Activation of Background Knowledge For

This article was downloaded by: [Istanbul Universitesi Kutuphane ve Dok]
On: 22 April 2014, At: 08:39

Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Scientific Studies of Reading

Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/hssr20
Activation of Background Knowledge for

Inference Making: Effects on Reading
Comprehension
a b
Carsten Elbro & Ida Buch-Iversen
a
University of Copenhagen
b
University of Stavanger
Published online: 09 May 2013.
To cite this article: Carsten Elbro & Ida Buch-Iversen (2013) Activation of Background Knowledge for
Inference Making: Effects on Reading Comprehension, Scientific Studies of Reading, 17:6, 435-452,
DOI: 10.1080/10888438.2013.774005
To link to this article: http://dx.doi.org/10.1080/10888438.2013.774005
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Scientific Studies of Reading, 17:435–452, 2013
Copyright © 2013 Society for the Scientific Study of Reading
ISSN: 1088-8438 print/1532-799X online
DOI: 10.1080/10888438.2013.774005
Activation of Background Knowledge for Inference

Making: Effects on Reading Comprehension
Carsten Elbro
Downloaded by [Istanbul Universitesi Kutuphane ve Dok] at 08:39 22 April 2014
University of Copenhagen
Ida Buch-Iversen
University of Stavanger
Failure to activate relevant, existing background knowledge may be a cause of poor reading com-
prehension. This failure may cause particular problems with inferences that depend heavily on prior
knowledge. Conversely, teaching how to use background knowledge in the context of gap-filling
inferences could improve reading comprehension in general. This idea was supported in an exper-
imental study comprising 16 sixth-grade classes (N = 236) randomly assigned to experimental or
control conditions. In the experimental condition, students’ contribution to “gap-filling” inferences
with expository texts were made explicit by means of graphic models and inference-demanding ques-
tions. After eight 30-min sessions, a large training effect was found on students’ inference making
skills with a substantial and sustained transfer effect to a standard measure of reading comprehension.
The effects were not mediated by students’ motivation, decoding ability, vocabulary, or nonverbal IQ.
Texts cannot be understood without contributions from readers. Texts provide instructions to
readers about how to use relevant knowledge and experience to build an understanding of the
text. This building of understanding depends heavily on the reader’s ability to draw inferences
(e.g., Bransford & McCarrell, 1974; Graesser, Singer, & Trabasso, 1994; Kintsch, 1998). Most
of these inferences during reading are completed automatically, for example, “Moses received
the tablets. He needed a glass of water to swallow them.” The reader automatically infers that
he must be Moses and them the tablets. On closer inspection, it is evident that the reader must
also assume that Moses will use the glass of water to wash the tablets down. The text does not
specify the precise use of the glass of water, so the reader has to infer this use to make a coherent
interpretation. If the reader does a double take with the second sentence, it may be because the
mention of Moses in the first sentence has made the reader automatically infer that the relevant
meaning of tablets is “large, inedible slates.”
In some cases, the reader does not have the knowledge necessary to form the inferences
needed to comprehend a text (Hirsch, 2003). For instance, the reader may not know enough about
Moses or how to drink water with tablets to draw the aforementioned inferences. The reader’s
Correspondence should be sent to Ida Buch-Iversen, National Centre for Reading Education and Research, University
of Stavanger, Stavanger, Norway. E-mail: ida.buch.iversen@gmail.com
436 ELBRO AND BUCH-IVERSEN
background knowledge may also be wrong and lead to erroneous inferences and poor memory of
the text (Kendeou & van den Broek, 2007). However, a third reason for comprehension problems
may be that the reader has the relevant knowledge but does not use it (Cain, Oakhill, Barnes, &
Bryant, 2001; Oakhill & Cain, 2007). The present study explored this latter possibility. It exam-
ined the effects of teaching inference making to sixth graders during reading of expository texts.
The point of the teaching was to elucidate the precise contribution from the reader’s knowledge
to inferences that are necessary in order to understand texts.
Longitudinal studies have provided evidence for the importance of inference making for the
development of reading comprehension (Kendeou, Bohn-Gettler, White, & van den Broek, 2008;
Kendeou, van den Broek, White, & Lynch, 2007; Oakhill & Cain, 2007). Findings from such
studies have indicated that the ability to make inferences contributes to reading comprehension
even when other abilities and knowledge are controlled, such as decoding, vocabulary, awareness
of text structure, comprehension monitoring, and verbal IQ (Oakhill & Cain, 2012). Hence, there
may be a specific link between inference-making ability and reading comprehension. However,
longitudinal studies do not allow for strong conclusions about causality.
Possible causal links between inference-making ability and reading comprehension have been
investigated in several experimental studies, many of which are from the 1980s. However, most
experimental studies have aimed broadly to encourage students to reflect about texts. Inference
making has not been taught exclusively, but as part of a broader training program (e.g., Carr,
Dewitz, & Patberg, 1983; Dewitz, Carr, & Patberg, 1987; Reutzel & Hollingsworth, 1988;
Sundbye, 1987; Winne, Graham, & Prock, 1993). For example, inference making has been trained
in combination with imagery (Moore & Kirby 1988), with comprehension monitoring (Carr et al.,
1983; Yuill & Oakhill, 1988), and with question generation and prediction (McGee & Johnson,
2003). More recently, McNamara and colleagues have found support for the effectiveness of
a reading strategy intervention, Self-Explanation Reading Training (SERT and iSTART), with
high school students and college students reading science texts (McNamara, 2004; McNamara,
Levinstein, & Boonthum, 2004). With this intervention, students are taught various reading strate-
gies to support their reading comprehension, including linking ideas in the text by means of
bridging inferences, making predictions, and using prior knowledge to elaborate upon text con-
tent. Although these strategies involve inferences, they clearly also involve other components of
reading comprehension such as elaborating and predicting.
Such combined strategy approaches to inference training are in the majority and perhaps a
reason why the report from the National Reading Panel (NRP; 2000) did not include inference
making as a separate category. Although studies of combined approaches may be of great impor-
tance for both theoretical and practical reasons, they do not allow for assessments of specific
effects of inference training because one or several other components of the training could be
responsible for a part of or even all of the positive effects. The present study focused exclusively
on inferences during reading that require integration of background knowledge with information
from the text to form a coherent representation of the meaning. It did not include “elabora-
tive” (e.g., predictive) inferences that lead to conclusions beyond what is strictly necessary to
understand the text. Neither did it include “bridging” inferences, which are necessary to maintain
local coherence, for example, between pronouns and antecedents, or between co-referential but
different lexical items.
The current study focused specifically on such inferences, which are both dependent on prior
knowledge and crucial for maintaining global coherence in text comprehension. Such inferences
ACTIVATION OF BACKGROUND KNOWLEDGE 437
have been termed “gap-filling” inferences (Baker & Stein, 1981). Here is an example from the
present study:
During the 20th century, fishing boats became hugely more efficient so that it was possible to catch
large quantities of fish in a short time. Towards the end of the century it became necessary to regulate
fishing, for example by setting limits (quotas) on the catches of each fisherman or boat.
To make the text cohere, the reader must supply knowledge. In this example, there is a “gap” in
the argument, a missing premise, to be supplied by the reader: the amount of fish to be caught is
limited. Without any limitation, there would be no need for regulation. The reader’s knowledge
fills the gap in the text between the one explicit premise and the conclusion. One purpose of
the present study was to extend previous work on inference making by focusing exclusively on
knowledge-demanding “gap-filling” inferences. The hypothesis was that training examples of
such inferences would generalize across texts to similar inferences in untrained texts.
To elucidate the contributions from the reader’s knowledge, graphic models were employed
in a questioning format. For the previous example, the question would be concerned with the
missing premise (e.g., Why did it become necessary to regulate fishing?). Students would work
with graphic models, such as the one in Figure 1, and fill in the box with information from the
students’ background knowledge. To the best of our knowledge, the effects of this kind of explicit
work with knowledge-demanding inferences have not been reported before.
Expository texts were chosen for the present study because background knowledge may
be more important for comprehension of expository text than for narratives (Best, Floyd, &
McNamara, 2008). As pointed out by Best and colleagues (Best, Rowe, Ozuru, & McNamara,
2005), new topics and concepts are often introduced to students through expository reading.
Such new concepts have to be linked with relevant background knowledge to be understood.
In support of this idea, Kendeou and van den Broek (2007) found that students who have poor
background knowledge in science (e.g., incoherent ideas and misconceptions related to the topic
of a text) made significantly fewer correct inferences during reading science texts as compared
with students with better background knowledge. The decision to use expository texts was made
From the text: From the reader:
"Fishing boats
became hugely more
efficient"
"It became necessary

