Cultural Diversity and Ethnic Minority Psychology

2014, Vol. 20, No. 4, 570 582

2014 American Psychological Association

1099-9809/14/$12.00 DOI: 10.1037/a0036503

Exploring the Various Interpretations of Test Bias

Russell T. Warne

Myeongsun Yoon

Utah Valley University

Texas A&M University

Chris J. Price

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Utah Valley University

Test bias is a hotly debated topic in society, especially as it relates to diverse groups of examinees who
often score low on standardized tests. However, the phrase test bias has a multitude of interpretations
that many people are not aware of. In this article, we explain five different meanings of test bias and
summarize the empirical and theoretical evidence related to each interpretation. The five meanings are
as follows: (a) mean group differences, (b) differential predictive validity, (c) differential item functioning, (d) differing factor structures of tests, and (e) unequal consequences of test use for various groups.
We explain in this article why meanings (a) and (e) are not actual forms of test bias and that there are
serious concerns about (b). In our conclusion, we discuss the benefits of standardized testing for diverse
examinees and urge readers to be careful and precise in their use of the phrase test bias.
Keywords: test bias, differential item functioning, item bias, group differences, standardized tests

readers of Cultural Diversity and Ethnic Minority Psychology.

Therefore, we also discuss in this article the benefits of standardized educational and psychological tests for diverse examinees and
those who advocate for diverse populations. We hope that our
discussion of test bias will empower advocates of marginalized
groups and improve the quality of discourse about the use of tests
in psychology and education.

Thats a great deal to make one word mean, Alice said in a

thoughtful tone.
When I make a word do a lot of work like that, said Humpty
Dumpty, I always pay it extra. (Carroll, 1871/1917, p. 100)

In the English language there are many words and phrases that
receive extra pay from those wholike Humpty Dumpty give
them many different meanings. One phrase in particular that has
received a great deal of extra pay is test bias, which we found to
have a multitude of meanings (Reynolds & Lowe, 2009). The
purpose of this article is to explore the nature of test bias from the
perspective of psychometrics (i.e., the science of mental testing).
We intend to technically define the phrase and expound upon five
different ways that test bias is often used, discuss the nature of
item content as it relates to bias, and the benefits of standardized
testing for diverse examinees especially in the realm of education. We believe that the exploration of the phrase is important
because test bias is a hotly debated topic in education and psychology, but some of these debates have not been productive
because of those expressing opposing sides are often using the
phrase differently (e.g., the exchange between Mercer, 1979, and
Clarizio, 1979).
Although an article about semantics and terminology would
itself be useful, it would probably be of limited interest to the

What Is a Standardized Test?

Before discussing the phrase test bias, it is imperative to define
the phrase standardized test. In the popular lexicon the phrase
refers to a government-mandated multiple-choice test in which the
examinees (usually students) are required to give their responses
by filling in a bubble on an answer sheet with a No. 2 pencil. This
definition is, however, limiting. The word standardized merely
refers to the fact that administration, format, and scoring of a test
are the same for all examineeswhich is an essential requirement
to producing interpretable data (Sireci, 2005). There is nothing
about this standardization that dictates that tests must be multiple
choice format, or that a bubble sheet must be used, or that a
standardized test be required or administered by a government
entity. Indeed, standardized tests have many item formats, require
examinees to make a wide variety of responses, and are administered by many persons and organizations.
Standardized tests have diverse formats. Some, such as licensure
exams for architects or physicians, require examinees to create a
product or perform a task under a uniform set of constraints.
Otherssuch as the Torrance Test of Creative Thinkingrequire
examinees to produce their own responses, which are then graded
through a strict scoring rubric. Indeed, because consistency of
format, scoring, and administration are the only requirements for a
test to be standardized, one could argue that many life tasks are
standardized. Despite the broad definition of the phrase standardized test, most authors who raise concerns about test bias are doing

Russell T. Warne, Department of Behavioral Science, Utah Valley

University; Myeongsun Yoon, Department of Educational Psychology,
Texas A&M University; Chris J. Price, Department of Behavioral Science,
Utah Valley University.
Correspondence concerning this article should be addressed to Russell T.
Warne, Department of Behavioral Science, Utah Valley University, 800 W.
University Parkway, MC 115Orem, UT 84058. E-mail: rwarne@uvu
.edu
570

EXPLORING TEST BIAS

so in the context of high-stakes standardized tests, such as college

admissions tests, high school exit exams, intelligence tests, and
employment tests.

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Definitions of Test Bias

According to the standards set by the American Educational
Research Association (AERA), the American Psychological Association (APA), and the National Council on Measurement in
Education (NCME), test bias . . . is said to arise when deficiencies
in a test itself or the manner in which it is used result in different
meanings for scores earned by members of different identifiable
subgroups (American Educational Research Association, American Psychological Association, & National Council on Measurement in Education, 1999, p. 74). Although this definition is helpful, it is also quite broad and open to a variety of interpretations.
In our experiences with the topic, we have found five ways that the
phrase test bias has been interpreted in the literature:
1. Score gaps between groups which result from members of one
group on averagescoring higher than members of another
group.
2. Differences in the ability of scores to predict outcomes for
examinees.
3. Items functioning differently for examinees who belong to
different groups.
4. Differences in the intercorrelations and groupings (i.e., factor
structure) of items.
5. Consequences of test use or interpretation that create or
perpetuate social inequalities between groups.
These methods of observing test bias are presented in the
approximate order that these arguments first appeared in peerreviewed journals because often later conceptualizations of test
bias were often created to compensate for difficulties or shortcomings associated with earlier interpretations of test bias. We
also think that the order of interpretation of these definitions is
pedagogically appropriate because (except for Interpretation 5) the
explanations and examples of the test bias become increasingly
more technical and complex as one moves down the list.
In this section of the article, we will examine each of these
interpretations and explain mainstream thought among psychometricians concerning each and how they relate to AERA et al.s,
1999 broad definition. It is important for readers to realize as they
read this article that the groups referred to in the list above do not
have to be racial groups, although racial and ethnic groups get the
most attention in conversations about test bias. Other possible
groups may be gender, religious, age, economic, education, cultural, national, or any other type of group imaginable.

