This article was downloaded by: [Tufts University]On: 12 November 2014, At: 13:03Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Educational AssessmentPublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/heda20
Discrepant SAT Critical Reading andWriting Scores: Implications for CollegePerformanceEmily J. Shaw a , Krista D. Mattern a & Brian F. Patterson aa The College BoardPublished online: 08 Sep 2011.
To cite this article: Emily J. Shaw , Krista D. Mattern & Brian F. Patterson (2011) Discrepant SATCritical Reading and Writing Scores: Implications for College Performance, Educational Assessment,16:3, 145-163, DOI: 10.1080/10627197.2011.604241
To link to this article: http://dx.doi.org/10.1080/10627197.2011.604241
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 tothe accuracy, completeness, or suitability for any purpose of the Content. Any opinionsand 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 Contentshould not be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions, claims,proceedings, demands, costs, expenses, damages, and other liabilities whatsoever orhowsoever caused arising directly or indirectly in connection with, in relation to or arisingout of the use of the Content.
This article may be used for research, teaching, and private study purposes. Anysubstantial 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
Educational Assessment, 16:145–163, 2011
Copyright © Taylor & Francis Group, LLC
ISSN: 1062-7197 print/1532-6977 online
DOI: 10.1080/10627197.2011.604241
Discrepant SAT Critical Reading and WritingScores: Implications for College Performance
Emily J. Shaw, Krista D. Mattern, and Brian F. PattersonThe College Board
Despite the similarities that researchers note between the cognitive processes and knowledge
involved in reading and writing, there are students who are much stronger readers than writers
and those who are much stronger writers than readers. The addition of the writing section to the
SAT provides an opportunity to examine whether certain groups of students are more likely to
exhibit stronger performance in reading versus writing and the academic consequences of this
discrepant performance. Results of this study, based on hierarchical linear models of student
performance, showed that even after controlling for relevant student characteristics and prior
academic performance, an SAT critical reading–writing discrepancy had a small effect on 1st-
year grade point average as well as English course grades in college. Specifically, students who
had relatively higher writing scores as compared to their critical reading scores earned higher
grades in their 1st year of college as well as in their 1st-year English course(s).
The addition of the writing section to the SAT in March 2005 not only allowed admission
officers the opportunity to better understand students’ writing skills but also allowed students
to show another aspect of their knowledge, skills, and abilities—those related to writing—that
they will need and surely use in their 1st year of college. Although there were a number of
proponents of this change, including the president of the University of California system at
the time (Atkinson, 2002), numerous critics were skeptical of the new section, claiming that
it would only increase test-taking time while adding little value to the meaning of their scores
(Baron, 2005; Perelman, 2005). Many thought the new section might be redundant or capture
many of the similar cognitive dimensions that were already being examined on the critical
reading section. Although reading and writing do rely on many of the same cognitive skills, they
are essentially utilized in different ways within these two domains (Kucer, 1987, 2005; Langer
& Flihan, 2000; Rosenblatt, 1994; Shanahan & Lomax, 1986; Tierney & Shanahan, 1991).
Therefore, it is not surprising that some students have much higher writing than critical reading
Correspondence should be sent to Emily J. Shaw, Research & Development, The College Board, 45 Columbus
Avenue, New York, NY 10023. E-mail: [email protected]
145
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014
146 SHAW, MATTERN, PATTERSON
scores, and vice versa. Very few studies, however, have effectively analyzed the discrepant
reading and writing performance of high school students, despite Stotsky’s (1983) call for
such research more than two decades ago. The addition of the writing section to the SAT
provides a fitting opportunity to examine this performance discrepancy and to understand the
student characteristics and academic consequences associated with much stronger performance
in critical reading than writing, and vice versa.
A practical understanding of the consequences of discrepant performance in these two
domains is particularly useful for admission and enrollment professionals who examine student
test scores in conjunction with other student information on an application. Having access to
empirical research on the impact of discrepant critical reading and writing performance and its
relationship with 1st-year college performance would provide context and deeper understanding
when highly discrepant SAT critical reading and writing scores are present during holistic or
“whole folder” reviews of applicants (Rigol, 2003, p. 9).
This study examines whether certain groups of students are more likely to have higher
writing performance as compared to critical reading performance, and vice versa. In addition,
this study examines the relationship between discrepant SAT critical reading and writing
performance and college outcomes, including 1st-year grade point average (FYGPA) and 1st-
year English grade point average (FY English GPA). This predictive model includes relevant
variables such as academic performance (i.e., SAT total score, high school grade point average
[HSGPA]) and demographic variables (i.e., gender, best spoken language, and race/ethnicity) to
determine the independent impact of the discrepancy on 1st-year grades. Finally, the interaction
between the demographic variables and the critical reading and writing performance discrepancy
will be included in the model to determine whether discrepant performance has a different
relationship with college outcomes for certain subgroups of students.
REVIEW OF THE LITERATURE
It is largely accepted that reading and writing are highly related processes. Although there are
clear overlaps in many of the component skills and knowledge bases (Shanahan, 1984, 1987;
Stotsky, 1983), there are also differences between reading and writing that are sometimes not
as clear (Fitzgerald & Shanahan, 2000; Kucer, 1985; Langer, 1986a, 1986b; Shanahan, 1984,
1987; Stotsky, 1983; Tierney & Shanahan, 1991). Studies primarily examining elementary and
middle school students have found that when writing, students tend to be more concerned with
bottom-up issues such as syntax, mechanics, and lexical choices than when reading (Gleason,
1995; Kucer, 1985; Langer, 1986b). Students are also more likely to set goals and be more
cognizant of the strategies employed while writing versus reading. When reading, however,
students are more focused on the content and validation of their understanding of the meaning
behind the text.
The few studies that have examined discrepant reading and writing performance in depth
have looked at this issue among relatively small samples of elementary and middle school
students utilizing discrepant performance on local measures to primarily classify students
as good readers/good writers, good readers/poor writers, or poor readers/poor writers (e.g.,
Honeycutt, 2002; Jordan, 1986; Thacker, 1990, 1991). Jordan (1986), for example, looked at
the differences between good readers/good writers’ and good readers/poor writers’ composing
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014
DISCREPANT SAT CRITICAL READING AND WRITING 147
processes using think-aloud protocols and used a set of descriptive categories to code reading
and writing behaviors. This research showed that good readers/good writers (a) were able to
abstract content from a reading passage and write about what they read in their own words,
(b) were more aware of the structural features of sentences, and (c) spent much more time
planning prior to writing. Good readers/poor writers struggled with each of these aspects.
Similarly, based on 90 ninth-grade students divided equally into groups of good readers/good
writers, good readers/poor writers, and poor readers/poor writers, Thacker (1990, 1991) studied
students’ ability to understand and recognize varying degrees of text organization when reading.