to regulate fishing"
FIGURE 1 The teaching of gap-filling inferences used graphic mod-

els to elucidate the contributions from both the text—and the reader, for
example, “There is a limited amount of fish to catch.”
even though young students may find expository texts more difficult to understand than narra-
tive texts (Best et al., 2008; O’Reilly & McNamara, 2002). Another potential difficulty was that
students normally make fewer inferences when reading expository texts than when reading nar-
rative texts (Kucan & Beck, 1996). Even so, for the reasons just given, it is likely that making the
right knowledge-based inferences is more fundamental for making sense of expository than for
narrative texts—at least for children at the end of primary school.
The present study comprised Grade 6 students (11-year-olds) because they face an increasing
amount of expository texts from which they are expected to acquire new knowledge. Hence, they
are aware that expository texts can be challenging and that their success at school can depend on
how well they understand such texts.

Reading comprehension involves several component processes, and the reader’s prior knowl-
edge may be important for many of these processes. This means that it may be difficult to detect
general effects on comprehension of training that focus on just one component process (cf. NRP,
2000, pp. 4–47). The training effects may be diluted and too small to reach statistical significance
and even further from practical interest. In addition, reviews of reading comprehension instruc-
tion (Rosenshine & Meister, 1994; Rosenshine, Meister, & Chapman, 1996) have detected much
stronger effects measured with experimenter-designed tests than with standard comprehension
tests that span a range of text types and reading requirements. However, we hypothesized that
inference training could be done in a manner—using expository texts with inference-demanding
questions and graphic models—that would be a powerful way to make students activate back-
ground knowledge in text comprehension. Such a powerful demonstration might lead to better
use of background knowledge in general, that is, also for other comprehension processes than the
particular inferences that were trained. For example, activation of prior knowledge is important
for disambiguation of lexical items with multiple meanings, and for comprehension monitoring
to detect possible conflicts between the text and prior knowledge. Hence, a second purpose of
the study was to contribute insights into the generalized effects of training of knowledge-based
inferences—both at the end of the program and at a later point. To serve this purpose, a standard
measure of comprehension was employed.
A few studies have found that specific inference training with expository texts is associ-
ated with generalized effects on broad, standard tests of reading comprehension (Hansen, 1981;
Hansen & Pearson, 1983; Holmes, 1985). Hansen (1981) exposed two groups of second-grade
students to “strategy” instruction on integrating text information and prior knowledge before
reading. During subsequent question-guided reading, one group was asked primarily literal ques-
tions, and the other group was asked exclusively inference-demanding questions. Posttesting with
a standard test of reading comprehension (Stanford Achievement Test) revealed significant advan-
tages for the strategy+practice group over both the strategy-only group and an untrained control
group. Similar results were reported in groups of poor readers in Grades 4 and 5 by Holmes
(1985), suggesting that practice with inference-demanding questions is a necessary component.
Hansen and Pearson (1983) compared effects of strategy+practice intervention in groups of good
and poor readers in Grade 4. Positive effects were found for the poor readers with reading-level
appropriate materials, but only limited gains were present in the good readers. However, the
experimental groups were very small in these studies, and uncontrolled differences may have
contributed to the differences reported. Hence, a third purpose of the present study was to assess
the robustness of possible training effects—across different student abilities, and across variation
in the implementation of the program.
In sum, the present study had the following rationale and aims. A cause of poor reading com-
prehension may be a failure to activate relevant, existing background knowledge (Cain et al.,
2001). This failure to activate background knowledge may be particularly problematic and appar-
ent during reading of inference-demanding expository texts. If this is the case, we hypothesized
that middle-school students would benefit from an intervention that highlights and practices their
active use of background knowledge. The study focused exclusively on knowledge-demanding
gap-filling inferences, because they are necessary for building a coherent representation of texts
and because they may work as an ideal showcase for the use of background knowledge.
The first purpose of the study was to extend previous work on inference making by assessing
the possible effects of training knowledge-demanding gap-filling inferences. The first research
question was whether such focused training would generalize across texts to similar inferences in
untrained texts with no prompts to activate background knowledge. A generalization effect would
suggest that students can be taught to activate their background knowledge spontaneously when
it is needed for knowledge-demanding inferences.
The second purpose was to study whether primary training effects would generalize to reading
comprehension more broadly as measured with a standard test of reading comprehension—and
whether the effect would be sustained. Generalization would support the hypothesis that activat-
ing readers’ knowledge for inference making can have a beneficial effect on text comprehension
in general.
The third purpose was to assess the robustness of possible effects. The third research question
was whether possible effects would be mediated by student’s abilities prior to training, or perhaps
by variation in the completion of the program or in class management.
METHOD
Design
The study employed two conditions (experimental and control) and measures at pretest, posttest,
and delayed posttest (follow-up). Sixteen sixth-grade classes, the teachers of which volunteered
to participate in the study, were randomly assigned to either the experimental or the control con-
dition. The assignment was done by school to minimize interaction between teachers in the two
conditions. The schools in the experimental condition had 10 classes, and the two control schools
had six classes. In the experimental condition, eight lessons in knowledge-based inference mak-
ing replaced a similar amount of teaching of mother tongue (Norwegian) language and literature,
whereas the students in the control condition received ordinary teaching. The design was quasi-
experimental (Shadish, Cook, & Campbell, 2002) in the sense that it was not double-blind: The
participating teachers were obviously aware of their assignment to either experimental or control
conditions. In an attempt to control for general expectancy (or Hawthorne) effects, measures of
motivation and maths abilities were taken at both pre- and posttest.
To be able to detect causal links between inference making and reading comprehension, effects
of the experimental teaching were measured with respect to both inference making and read-
ing comprehension. The latter was measured with a standard test that has high reliability and
validity. Smaller subgroups of participants were followed up after the study to see whether any
effects persisted. Control measures at pretest included measures of vocabulary, word decoding,
and nonverbal IQ. These measures were taken to ensure that the two groups had equal basic
abilities and to allow for a search for abilities that might determine individual outcomes of the
training—and challenge the robustness of the training effects.
Participants
The study comprised a total of 236 sixth graders from six schools located in the southwest of
Norway. The area is characterized by a mix of towns and rural districts. The socioeconomic lev-
els are homogenous at or slightly above the national average. No specific inclusion or exclusion
criteria were used. The age range of the students was 10;5 to 11;10 years (average = 11;2 years).
There was an equal number of girls and boys; 9.7% of the students were bilingual. The experi-
mental group consisted of 151 students in 10 classes, and the control group had 85 students in
six classes. The groups were equal in terms of participants’ age, gender, and teachers’ experience
(average = 9 years). A few students were absent either at pretest or posttest, but pre–posttest
analyses were conducted with data from at least 214 of the students.
The Inference Training Program
The materials for the eight training sessions consisted of short expository texts and questions that
required the students to make gap-filling inferences. In the first session, students were introduced
to the program by a mix of short narrative and expository passages (typically two to four sen-
tences). A comprehension question was asked after each passage that required that the students