Interpretation #1: Mean Score Differences

One of the most consistentand frustratingfindings in
quantitative educational studies is the pervasive score differences among racial and ethnic groups. On many academic
achievement and aptitude tests, Asian American students score
higher than White students, who then in turn score higher than
Hispanics and African American students. These results have
been found on intelligence tests (e.g., Gottfredson, 1997; Neisser et al., 1996; Roid, 2003), academic aptitude tests (e.g.,
Lohman, 2005), tests for identifying gifted children (e.g.,

571

Olszewski-Kubilius & Lee, 2011) and children with special

needs (e.g., Morgan, Farkas, Hillemeier, & Maczuga, 2012),
high school exit exams (e.g., Nichols, 2003), standardized academic achievement tests (e.g., Hoover et al., 2003; Forsyth et
al., 2003; Lee, 2002), college admissions tests (e.g., Flowers,
2008; Posselt, Jaquette, Bielby, & Bastedo, 2012), and employment tests (OBoyle & McDaniel, 2009). Moreover, these score
gaps especially those between African American and White
examineesare long-standing and have been observed in every
generation since the beginning of modern mental testing in the
United States (e.g., Cleary, Humphreys, Kendrick, & Wesman,
1975; Lee, 2002; Terman, 1928; Yoakum & Yerkes, 1920).
Score gaps among racial and ethnic groups are so pervasive that
their existence has been discussed at length in official APA
publications (Cleary et al., 1975; Neisser et al., 1996) and has
been called . . . not a debatable issue (Kaplan & Saccuzzo,
2009, p. 512; see Rushton & Jensen, 2005, p. 236, for almost
identical language).
While the existence of score gaps on tests among demographic
groups is well established, the causes of such group differences is
still strongly debated (Reynolds, 2000). One common explanation
for score differences on academic tests among demographic
groups is test bias, with critics contending that deficiencies in the
test cause the score gaps and make scores incomparable across
demographic groups (as per AERA et al. 1999, definition of test
bias). Richert, for example, wrote,
Measures of academic achievement that are most often used by
schools [to identify gifted students], including teacher recommendations, grades, and especially standardized tests, have been amply
demonstrated to have cultural biases . . . (Richert, 2003, pp. 150
151, emphasis added).

According to Richert, not only are tests biased, but the evidence
is overwhelming that bias is an inherent characteristic of standardized tests. Similarly, Salend and his colleagues stated, Research
indicates that norm-referenced standardized tests are culturally
and socially biased . . . (Salend, Garrick Duhaney, & Montgomery, 2002, p. 290, emphasis added). Beliefs about the inherently
biased nature of tests are also found among other authors (e.g.,
Mensch & Mensch, 1991), including some that are highly respected in their fields (e.g., Ford, 2003; Gould, 1981). Such claims
are common in the journalistic media, too (Cronbach, 1975; Gottfredson, 1994; Phelps, 2003; Reynolds, 2000).
Others have a much more sinister view of standardized testing.
Moss described her experience teaching at a high school where,
Most of my students were poor and African American . . . (p.
217). She stated,
By the end of 13 years of experience, I became convinced that it did
not matter how successful students of color became, the test would be
revised to insure we start over in the cyclical process of teaching
students how to demonstrate their ability to take culturally biased
standardized tests. (Moss, 2008, p. 217)

For Moss, standardized tests are not just biased as some accident
of their creation. Rather, the writers of the tests her students took
were nefarious in their work, and the test creators intended to use
the tests to discriminate against her students (see Carter & Goodwin, 1994; Mercer, 1979; and Smith, 2003; for a similar viewpoint
of standardized tests).

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

572

WARNE, YOON, AND PRICE

What all of these examples have in common is that the writers

believe that because the average scores for some groups is consistently below the average scores of other groups, the test must be
biased and the scores from different demographic groups therefore
have different meanings across groups. The group mean differences interpretation of test bias is probably the most popular
interpretation of the five discussed in the article. This interpretation is illustrated in Figure 1, which shows two normal distributions. Although the two score distributions overlap, there is a
difference of roughly half a standard deviation between the average scores in these groups. Because of this mean difference, there
could be detrimental consequences to disadvantaged group members if the test is used to select students for an educational or
therapeutic program.
Nevertheless, this interpretation of test bias is unanimously
disregarded by those with training in psychological testing. Reynolds, for example, stated, The mean differences definition of test
bias is by far [emphasis added] the most widely rejected of all
definitions of bias by psychometricians who study these issues
(Reynolds 2000, p. 146). Many other testing experts agree (e.g.,
AERA et al., 1999; Camara, 2009; Clarizio, 1979; Flaugher, 1978;
Linn & Drasgow, 1987; Reynolds & Lowe, 2009).
Some advocates for diverse populations may be perplexed by
this stance among psychometricians. If these tests consistently
produce score gaps among groups, how could the tests not be
biased? A simple thought experiment should answer this question.
Instead of Figure 1 referring to two demographic groups score
distributions on an academic test, one should imagine that it refers
to two groups score distributions on a test of job satisfaction. The
reader should then imagine that the lower distribution represents
scores on a test of job satisfaction obtained from medical interns
and the higher distribution should represent scores obtained from
tenured college professors. A typical interns schedule includes 80
hr of work per week, nights on call, and very stressful working
conditions, while tenured university faculty have a great deal of
work flexibility, high job security, and tend to enjoy their jobs.
Under these circumstances, the professors should outscore the
medical interns on a test of job satisfaction. Any other results
would lead a reasonable observer to strongly question the validity

Figure 1. Score distributions for two groups. The mean difference between the two groups scores is .5 SD.

of the test results. The lesson from this thought experiment is that
mean score gaps are not evidence of test bias because there may be
other explanations of score gaps. In fact, score gaps may indicate
that the test is operating exactly as it should and measures real
differences among groupsas is the case with this hypothetical
test of job satisfaction (Clarizio, 1979; Linn & Drasgow, 1987).