He found that good readers/good writers and good readers/poor writers were both skilled at
distinguishing between well and poorly organized text. However, good readers/poor writers
seemed to lack an awareness of how cohesive ties can bring meaning to disorganized text
and would likely benefit from greater instructional focus on cohesive relationships and the
effective organization of their own written responses to material. Given that there are students
with highly discrepant reading and writing performance and that these students do tend to
approach the two related tasks in different ways than students with more consistent reading
and writing performance, it would seem useful to study the impact of discrepant reading and
writing performance at the postsecondary level.
Although discrepant SAT critical reading and writing performance has not yet been studied
in relation to postsecondary performance, other analyses related to discrepant SAT performance
patterns have been conducted. For example, a recent study by Mattern, Camara, and Kobrin
(2007) showed that there are sizeable groups of students considered to have discrepant critical
reading and writing scores on the SAT. Mattern et al. standardized students’ critical reading
and writing scores across the 2006 College Bound Seniors cohort1 to examine the difference
between these scores. Those students with critical reading and writing scores that differed by
1 or more standard units were considered to be discrepant. There were 49,356 students (3.6%
of the cohort of test takers) who scored 1 or more standard units higher on writing than critical
reading (referred to as better at writing), and there were 50,336 students (3.7% of the cohort
of test takers) who scored 1 or more standard units higher on critical reading than writing
(referred to as better at critical reading).
Mattern et al. (2007) also investigated whether there were any differences between the
performance groups with regard to gender and/or racial/ethnic composition. They found that
the better at writing group was comprised of almost twice as many female as male students.
Conversely, the better at critical reading group had almost twice as many male as female
students. With regard to race/ethnicity, the only differences noted were that the percentages of
White and American Indian/Alaskan Native students were higher in the better at critical reading
group, whereas the percentage of Asian, Asian American, and Pacific Islander students was
higher in the better at writing category. The difference in HSGPA among the better at critical
reading group, better at writing group, and students who scored similarly on both sections was
also investigated. Mattern et al. found that HSGPA was the highest for the better at writing
group and lowest for the better at critical reading group, with significant differences (p < .05)
among all three groups based on analysis of variance (ANOVA) results.
In addition to noting gender and racial/ethnic differences in discrepant critical reading and
writing performance on the SAT, in a different study, Shaw (2007) also found that students
1Comprised of students with an SAT or SAT Subject Test score that reported to graduate from high school in 2006.
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014
148 SHAW, MATTERN, PATTERSON
that were much stronger in SAT writing than SAT critical reading were significantly more
likely (p < .05) to have taken English as a Second Language coursework in high school than
students who were much stronger in SAT critical reading than SAT writing. Shaw speculated
that perhaps this much stronger writing than reading performance for students with English as
a Second Language experience was related to issues of biliteracy, noting that Holm and Dodd
(1996) found that students from nonalphabetic written language backgrounds tend to struggle
with new or unfamiliar words when attending universities where English is the medium of
instruction. There are likely many unfamiliar words to students on the SAT critical reading
section, whereas the writing multiple-choice section tends to be more rule based and related
to grammar and the SAT essay is student produced, free from most of the constraints placed
on a reader by the author of an existing text.
These demographic and academic (HSGPA) differences between the discrepant groups signal
that there may also be differences in how these students perform in college, or specifically how
the magnitude and direction of the critical reading and writing discrepancy might impact 1st-
year college performance, particularly in English coursework. Despite the importance and
value in integrating writing across the disciplines, many content areas require few or no
writing assignments, nor do they offer the corresponding writing instruction that would foster
success on the writing assignments (Lavelle, 2003). Writing scholars have remarked that
training in writing has all too often become the sole responsibility of the freshman English or
composition course (Lavelle, 2003; Moore, 2003). Understanding the connection between the
critical reading and writing discrepancy with English course performance would be particularly
interesting, as both reading and writing activities are both considered central to freshman
English coursework (El-Hindi, 1997; Flower et al., 1990). However, Bosley (2008) commented
that although there is a large body of literature that documents the value in teaching critical
reading and writing reciprocally in the classroom, most college composition courses do not
explicitly cover critical reading strategies or effectively integrate reading into the writing
lessons and assignments. Further complicating our understanding of undergraduate English
coursework and performance, Lavelle (2003) observed that 1st-year composition grades do
not always reflect the students’ writing skills (such as analysis, synthesis, transcription, and
revision) but includes contaminating factors like attendance, promptness, or public speaking
skills.
This model-based study allows for the systematic investigation of differences in 1st-year
college performance, in addition to descriptive information about discrepant reading and writing
performance by different subgroups. Unlike the few previous studies that have focused on
describing students with discrepant reading and writing performance, the current research
focuses on the academic consequences of discrepant reading and writing performance in
college. Moreover, studying discrepant reading and writing performance on the SAT allows
for the examination of this issue on a much larger, national scale than has been studied in the
past. Also different from previous studies on discrepant SAT reading and writing performance,
which developed categorical groups of discrepant performance for analysis (i.e., Mattern
et al., 2007; Shaw, 2007), this study uses a continuous discrepancy measure to avoid losing
any measurement precision associated with dichotomizing continuous variables (MacCallum,
Zhang, Preacher, & Rucker, 2002). More than 1.5 million students take the SAT each year
(College Board, 2010), and the test is used in admission decisions at the large majority of 4-
year colleges and universities, rendering the results of the current study to be both theoretically
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014
DISCREPANT SAT CRITICAL READING AND WRITING 149
and practically useful to various stakeholders. In a similar vein, much of the research in this
area is based on small samples, thereby potentially limiting the generalizability of the results,
whereas the current study includes a sample of more than 140,000 attending a diverse set of
109 postsecondary institutions.
METHOD
Participants
The sample for this study is based on the students in the national SAT Validity Study sample
(for details, see Kobrin, Patterson, Shaw, Mattern, & Barbuti, 2008). For the national SAT
Validity Study, a wide range of 4-year institutions in the United States submitted 1st-year
college performance data to the College Board on the first-time, 1st-year students who entered
their institutions in the fall of 2006. The final sample in the current study included 140,919
students—all with valid SAT scores, self-reported high school GPAs, and FYGPAs provided
by their college/university, from 109 four-year institutions in the United States.
Measures
FYGPA. Each participating institution supplied FYGPA values for their 2006 first-time,
1st-year students. The range of FYGPA across institutions was 0.00 to 4.27.2
FY English GPA. Each participating institution supplied grades for all of the courses
taken by their 1st-year, first-time students during the 2006–2007 school year. All coursework
was coded for the subject area of the course. Those courses coded as English courses taken in
the 1st year of college3 were averaged for each student and considered to be the students’ FY
English GPA. Of the 140,919 students in the sample, 101,765 took at least one English course;
therefore, analyses based on FY English GPA are based on that subset of students.