supply background knowledge to draw simple inferences, for example, The bank did not give Ole
a loan for a new boat. He began to look for a spare-time job. Why did Ole want a spare-time job?
(Expected: He had to earn the money for the boat.)
In each of the subsequent seven sessions, students read two or three short texts of 100 to
200 words and answered inference-demanding questions about them. The training program com-
prised 15 such texts. They were expository prose written for the program. The topics of the
texts were selected from biology, geography, technology, sociology, and history. Topics were
selected so that the necessary background knowledge would be well within the students’ knowl-
edge. In Sessions 2 to 6, all texts were supported by graphic organizers. The organizers had empty
boxes into which the students were asked to put the relevant pieces of information stated in the
texts or supplied from their background knowledge. The purpose of the graphic organizers was
to make the “gaps” in the texts visible and thus make the students aware of the fact that they had
to contribute information and precisely what they needed to contribute to answer the questions.
The information stated in the texts was not sufficient. A brief example is displayed in Figure 1.
An example from the materials is shown in the appendix (Figure 5).
During the first sessions, some of the boxes of the organizers were prefilled and the students
only had to fill in the remaining empty boxes. As students grew familiar with knowledge-
demanding inferences and the graphic organizers, they filled in all the boxes. In the last two
sessions, students read the same kind of texts and answered knowledge-demanding inference
questions about them but without the support from graphic organizers.
All teaching in the experimental program was provided by the classes’ ordinary teachers dur-
ing eight sessions of approximately 30 min each. At the beginning of each session, the teacher
would present the students with an overview of the session and explain any difficult words in the
texts. During the first three sessions, the teacher provided thorough explanations and modeling
for the whole class. During later sessions, explanations and guidance were given only when nec-
essary. All sessions ended with a review of the correct answers led by the teacher. The students
worked in pairs or small groups in most sessions. During the last two sessions, the students were
encouraged to answer the inference questions individually.
The Control Condition
The experimental condition replaced ordinary mother tongue (Norwegian) lessons in the control
classes. The control group did not see any of the experimental materials. Mother tongue teach-
ing for fifth to seventh grade follows national guidelines. According to these guidelines, students
should acquire a broad range of reading and writing abilities such as reading and expressing com-
prehension of texts of different genres and making use of reading strategies. Hence, the control
group engaged in a broad spectrum of reading and writing activities between pre and posttesting,
including comprehension-focused activities. However, the national guidelines do not mention
systematic or explicit teaching of the use of background knowledge for inference making, and
there is no such tradition in practical teaching.
Implementation and Fidelity of the Teaching
All participating teachers were given a brief introduction to the aims and plan of the study. After
teachers and classes had been randomly assigned to the experimental or the control condition, the
teachers in the experimental condition received a 1-hr introduction to the experimental program.
During this introduction, teachers were provided with all the teaching materials needed, including
all students’ worksheets, and with a written step-by-step teaching manual that explained the goals
and procedures of each training session. In this way the experimental teaching was relatively
tightly controlled—to ensure fidelity of the teaching.
All experimental classes were observed during one inference training session, and control
classes were observed during one language and literature lesson. The observations were made
mainly to ensure that the classes in the experimental condition were following the program and
that students’ working conditions were comparable across conditions.
Additional information about the implementation of the experimental teaching was collected
by means of teachers’ written reports. At the end of each training session, teachers recorded the
duration of the session, the number of students present, the number of students (all, more than
half, less than half or no one) who completed all exercises, and any other comments about the
implementation.
Measures
Inference making. To assess the development of the directly trained abilities, a measure of
the students’ abilities to make inferences during reading was developed. The measure comprised
texts and questions similar to those in the inference training. There were five short, expository
texts and 16 questions that required the students to make gap-filling inferences. Following each
question, participants were asked to underline the words in the text that helped them answer the
question. Correct answers were awarded 2 points; answers that referred only to information stated
in the text were given 1 point. An additional point was given when the participant had underlined
the relevant words in the text, so that the maximum score per question was 3 and the maximum
score on this measure was 48. Given the open-ended nature of the task, the scoring followed
detailed guidelines. In addition, 20% of all responses were scored twice and the agreement was
above 90%. Reliability was .82 (Cronbach’s alpha) in the present sample at both pre- and posttest.
Reading comprehension. In the absence of a Norwegian, standardized measure of read-
ing comprehension, the Diagnostic Reading Analysis (DRA; Crumpler & McCarty, 2004) was
included to assess transfer effects to reading comprehension. Although this test was not standard-
ized in Norwegian, it was regarded as a “standard” test in the sense that it is not experimenter
devised and that it covers a broad range of text types and reading requirements. First, the test con-
sists of short texts and poses relatively small demands on students’ decoding skills. Second, half
of the texts in this test are classed as fiction and half nonfiction, which made it possible to inves-
tigate whether training effects would transfer across genres. Third, half the questions in the DRA
are literal, whereas the other half are inferential/summative, predictive, or vocabulary questions,
which all require some kind of interpretation. This diversity made it possible to assess whether
any effects would be limited to interpretation questions or whether training would also transfer to
literal questions. The test has two parallel forms. Form A was used at pretest and delayed posttest,
whereas Form B was administered at posttest. When the test was translated into Norwegian, it
was made suitable for group administration (with written rather than oral responses) because of
limited resources. Even with the group format, reliability was acceptable in the present sample,
Cronbach’s alpha for Form A was .84 (pretest) and .76 for Form B (posttest).
Receptive vocabulary. A measure of receptive vocabulary was developed for group testing.
With each item, a spoken word or phrase was read to the participants who were asked to select
and circle a synonym from three written alternatives, for example, “Is ‘a collection of stars’ the
same as an asteroid, a meteor, or a galaxy?” (Expected answer: galaxy). To reduce the influence of
decoding skills, all words were read aloud by the experimenter. The score was number of correct
of the 20 items. Cronbach’s alpha was .65.
Word decoding. A word-chain test was used as a measure of word decoding (Høien &
Tønnesen, 1997). Each item of this test has four words with no spaces (e.g., redbusearhat).
Participants are asked to divide up the chains into words by inserting slashes (red/bus/ear/hat).
The test has 90 chains, and participants are allowed 4 min to divide up as many chains as possible.
The score is the number of correctly divided chains within the time limit. Previous studies have
indicated that the word-chain test is a reliable and valid test of word decoding ability (Jacobson,
1999; Miller-Guron, 1999).
Nonverbal IQ. Raven’s Standard Progressive Matrices (Raven, 1958) were used as a mea-
sure of nonverbal IQ. The participant is presented with five series of 12 matrices. Each matrix
consists of a figure with a missing piece and six pieces to choose between. The participants are
asked to choose the piece that completes the figure. No time limits are imposed. The raw score is
the number of correctly solved matrices. The test has been standardized several times.
Mathematics abilities. A measure of the participants’ maths abilities was included as a con-
trol for general expectancy (Hawthorne) effects. The measure was part one of the M5-test (Tornes,
1996), a standardized test. The test comprises five subcategories of items: understanding of
numbers, arithmetical operations (subtraction, addition, multiplication and division), inequalities