Interpretation #2: Differential Predictive Validity

Many educational tests are used to predict an outcome, such as
success in an educational program. If a test score is able to predict
some future outcome (called a criterion), it is said to have predictive validity. For example, the SAT and ACT are designed to
predict a students probability of succeeding in college. Although
details vary from study to study, both tests tend to be moderately
strong predictors of college grades (r .40), although the predictive power of these tests increases when combined with information about the students high school grades (Camara, 2009; Kaplan
& Saccuzzo, 2009; Maruyama, 2012; Zwick, 2006). College admissions test score correlations with other criteria are less consistent because other outcomes (e.g., income, career success, or
whether a student will graduate with a bachelors degree) occur
many years after a students score is obtainedwhich means many
circumstances can intervene before the outcome is observed. Restriction of range of data has also been shown to reduce the
strength of the correlations between test scores and outcomes
(Camara, 2009; Zwick, 2006, 2007).
If mean score gaps themselves are not evidence of bias, perhaps
a test could be biased if it is better at predicting outcomes for some
groups and worse at predicting outcomes for other groups, a
situation called differential predictive validity (Camilli, 2006). In
other words, if the predictive validity of the test score varies from
group to group, then it is an indication that it is not appropriate to
use the test to make predictions for at least some examinees and
scores may have different meanings across groups, making the test
biased (AERA et al., 1999, p. 79). A few possible visual representations of differential predictive validity across groups are
shown in Figure 2, a c. Ideally, if a test has equal predictive
validity for both groups, then the same regression line will apply
to both groups, as demonstrated by Figure 2a. As is apparent in
Figure 2a, one group scores higher than the other on average, but
test scores predict outcomes for both groups equally well. In this
case, no test bias would exist (Reynolds & Lowe, 2009).
However, if mean differences in scores exist between two
groups, the schematic shown in Figure 2b is also possible. In the
figure, there are two scatterplots. However, the two groups cannot
be represented by the same regression line. Instead, there are two
parallel regression lines that each separately represent the relationship between the test score and the outcome for each group. The
dotted line halfway between the two main regression lines represents the regression line that would apply to both groups combined. As is apparent, the overall regression line represents the
predicted outcomes for neither group very well. This would be an
example of differential predictive validity (Cleary, 1968).
The difference in predictive validity shown in Figure 2b occurs
somewhat frequently in academic testing, such as intelligence tests
(Reschly & Sabers, 1979) and preschool assessment batteries
(Reynolds, 1980). However, most testing experts are not terribly
concerned when this differential predictive validity occurs because

573

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

EXPLORING TEST BIAS

Figure 2. (a) Scatterplots for the scores of two groups that share the same regression line. Notice that Group A on
average obtains lower scores than members of Group B. (b) Scatterplots for the scores of two groups with different,
but parallel, regression lines. The middle dashed line represents the regression line for both groups combined. Notice
that Group A on average obtains lower scores, but the combined regression line predicts more favorable outcomes than
would be expected from a regression line solely based on Group As data. (c) Scatterplots for the scores of two groups
with different, nonparallel regression lines. The middle dashed line represents the regression line for both groups combined.

it does not hurt disadvantaged groups. This is because if outcomes

are predicted using the regression equation from both groups
combined, then the predictions for disadvantaged group members
will systematically be more favorable than the predictions based

on the disadvantaged groups regression data alone (Clarizio,

1979; Crocker & Algina, 2008; Kaplan & Saccuzzo, 2009). This
can be seen in Figure 2b, which shows that for a given test score
the predicted outcome for a disadvantaged examinee will always

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

574

WARNE, YOON, AND PRICE

be higher using the regression equation for the combined groups

than for the regression equation based on just the disadvantaged
groups data.
In the real world of educational testing, this is exhibited by data
from the SAT. Young (2004) has shown that predictions of African
American and Hispanic students college grades when using data
from the entire body of SAT examinees are systematically higher
than diverse students actual college grades. In other words, by
using a test on which they score lower than other groups to decide
who is admitted to college, African American and Hispanic students actually benefit and are admitted to college at greater rates
than they would be if an admissions test were designed specifically
for these diverse groups. This paradox has been long recognized by
testing experts (e.g., Cleary, 1968; Cronbach, 1980).
Another reason differential prediction validity does not bother
many experts is because some outcomes are influenced by a wide
variety of independent variables (not just test scores), and these
other variables may impact groups of examinees in different ways.
For example, the SAT exhibits differential predictions for college
graduation of White and Black examinees, with the latter being
predicted to graduate at lower rates than White examinees, even if
the students obtain the same SAT score (Zwick, 2006). This may
be because of test problems, but it may also be because Black
students face more barriers to graduation than White students. It
has been suggested that a hostile campus environment, greater
financial problems, greater anxieties about academic abilities, and
other life challenges may make graduation more difficult for Black
students compared with their White classmates who obtain the
same score on the SAT (Zwick, 2007). Therefore, differential
validity as displayed in Figure 2blike mean score differences
among groups by itself is not evidence of test bias because it can
have causes other than problems with the test.
Finally, there is some evidence that the type of differential
predictive validity shown in Figure 2b is merely a statistical
artifact. Traditional ordinary least squares (OLS) regression operates under the assumption that the predictor variable is measured
perfectly (Young, Kane, Monfils, Li, & Ezzo, 2013). However,
this is definitely not true with test scores, which always have some
error (Kaplan & Saccuzzo, 2009) that can distort predictions made
with OLS regression (Kane & Mroch, 2010). Indeed, in multiple
studies (e.g., Kane & Mroch, 2010; Young et al., 2013) when
researchers used an alternative prediction method that took into
account test score error, the differential predictive validity of the
SAT for two different racial groups disappeared. These results
indicate that differential predictive validity may sometimes be a
result of the statistical model that a researcher choosesnot an
actual difference among examinee groups in test performance.
A more troubling example of differential predictive validity,
however, is found in Figure 2c. Like Figure 2b, it shows differing
regression lines for two groups of examinees. However, the regression lines in this example are radically different because they
are not parallel. This leads to inaccurate predictions for all students
when a single regression equation based on all examinees is used
(like in Figure 2b). In this case, the test score does not make the
same predictions for the two groups (Kaplan & Saccuzzo, 2009).
Thus, combining their data into one general regression line is
theoretically untenable. This form of differential prediction validity is insurmountable and has no other solution than to interpret
group members scores separately, although how to use those

scores may be complicated and unclear (Allen & Yen, 1979).

However, the differential predictive validity results shown in Figure 2c are rare and almost never encountered with real data from
professionally developed tests (Kaplan & Saccuzzo, 2009) because
the differences in correlations between test scores and criteria
rarely differ across groups by more than what would be expected
from regular sampling error (Jensen, 1998).

