaphors yielded faster
reading times than incorrect high-related anaphors, and incorrect high-
related anaphors yielded faster reading times than incorrect low-related ana-
phors. And consistent with the first assumption of RI-Val that processing
runs to completion, the difficulty associated with the incorrect anaphors
continued into the spillover sentence that followed the target sentence.
Although both incorrect anaphors resulted in initial processing difficulty,
it was mediated by the underlying semantic relation between the anaphor
and the antecedent. Cook’s findings demonstrate that validation is mediated
by factors that influence reactivation of information (ie, underlying semantic
relation between a cue and a target), as well as the degree of contextual sup-
port for that underlying relation in the text.
The same factors that influence reactivation should mediate subsequent
processes. This implies that information in long-term memory (either earlier
portions of the discourse model or general world knowledge) that is strongly
related to incoming information should be reactivated quickly and should
have an earlier and stronger influence on validation than information that
is weakly related to incoming information. Weakly related information
may influence validation, but these effects may not be evident immediately;
instead, the impact of weakly related information is likely to be delayed until
after the reader has already moved on in the textdthat is, until after the
coherence threshold has been met. Consider the example passage in Table 3.
Using the inconsistency paradigm developed by O’Brien and colleagues (eg,
Albrecht & O’Brien, 1993; O’Brien & Albrecht, 1992), Cook and O’Brien
(2014) manipulated the underlying relation between a critical characteristic
Introduction
Terry loved classical music. She spent most of her waking hours listening to it, either
in her room or in the car.
Correct Antecedent
Terry decided that it would be fun to teach herself how to play. She drove to a
music shop located in the next town. As she entered the store she saw a beautiful
cello. The large instrument was almost bigger than she was. Terry decided she
wanted to learn how to play it. She imagined herself sitting down to play the
heavy instrument. Terry asked the salesman for a price. After thinking for a few
minutes, she decided to buy it that afternoon. Just then, Jill walked over to where
Terry was standing.
IncorrecteHigh-Overlap Condition
Terry decided to teach herself how to play. She and her friend Jill drove to a music
shop located in the next town. As they entered the store, Terry saw a beautiful
violin. The small instrument fit perfectly between her chin and shoulder. Terry
decided she wanted to learn how to play it. She imagined herself dancing as she
played the lightweight instrument. After thinking for a few minutes, she decided
to buy it. Just then, Jill walked over to where Terry was standing.
IncorrecteLow-Overlap Condition
Terry decided to teach herself how to play. She and her friend Jill drove to a music
shop located in the next town. As they entered the store, Terry saw a beautiful
oboe. The keys were bright and shiny, and the case was lined in black velvet.
Terry decided she wanted to learn how to play it. She imagined herself fingering
the keys to create perfect notes. After thinking for a few minutes, she decided to
buy it. Just then, Jill walked over to where Terry was standing.
Background
Jill told Terry she wanted to stop at the craft store on the way home. She had been
making a bunch of candle holders and had run out of supplies. She ran down her
list and told Terry that she needed some paint, glue, glitter, and brushes. She also
wanted to look to see if they had any stencils. After Jill finished her shopping,
they went home and unloaded all of their purchases.
Reinstatement and Spillover Sentences
Terry showed Jill the cello she bought.
She even tried to play a few notes.
Closing
Terry told Jill that she was going to start practicing that very evening.
of the protagonist (eg, Mary was a vegetarian) and information that was
inconsistent with this characteristic in the target sentence. In the high-
related condition, Mary ordered a cheeseburger; which is highly related to
the concept of “being a vegetarian.” In the low-related condition, Mary
orders a tuna salad, which is arguably more weakly related to the concept
Introduction
Today Mary was meeting a friend for lunch. She arrived early at the restaurant and
decided to get a table. After she sat down, she started looking at the menu.
Consistent Elaboration
This was Mary’s favorite restaurant because it had fantastic meat dishes. She enjoyed
eating anything that was quick and easy to fix. In fact, she ate at McDonald’s at
least 3 times a week. She never worried about her diet and saw no reason to limit
what she ate.
Inconsistent Elaboration
This was her favorite restaurant because it had fantastic health food. She, a health
nut, had been a strict vegetarian for 10 years. Her favorite food was cauliflower.
She was so serious about her diet that she refused to eat anything which was fried
or cooked in grease.
Background
After about ten minutes, Mary’s friend arrived. It had been a few months since they
had seen each other. Because of this they had a lot to talk about and chatted for
over a half hour. Finally, Mary signaled the waiter to come take their orders.