(>, <), equations, and the decimal system. The raw score (number correct) was used.
Motivation. A measure of reading motivation was included as another type of measure
of possible expectancy effects. The two constructs “curiosity” and “involvement,” nine items
altogether, from the Motivation for Reading Questionnaire (Guthrie, Wigfield, & VonSecker,
2000) was translated into Norwegian and used. For the Norwegian translation, Cronbach’s alpha
was .68.
Procedure
Pretest. The measures were taken with intact classes in three separate sessions of approx-
imately 45 min, each in the following fixed order: nonverbal IQ, inference making, reading
comprehension, receptive vocabulary, word decoding, maths abilities, and motivation. The
Reading Motivation Questionnaire was administered by the teachers. Because of illness among
teachers, the questionnaire was administered in only nine experimental classes (N = 128) and
three control classes (N = 40).
Inference training. The eight sessions of the training program were completed in 5 to
8 weeks (including 1 week of school holidays).
Posttest. All posttesting was conducted about a week after the completion of the program
in each class. It took about two lessons of 45 min each. The measures were administered to
whole classes in the following order: inference making, reading comprehension, maths abilities,
and motivation. Because of illness, one experimental class did not receive the measure of maths
abilities and motivation, reducing the total sample size to 113 from 135.
Delayed posttest. Measures of inference making and reading comprehension were admin-
istered in two sessions of approximately 45 min each in experimental and control classes 5 weeks
after the completion of the inference training. Because of limited resources, the delayed posttest
comprised only two plus two classes randomly selected from each of the conditions.
Data transformations. It is difficult to assess changes in accuracy scores over time. Students
who start with a low score of, say, 10% correct, have a large 90% scope for improvement; whereas
students who start with a 90% score only have a 10% scope for improvement. To compensate for
the limitations imposed by finite accuracy scales, accuracy scores were log-odds (lod) trans-
formed, that is, to log(score / (max - score)), before further data analyses (see Allerup & Elbro,
1998). Problems with extreme scores were few. Only three students with a zero score with infer-
ence making at pretest were excluded from the analysis of gains in inference making because
progress over time could not be reliably measured.
RESULTS
Descriptive statistics are presented in Table 1. All score distributions were approximately nor-
mal with all absolute values of skewness and kurtosis smaller than 1 (the only exception was
nonverbal IQ which was moderately skewed; g1 = –1.3). The group differences at pretest were
TABLE 1
Descriptive Statistics of Measures by Condition and Time
Group
Experimental Group Control Group Difference
Measure M SD M SD d
Pretest
Inference making (max. 48) 14.13 8.57 14.17 7.07 ns
Reading comprehension (max. 34) 22.13 5.55 21.63 5.55 ns
Receptive vocabulary (max. 20) 12.59 3.08 11.92 2.92 ns