Interpretation #3: Differences in Group Performance

on Specific Items
The first two interpretations of test bias discussed in this articlemean score gaps and differential predictive validityare
concerned with total test scores. However, tests are comprised of
individual items. Therefore, the question of bias is perhaps best
examined at the item level. Experts who subscribe to this belief
have developed procedures to measure differential item functioning (DIF, also called item bias), which occurs when an item
performs differently for different groups. For DIF to be present
two examinees who belong to different groups but with equal
levels of individual ability must have different probabilities of
correctly answering an item (AERA et al., 1999; Camilli, 2006;
Cleary & Hilton, 1968; McDonald, 1999; Swaminathan & Rogers,
1990). Items may display DIF for a variety of reasons. For example, an item
. . . may be measuring something different from the remainder of the
test or it may be measuring with different levels of precision for
different subgroups of examinees. Such an item may offer a valid
measurement of some narrow element of the intended construct, or it
may tap some construct-irrelevant component that advantages or
disadvantages members of one group. (AERA et al., 1999, pp. 77, 78)

Procedures for examining DIF are very complex and technical.

Put in simple terms, they usually split examinees into groups based
on their total test score. Then for each item, the probability that
members of each group of interest (e.g., racial groups, gender
groups, cultural groups, socioeconomic groups) will correctly answer an item correct is calculated. If the null hypothesis that the
two groups have an equal probability of answering the item correctly is rejected (i.e., p .05, .01, or some other predetermined
value for ), then DIF is said to be present and the test scores are
biased in the AERA et al. (1999) sense of the phrase in that the
scores do not have the same meaning for both groups. For example, if average-scoring low-socioeconomic status (SES) and highSES examinees have the same probability of answering an item
correctly, then DIF is not present for that particular item. This
summary of DIF suffices for the purposes of this article, but
readers should recognize that DIF analysis takes many forms. The
reader is encouraged to examine more thorough and technical
treatments of DIF, some of which do not rely on null hypothesis
testing (e.g., Camilli, 2006; Crocker & Algina, 2008, chapter 16;
Embretson & Reise, 2000, pp. 249 263; Reynolds & Lowe, 2009,
pp. 345351; Swaminathan & Rogers, 1990).
DIF is quite useful for helping researchers and test developers
find items that do not function the same way across demographic
or social groups. Past research has shown, for example, that DIF
items that favor diverse students occur when the items contain
content that is of interest or especially relevant to members of their
demographic groupsuch as reading passages about the civil

EXPLORING TEST BIAS

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

rights movement for Black students (Schmitt & Dorans, 1990;

Zwick, 2007). Similarly, Hispanic students tend to find verbal
analogy items that use words that have Spanish cognates (i.e.,
words in Spanish that are spelled similarly to English words with
the same meaning) easier than White students (Zwick, 2006,
2007). For math items, basic computation items tend to favor
Black students, while White students perform better on story
problems (Hunter & Schmidt, 2000; Zwick, 2007).
DIF across gender groups has also been shown in certain types
of items on academic tests. Zwick (2007, p. 26) explained,
Women tend not to do as well as a matched group of men on verbal
SAT items . . . about scientific topics or about stereotypically male
interests, like sports or military activities . . . On the other hand,
women tend to perform better than their male counterparts on questions about human relationships or questions about the arts . . .

However, these tendencies are the exception to the rule; usually

it is not clear why an item displays DIF (AERA et al., 1999;
Camilli, 2006; Flaugher, 1978; Reynolds & Lowe, 2009; Schmitt
& Dorans, 1990; Zwick, 2007).
If an item displays DIF, then it may be eliminated from a test,
revised, or be balanced with DIF that cancels out the DIF in the
first item (Hunter & Schmidt, 2000; Zwick, 2006; see Hoover et
al., 2003 for an example of these procedures being used on an
academic achievement test). If many items are found to exhibit
DIF that consistently favors one group over another, then the entire
test score may systematically favor one group solely because of
group membership (and not the relevant characteristic being tested,
such as academic achievement). Such results would be a strong
indication that the test cannot be interpreted in the same way for
both groups.
How common is DIF? With the plethora of tests in the world,
most of which consist of many items, it should be unsurprising that
many tests have at least a few items that display DIF. However, in
professionally developed testssuch as the high-stakes educational tests that are the subject of conversations about test bias
items with large DIF are not present. This is because professional
test standards of practice and ethics to which test developers are
bound require that items with substantial DIF be removed from
tests (AERA et al., 1999). It is common, however, for items with
modest DIF to be found on professional tests. For example, on
Form A of the Iowa Tests of Educational Development there are
1,134 items across all grade levels. Of these, 5 (0.4%) favor
females, 13 (1.1%) favor males, 3 (0.3%) favor Black students, 1
favors White students over Black students (0.1%), none favor
Hispanic students, and 2 (0.2%) favor White students over Hispanic students (Forsyth et al., 2003, p. 86). As is apparent in this
example, items displaying DIF on professionally created tests are
usually so few and the DIF so small that the impact on total scores
is negligible (Camara, 2009; Goodman & Hambleton, 2005; Reynolds, 2000; Reynolds & Lowe, 2009).
DIF procedures are probably the most common way that tests
are examined for bias today. However, the procedure is not without its problems. DIF is inherently atheoretical and subjected to
statistical results that incorrectly flag acceptable items as having
DIF (Hunter & Schmidt, 2000). It is also fundamentally circular; in
DIF procedures based on total test scores, examinees must be
matched on their total score in order to examine a particular item
for bias. But the total score usually includes the item being

575

screened for DIF. Therefore, one must accept that the total test
score is unbiased as a prerequisite to test individual items for bias;
then, once DIF is not found in any of the items, a researcher can
assert that the total test score is not biased. This circular reasoning
(i.e., where the test score must be assumed to be unbiased in order
to determine that items are unbiased) has been correctly criticized
by researchers (e.g., Navas-Ara & Gmez-Benito, 2002).
Many DIF procedures also create problems statistically; if one
item with DIF is found but the total test score used to match
examinees across groups, then the advantage that a group receives
from the DIF item must be balanced out by other items that favor
the other group(s). This creates statistical artifacts in which items
are labeled as incorrectly having DIF (Andrich & Hagquist, 2012).
These are issues that the testing field is still grappling with, and
fully satisfactory solutions have not yet been found. DIF procedures based on latent variable methods such as confirmatory factor
analysis and item response theory are promising and have reduced
the severity of these statistical and logical problems. Nevertheless,
DIF has shed light on important questions about test construction
and has undoubtedly made tests fairer today than they were in
previous generations.