Mary checked the menu one more time. She had a hard time figuring out what
to have for lunch.
Target Sentences
High-related: Mary decided to order a cheeseburger.
Low-related: Mary decided to order a tuna salad.
She handed the menu back to the waiter.
Closing
Her friend didn’t have as much trouble deciding what she wanted. She ordered and
they began to chat again. They didn’t realize there was so much for them to catch
up on.
reached the coherence threshold, but that the inconsistency effect for the
low-related condition may not appear until after the reader had already sur-
passed the coherence threshold. Our results supported these predictions.
The inconsistency effect in the high-related condition was stronger and
appeared earlier (ie, on the target sentence), whereas the inconsistency effect
in the low-related condition was weaker and appeared later (ie, after the
reader had already moved on to the spillover sentence).
In Section 2.1, we argued that validation can be influenced by informa-
tion from both general world knowledge and the episodic discourse repre-
sentation (Cook & Myers, 2004; Rizzella & O’Brien, 2002) and that the
influences of each depend on the degree to which the overall passage
context supports underlying relations in general world knowledge (Cook,
2014). We have made the point in Section 2.2 that because processing is
continuous, effects may be observed either immediately or after a delay
(Cook & O’Brien, 2014). However, the studies supporting these assump-
tions all represented cases in which validation appeared to be complete;
that is, they represented situations in which validation ultimately resulted
in the detection (whether conscious or not) of inconsistent information.
In Section 2.3, we address a scenario in which validation is often considered
incompletedsemantic anomalies.
Sanford, Sanford, Molle, & Emmott, 2006; Sturt, Sanford, Stewart, &
Dawydiak, 2004; Ward & Sturt, 2007), or short texts (Barton & Sanford,
1993; Bohan & Sanford, 2008; Daneman, Hannon, & Burton, 2006;
Hannon & Daneman, 2004; Sanford, Leuthold, Bohan, & Sanford, 2011)
to demonstrate that readers often fail to detect anomalies (ie, errors) unless
the anomalies represent gross distortions of meaning and/or are highlighted
via some distinguishing characteristic (eg, capitalization, syntactic focus,
etc.). Furthermore, the tasks used in these studies typically require partici-
pants to consciously detect an anomaly (ie, provide a dichotomous
response); the measure is discrete and does not allow for the possibility
that information leading to detection could continue to accrue after a
response has been made. Within the RI-Val model, the validation process
continues even after the reader has passed their coherence threshold, allow-
ing for the possibility that processing difficulty resulting from anomalies
might only become evident after a delay (cf. Bohan & Sanford, 2008;
Sanford et al., 2011).
Williams, Cook, and O’Brien (2016) developed narrative passages that
contained semantic anomalies similar to the Moses Illusion (see Table 4).
However, our task involved reading for comprehension rather than anomaly
detection. We were thus able to measure processing on the erroneous term,
but also to measure delayed processing by assessing reading on the subse-
quent line of text (ie, spillover processing). We also manipulated the extent
to which the passage provided contextual support for the underlying relation
between the erroneous (Moses) and correct (Noah) terms. When this
contextual support was high (see High ContextdUnfocused Condition
in Table 4), there were several allusions to shared features between the erro-
neous and correct terms (eg, religious figure, Old Testament, etc.). When
contextual support was low, there were only a couple of allusions to their
shared features.
We assume that when the target sentence is encoded, related informa-
tion will be reactivated from memory. The curves in Fig. 2A and B depict
the relative influences of contextual information from the text as well as
general world knowledge on the validation process for both the high-
and low-context conditions. In the high-context condition (Fig. 2A), we
assume that contextual information (ie, shared features between Moses
and Noah; solid line) would initially dominate the validation process,
with the influence of general world knowledge (dotted line) trailing but
gaining additional influence over time. Thus the first information to be
reactivated and made available for integration and validation would include
Introduction
Tom was an avid reader who enjoyed many different kinds of magazines and books.
Each morning, Tom started his day by eating a bowl of cereal and reading the
morning paper. He found this time to be both relaxing and informative. When
Tom had any free time, he would sneak away to his den and read as much as he
could. He took great pride in being well read and saw reading as a way to broaden
his world knowledge.