Word decoding (max. 90) 35.25 8.87 31.24 8.08 0.47∗∗
Motivation 25.58 4.08 25.32 3.74 ns
Nonverbal IQ (max. 60) 39.46 8.91 38.94 10.22 ns
Mathematics (max. 74) 51.85 12.12 53.17 11.47 ns
Posttest
Inference making (max. 48) 23.22 8.43 16.86 7.13 0.81∗∗
Reading comprehension (max. 34) 24.47 4.84 21.06 4.25 0.75∗∗
Motivation 25.32 4.13 25.40 3.41 ns
Mathematics (max. 74) 56.35 11.23 56.14 11.19 ns
∗∗ p < .01.
all nonsignificant—with the exception of a significant difference in word decoding at pretest in

favour of the experimental group, t(231) = 3.39, p = .001. The possible effects of this differ-
ence were controlled in separate analyses of the training effects (next). The average nonverbal IQ
score (Raven) was close to 40, which corresponds to the 50th percentile in 11-year-olds in the
U.S. 1978 standardization.
Classroom observations revealed that all teachers in the experimental condition followed the
program (at least while the observer was present). The observations also found variation between
classes. Some classes took 5 to 10 min longer than others to calm down and be ready for teaching
after breaks, and in some classes, students were more likely than in other classes to be late for
teaching, to interrupt teachers during instructions, and to interrupt each other. However working
conditions were fairly similar across classes and conditions. According to the teachers’ reports,
14 of the 15 texts were completed in all experimental classes. Four teachers reported that they did
not have time enough to complete work with one (the same) text in the sixth lesson.
To control for expectancy effects, gains (from pre- to posttest) in math ability and reading
motivation were analysed. No significant effects were found in repeated analyses of variance
(ANOVAs) on either math abilities or motivation with the exception of a main effect of time on
math, F(1, 211) = 90.4, p < .01. The lack of significant interactions between time and group
goes against a general expectancy effect.
Figures 2 and 3 show the development of inference making and reading comprehension from
pre- to posttest. A log odds (lod) score of zero is equal to 50% correct answers on the test.
Possible differences between the gains in inference making in the two groups were first
analysed in an ANOVA with 2 time points × 2 conditions × 16 classes nested in the two condi-
tions with repeated measures on participants. The analysis yielded the following main effects: a
significant main effect of time, F(1, 201) = 184.6, p < .001; a nonsignificant effect of condition,
0.1
–0.1
Lod-score
–0.2
Experiment
Control
–0.3
–0.4
–0.5
–0.6
Pretest Posttest
FIGURE 2 The development of inference making in reading in the

experiment and control groups (color figure available online).
0.6
0.5
0.4
Lod-score
Experiment
0.3
Control
0.2
0.1
0
Pretest Posttest
FIGURE 3 The development of reading comprehension in the experi-

ment and control groups (color figure available online).
F(1, 14) = 3.3, ns; and a nonsignificant effect of class within conditions, F(14, 201) = 1.3, ns.
It is important to note that a highly significant interaction effect was found between time and
condition, F(1, 14) = 22.7, p < .001, the size of this (training) effect was large (d = .92; Cohen,
1988). In addition, a small but significant interaction effect was found of Time × Classes within
conditions, F(14, 201) = 1.9, p < .05.
Transfer to Reading Comprehension
Reading comprehension was measured with different forms of the same test. This may explain
the apparent slight decline in the control group scores from pre- to posttest (Figure 3). The
transfer effect of the program to reading comprehension was investigated in an ANOVA with
2 time points × 2 conditions × 16 classes nested in the two conditions with repeated measures
on participants. The analysis yielded the following main effects: a significant main effect of time,
F(1, 201) = 11.3, p < .01; a significant effect of condition, F(1, 14) = 6.7, p < .05; and a non-
significant effect of class within conditions, F(14, 201) = 1.4, ns. A highly significant interaction
effect was found between time and condition, F(1, 14) = 23.0, p < .001. The interaction effect
of Time × Classes Within Conditions was insignificant, F(14, 201) = 1.2, ns. The size of the
training effect (Time × Condition) was medium to large, d = .69. The effects with the DRA for
fiction and nonfiction texts were both medium, d = .46 and d = .57, respectively. With regard
to question types, the effect on literal questions was medium, d = .45, whereas the effect on
interpretation questions was medium to large, d = .73. Possible differences in effect sizes went
in the expected direction, though no significant three-way interactions were found of Time ×
Condition × Question Type or Time × Condition × Text Genre.
In addition, analyses of the training effects were conducted with controls for word decoding,
vocabulary, and nonverbal IQ. These analyses were one-way, between-groups analysis of covari-
ance. The first analysis was carried out on posttest inference making with teaching condition as
independent variable and control for pretest inference making, decoding, vocabulary, and non-
verbal IQ. Assumptions for analysis of covariance were all met including linearity, normality,
homogeneity of variances, and regression slopes. The analysis showed a significant difference
between the two groups, F(1, 199) = 69.4, p < .001, partial η2 = .26. A parallel analysis on
posttest reading comprehension controlling for pretest reading comprehension, decoding, vocab-
ulary, and nonverbal IQ provided a significant training effect, F(1, 204) = 28.4, p < .001, partial
η2 = .12.
A reduced, random sample, comprising 27 students from the inference training condition and
26 students from the control condition, was used to assess the effects of condition 5 weeks after
completion of the inference training. With respect to gains in inference making, the effect of con-
dition across the three time points was assessed using a mixed between-within subject ANOVA
with repeated measures on participants. The analysis yielded a significant main effect of time,
F(2, 102) = 14.4, p < .001, partial η2 = .22, and condition, F(1, 51) = 6.9, p < .05, partial η2 =
.12, and a significant interaction effect, F(2, 102) = 6.4, p < .01, partial η2 = .11. Post hoc com-
parisons indicated that the pre-to-post test interaction with condition was significant, F(1, 51) =
6.8, p < .05, partial η2 = .12, as in the total sample, as was the pretest-to-follow-up interaction
with condition, F(1, 51) = 9.6, p < .01, partial η2 = .16. However, there was no significant inter-
action involving posttest-to-follow-up differences. These results indicated that the training effect
occurred during the training period and that it was sustained after the completion of the training.
Analogous analyses for reading comprehension provided a similar picture, though with
smaller effects (Figure 4). The analysis yielded nonsignificant main effects of time, F(2, 102) =
2.7, p = .07, and condition, F(1, 51) ≤ 1, ns, and a marginally significant interaction effect,
F(2, 102) = 3.8, p = .06. However, there was a significant interaction effect between time and
condition when only pre and posttest points were included, F(1, 51) = 4.3, p < .05, partial η2 =
.08, and, of importance, a significant interaction between time and condition when only pretest
0.6
0.5
0.4
Lod-score
Experiment
0.3
Control
0.2
0.1
0
Pretest Posttest Follow-up
FIGURE 4 Development of reading comprehension over time in the