Interpretation #4: Differing Factor Structures

Test items are rarely interpreted in isolation, but rather items are
combined together in order to produce a score that is subject to
interpretation. In order for the interpretation or scoring of a set of
items to make sense, the items must intercorrelate with one another
to form a coherent group; these groups of items are assumed to be
related to one another because similar responses are caused by an
underlying factor, such as verbal ability, agreeableness, or
mathematics achievement. Most tests are created with a theory
concerning the number of factors and which items belong to each
factor (Kaplan & Saccuzzo, 2009). This theory of the makeup of
the factors is essential for interpretation because if items do not
intercorrelate according to theory, then the test score(s) may not
make sense and may be uninterpretable.
Factors are estimated and interpreted by researchers using a
multivariate statistics method called factor analysis (Gorsuch,
2003). The results of a factor analysis show the number of factors
in a set of test items and which items belong to each factor. Figure
3a shows an example of seven items (represented by the seven
rectangles) that form two factors (represented by the two circles).
As is apparent in the figure, one factor consists of four items and
the other factor consists of three items. The double-headed arrow
between the two circles shows that the two factors are correlated.
Factor analysis is relevant to the argument of test bias because
it is possible that the items may form different factors for different
groups (Reynolds & Lowe, 2009). This is tested through a method
called a test of factorial invariance, which is a complex statistical
process of testing whether the test items have the same factor
structure for both groups (Meredith, 1993; Millsap & Everson,
1993; Reise, Widaman, & Pugh, 1993; Schmitt & Kuljanin, 2008).
If items on a test have the same factor structure for both groups,
then no bias is present (Reynolds & Lowe, 2009).
However, ifas a hypothetical exampletest items for one
group have the factor structure in Figure 3a and the same items
have the factor structure in Figure 3b in another group, then test
bias. This would signify that the items from the test do not

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

576

WARNE, YOON, AND PRICE

Figure 3. (a) A 7-item test with two factors: four items load onto the first factor and three items load onto the
second factor. (b) A 7-item test with all items loading on a single factor.

necessarily behave the same way for both groups of subjects and
test biasas defined by AERA et al. (1999)may be present
because the scores from the two groups would have different
meanings. Differing factor structures may indicate a number of
possibilities, including the following:
The test items may be interpreted differently by the two
different groups.
The psychological construct (e.g., depression, personality,
intelligence, language arts achievement) may have different structures for the two groups. The nature of the construct may vary
across groups because of cultural, developmental, or other differences.
The test may measure completely different constructs for the
two groups.

Groups may use different mental processes to respond to the

items.
Examinees in different groups use different strategies when
responding to test questions.
If two different groups of examinees have different factor structures for a group of items, it is not appropriate to interpret the items
in the same way for both groups and scores cannot be compared
across groups (Meredith, 1993; Reynolds & Lowe, 2009), a situation that meets AERA et al.s (1999) definition of test bias.
Examples abound of tests that have differing factor structure items
across demographic groups, especially for instruments created by
nonpsychometricians (e.g., Li et al., 2009; Warne, 2011). For most
professionally developed intelligence, academic, and aptitude
tests, tests of invariance usually indicate that the factor structure is

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

EXPLORING TEST BIAS

the same across groups (e.g., Beaujean, McGlaughlin, & Marguiles, 2009; Benson, Hulac, & Kranzler, 2010; Dolan, 2000).
Evaluations of factor structure are especially complex and require large datasets (Meredith, 1993). However, they are generally
agreed to be the best methods of evaluating test bias (Borsboom,
2006). Because tests of invariance are somewhat new, there are
still several aspects of them that remain unresolved. First, the
results of a test of invariance are frequently not as clear and
unambiguous as the hypothetical example shown in Figure 3, a and
b. Rather, differences among factor structure are often a matter of
degree, not of kind. It is often difficult to know what to do with a
test when some parts of it operate the same across demographic
groups and other parts do not (Millsap & Kwok, 2004). Alsojust
like DIFit is not always clear why factor structures vary across
tests or why parts of a test function differently for different groups
(Schmitt & Kuljanin, 2008).

Interpretation #5: Unequal Consequences of Test Use

Across Groups
Others claim that a test is biased if its use disadvantages some groups
in society compared with others. Such a test is said to have poor
consequential validity because the consequences of using the test
are socially undesirable. For example, women on average score
consistently higher on the Beck Depression Inventory (BDI) than
men, indicating higher levels of depression (Santor, Ramsay, &
Zuroff, 1994). Therefore, women may be more likely to be diagnosed as depressed and be administered therapy and prescription
drugs. Some may believe that the negative consequences for
women of using the BDI to diagnose depression are undesirable
and damaging to American society because they may foster sexism, stigma, and discrimination in individual examinees lives.
These people say that the consequences of using the BDI are too
great and that the test is biased because it has poor consequential
validity. The consequential validity perspective of test bias is
unique because it is a subjective judgment based on ethical and
moral values, whereas the other interpretations of test bias are
based onand can be evaluated withstatistical data.
Those who make the argument that standardized tests that
produce negative consequences for some segments of society are
making the argument that using the tests is unfair. Because of the
moral nature of fairness, the question of whether it is fair to use a
standardized test even a high-quality, professionally developed
standardized test that exhibits no DIF or differential predictive
validity and has the same factor structure for all examinees
cannot be settled scientifically. Rather, such judgments are best
made by society through the public mechanisms that exist in
democratic nations to make decisions where competing values
among people lead to tensions. These mechanisms include state
and federal legislatures, the ballot box, the court system, and the
public meetings of elected and appointed officials in which administrative rules are created. Expert analysis of these public
discussions (e.g., Buckendahl & Hunt, 2005; Kaplan & Saccuzzo,
2009; Phelps, 2003; Phillips, 2000; Phillips & Camara, 2006;
Schafer, 2000; Ward, 2000) often leads to fascinating understandings of societys current and past values as they relate to standardized testing especially in employment and education.
We believe that it is easy to make a very compelling case that it
is unfair to use a test to gather information to determine whether a

577

person goes to college, is diagnosed with a mental illness, gets

hired for a job, graduates from high school, and so forth. However,
wealong with many other testing experts (e.g., AERA et al.,
1999; Camilli, 2006; Kaplan & Saccuzzo, 2009)strongly believe
that test bias and fairness are not synonymous. Mainstream professional opinion about this topic is that test bias is a purely
technical term defined by Interpretations 2 4 concerned with
actual test score interpretation (Popham, 1997), which is the crux
of the AERA et al. (1999) definition of test bias that we subscribe
to. Therefore, moral or ethical judgments concerning the unfairness of test use are not evidence that the test is biased, although the
judgments may be used to construct persuasive arguments that a
test is unfair. This important distinction leads to the widely recognized (among psychometricians) fact that an absence of test bias
is a prerequisite for fairness (AERA et al., 1999; Kaplan &
Saccuzzo, 2009). However, it is possible for a test to demonstrate
no statistical evidence of bias, yet still be unfair in the eyes of some
members of society.