SAT critical reading section. The SAT critical reading section, scored on a scale ranging
from 200 to 800, consists of 67 items in two 25-min sections and one 20-min section. The
SAT critical reading section measures a student’s ability to read and think carefully based on
sentence completions and items related to passages ranging in length from 100 to approximately
850 words and on topics from literary fiction to natural sciences. There are 19 sentence
completion items and 48 passage-based reading items, all of which fall into three general
content categories: extended reasoning (42–50 items), literal comprehension (4–6 items), and
vocabulary in context (12–16 items).
2Although a few institutions’ GPA scales ranged from 0.00 to 4.33, most had a maximum of 4.00.3Shaw and Patterson (2010) examined 1st-year college coursework in English across a national sample of 4-year
institutions and found that the vast majority of English courses are composition courses (72%).
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014
150 SHAW, MATTERN, PATTERSON
SAT writing section. The SAT writing section consists of one 25-min essay, one 25-min
multiple-choice section, and one 10-min multiple-choice section with a total of 60 items. The
SAT writing section measures a student’s ability to improve sentences, identify sentence errors,
improve paragraphs, and write an essay that will assess a student’s ability to think critically and
write effectively in response to a prompt adapted from an authentic text, under time constraints
similar to those encountered in essay tests in college courses. The essay score scale ranges
from 0 to 12 and is based on the ratings of two trained essay readers. The multiple-choice
writing section counts for approximately 70%, and the essay counts for approximately 30% of
the total raw score, which is used to calculate the 200 to 800 scale score for the section.
SAT Questionnaire. The SAT Questionnaire is a survey administered to all students
when they register for the SAT, either online or by mail. It consists of 42 questions about
the student’s background, high school experiences, and plans for college. Self-reported gender,
race/ethnicity, and best language spoken were obtained from the SAT Questionnaire.
Design and Procedure
This study focused on discrepant critical reading (CR) and writing (W) performance as indexed
by the difference in the two SAT section scores (CR-W Discrepancy D SAT CR–SAT W).
Unlike previous research, which created categorical groups based on a difference score (i.e.,
Mattern et al., 2007; Shaw, 2007), the continuous CR-W Discrepancy variable was used in
subsequent analyses to avoid any loss of information. The CR-W Discrepancy scores ranged
from �320 to 530 with a mean of 6.16 and a standard deviation of 60.58, indicating that, on
average, students’ SAT critical reading scores were 6 points higher than their writing scores. The
first set of analyses examined whether the magnitude of CR-W Discrepancy differed for specific
subgroups, and if so, in what direction. That is, the average CR-W Discrepancy score was
computed overall and by gender, race/ethnicity, and best language subgroups. Student’s t tests
were conducted to examine whether the average CR-W Discrepancy score was significantly
different from zero (p < .01), where a zero value indicates that, on average, students earn
the same SAT critical reading and writing score. This information is useful in highlighting
the commonalities and differences that characterize students exhibiting discrepant critical
reading and writing performance on the SAT and can be particularly helpful in the design
and development of specific educational interventions for their weaker area.
In addition, this study examined whether a student’s CR-W Discrepancy score is related
to subsequent college performance (i.e., FYGPA and FY English GPA), above and beyond
traditional measures of academic performance and student characteristics. Such analyses can
inform whether discrepant performance has a positive, negative, or insignificant effect on
college success. Also, whether the direction of the discrepancy (i.e., higher writing performance
vs. higher critical reading performance) mattered in relation to future college performance
was examined. Finally, the impact of discrepant reading and writing performance on college
performance for different student subgroups, such as English Language Learners, was analyzed.
To study these research questions, hierarchical linear modeling (HLM) techniques were
employed because of the inherent nested structure (i.e., students within colleges) of the data
(Raudenbush & Bryk, 2002). In the first step, a model with student-level demographic charac-
teristics (i.e., gender, race/ethnicity, best spoken language) and academic performance measures
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014
DISCREPANT SAT CRITICAL READING AND WRITING 151
(i.e., HSGPA, SAT composite score) was estimated. Next, the CR-W Discrepancy variable was
added and the change in model fit was examined using the Akaike Information Criterion (AIC).
Finally, interactions among student subgroups and the CR-W Discrepancy were added to the
model, and again the change in AIC was computed to examine the change in model fit.
RESULTS
Descriptive Statistics
The sample size, mean, and standard deviation for each academic performance measure for the
total sample and by gender, racial/ethnic, and best spoken language subgroups are provided
in Table 1. In general, female participants earned higher grades in high school and college
and had slightly higher SAT writing scores as compared to male participants who had higher
SAT math and critical reading scores. With regard to racial/ethnic comparisons, White and
Asian students outperformed the other subgroups on all academic indicators. Students whose
best spoken language was English earned the highest grades in their 1st-year English courses;
however, students whose best spoken language was something other than English had the
highest FYGPA. It should be pointed out that not all students took a 1st-year English course, and
therefore the sample sizes for that measure are smaller than for the other academic indicators.
To determine whether certain students performed more discrepantly on the SAT critical
reading and writing sections than other students, the mean CR-W Discrepancy score overall
and by student subgroups is provided along with the corresponding t-test results in Table 2. As
mentioned previously, the average CR-W Discrepancy score was 6.16 (SD D 60.58), indicating
that, on average, students in this sample had critical reading scores that were six points higher
than their writing. This is similar to the national results, where the average critical reading
TABLE 1
Descriptive Statistics of Study Variables by Student Characteristics
HSGPA SAT CR SAT M SAT W FYGPA FY English GPA
Subgroup n M SD M SD M SD M SD M SD n M SD
Overall 140,919 3.61 0.50 560 95 579 96 553 94 2.97 0.71 101,765 3.12 0.84
Gender
Female 75,940 3.65 0.48 556 95 559 93 556 93 3.05 0.67 56,064 3.21 0.78
Male 64,979 3.55 0.52 564 95 602 95 550 94 2.89 0.74 45,701 3.01 0.89
Race/Ethnicity
American Indian 772 3.53 0.54 544 87 555 89 529 88 2.78 0.76 538 2.97 0.92
Asian 13,775 3.67 0.47 563 104 624 97 563 101 3.05 0.66 9,426 3.15 0.79
Black 9,944 3.40 0.55 508 88 505 87 499 87 2.64 0.73 8,187 2.80 0.94
Hispanic 10,338 3.59 0.51 525 93 538 94 520 90 2.74 0.77 7,123 2.83 0.97
Other 4,372 3.58 0.50 559 98 573 98 554 97 2.96 0.71 3,217 3.08 0.83
White 101,718 3.62 0.49 568 92 585 92 561 91 3.02 0.69 73,274 3.18 0.80
Best spoken language
Another 1,605 3.62 0.52 464 99 606 114 480 102 3.05 0.68 1,139 3.09 0.84
English and another 7,033 3.62 0.49 531 100 570 108 534 101 2.90 0.72 4,959 2.97 0.89
English only 132,281 3.61 0.50 562 94 579 95 555 93 2.98 0.70 95,667 3.12 0.83
Note. HSGPA D high school grade point average; CR D critical reading; M D math; W D writing; FYGPA D 1st-year grade
point average.