smaller student samples (color figure available online).
and follow-up points were considered, F(1, 51) = 4.5, p < .05, partial η2 = .08. Again, the
training effect occurred during the training and was sustained at follow-up.
Robustness of the Effects
To investigate possible participant-related moderators of the training effect, a series of three-way

ANOVAs was conducted on both inference making and reading comprehension. The third factor
in the analyses was, in turn, gender, word decoding, vocabulary, nonverbal IQ, and motivation.
The analyses were all of the same format, for example, for gender the format was 2 conditions
(experiment vs. control) × 2 time points (pre- vs. posttest) × 2 genders (female vs. male). To form
variables suitable for these analysis, median splits were used for the pretest measures of word
decoding, vocabulary, nonverbal IQ, and motivation. None of these analyses (10 in all) showed
any significant three-way interaction (two-way interactions were not studied because they did not
pertain to the research questions). These results suggest that the effects of the training program
were not selectively mediated by participant variation of the types included in the study.
With only 10 classes in the experimental condition, the study had few degrees of freedom
for between-class analyses of effects of differences in the implementation. The following results
should, thus, be interpreted with caution. The effects of two intraclass variables were studied: the
number of students in each class, and the average length of the lessons spent on the experimental
program. The number of students (13 to 23) per class correlated negatively with gains in reading
comprehension (Spearman ρ = –.89, p < .05), indicating that smaller classes were associated with
larger gains. Average time per session (range 20 to 44 min) spent on the experimental program
correlated positively with average class gains in reading comprehension (Spearman ρ = .76,
p < .05). However, these correlations were not backed up by similar, significant correlations with
inference making, so the correlations are not easy to interpret.
DISCUSSION
The results of the study provide some relatively straightforward answers to the research ques-
tions. First, it was indeed possible to help 11-year-olds improve their ability to make gap-filling
inferences in a short, eight-lesson program that focused on the contribution of the reader’s back-
ground knowledge to text comprehension. The effect size was almost a whole standard deviation.
The training effect remained large when initial abilities in word decoding, receptive vocabulary,
and verbal IQ were controlled. There was no associated effect of training on math abilities or
motivation, which suggests that the training effect was specific to reading and not mediated by
motivation or a general expectancy effect. The positive outcome of training supports the idea that
students can be taught to activate their background knowledge spontaneously when it is needed
for knowledge-demanding inferences.
Second, training was also associated with a significant advance in general reading
comprehension—for both fiction and nonfiction, and for both literal and nonliteral questions.
The average effect size (d = .69) was considerable given the limited amount of training. The
effect was sustained 5 weeks after the termination of the experimental teaching. The generalized
effect supports the hypothesis that activating readers’ knowledge for inference making can have
a beneficial effect on text comprehension in general.
Third, the effects of the experimental training were found to be robust. The effects were found
to be independent of students’ characteristics and abilities assessed prior to participation in the
experimental program: gender, vocabulary, decoding fluency, or nonverbal IQ. Neither did they
interact with the observed differences in the administration of the program. The effects were fairly
uniform across classes within each condition, though a small but significant class effect was found
for inference making but not for reading comprehension. The effect differences between classes
were not really surprising because the program was administered by different teachers to different
classes. Possible sources of class-related outcomes might include (a low) number of students and
(more) time spent on the program. Correlations in support of these factors were found at the class
level, though not for progress with the directly trained contents.
The results of the study concur with earlier studies (e.g., Hansen, 1981; Holmes, 1985) in
finding positive effects of informed practice with inference making for reading comprehension.
The results are also in line with those of a recent study by McMaster et al. (2012). These authors
reported positive effects of inference training using a questioning format with poor comprehen-
ders in Grade 4. Of interest, a subgroup of students who made inaccurate or invalid inferences
benefitted particularly from work with causal inferences, whereas other poor comprehenders
(“paraphrasers”) benefitted particularly from training with more general questions. Our study also
used a questioning format, and many of the inferences in the training were causal (as reflected in
the examples). Perhaps causal inferences are particularly powerful exemplars of how to connect
text with background knowledge for students in the middle school years.
The present study does not shed light on the relative merits of “strategy” versus “contents”
approaches (cf. McKeown, Beck, & Blake, 2009). The NRP (2000) report classified previous
studies of inference making as “strategy” approaches. Both previous studies and the present one
employed a strategy approach in the sense that they taught a specific component and procedure
directly, that is, inference making. However, in the present study, the teaching of the use of back-
ground knowledge for inference making always took place in the context of reading for meaning,
as is characteristic of a “contents” approach. Hence, we would object to a classification of the
experimental training of the current study as either a strategy or a contents approach. We suggest
“informed practice” as a more fitting category.
The study had a number of limitations that should be considered when drawing conclusions.
We consider a methodological weakness first. Neither the teachers nor the experimenter were
blind to group membership. Teachers and their students knew whether they were trying out some-
thing new. It is possible that teachers in the experimental condition delivered the training program
with more enthusiasm than the control teachers. However, if such differences had had a general
effect on the students, we would have expected to see increased motivation and abilities across
the board, which we did not. The experimental classes did not become more motivated than the
control classes in the course of the experiment. On the other hand, we cannot exclude the possi-
bility that the teachers in the experimental condition were teaching the experimental contents so
enthusiastically that the results were obtained more quickly than they would have been under less
inspiring conditions. In the present study, a primary training effect of nearly 1 standard deviation
was obtained over a training period of only eight sessions of 30 min. So even if the efficiency
drops under less inspiring conditions, substantial gains would still be likely after a relatively low
number of teaching sessions. A closer monitoring of the teaching in both conditions might have
revealed unexpected differences—and possible threats to the fidelity of the implementation of the
experimental program.
It is an important limitation that the activation of background knowledge was encouraged in
only one way, that is, in the context of gap-filling inferences with expository texts. There is no way
we can separate the target—activation of background knowledge—from the context—gap-filling
inference making. It is impossible to say to what extent the observed effects were caused by one
or the other. Consequently, further studies are needed to assess other specific ways of activating
background knowledge and to assess other ways in which to support gap-filling inference making.
Likewise, the study was conducted with middle school children, and they may be at a point in
their development when better use of background knowledge is particularly beneficial for their
reading development. It should also be noted that follow-up testing was conducted with only a
subsample so that generalisation may be an issue.
The conclusions also rest on the characteristics and quality of the outcome measures. As men-
tioned in the introduction, many similar training studies do not employ standard measures. The
present one did so and observed positive training effects with it. It would be an exaggeration to
claim that the measure, the DRA, or any one measure, reliably captures all aspects of reading
comprehension. For example, reading comprehension tests tap decoding and listening com-
prehension to different degrees (see Keenan, Betjemann, & Olson, 2008). However, the DRA
does comprise a variety of text types and questions—including many that were not part of the
training, and effects were present even when differences in decoding, vocabulary and IQ were
controlled.
Finally, one might have wished for continued, exceptional growth in reading comprehension in
the experimental group after the end of the training. Such growth would have provided a strong
indication of transfer to other reading situations. However, the observed picture is the typical