But Look at the Items! Its Obvious That

Theyre Biased!
Some critics claim that many standardized tests are biased use
test items themselves as the basis for their arguments against tests.
These people claim that the test items ask questions about cultural
information that diverse examinees cannot relate to and/or are
much less likely to be exposed to. For example, in the civil rights
case Larry P. v. Wilson Riles (1979) the plaintiffs lawyerswho
were advocating for African American students in California public schools examined the revised version of the Wechsler Intelligence Scale for Children (WISCR) and found items that they
claimed were biased against African American schoolchildren.
According to Elliott (1987), one of these items was, Who wrote
Romeo and Juliet? The lawyers argued that such cultural information was irrelevant to diverse examinees who may have had
fewer opportunities than White students to learn the answer to this
test question. Therefore, it was discriminatory to use the item to
test students intelligence. In the technical language of testing, it
would be said that this item would lack face validity for testing the
intelligence of diverse children. Such face validity arguments are
made solely on the basis of whether an item appears to examine
what test creators purport it examines (Carlson & Geisinger, 2009;
Gottfredson, 2009; Kane, 2006).
Such an argument about test bias is so self-evident to critics of
standardized tests that merely the existence of such items is
enough evidence that the tests are biased (Elliott, 1987). However,
this is a problematic stance to take for multiple reasons. First, such
items are somewhat rare on tests, and like the number of items
displaying DIF, are not numerous enough to explain the score
differences that exist among demographic groups on most standardized tests (Flaugher, 1978). Second, such arguments against
standardized tests based solely on face validity judgments are
anecdote wholly unsupported by evidence (Elliott, 1987, p. 123).
Clarizio (1979), agreed stating, Subjective, armchair analysis of
biasthe primary method used by critics who charge cultural
biasis no substitute for item statistics (p. 81; see also Mercer,
1979). We contend that critics who make arguments about test bias
based on single items forget Aristotles maxim that, One swallow
does not a summer make (as quoted in Johansen, 1998, p. 382).

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

578

WARNE, YOON, AND PRICE

In other words, a single incidentin this case one unfair itemis

not indicative of a trend of a biased test (Carlson & Geisinger,
2009; Reschly, 1980).
Third, it is notoriously difficult to pick out items that function
differently across groups through face validity judgments. In
one well known attempt at trying to eyeball biased items,
Judge John Grady in the PASE v. Hannon case examined every
item on the WISCR and Stanford-Binet Intelligence Scale and
found seven items on the WISCR to be biased and one item on
the Stanford-Binet to exhibit bias (Bersoff, 1981). In later DIF
analyses, however, it was found that none of the items that
Judge Grady had identified as being biased actually exhibited
DIF (Koh, Abbatiello, & McLoughlin, 1984). This demonstrates a well-known principle of psychometrics: subjective
face validity judgments are prone to erroneous conclusions
(Reschly, 1980) and that, Face validity . . . does not offer
evidence to support conclusions drawn from test scores (Kaplan & Saccuzzo, 2009, p. 136).
Although subjective judgments about bias on the basis of face
validity are inadequate for identifying test bias, examinations of
experts are still an important part of the process to make tests
as fair as possible for all examinees. Current ethical standards
for testing demand that, Test developers should strive to
identify and eliminate language, symbols, words, phrases, and
content that are generally regarded as offensive by members of
racial, ethnic, gender, or other groups . . . (AERA et al., 1999,
p. 82). The most typical method of eliminating such offensiveness is through a sensitivity panel, which is an independent
group of trained reviewers who examine every component of a
test for language that might be offensive, threatening, controversial, stereotypical, sexist, insensitive, condescending, inflammatory, or distracting to any subgroup of examinees (Camilli, 2006; Reynolds, 2000; Reynolds & Lowe, 2009).
Sensitivity panels also look for language or vocabulary that
would have a different meaning for any subgroup (AERA et al.,
1999) and therefore change the meaning of the item for some
examinees. Sensitivity panels are usually made up of experts
who are often themselves members of diverse racial, ethnic,
gender, religious, or disability groups, and are carefully trained
in identifying problematic language that would disadvantage an
examinee.
Although not able to identify items that show DIF (Flaugher,
1978; Reynolds, 2000), sensitivity panels are an important
aspect of test development that ensure that tests are as fair as
possible for many groups of examinees. This is because sensitivity panels ensure that examinees are not exposed to distracting languagean important function in our opinion because we
believe that every examinee has a right to be tested free from
unnecessary distraction. We also think that any test that contains insensitive language is unfair, regardless of what statistical analyses may say about the items. Thankfully, the prevailing
ethical standards of the testing field agree with us (AERA et al.,
1999), and it is probably impossible today for a commercial
testing company to sell an educational or employment test today
without all components of a test being screened by a sensitivity
panel (for examples of documentation of the work of sensitivity
panels, please see ACT, 2007; Forsyth et al., 2003; and Hoover
et al., 2003).

Benefits for Diverse Populations of

Standardized Testing
Now that we have established various interpretations of the
phrase test bias and discussed the strengths and weaknesses of
each, we believe that it is important to discuss the benefits of
standardized testing for diverse examinees because of the tendency
of some advocates for diverse populations to either dismiss standardized tests as inherently biased (e.g., Ford, 2003; Richert, 2003;
Smith, 2003) or as tools of powerful racial and economic groups to
maintain their dominance over diverse populations (e.g., Carter &
Goodwin, 1994; Mercer, 1979; Moss, 2008). We reject these
positionsas do other testing experts (e.g., Gottfredson, 2009;
Reynolds, 2000)and believe that standardized tests can be quite
useful in promoting the efforts of diverse populations in psychological, employment, and educational testing for two reasons: (a)
standardized tests can be used as a measurement of social equality,
and (b) standardized tests are less problematic than other alternatives. Each of these issues will be discussed briefly in the succeeding subsections.