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014
152 SHAW, MATTERN, PATTERSON
TABLE 2
SAT CR-W Discrepancy by Student Characteristics
Subgroup n M SD t p
Overall 140,919 6.16 60.58 38.15 <.001
Gender
Female 75,940 �0.50 59.24 �2.33 .020
Male 64,979 13.94 61.19 58.07 <.001
Race/Ethnicity
American Indian 772 15.38 58.69 7.28 <.001
Asian 13,775 0.39 62.06 0.73 .465
Black 9,944 8.54 60.91 13.98 <.001
Hispanic 10,338 4.49 59.58 7.65 <.001
Other 4,372 4.29 60.85 4.66 <.001
White 101,718 6.89 60.39 36.36 <.001
Best spoken language
Another 1,605 �16.65 63.43 �10.51 <.001
English and another 7,033 �2.92 61.60 �3.97 <.001
English only 132,281 6.92 60.40 41.65 <.001
Note. CR D critical reading; W D writing.
score was six points higher than the average writing score for the 2006 College Bound Seniors
cohort (SAT CR D 503, SAT W D 497; College Board, 2006). These results suggest that most
students had similar scores on the two sections of the SAT; however, there was much variability
across students with CR-W Discrepancy scores ranging from �320 to 530. See Figure 1 for
the distribution of CR-W Discrepancy scores in the sample.
The mean CR-W Discrepancy score also varied across student subgroups. For example,
female students had similar critical reading and writing scores (M D �0.50, t D �2.33,
p D .020) whereas male students had critical reading scores that were roughly 14 points
higher (M D 13.94, t D 58.07, p < .001). In addition, ANOVA results revealed that CR-W
Discrepancy scores were significantly higher for male than female students (F D 1401.18,
p < .001). These results parallel the pattern of scores for the national population of SAT test
takers in 2006, where female participants had both an average critical reading and writing
score of 502 and male participants had an average critical reading score of 505 and writing
score of 491 (College Board, 2006). When the results were parsed by race/ethnicity, all CR-W
Discrepancy score averages were positive, but the magnitude varied across subgroups, with
American Indian students followed by Black students having the most discrepant performance.
All values were significantly different from zero (p < .01) except for the Asian students, who
had a mean CR-W Discrepancy score of 0.39 (t D 0.73, p D .465). ANOVA results revealed
significant differences among racial/ethnic groups in CR-W Discrepancy scores (F D 29.91,
p < .001). Of the 15 post hoc pairwise comparisons, six were not significant at p < .01.
Four of the nonsignificant comparisons included White students when compared to American
Indian, Black, Hispanic, and Other students. The American Indian—Black and Hispanic—
Other contrasts were also nonsignificant. Finally, the results by best language spoken revealed
that students who indicated that their best language was not English earned higher writing
scores than critical reading scores, whereas the opposite was true for students who indicated
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014
DISCREPANT SAT CRITICAL READING AND WRITING 153
FIGURE 1 Distribution of students’ CR-W discrepancy scores.
that their best spoken language was English. Results revealed significant differences among
best language groups in CR-W Discrepancy scores (F D 146.46, p < .001) with all post hoc
pairwise comparisons significant at p < .01. The results are consistent with the findings from
Mattern et al. (2007).
HLM Results
FYGPA. In the first set of HLM analyses, the effect of CR-W Discrepancy scores on
FYGPA was examined, modeling student-level effects and controlling for random college-level
variation. The reference group was defined as White, male students whose best spoken language
was English. Prior to model estimation, HSGPA, SAT composite score, and CR-W Discrepancy
scores were grand mean centered—across all colleges and using sample means (HSGPA M D
3.61; SAT composite M D 1692)—in order for the intercept to be more meaningful (i.e., the
expected FYGPA for a student in the reference group with quantitative predictors equal to
the sample mean). In addition, the SAT composite score was divided by 600 and the CR-W
Discrepancy score was divided by 300 to make the scale of the predictors more comparable
and for significance test results to be more meaningful. The model estimates and corresponding
p values are provided in Table 3.
In the first step, a model was estimated that included gender, race/ethnicity, best language
spoken, HSGPA, and SAT total score as predictors of FYGPA. For Level 1, all of the parameter
estimates for the demographic variables were significant (p < .01) except for the indicator for
Asian students and for students whose best spoken language was English and another language.
HSGPA and SAT scores were both positively related to FYGPA, which is congruent with
previous research (Kobrin et al., 2008).
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014
154 SHAW, MATTERN, PATTERSON
TABLE 3
Hierarchical Linear Model Results for 1st-Year Grade Point Average
Model 1 Model 2 Model 3
Effect Est. p Est. p Est. p
Intercept 2.964 <.001 2.965 <.001 2.966 <.001
Gender
Female 0.151 <.001 0.148 <.001 0.148 <.001
Race/Ethnicity
American Indian �0.151 <.001 �0.149 <.001 �0.152 <.001
Asian �0.026 .018 �0.026 .015 �0.026 .016
Black �0.163 <.001 �0.162 <.001 �0.162 <.001
Hispanic �0.118 <.001 �0.116 <.001 �0.116 <.001
Other �0.049 <.001 �0.050 <.001 �0.050 <.001
Best language
Another 0.213 <.001 0.208 <.001 0.210 <.001
English and another 0.007 .576 0.005 .652 0.005 .691
HSGPAa 0.400 <.001 0.399 <.001 0.399 <.001
SAT CR C M C Wa;b 0.487 <.001 0.489 <.001 0.488 <.001
SAT CR � Wa;c �0.069 <.001 �0.107 <.001
SAT CR � Wa;c � Female 0.060 <.001
SAT CR � Wa;c� American Indian 0.090 .414
SAT CR � Wa;c� Asian 0.023 .415
SAT CR � Wa;c � Black 0.056 .073
SAT CR � Wa;c � Hispanic 0.006 .863
SAT CR � Wa;c� Other 0.018 .699
SAT CR � Wa;c� Another language 0.010 .894
SAT CR � Wa;c� English and another language �0.032 .394
Covariance Parameter Estimates With Approximate p Values
Group Parameter Est. p Est. p Est. p
College Intercept 0.016 <.001 0.016 <.001 0.016 <.001
College Gender 0.001 <.001 0.001 <.001 0.001 <.001
College Race/Ethnicity 0.003 <.001 0.003 <.001 0.003 <.001
College Best language 0.003 .001 0.003 .001 0.003 .001
College HSGPAa 0.010 <.001 0.010 <.001 0.010 <.001
College SAT CR C M C Wa;b 0.012 <.001 0.012 <.001 0.012 <.001
College SAT CR-Wa;c 0.002 .057 0.001 .083
Residual 0.348 <.001 0.348 <.001 0.348 <.001
Par. AIC Par. AIC Par. AIC
18 252,441 20 252,361 28 252,357
Note. N D 140,919, k D 109. Reference group: White male students whose best spoken language is English alone, with high school
grade point average (HSGPA) D 3.61 and SAT critical reading (CR) C SAT math (M) C SAT writing (W) D 1692. Models estimated
under full-information maximum likelihood with random college effects for all predictors. AIC D Akaike Information Criterion.aVariable was grand mean centered based on the sample included in the 1st-year grade point average analysis. bVariable was
divided by 600 to make the scale of predictors more comparable. cVariable was divided by 300 to make the scale of predictors more
comparable.