High ContexteUnfocused Condition
Because of his desire to understand other people’s culture, Tom decided to read
some of the bible. He began with the Old Testament and the many stories found
within its chapters. Tom read about the beginning of mankind and a great flood
that God had used to punish his people. Although Tom was not devotedly
religious, he thought that by reading about prominent religious figures he would
better understand other people’s views. One new piece of information Tom
learned was that
Low Context
Because he had the day off from work, Tom decided that he was going to spend the
majority of the day reading. Tom was excited for some quiet time. After reading
the latest sports news on the internet, he happily settled down in his study with a
book on world culture and religion. The first few chapters Tom read were
written about some of the opening stories of the bible. One new piece of
information Tom learned was that
High ContexteFocused Condition
Because of his desire to understand other people’s culture, Tom decided to read
some of the bible. He read about a great flood that God had used to punish the
evils of mankind and about the massive ship that was constructed to save the
animals from the flood. Although Tom was not devotedly religious, he thought
his readings would help him to understand other people’s views and better relate
to them. One new piece of information Tom learned was that
Target Sentence
Noah/Moses brought two animals of each kind on the ark.
Spillover Sentence
Tom enjoyed learning about the ark in the bible.
Closing
Eventually, Tom’s eyes grew tired and he got up to take a break. He made himself a
sandwich and thought about all the interesting things he had read about.
Figure 2 (2A and B) Relative influences of context and general world knowledge on
validation over time as a function of context condition. Immediate influences on pro-
cessing are shown to the left of the coherence threshold, whereas spillover effects
are shown to the right of the coherence threshold. Fig. 2A depicts these influences
over time in the high-context condition, and Fig. 2B depicts the influences in the
low-context condition. In both figures, CT stands for coherence threshold.
an anomaly before the reader has reached their coherence threshold. How-
ever, activation, integration, and validation processes would continue to run
even after the reader’s coherence threshold has been met and they have
moved on in the text. Thus, it is likely that other more distinguishing infor-
mation from general world knowledge would still be contacted and reacti-
vated, integrated, and validated. Because this latter information would
contain increasing amounts of distinguishing information from general
processes, such as eye movement control, have long viewed spillover effects
as an important window into the nature of early occurring processes during
reading (eg, EZ Reader Model; Reichle, Rayner, & Pollatsek, 2003).
The Williams et al. (2016) experiments just described assume that readers
require a moderate level of match between reactivated content and
incoming information in order to meet the coherence threshold and
move on in the text. That is, they may wait for some, but not all, related
information to accrue and influence processing before shifting attention to
subsequent information. This coherence threshold should be somewhat
flexible in nature, such that readers may wait for relatively little information
to accrue before moving on, or they may wait for a great deal of information
to accrue before continuing in the text. This issue is considered in the next
section.
Figure 3 (3A and B) Relative influences of context and general world knowledge on
validation over time as a function of context condition and coherence threshold manip-
ulation. In both figures, CT stands for Coherence Threshold. The Low CT line represents
influences of information on immediate and delayed processing when the coherence
threshold was low, and the High CT line represents these influences when the coher-
ence threshold was high. As in Fig. 2A and B, the dotted line represents the change
in influence of general world knowledge over time, and the solid line represents
contextual information.
be delayed until the spillover sentence (or possibly missed altogether) inde-
pendent of context. In a fourth experiment, Williams et al. (2016) lowered
the coherence threshold by reducing the number of comprehension ques-
tions to one every few passages. In the high-context condition (Fig. 3A),
readers no longer experienced any observable processing difficulty on either
the target or the spillover sentence. In the low-context condition (Fig. 3B),
processing difficulty was delayed until after readers had already passed the
coherence threshold and moved on to the spillover sentence.
Studying the role of the coherence threshold in validation is a fertile area
for research in comprehension, as relatively little is known about this stage of
processing compared to the reactivation and integration stages. Although
Williams et al. (2016) varied the coherence threshold with a relatively subtle
manipulation, it is also likely to be influenced by stronger manipulations of
task, text, or reader variables (eg, standards of coherence; van den Broek
et al., 1995, 2002, 2011). The final section of this chapter will discuss broad
implications of the RI-Val model and related questions for future research in
discourse comprehension.
4. CONCLUSIONS
In conclusion, the RI-Val model (Cook & O’Brien, 2014; O’Brien &
Cook, 2016) builds on previous two-stage activation þ integration models
(Kintsch, 1988, 1998; Long & Lea, 2005; Rizzella & O’Brien, 1996,
2002; Sanford & Garrod, 1989) by mapping on an additional validation
stage. We have relied on passive mechanisms to explain effects across the
time course of processing, without appealing to strategic actions on the
part of the reader. We consider the addition of a validation stage to common
REFERENCES
Albrecht, J. E., & O’Brien, E. J. (1991). Effects of centrality on retrieval of text-based concepts.