Standardized Tests Can Measure Social Inequality

Many expertsincluding ourselvesare greatly concerned
about the inequalities that exist in education and economic
outcomes among different groups in society. We also believe
that education can be a means of reducing inequalities among
racial, gender, and other demographic groups. Standardized
tests can be a yardstick to help advocates understand the degree
of inequality among groups and how much progress society has
made toward reducing these inequalities (Flaugher, 1978). As
recognized by the authors of the current ethical standards for
educational and psychological testing, Properly designed and
used, tests can and do further societal goals of fairness and
equality of opportunity (AERA et al., 1999, p. 73).
An example of the use of standardized tests as a measurement
of inequality is the National Assessment of Educational Progress (NAEP), a testing program administered by the federal
government. NAEP is important to educational researchers and
policymakers because it is the only standardized test that produces results which permit comparisons across demographic
groups and state lines (Lane et al., 2009). Moreover, federal law
makes NAEP a sanctioned instrument for measuring educational progress (Koretz, 2003). Recent NAEP results displayed
in Figure 4 show that since 1971 reading achievement has
increased for White, Hispanic, and Black students. As encouraging as these results are, a more encouraging fact is that in that
time the achievement gap between White students and the other
two groups on NAEP reading assessments has dramatically
narrowed (National Center for Educational Statistics, 2009).
Similar results have also been found on NAEP mathematics
scores since 1973 (National Center for Educational Statistics,
2009). Although progress has been made, it is clear from NAEP
data that substantial score gaps in achievement still exist and
that equality in achievement has not yet been obtained. This is
information that would not be available without standardized
tests.

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

EXPLORING TEST BIAS

579

Figure 4. Trend in WhiteBlack National Assessment of Educational Progress (NAEP) reading average scores
and score gaps for 9-year-old students. Source: National Center for Education Statistics (2009, p. 14).

Standardized Tests Are Less Problematic

Than Alternatives
Some critics of standardized tests wish to replace the tests with
alternatives (such as diagnostic interviews, teacher ratings, grade point
averages, and portfolio assessments) that they claim would be less
discriminatory against diverse examinees. The quality of these alternatives is often greatly overstated and replacing standardized tests
with them would cause far more problems than they would solve
(Clarizio, 1979; Flaugher, 1978; OBoyle & McDaniel, 2009; Gottfredson, 2009; Phelps, 2003). For example, some claim that using
high school grade point averages (GPAs) alone in college admissions
would be fairer than SAT scores. However, research has shown that
using high school GPA as a predictor for college outcomes (such as
GPA) results in greater prediction errors than using SAT scores as a
predictor (Mattern, Shaw, & Kobrin, 2011; Zwick, 2007). Moreover,
the recent trends in grade inflation (e.g., Camara, 2009; Posselt et al.,
2012) and the strong tendency for teachers to be subjective and
idiosyncratic in their awarding of grades (e.g., Cizek, Fitzgerald, &
Rachor, 1995) have been well documented. These problemsand
others that have been explained at length elsewhere (e.g., Camara,
2009; Zwick, 2006, 2007)are much less severe with standardized
tests than with any other alternative that has yet been suggested
(Camara, 2009; Clarizio, 1979; Phelps, 2003; Reynolds & Lowe,
2009). Stated quite simply, standardized tests are the most efficient,
cheapest, and least biased way of assessing a large number of people.
This has been true since the 1920s.

Conclusion
We now end this article the same way we began itwith an
excerpt from Through the Looking-Glass:
When I use a word, Humpty Dumpty said, in rather scornful tone,
it means just what I choose it to meanneither more nor less.
The question is, said Alice, whether you can make words mean so
many different things. (Carroll, 1871/1917, p. 99)

Our intention in this article is not to play the role of Humpty

Dumpty and arbitrarily choose what test bias does and does not mean.

Instead, we hope in this article to clarify the meaning of a phrase that

often receives extra pay because it means a great deal (Carroll,
1871/1917, p. 100). As defined by AERA et al., (1999), test bias
exists when test scores have different meanings for different groups of
examinees. As we have demonstrated, there are at least five common
interpretations for the phrase test bias: (1) mean score differences
between groups, (2) differential predictive validity, (3) differential
item functioning, (4) differences in item factor structure, and (5)
consequences of a test that disadvantage members of some demographic groups. This multitude of interpretations introduces a degree
of impenetrability to conversations on the topic of test bias which we
hope to clarify.
During the course of this article we have shown that some of these
meanings of test bias are not supported: interpretation (1) is not
supported by psychometricians because sometimes group scores differences are expected, while interpretation (5) is not concerned with
test score meaning, so some experts do not consider it a manifestation
of test bias. However, interpretations (2), (3), and (4) are potential
manifestations of test bias (as defined by AERA et al., 1999) because
they are possible ways that differences in score meaning can be
detected statistically. We hope that we have explained the merits and
drawbacks of each one and have added to the conversation about
testing and will foster a more productive conversation about test use
in society.
We encourage readers to use the information in this article for two
purposes: first, to be more careful in their use of the phrase test bias.
Although test bias does have each of the five meanings we have
elucidated, haphazard use of the phrase introduces confusion. Careless
or vague use of language is a sign of sloppy thinking (Woodford,
1967), which can only muddle dialogues among professionals about
tests and their uses. We encourage readers to be precise with their
language and to perhaps use terms like differential predictive validity, differential item functioning, mean group differences, and so
forth, in leiu of the term test bias. Second, we hope that readers of
this article will use the explanations we have given to use tests to
advocate for diverse groups in society in a thoughtful and productive
way. We believe that such advocacy efforts would be far more
productive than attacking standardized tests.