In the next step, the students’ CR-W Discrepancy score was added to the model. The
parameter estimate was �0.069 (p < .001), indicating that students with higher SAT writing
scores than critical reading scores earned higher grades in college, controlling for the students’
demographic characteristics and academic credentials (i.e., HSGPA, SAT total score). The
addition of the two parameters (i.e., the student-level effect for CR-W Discrepancy and its
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014
DISCREPANT SAT CRITICAL READING AND WRITING 155
college-level variance) indicates an improvement in model fit as illustrated by the reduction in
AIC.
Finally, the interaction terms between a student’s discrepancy score and demographic vari-
ables were included in the model to examine whether the relationship between CR-W Discrep-
ancy score and FYGPA differed for various subgroups of students. Only the parameter estimate
for CR-W Discrepancy � Gender interaction was significant (0.060, p < .001) indicating that
discrepant performance had a smaller effect on FYGPA for female participants as compared to
male. In Table 4, the expected FYGPA for various combinations of CR-W Discrepancy scores
are provided to illustrate the effect of discrepant performance on FYGPA. For example, for
White male students whose best spoken language is English and who have a HSGPA of 3.61
and an SAT total score of 1500, students with a �100 CR-W Discrepancy have an expected
FYGPA of 2.84 as compared to 2.77 for students with a C100 CR-W Discrepancy. However,
for White female students whose best spoken language is English and who have a HSGPA
of 3.61 and an SAT total score of 1500, students with a �100 CR-W Discrepancy have an
expected FYGPA of 2.97 as compared to 2.94 for students with a C100 CR-W Discrepancy.
The difference in FYGPA was 0.03 for female students as compared to 0.07 for male students.
Refer to Table 4 for the expected FYGPA for other combinations of SAT scores, SAT CR-W
Discrepancy scores, and student subgroups.
FY English GPA. Given that students take a variety of courses during their college
experience, grades in English courses were also examined as a more focused investigation of
the consequences of discrepant performance on the SAT writing and critical reading sections.
Only students who took at least one English course during their 1st year of college were
included in the analyses, reducing the sample to 101,765 students. For students who took more
than one English course, the average English grade was used as the outcome of interest.
Parallel to the FYGPA analyses, gender, race/ethnicity, best language spoken, HSGPA, and
SAT composite scores were included in the first step as predictors of FY English GPA. For the
student-level effects, gender and all of the race/ethnicity indicators except for Asian students
were significant (refer to Table 5). Unlike the FYGPA results where stating a language other
than English was their best language was significant (p < .01), neither of the best language
effects were significantly different from zero for FY English GPA. The parameter estimates
for both HSGPA and SAT scores were positive, indicating that students with higher HSGPAs
and SAT scores earned higher grades in their 1st-year English course(s).
In the next step, the students’ CR-W Discrepancy score was added to the model. The
parameter estimate was �0.134 (p < .001) indicating that students with higher SAT writing
scores than critical reading scores earned higher grades in their 1st-year English course(s),
controlling for the students’ demographic characteristics and academic credentials. Perhaps not
surprising, the effect is more than twice as large as what was found for FYGPA, which is
likely a function of matching the content area of the criterion with that of the predictors. In
terms of model fit, the addition of the two CR-W Discrepancy parameters (i.e., the fixed- and
random-effects) indicates an improvement as illustrated by the reduction in AIC.
Finally, the interaction terms between a student’s discrepancy score and demographic vari-
ables were included in the model to examine whether the relationship between the CR-W
Discrepancy score and FY English GPA differed for various subgroups of students. Similar to
the FYGPA results, only the parameter estimate for the CR-W Discrepancy � Gender interaction
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014
156 SHAW, MATTERN, PATTERSON
TABLE 4
Expected 1st-Year Grade Point Average by SAT CR-W Discrepancy Score
White, Male, English-Only Students With HSGPA D 3.61
SAT CR-W
SAT CR C M C W �100 �50 0 50 100
1500 2.84 2.82 2.80 2.79 2.77
1600 2.92 2.90 2.88 2.87 2.85
1700 3.00 2.98 2.97 2.95 2.93
White, Female, English-Only Students With HSGPA D 3.61
SAT CR-W
SAT CR C M C W �100 �50 0 50 100
1500 2.97 2.96 2.96 2.95 2.94
1600 3.05 3.04 3.04 3.03 3.02
1700 3.13 3.13 3.12 3.11 3.10
White, Male, Another–Language-Only Students With HSGPA D 3.61
SAT CR-W
SAT CR C M C W �100 �50 0 50 100
1500 3.05 3.03 3.01 3.00 2.98
1600 3.13 3.11 3.09 3.08 3.06
1700 3.21 3.19 3.18 3.16 3.14
White, Female, Another–Language-Only Students With HSGPA D 3.61
SAT CR-W
SAT CR C M C W �100 �50 0 50 100
1500 3.18 3.17 3.17 3.16 3.15
1600 3.26 3.25 3.25 3.24 3.23
1700 3.34 3.33 3.33 3.32 3.32
Note. Based on the parameter estimates from Model 3. CR D critical reading; M D math; W D writing; HSGPA
D high school grade point average.