Journal of Experimental Psychology: Learning, Memory, and Cognition, 17, 932e939.
Albrecht, J. E., & O’Brien, E. J. (1993). Updating a mental model. Journal of Experimental Psy-
chology: Learning, Memory, & Cognition, 19, 1061e1070.
Barton, S. B., & Sanford, A. J. (1993). A case study of anomaly detection: shallow semantic
processing and cohesion establishment. Memory & Cognition, 21(4), 477e487.
Bohan, J., & Sanford, A. (2008). Semantic anomalies at the borderline of consciousness: an eye-
tracking investigation. The Quarterly Journal of Experimental Psychology, 61(2), 232e239.
van den Broek, P., Bohn-Gettler, C. M., Kendeou, P., Carlson, S., & White, M. J. (2011).
When a reader meets a text: the role of standards of coherence in reading comprehen-
sion. In M. T. McCrudden, J. Magliano, & G. Schraw (Eds.), Text relevance and learning
from text (pp. 123e139). Greenwich, CT: Information Age Publishing.
van den Broek, P., Risden, K., Fletcher, C. R., & Thurlow, R. (1996). A “landscape” view
of reading: fluctuating patterns of activation and the construction of a stable memory rep-
resentation. In B. K. Britton, & A. C. Graesser (Eds.), Models of understanding text (pp.
165e187). Mahwah, NJ: Erlbaum.
van den Broek, P., Risden, K., & Husebye-Hartmann, E. (1995). The role of readers’ stan-
dards for coherence in the generation of inferences during reading. In R. F. Lorch, &
E. J. O’Brien (Eds.), Sources of coherence in reading (pp. 353e373). Mahwah, NJ: Lawrence
Erlbaum Associates.
van den Broek, P., Virtue, S., Everson, M. G., Tzeng, Y., & Sung, Y. C. (2002). Compre-
hension and memory of science texts: inferential processes and the construction of a
mental representation. In J. Otero, J. Leon, & A. C. Graesser (Eds.), The psychology of sci-
ence text comprehension (pp. 131e154). Mahwah, NJ: Erlbaum.
Colbert-Getz, J., & Cook, A. E. (2013). Revisiting effects of contextual strength on the sub-
ordinate bias effect: evidence from eye movements. Memory & Cognition, 41, 1172e1184.
Cook, A. E. (2014). Processing anomalous anaphors. Memory & Cognition, 42, 1171e1185.
Cook, A. E., Colbert-Getz, J., & Kircher, J. C. (2013). Number-of-Feature effects during
reading: evidence from eye movements. Discourse Processes, 50(3), 210e225.
Cook, A. E., & Guéraud, S. (2005). What have we been missing? The role of general world
knowledge in discourse processing. Discourse Processes, 39, 365e378.
Cook, A. E., Halleran, J. G., & O’Brien, E. J. (1998). What is readily available during reading?
A memory-based text processing view. Discourse Processes, 26, 109e129.
Cook, A. E., Lassonde, K. A., Splinter, A., Guéraud, S., Steigler, J., & O’Brien, E. J. (2014).
The role of relevance in the activation and instantiation of predictive inferences. Language
and Cognitive Processes, 29, 244e257.
Cook, A. E., & Myers, J. L. (2004). Processing discourse roles in scripted narratives: the in-
fluences of context and world knowledge. Journal of Memory and Language, 50, 268e288.
Cook, A. E., & O’Brien, E. J. (2014). Knowledge activation, integration, and validation dur-
ing narrative text comprehension. Discourse Processes, 51, 26e49.
Daneman, M., Hannon, B., & Burton, C. (2006). Are there age-related differences in shallow
semantic processing of text? Evidence from eye movements. Discourse Processes, 42(2),
177e203.
Erickson, T. D., & Mattson, M. E. (1981). From words to meaning: a semantic illusion. Jour-
nal of Verbal Learning and Verbal Behavior, 20, 540e551.
Ferguson, H. J. (2012). Eye movements reveal rapid concurrent access to factual and coun-
terfactual interpretations of the world. The Quarterly Journal of Experimental Psychology, 65,
939e961.
Ferguson, H. J., & Sanford, A. J. (2008). Anomalies in real and counterfactual worlds: an eye-
movement investigation. Journal of Memory and Language, 58, 609e626.
Ferguson, H. J., Sanford, A. J., & Leuthold, H. (2008). Eye-movements and ERPs reveal the
time course of processing negation and remitting counterfactual worlds. Brain Research,
1236, 113e125.