one: Training had particular effect while it lasted, and the effect was sustained after training.
As mentioned in the introduction, there are several components in reading comprehension, and
it is unlikely that just one is the key to continuous development. We find it much more likely
that the students in the experimental learned to take more of their background knowledge into
consideration during reading but that further progress would require higher quality of background
knowledge, more advanced inference making, and/or development of several other component
processes.
In conclusion, the results of this study indicate that it is possible to teach middle school stu-
dents to make more active use of background knowledge for inference making during reading
of expository texts. Even a short program can have relatively strong effects: The students can
learn to make better inferences on their own—and this ability can generalize to better reading
comprehension across a wide range of texts and tasks. It may be that some poor comprehenders
have particularly severe problems with knowledge-demanding inferences, but they may not be
the only ones to profit from intervention.
ACKNOWLEDGMENTS
The study was supported by a grant to Ida Buch-Iversen from the University of Stavanger.
We thank Anne Charlotte Begnum and Jane Oakhill for valuable comments on both the study
and this article. The engagement and solid work of the participating teachers and their students
deserve recognition.
REFERENCES
Allerup, P., & Elbro, C. (1998). Comparing differences in accuracy across conditions or individuals: An argument for
the use of log odds. Quarterly Journal of Experimental Psychology Section a–Human Experimental Psychology, 51,
409–424.
Baker, L., & Stein, N. (1981). The development of prose comprehension skills. In C. M. Santa & B. L. Hayes (Eds.),
Children’s prose comprehension: Research and practice (pp. 7–43). Newark, DE: International Reading Association.
Best, R. M., Floyd, R. G., & McNamara, D. S. (2008). Differential competencies contributing to children’s comprehension
of narrative and expository texts. Reading Psychology, 29, 137–164.
Best, R. M., Rowe, M., Ozuru, Y., & McNamara, D. S. (2005). Deep-level comprehension of science texts: The role of
the reader and the text. Topics in Language Disorders, 25, 65.
Bransford, J. D., & McCarrell, N. S. (1974). A sketch of a cognitive approach to comprehension. In W. Weimer & D.
Palermo (Eds.), Cognition and the symbolic process (pp. 189–229). Hillsdale, NJ: Erlbaum.
Cain, K., Oakhill, J., Barnes, M. A., & Bryant, P. E. (2001). Comprehension skill, inference-making ability, and their
relation to knowledge. Memory & Cognition, 29, 850–859.
Carr, E. M., Dewitz, P., & Patberg, J. (1983). The effect of inference training on children’s comprehension of expository
text. Journal of Reading Behavior, 15 (3), 1–18.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.
Crumpler, M., & McCarty, C. (2004). Diagnostic reading analysis. London, UK: Hodder Education.
Dewitz, P., Carr, E. M., & Patberg, J. P. (1987). Effects of inference training on comprehension and comprehension
monitoring. Reading Research Quarterly, 22, 99–121.
Graesser, A. C., Singer, M., & Trabasso, T. (1994). Constructing inferences during narrative text comprehension.
Psychological Review, 101, 371–395.
Guthrie, J. T., Wigfield, A., & VonSecker, C. (2000). Effects of integrated instruction on motivation and strategy use in
reading. Journal of Educational Psychology, 92, 331.
Hansen, J. (1981). The effects of inference training and practice on young children’s reading comprehension. Reading
Research Quarterly, 16, 391–417.
Hansen, J., & Pearson, P. D. (1983). An instructional study: Improving the inferential comprehension of good and poor
fourth-grade readers. Journal of Educational Psychology, 75, 821–829.
Hirsch, E. D. (2003). Reading comprehension requires knowledge—of words and the world: Scientific insights into
the fourth-grade slump and the nations stagnant reading comprehension scores. American Educator, 27, 10–22,
28–29, 48.
Holmes, B. C. (1985). The effects of a strategy and sequenced materials on the inferential comprehension of disabled
readers. Journal of Learning Disabilities, 18, 542–546.
Høien, T., & Tønnesen, G. (1997). Ordkjedetesten [The Word Chain Test]. Stavanger, Norway: Stiftelsen
dysleksiforskning.
Jacobson, C. (1999). How persistent is reading disability? Individual growth curves in reading. Dyslexia, 5, 78–93.
Keenan, J. M., Betjemann, R. S., & Olson, R. K. (2008). Reading comprehension tests vary in the skills they assess:
Differential dependence on decoding and oral comprehension. Scientific Studies of Reading, 12, 281–300.
Kendeou, P., Bohn-Gettler, C., White, M. J., & van den Broek, P. (2008). Children’s inference generation across different
media. Journal of Research in Reading, 31, 259–272.
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, 1567–1577.
Kendeou, P., van den Broek, P., White, M. J., & Lynch, J. (2007). Comprehension in preschool and early elementary
children: Skill development and strategy interventions. In D. S. McMamara (Ed.), Reading comprehension strategies:
Theories, interventions, and technologies (pp. 27–45). Mahwah, NJ: Erlbaum.
Kintsch, W. (1998). Comprehension: a paradigm for cognition. Cambridge, UK: Cambridge University Press.
Kucan, L., & Beck, I. L. (1996). Four fourth graders thinking aloud: An investigation of genre effects. Journal of Literacy
Research, 28, 259–287.
McGee, A., & Johnson, H. (2003). The effect of inference training on skilled and less skilled comprehenders. Educational
Psychology, 23, 49–59.
McKeown, M. G., Beck, I. L., & Blake, R. G. K. (2009). Rethinking reading comprehension instruction: A comparison
of instruction for strategies and content approaches. Reading Research Quarterly, 44, 218–253.
McMaster, K. L., den Broek, P. v. A., Espin, C., White, M. J., Rapp, D. N., Kendeou, P., . . . Carlson, S. (2012). Making
the right connections: Differential effects of reading intervention for subgroups of comprehenders. Learning and
Individual Differences, 22, 100–111.
McNamara, D. S. (2004). SERT: Self-Explanation Reading Training. Discourse Processes, 38, 1–30.
McNamara, D. S., Levinstein, I. B., & Boonthum, C. (2004). iSTART: Interactive strategy training for active reading and
thinking. Behavior Research Methods Instruments & Computers, 36, 222–233.
Miller-Guron, L. (1999). Wordchains. Windsor, UK: NFER-Nelson.
Moore, P. J., & Kirby, J. R. (1988). Comprehension training and reading performance. Reading, 22, 126–136.
National Reading Panel. (2000). Teaching children to read: An evidence-based assessment of the scientific research
literature on reading and its implications for reading instruction. Washington, DC: National Institute for Child Health
and Human Development.
Oakhill, J., & Cain, K. (2007). Issues of causality in children’s reading comprehension. In D. McNamara (Ed.), Reading
comprehension strategies. Theories, interventions and technologies (pp. 47–71). London: Erlbaum.
Oakhill, J. V., & Cain, K. (2012). The precursors of reading ability in young readers: Evidence from a four-year
longitudinal study. Scientific Studies of Reading, 16, 91–121.
O’Reilly, T., & McNamara, D. (2002). What’s a science student to do? In W. D. Gray & C. D. Schunn (Eds.), Proceedings
from the 26th annual meeting of the Cognitive Science Society. Mahwah, NJ: Erlbaum.
Raven, J. C. (1958). Standard progressive matrices: Sets A, B, C, D and E. London, UK: H.K. Lewis.
Reutzel, D. R., & Hollingsworth, P. M. (1988). Highlighting key vocabulary—a generative–reciprocal procedure for
teaching selected inference types. Reading Research Quarterly, 23(3), 358–378.
Rosenshine, B., & Meister, C. (1994). Reciprocal teaching—A review of the research. Review of Educational Research,
64, 479–530.
Rosenshine, B., Meister, C., & Chapman, S. (1996). Teaching students to generate questions: A review of the intervention
studies. Review of Educational Research, 66, 181–221.
Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-experimental designs for generalized
causal inference. Boston, MA: Houghton Mifflin.
Sundbye, N. (1987). Text explicitness and inferential questioning—Effects on story understanding and recall. Reading
Research Quarterly, 22, 82–98.
Tornes, J. (1996). Kartleggingsprøve: Matematikk: M4, M5, M6, M7: Lærerveiledning [Screening test: Mathematics
skills: M4, M5, M6, M7: Teacher manual]. Jaren, Norway: PP-tjenestens materiellservice.
Winne, P. H., Graham, L., & Prock, L. (1993). A model of poor readers text-based inferencing—Effects of explanatory
feedback. Reading Research Quarterly, 28, 52–66.
Yuill, N., & Oakhill, J. (1988). Effects of inference awareness training on poor reading-comprehension. Applied Cognitive
Psychology, 2, 33–45.
APPENDIX
Grasshoppers
Most people have once heard grasshoppers sing in the summer. The grasshoppers’ song is very
high pitched. Therefore you need a good sense of hearing in order to hear the song. Actually,
it is not correct to say that grasshoppers sing because they do not use their voice to sing. The
grasshopper makes the song, or the sound, by rubbing its legs against a sharp edge on the wings.
It is not easy to spot a grasshopper as its body is almost invisible when it is surrounded by plants.
It blends in with the surroundings. If you want to find a grasshopper, you ought to follow the
sound. Remember to listen to the grasshopper while you are young, because you might not hear
the grasshopper when you get older.
Why might you not hear the grasshopper when you get older?
Write your answer in the empty box.
Information from the text Information from the reader
“you need a good sense of hearing