was significant (0.095, p < .001), indicating that discrepant performance had a smaller effect
on FY English GPA for female students as compared to male students. Analogous to Table 4
for FYGPA, Table 6 provides the expected FY English GPA for various combinations of CR-
W Discrepancy scores to illustrate the effect of discrepant performance on English course
performance. For example, for White male students whose best spoken language is English
and who have a HSGPA of 3.56 and a SAT total score of 1500, students with a �100 CR-W
Discrepancy score have an expected FYGPA of 2.98 as compared to 2.85 for students with
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014
DISCREPANT SAT CRITICAL READING AND WRITING 157
TABLE 5
Hierarchical Linear Model Results for 1st-Year English Grade Point Average
Model 1 Model 2 Model 3
Effect Est. p Est. p Est. p
Intercept 3.040 <.001 3.045 <.001 3.046 <.001
Gender
Female 0.208 <.001 0.202 <.001 0.202 <.001
Race/Ethnicity
American Indian �0.127 <.001 �0.124 <.001 �0.125 <.001
Asian �0.029 .030 �0.031 .023 �0.030 .028
Black �0.143 <.001 �0.141 <.001 �0.141 <.001
Hispanic �0.114 <.001 �0.112 <.001 �0.112 <.001
Other �0.043 .009 �0.043 .009 �0.043 .008
Best language
Another 0.062 .018 0.056 .033 0.053 .053
English and another �0.033 .038 �0.035 .026 �0.035 .028
HSGPAa 0.345 <.001 0.343 <.001 0.343 <.001
SAT CR C M C Wa;b 0.421 <.001 0.425 <.001 0.424 <.001
SAT CR � Wa;c �0.134 <.001 �0.201 <.001
SAT CR � Wa;c � Female 0.095 <.001
SAT CR � Wa;c� American Indian 0.067 .681
SAT CR � Wa;c� Asian 0.102 .016
SAT CR � Wa;c � Black 0.022 .624
SAT CR � Wa;c � Hispanic 0.036 .454
SAT CR � Wa;c� Other 0.077 .236
SAT CR � Wa;c� Another language �0.094 .374
SAT CR � Wa;c� English and another language �0.021 .712
Covariance Parameter Estimates With Approximate p Values
Group Parameter Est. p Est. p Est. p
College Intercept 0.034 <.001 0.034 <.001 0.034 <.001
College Gender 0.002 <.001 0.002 <.001 0.002 <.001
College Race/Ethnicity 0.003 <.001 0.003 <.001 0.003 <.001
College Best language 0.003 .018 0.003 .020 0.003 .018
College HSGPAa 0.016 <.001 0.016 <.001 0.016 <.001
College SAT CR C M C Wa;b 0.041 <.001 0.041 <.001 0.041 <.001
College SAT CR-Wa;c 0.013 <.001 0.012 .001
Residual 0.540 <.001 0.539 <.001 0.539 <.001
Par. AIC Par. AIC Par. AIC
18 257,162 20 226,993 28 226,985
Note. N D 101,765, k D 109. Reference group: White male students whose best spoken language is English alone, with high school
grade point average (HSGPA) D 3.56 and SAT critical reading (CR) C SAT math (M) C SAT writing (W) D 1666. Models estimated
under full-information maximum likelihood with random college effects for all predictors. AIC D Akaike Information Criterion.aVariable was grand mean centered based on the sample included in the 1st-year grade point average analysis. bVariable was
divided by 600 to make the scale of predictors more comparable. cVariable was divided by 300 to make the scale of predictors more
comparable.
a C100 CR-W Discrepancy score. However for White female students whose best spoken
language is English and who have a HSGPA of 3.56 and a SAT total score of 1500, students
with a �100 CR-W Discrepancy score have an expected FYGPA of 3.16 as compared to 3.09
for students with a C100 CR-W Discrepancy score. As was the case for FYGPA, the difference
in expected FY English GPA (given the sample mean HSGPA, SAT composite of 1500, and
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014
158 SHAW, MATTERN, PATTERSON
TABLE 6
Expected 1st-Year English Grade Point Average by SAT CR-W Discrepancy Score
White, Male, English-Only Students With HSGPA D 3.56
SAT CR-W
SAT CR C M C W �100 �50 0 50 100
1500 2.98 2.95 2.92 2.88 2.85
1600 3.05 3.02 2.99 2.95 2.92
1700 3.12 3.09 3.06 3.02 2.99
White, Female, English-Only Students With HSGPA D 3.56
SAT CR-W
SAT CR C M C W �100 �50 0 50 100
1500 3.16 3.14 3.12 3.11 3.09
1600 3.23 3.21 3.19 3.18 3.16
1700 3.30 3.28 3.27 3.25 3.23
White, Male, Another-Language-Only Students With HSGPA D 3.56
SAT CR-W
SAT CR C M C W �100 �50 0 50 100
1500 3.06 3.01 2.96 2.91 2.87
1600 3.13 3.08 3.03 2.99 2.94
1700 3.20 3.15 3.10 3.06 3.01
White, Female, Another–Language-Only Students With HSGPA D 3.56
SAT CR-W
SAT CR C M C W �100 �50 0 50 100
1500 3.24 3.20 3.17 3.14 3.10
1600 3.31 3.27 3.24 3.21 3.17
1700 3.38 3.35 3.31 3.28 3.25
Note. Based on the parameter estimates from Model 3. CR D critical reading; M D math; W D writing; HSGPA
D high school grade point average.
a change of �100 to C100 CR-W Discrepancy) was smaller for female students (0.07) as
compared to male students (0.13).
DISCUSSION
The intent of this study was to determine whether there were distinct differences in 1st-
year college performance, by discrepant reading and writing performance, overall, and for
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014
DISCREPANT SAT CRITICAL READING AND WRITING 159
different subgroups among a large, national sample of students, as well as to provide descriptive
information by subgroup. A primary contribution of this research is that it is the first study
to examine the academic consequences (e.g., FYGPA, grades in 1st-year English courses) of
discrepant reading and writing performance in higher education.
Results of the current study showed that, indeed, there were differences in the magnitude and
direction of the reading and writing discrepancy by demographic subgroups. For example, on
average, male students perform 14 points higher on critical reading than writing, whereas female
students, on average, perform quite similarly on both SAT sections. Also of note, American
Indian students perform, on average, 15 points higher on critical reading than writing, whereas
students, whose best spoken language is not English, perform almost 17 points higher on
writing than critical reading.
These performance differences by subgroup are interesting because they call into question
various explanations for systematic performance differences in two highly related domains. As
previously mentioned, though reading and writing are highly related, they entail the different use
of similar but not exactly the same knowledge, skills, and abilities. The discrepancy favoring
reading over writing for American Indian students may at least be partially explained by
cultural factors that have historically placed a greater emphasis on the oral tradition with a
lesser focus on writing skills (Ingalls, Hammond, Dupoux, & Baeza, 2006; Pearce & Gayle,
2009). Although there has been a substantial amount of research documenting stronger writing
performance for female students over male students (e.g., Engelhard, Gordon, Siddle Walker,
& Gabrielson, 1994; Halpern, 2004), there has been far less research as to why male students
might be stronger at reading than writing. The stronger writing than reading performance for
students who report their best language as other than English may be related to issues of
biliteracy or the transfer of literacy-processing skills from students’ first language to English
(Holm & Dodd, 1996). Shen (2009) remarked that for English Language Learners, reading and
writing are often taught as such separate skills and that the emphasis is typically on grammar
and mechanics, which could help to explain the performance difference favoring SAT writing
versus critical reading for this group of students.
When the academic consequences of discrepant critical reading and writing performance
were examined, including a measure of this discrepancy in a model of general 1st-year college
performance and controlling for relevant demographic and academic variables, there was an
improvement in the model fit. There was even greater improvement of model fit by the
discrepancy measure when the outcome of interest was 1st-year English GPA. The findings
from these analyses indicate that there appears to be a slight academic advantage in college to
being stronger at writing than reading. A cognitive explanation may be related to the strengths
and weaknesses found in earlier studies of discrepant reading and writing performance (e.g.,
Honeycutt, 2002; Jordan, 1986; Langer, 1986a, 1986b; Palmer, 1986; Thacker, 1990, 1991).