Ferreira, F., Ferraro, V., & Bailey, K. G. D. (2002). Good-enough representations in language
comprehension. Current Directions in Psychological Science, 11, 11e15.
Ferreira, F., & Patson, N. D. (2007). The ‘good enough’approach to language comprehension.
Language and Linguistics Compass, 1(1e2), 71e83.
Ferretti, T. R., Singer, M., & Harwood, J. (2013). Processes of discourse integration: evi-
dence from event-related brain potentials. Discourse Processes, 50, 165e186.
Garrod, S., & Terras, M. (2000). The contribution of lexical and situational knowledge to
resolving discourse roles: bonding and resolution. Journal of Memory and Language, 42(4),
526e544.
Graesser, A. C., Singer, M., & Trabasso, T. (1994). Constructing inferences during narrative
text comprehension. Psychological Review, 101, 371e395.
Guéraud, S., Harmon, M. E., & Peracchi, K. A. (2005). Updating situation models: the
memory-based contribution. Discourse Processes, 39, 243e263.
Hannon, B., & Daneman, M. (2004). Shallow semantic processing of text: an individual-
differences account. Discourse Processes, 37(3), 187e204.
Hess, D. J., Foss, D. J., & Carroll, P. (1995). Effects of global and local context on lexical pro-
cessing during language comprehension. Journal of Experimental Psychology: General,
124(1), 62e82.
Hinze, S. R., Slaten, D. G., Horton, W. S., Jenkins, R., & Rapp, D. N. (2014). Pilgrims sail-
ing the Titanic: plausibility effects on memory for misinformation. Memory & Cognition,
42, 305e324.
Isberner, M. B., & Richter, T. (2014). Does validation during language comprehension
depend on an evaluative mindset? Discourse Processes, 51, 7e25.
Jacovina, M. E., Hinze, S. R., & Rapp, D. N. (2014). Fool me twice: the consequences of
reading (and rereading) inaccurate information. Applied Cognitive Psychology, 28, 558e568.
Kamas, E. N., & Reder, L. M. (1995). The role of familiarity in cognitive processing. In
R. F. Lorch, & E. J. O’Brien (Eds.), Sources of coherence in reading (pp. 177e202). Hillsdale,
NJ: Erlbaum.
Kamas, E. N., Reder, I. M., & Ayers, M. S. (1996). Partial matching in the Moses illusion:
response bias not sensitivity. Memory & Cognition, 24, 687e699.
Kendeou, P., & van den Broek, P. (2007). The effects of prior knowledge and text structure
on comprehension processes during reading of scientific texts. Memory & Cognition, 35,
1567e1577.
Kendeou, P., & O’Brien, E. J. (2014). 16 the knowledge revision components (KReC)
framework: processes and mechanisms. In D. N. Rapp, & J. L. G. Braasch (Eds.), Process-
ing inaccurate information: Theoretical and applied perspectives from cognitive science and the educa-
tional sciences (pp. 353e377). Cambridge, MA: MIT Press.
Kendeou, P., Smith, E. R., & O’Brien, E. J. (2013). Updating during reading comprehen-
sion: why causality matters. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 39, 854e865.
Kendeou, P., Walsh, E. K., Smith, E. R., & O’Brien, E. J. (2014). Knowledge revision pro-
cesses in refutation texts. Discourse Processes, 51, 374e397.
Rapp, D. N., Hinze, S. R., Slaten, D. G., & Horton, W. S. (2014). Amazing stories: acquiring
and avoiding inaccurate information from fiction. Discourse Processes, 51, 50e74.
Reder, L. M., & Cleeremans, A. (1990). The role of partial matches in comprehension: the
Moses illusion revisited. In A. C. Graesser, & G. H. Bower (Eds.), The psychology of
learning and motivation (Vol. 25, pp. 233e258). San Diego, CA: Academic Press, Inc.
Reder, L. M., & Kusbit, G. W. (1991). Locus of the Moses illusion: imperfect encoding,
retrieval, or match? Journal of Memory and Language, 30, 385e406.
Reichle, E. D., Rayner, K., & Pollatsek, A. (2003). The EZ Reader model of eye-movement
control in reading: comparisons to other models. Behavioral and Brain Sciences, 26(04),
445e476.
Richter, T. (2015). Validation and comprehension of text information: two sides of the same
coin. Discourse Processes, 52, 337e354.
Rizzella, M. L., & O’Brien, E. J. (1996). Accessing global causes during reading. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 22, 1208e1218.