in order to hear the song”
“you might not hear the

grasshopper when you get older”
FIGURE 5 An example of the training materials: The students are

asked a knowledge-demanding inference question. A distinction is made
between the information from the text and the information the students
have to contribute. The expected response is something like “Because
when you get older, your sense of hearing gets bad.”

Activation of Background Knowledge For

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Activation of Background Knowledge For

Uploaded by

Copyright:

Available Formats

This article was downloaded by: [Istanbul Universitesi Kutuphane ve Dok]

On: 22 April 2014, At: 08:39

Scientific Studies of Reading

Activation of Background Knowledge for

To link to this article: http://dx.doi.org/10.1080/10888438.2013.774005

PLEASE SCROLL DOWN FOR ARTICLE

Activation of Background Knowledge for Inference

From the text: From the reader:

"It became necessary

FIGURE 1 The teaching of gap-filling inferences used graphic mod-

how well they understand such texts.

The Inference Training Program

The Control Condition

Implementation and Fidelity of the Teaching

numbers, arithmetical operations (subtraction, addition, multiplication and division), inequalities

Receptive vocabulary (max. 20) 12.59 3.08 11.92 2.92 ns

all nonsignificant—with the exception of a significant difference in word decoding at pretest in

FIGURE 2 The development of inference making in reading in the

FIGURE 3 The development of reading comprehension in the experi-

Transfer to Reading Comprehension

FIGURE 4 Development of reading comprehension over time in the

Robustness of the Effects

To investigate possible participant-related moderators of the training effect, a series of three-way

Information from the text Information from the reader

“you need a good sense of hearing

“you might not hear the

FIGURE 5 An example of the training materials: The students are

You might also like