For example, Thacker (1990, 1991) found that good readers/poor writers, compared to good
readers/good writers, seemed to lack an awareness of how cohesive ties can bring meaning to
disorganized text. This weakness may be closely tied to performance in college coursework,
particularly in the 1st year. Another potential hypothesis to explore is whether students who are
stronger writers than readers are more likely to be conscientious, higher achieving students. Yet
another explanation may be that the outcome measures examined—FYGPA and FY English
GPA—are more closely related to writing skills and abilities than to reading skills and abilities
as measured by the SAT (Tierney & Leys, 1986). This would allude to a measurement effect
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014
160 SHAW, MATTERN, PATTERSON
of some sort, whereby writing is of greater import in the 1st year of college and therefore
very strong writing skills as measured by the SAT serve students extremely well. All of these
hypotheses require testing with further research.
In both the FYGPA and FY English GPA models, only the parameter estimates for the
CR-W Discrepancy � Gender interaction were significant (p < .01), indicating that discrepant
performance had a smaller effect on FYGPA and FY English GPA for female students as
compared to male students. More research is needed to better understand this finding. In a
study of the role of personality in college performance and individual course grades, Nguyen,
Allen, and Fraccastoro (2005) found that personality explains 7% of the variance in overall
college GPA and 16% of the variance in course grades and that gender often moderates the
personality–academic performance relationship. It is certainly possible that there are personality
characteristics that affect academic performance differently by gender, which may also be
related to discrepant reading and writing performance. Also, this finding may be linked to the
notion that the process of writing requires a great deal of cognitive planning and organization
(Thacker, 1990, 1991) and that grammar, in particular, is quite rule based. These metacognitive
strategies along with conscientiously following rules could be quite helpful in the 1st year of
college (R. L. Brennan, personal communication, September 12, 2008). Perhaps male students
who excel at writing, particularly over reading, are more conscientious and/or organized than
the average male student and are therefore slightly more successful in college or in a course.
An additional explanation may involve differences in the nature of the coursework taken by
male and female students in the 1st year of college, which could affect the relationship between
the CR-W Discrepancy and GPA by gender.
A few limitations of this research should be noted. First, reading and writing performance in
this study were defined only by scores on the critical reading and writing sections of the SAT,
respectively. Although the SAT critical reading and writing sections are neither complete nor
perfect measures of these domains, research has demonstrated a strong link between the skills
measured by the SAT introduced in March 2005 and high school and college curricula and
instructional practice in reading and writing (Milewski, Johnsen, Glazer, & Kubota, 2005). It
would be useful to study discrepant performance among other reading and writing measures in
the future. Similarly, the definition and operationalization of discrepant performance in reading
in writing is open to interpretation. Also, the students in this sample with outcome data from
colleges and universities were slightly higher performing students based on mean SAT scores
and HSGPA than the general SAT cohort. This was expected, however, because the sample
included only students attending 4-year colleges and universities, whereas the SAT cohort
includes all students who have taken the SAT in a particular year. However, based on the fact
that the sample SAT critical reading and writing gaps by subgroup mimicked those of the full
cohort, it is unlikely that there would vastly different results for a less able group of college
students.
There are a number of avenues for future research, as there is still much to learn about
students with discrepant reading and writing performance. For example, it would be interesting
to study the role of first and best language in discrepant reading and writing performance
and study whether there are certain languages or types of alphabets that make a student more
apt to perform much more strongly in writing than reading in English. Similarly, it could
be worthwhile to examine the role that English Language Learner’s curricula in the K–12
system may play in this discrepancy. Another area for future research is whether there may
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014
DISCREPANT SAT CRITICAL READING AND WRITING 161
be some high schools that cultivate discrepant performance by focusing on different areas
of English Language Arts in the curriculum more than others. Perhaps this could also be
investigated with hierarchical linear modeling to determine if there are high school effects in
discrepant performance. It would also be interesting to determine whether there are cultural or
personality factors related to discrepant performance and/or the relationship between discrepant
performance and college performance.
It would also be useful to examine the relationship between having discrepant SAT critical
reading and writing scores and more distal academic outcomes, such as 2nd- or 3rd-year GPA,
or retention to the 2nd or 3rd year. This would help to disentangle the potential effect of having
so many writing intensive courses required in the 1st year of college. In addition, does being
weaker in one domain or the other relate to one’s choice of major, with students selecting
majors that fit their academic strengths and weakness? It would also be interesting to learn
whether graduation rates vary systematically as function of a student’s CR-W discrepancy
score.
Another area for future study is related to the definitions of discrepant performance. The
current study considers discrepant performance to be based on relative differences between
each student’s own reading and writing performance. Using this method, students exhibiting
discrepant reading and writing performance could be quite skilled at both but have wide gaps
in performance. A different approach to studying discrepant performance would be to create
discrepant categories relative to the student’s standing on the measure (in this study, the SAT),
so that students performing x points above the mean on critical reading and x points below the
mean on writing would be categorized as stronger readers than writers, for example.
Finally, it would be worthwhile to explore whether reading and writing self-efficacy inter-
ventions such as modeling, goal setting, and progress feedback in the weaker domain may
be useful in building up performance in the weaker area, particularly because the student has
excelled and developed a strong foundation in a highly related domain.
The results from this study on discrepant SAT critical reading and writing performance
clearly show that the critical reading and writing sections provide unique and relevant informa-
tion about students’ knowledge, skills, and abilities in two related domains. Not only was there a
sizeable number of students who displayed discrepant critical reading and writing performance,
but discrepant performance favoring SAT writing over reading is related to higher FYGPAs
and FY English GPAs, even after controlling for relevant demographic and academic student
characteristics. Future research on students with discrepant reading and writing performance
is particularly important because differing performance in cognitively similar domains may
present an opportunity to build and develop skills and expertise in the weaker area by efficiently
building on the ample foundation in the stronger area.
REFERENCES
Atkinson, R. (2002, May). The changing world of college admissions tests. Speech presented at the Western Association
of College and University Business Officers, San Diego, CA. Retrieved from http://works.bepress.com/richard_atkin
son/26
Baron, D. (2005, May 6). The College Board’s new essay reverses decades of progress toward literacy. The Chronicle
of Higher Education, pp. B14–B15.
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014
162 SHAW, MATTERN, PATTERSON
Bosley, L. (2008). “I don’t teach reading”: Critical reading instruction in composition courses. Literacy Research and
Instruction, 47, 285–308.
College Board. (2006). 2006 college bound seniors: Total group profile report. New York: Author.
College Board. (2010). 2010 college bound seniors: Total group profile report. New York: Author.
El-Hindi, A. E. (1997). Connecting reading and writing: College learner’s metacognitive awareness. Journal of
Developmental Education, 21(2), 10–15.
Engelhard, G., Gordon, B., Siddle Walker, E., & Gabrielson, S. (1994). The influence of writing tasks and gender on
quality of writing for black and white students. Journal of Educational Research, 87, 197–209.