Rizzella, M. L., & O’Brien, E. J. (2002). Retrieval of concepts in script-based texts and nar-
ratives: the influence of general world knowledge. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 28, 780e790.
Rumelhart, D. E., & McClelland, J. L. (1986). Parallel distributed processing. Explorations in the
microstructure of cognition. Vol. 1: Foundations. Cambridge: MIT Press.
Sanford, A. J. (1990). On the nature of text-driven inference. In D. A. Balota, G. B. Flores
d’Arcais, & K. Rayner (Eds.), Comprehension processes in reading (pp. 515e535). Hillsdale,
NJ: Lawrence Erlbaum Associates, Inc.
Sanford, A. J. (2002). Context, attention and depth of processing during interpretation. Mind
& Language, 17(1e2), 188e206.
Sanford, A. J., & Emmott, C. (2012). Mind, brain and narrative. Cambridge University
Press.
Sanford, A. J., & Garrod, S. C. (1989). What, when, and how?: questions of immediacy in
anaphoric reference resolution. Language and Cognitive Processes, 4, 235e262.
Sanford, A. J., & Garrod, S. C. (1998). The role of scenario mapping in text comprehension.
Discourse Processes, 26, 159e190.
Sanford, A. J., & Garrod, S. C. (2005). Memory-based approaches and beyond. Discourse Pro-
cesses, 39(2e3), 205e224.
Sanford, A. J., & Graesser, A. C. (2006). Shallow processing and underspecification. Discourse
Processes, 42(2), 99e108.
Sanford, A. J., Leuthold, H., Bohan, J., & Sanford, A. J. (2011). Anomalies at the borderline
of awareness: an ERP study. Journal of Cognitive Neuroscience, 23(3), 514e523.
Sanford, A. J., Sanford, A. J., Filik, R., & Molle, J. (2005). Depth of lexical-semantic process-
ing and sentential load. Journal of Memory and Language, 53(3), 378e396.
Sanford, A. J., Sanford, A. J., Molle, J., & Emmott, C. (2006). Shallow processing and atten-
tion capture in written and spoken discourse. Discourse Processes, 42(2), 109e130.
Sanford, A. J., & Sturt, P. (2002). Depth of processing in language comprehension: not
noticing the evidence. Trends in Cognitive Sciences, 6(9), 382e386.
Singer, M. (2006). Verification of text ideas during reading. Journal of Memory and Language,
54, 574e591.
Singer, M. (2013). Validation in reading comprehension. Current Directions in Psychological Sci-
ence, 22, 361e366.
Singer, M., & Doering, J. C. (2014). Exploring individual differences in language validation.
Discourse Processes, 51, 167e188.
Singer, M., Graesser, A. C., & Trabasso, T. (1994). Minimal or global inference during
reading. Journal of Memory and Language, 33(4), 421e441.
Singer, M., Halldorson, M., Lear, J. C., & Andrusiak, P. (1992). Validation of causal bridging
inferences. Journal of Memory and Language, 31, 507e524.
Singer, M., Harkness, D., & Stewart, S. T. (1997). Constructing inferences in expository text
comprehension. Discourse Processes, 24, 199e228.
Singer, M., & Ritchot, K. (1996). Individual differences in inference validation. Memory &
Cognition, 24, 733e743.
Smith, E.R., & O’Brien, E.J. (2016). Enhancing memory access for less-skilled readers.
Manuscript under review.
Sturt, P., Sanford, A. J., Stewart, A., & Dawydiak, E. (2004). Linguistic focus and good-
enough representations: an application of the change-detection paradigm. Psychonomic
Bulletin & Review, 11(5), 882e888.
Van Berkum, J. J. A. (2008). Understanding sentences in context: what brain waves can tell
us. Current Directions in Psychological Science, 17, 376e380.
Van Dijk, T. A., & Kintsch, W. (1983). Strategies of discourse comprehension. New York:
Academic Press.
Ward, P., & Sturt, P. (2007). Linguistic focus and memory: an eye movement study. Memory
& Cognition, 35(1), 73e86.
Wei, W., & Cook, A. E. (2016). Semantic size and contextual congruency effects during
reading: Evidence from eye movements. Discourse Processes (in press).
Williams, C., Cook, A. E., & O’Brien, E. J. (2016). Examining the Moses illusion in reading
comprehension. Manuscript under preparation.
Zwaan, R. A., & Radvansky, G. A. (1998). Situation models in language comprehension and
memory. Psychological Bulletin, 123, 162e185.