Fitzgerald, J., & Shanahan, T. (2000). Reading and writing relations and their development. Educational Psychologist,
35, 39–50.
Flower, L., Stein, V., Ackerman, J., Kantz, M. J., McCormick, K., & Peck, W. C. (1990). Reading to write: Exploring
a cognitive & social process. New York: Oxford University Press.
Gleason, M. M. (1995). Using direct instruction to integrate reading and writing for students with learning disabilities.
Reading Research Quarterly, 11, 91–108.
Halpern, D. (2004). A cognitive-process taxonomy for sex differences in cognitive abilities. Current Directions in
Psychological Science, 13, 135–139.
Holm, A., & Dodd, B. (1996). The effect of first written language on the acquisition of English literacy. Cognition,
59, 119–147.
Honeycutt, R. L. (2002). Good readers/poor writers: An investigation of the strategies, understanding, and meaning that
good readers who are poor writers ascribe to writing narrative text on-demand. Unpublished doctoral dissertation,
North Carolina State University, Raleigh.
Ingalls, L., Hammond, H., Dupoux, E., & Baeza, R. (2006). Teacher’s cultural knowledge and understanding of
American Indian students and their families: Impact of culture on a child’s learning. Rural Special Education
Quarterly, 25, 16–25.
Jordan, E. C. (1986). The comprehending and composing processes of good writers who are good readers and poor
writers who are good readers. Dissertation Abstracts International, 47(08), 2972A (UMI No. 8627468).
Kobrin, J. L., Patterson, B. F., Shaw, E. J., Mattern, K. D., & Barbuti, S. M. (2008). Validity of the SAT for Predicting
First-Year College Grade Point Average. (College Board Research Report No. 2008–5). New York: The College
Board.
Kucer, S. B. (1985). The making of meaning: Reading and writing as parallel processes. Written Communication, 2,
317–336.
Kucer, S. B. (1987). The cognitive base of reading and writing. In J. Squire (Ed.), The dynamics of language learning:
Research in the language arts (pp. 27–51). Urbana, IL: National Conference on Research in English and ERIC
Clearinghouse on Reading and Communication Skills.
Kucer, S. B. (2005). Dimensions of literacy: A conceptual base for teaching reading and writing in school settings
(2nd ed.). Mahwah, NJ: Erlbaum.
Langer, J. A. (1986a). Children reading and writing: Structures and strategies. Norwood, NJ: Ablex.
Langer, J. A. (1986b). Reading, writing, and understanding: An analysis of the construction of meaning. Written
Communication, 3, 219–267.
Langer, J. A., & Flihan, S. (2000). Writing and reading relationships: Constructive tasks. In R. Indrisano & J. Squire
(Eds.), Perspectives on writing: Research, theory, and practice (pp. 112–139). Newark, DE: International Reading
Association.
Lavelle, E. (2003). The quality of university writing: A preliminary analysis of undergraduate portfolios. Quality in
Higher Education, 9, 87–93.
MacCallum, R. C., Zhang, S., Preacher, K. J., & Rucker, D. D. (2002). On the practice of dichotomization of quantitative
variables. Psychological Methods, 7, 19–40.
Mattern, K. D., Camara, W. J., & Kobrin, J. L. (2007). SAT® Writing: An overview of research and psychometrics to
date (College Board Research Note RN–32). New York: The College Board.
Milewski, G. B., Johnsen, D., Glazer, N., & Kubota, M. (2005). A survey to evaluate the alignment of the new
SAT writing and critical reading sections to curricula and instructional practices (College Board Research Report
2005–1). New York: The College Board.
Moore, J. (2003). Writing and the disciplines. Peer Review, 6(1), 4–7.
Nguyen, N. T., Allen, L. C., & Fraccastoro, K. (2005). Personality predicts academic performance: Exploring the
moderating role of gender. Journal of Higher Education Policy and Management, 27, 105–116.
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014
DISCREPANT SAT CRITICAL READING AND WRITING 163
Palmer, W. S. (1986, March). Good readers/poor writers: Some implications for classroom practices. Paper presented
at the National Council of Teachers of English Spring Conference, Phoenix, AZ.
Pearce, L., & Gayle, R. (2009). Oral reading fluency as a predictor of reading comprehension with American Indian
and white elementary students. School Psychology Review, 38, 419–427.
Perelman, L. (2005, May 29). New SAT: Write longly, badly, and prosper. Los Angeles Times, p. M5.
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods
(2nd ed.). Thousand Oaks, CA: Sage.
Rigol, G. (2003). Admissions decision-making models: How U.S. institutions of higher education select undergraduate
students. New York: The College Board.
Rosenblatt, L. (1994). The transactional theory of reading and writing. In M. R. Ruddell & H. Singer (Eds.), Theoretical
models and processes of reading (4th ed.). Newark, DE: International Reading Association.
Shanahan, T. (1984). Nature of the reading-writing relation: An exploratory multivariate analysis. Journal of Educa-
tional Psychology, 76, 466–477.
Shanahan, T. (1987). The shared knowledge of reading and writing. Reading Psychology: An International Quarterly,
8, 93–102.
Shanahan, T., & Lomax, R. G. (1986). An analysis and comparison of theoretical models of the reading–writing
relationship. Journal of Educational Psychology, 78, 116–123.
Shaw, E. J. (2007). The reading and writing self-efficacy beliefs of students with discrepant reading and writing
performance (Doctoral dissertation, Fordham University, 2007). Dissertation Abstracts International, 69(02) (UMI
No. 3302121).
Shaw, E. J., & Patterson, B. P. (2010). What should students be ready for in college? A look at first-year course work
in four-year postsecondary institutions in the U.S. (College Board Research Report No. 2010–1). New York: The
College Board.
Shen, M-Y. (2009). Reading–writing connection for EFL college learners’ literacy development. Asian EFL Journal,
11(1), 87–106.
Stotsky, S. (1983). Research on reading/writing relationships: A synthesis and suggested directions. Language Arts,
60, 627–642.
Thacker, P. R. (1990). Effects of text organization on reading in ninth-grade good and poor readers and writers.
Dissertation Abstracts International, 51(06), 1971A (UMI No. 9032467).
Thacker, P. R. (1991). Text organization in reading: What ninth grade good and poor readers and writers know.
Washington, DC: Office of Educational Research and Improvement.
Tierney, R. J., & Leys, M. (1986). What is the value of connecting reading and writing? In B. T. Peterson (Ed.),
Convergences: Transactions in reading and writing (pp. 15–29). Urbana, IL: National Council of Teachers of
English.
Tierney, R. J., & Shanahan, T. (1991). Research on the reading–writing relationship: Interactions, transactions, and
outcomes. In R. Barr, M. L. Kamil, & P. D. Pearson (Eds.), Handbook of reading research (pp. 246–280). New
York: Longman.
Dow
nloa
ded
by [
Tuf
ts U
nive
rsity
] at
13:
03 1
2 N
ovem
ber
2014