WARNE, YOON, AND PRICE

580

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

References
ACT. (2007). The ACT technical manual. Retrieved from http://www.act
.org/aap/pdf/ACT_Technical_Manual.pdf
Allen, M. J., & Yen, W. M. (1979). Introduction to measurement theory.
Long Grove, IL: Waveland Press.
American Educational Research Association, American Psychological Association, & National Council on Measurement in Education. (1999).
Standards for educational and psychological testing. Washington, DC:
American Educational Research Association.
Andrich, D., & Hagquist, C. (2012). Real and artificial differential item
functioning. Journal of Educational and Behavioral Statistics, 37, 387
416. doi:10.3102/1076998611411913
Beaujean, A. A., McGlaughlin, S. M., & Margulies, A. S. (2009). Factorial
validity of the Reynolds Intellectual Assessment Scales for referred students. Psychology in the Schools, 46, 932950. doi:10.1002/pits.20435
Benson, N., Hulac, D. M., & Kranzler, J. H. (2010). Independent examination of the Wechsler Adult Intelligence ScaleFourth Edition
(WAIS-IV): What does the WAIS-IV measure? Psychological Assessment, 22, 121130. doi:10.1037/a0017767
Bersoff, D. N. (1981). Testing and the law. American Psychologist, 36,
10471056. doi:10.1037/0003-066X.36.10.1047
Borsboom, D. (2006). The attack of the psychometricians. Psychometrika,
71, 425 440. doi:10.1007/s11336-006-1447-6
Buckendahl, C. W., & Hunt, R. (2005). Whose rules? The relation between
the rules and law of testing. In R. P. Phelps (Ed.), Defending
standardized testing (pp. 147158). Mahwah, NJ: Erlbaum.
Camara, W. J. (2009). College admission testing: Myths and realities in an
age of admissions hype. In R. P. Phelps (Ed.), Correcting fallacies about
educational and psychological testing (pp. 147180). Washington, DC:
American Psychological Association. doi:10.1037/11861-004
Camilli, G. (2006). Test fairness. In R. L. Brennan (Ed.), Educational
measurement (4th ed., pp. 221256). Westport, CT: Praeger.
Carlson, J. F., & Geisinger, K. F. (2009). Psychological diagnostic testing:
Addressing challenges in clinical applications of testing. In R. P. Phelps
(Ed.), Correcting fallacies about educational and psychological testing
(pp. 67 88). Washington, DC: American Psychological Association.
doi:10.1037/11861-002
Carroll, L. (1871/1917). Through the looking-glass and what Alice found
there. New York, NY: Rand McNally.
Carter, R. T., & Goodwin, A. L. (1994). Racial identity and education.
Review of Research in Education, 20, 291336. doi:10.2307/1167387
Cizek, G. J., Fitzgerald, S. M., & Rachor, R. A. (1995). Teachers assessment practices: Preparation, isolation, and the kitchen sink. Educational
Assessment, 3, 159 179. doi:10.1207/s15326977ea0302_3
Clarizio, H. F. (1979). In defense of the IQ test. School Psychology Review,
8, 79 88.
Cleary, T. A. (1968). Test bias: Prediction of Negro and White students in
integrated colleges. Journal of Educational Measurement, 5, 115124.
doi:10.1111/j.1745-3984.1968.tb00613.x
Cleary, T. A., & Hilton, T. L. (1968). An investigation of item bias.
Educational and Psychological Measurement, 28, 6175. doi:10.1177/
001316446802800106
Cleary, T. A., Humphreys, L. G., Kendrick, S. A., & Wesman, A. (1975).
Educational uses of tests with disadvantaged students. American Psychologist, 30, 15 41. doi:10.1037/0003-066X.30.1.15
Crocker, L., & Algina, J. (2008). Introduction to classical and modern test
theory. Mason, OH: Cengage Learning.
Cronbach, L. J. (1975). Five decades of public controversy over mental testing.
American Psychologist, 30, 114. doi:10.1037/0003-066X.30.1.1
Cronbach, L. J. (1980). Selection theory for a political world. Public
Personnel Management, 9, 3750.
Dolan, C. V. (2000). Investigating Spearmans hypothesis by means of
multi-group confirmatory factor analysis. Multivariate Behavioral Research, 35, 2150. doi:10.1207/S15327906MBR3501_2

Elliott, R. (1987). Litigating intelligence: IQ tests, special education, and

social science in the courtroom. Dover, MA: Auburn House.
Embretson, S. E., & Reise, S. P. (2000). Item response theory for psychologists. Mahwah, NJ: Erlbaum.
Flaugher, R. L. (1978). The many definitions of test bias. American
Psychologist, 33, 671 679. doi:10.1037/0003-066X.33.7.671
Flowers, L. A. (2008). Racial differences in the impact of participating in
Advanced Placement programs on educational and labor market outcomes. Educational Foundations, 22, 121132.
Ford, D. Y. (2003). Equity and excellence: Culturally diverse students in
gifted education. In N. Colangelo & G. A. Davis (Eds.), Handbook of
gifted education (pp. 506 520). Boston, MA: Allyn & Bacon.
Forsyth, R. A., Ansley, T. N., Feldt, L. S., & Alnot, S. D. (2003). Iowa
Tests of Educational Development guide to research and development.
Itasca, IL: Riverside Publishing.
Goodman, D., & Hambleton, R. K. (2005). Some misconceptions about
large-scale educational assessments. In R. P. Phelps (Ed.), Defending
standardized testing (pp. 91110). Mahwah, NJ: Erlbaum.
Gorsuch, R. L. (2003). Factor analysis. In J. A. Schinka & W. F. Velicer
(Eds.), Handbook of psychology: Vol. 2. Research methods in psychology (pp. 143164). Hoboken, NJ: Wiley.
Gottfredson, L. S. (1994). Egalitarian fiction and collective fraud. Society,
31, 5359. doi:10.1007/BF02693231
Gottfredson, L. S. (1997). Why g matters: The complexity of everyday life.
Intelligence, 24, 79 132. doi:10.1016/S0160-2896(97)90014-3
Gottfredson, L. S. (2009). Logical fallacies used to dismiss the evidence on
intelligence testing. In R. P. Phelps (Ed.), Correcting fallacies about
educational and psychological testing (pp. 11 65). Washington, DC:
American Psychological Association. doi:10.1037/11861-001
Gould, S. J. (1981). The mismeasure of man. New York, NY: Norton.
Hoover, H. D., Dunbar, S. B., Frisbie, D. A., Oberley, K. R., Ordman,
V. L., Naylor, R. J., . . . Shannon, G. P. (2003). Iowa Tests of Basic Skills
guide to research and development. Itasca, IL: Riverside Publishing.
Hunter, J. E., & Schmidt, F. L. (2000). Racial and gender bias in ability and
achievement tests: Resolving the apparent paradox. Psychology, Public
Policy, and Law, 6, 151158. doi:10.1037/1076-8971.6.1.151
Jensen, A. R. (1998). The g factor: The science of mental ability. Westport,
CT: Praeger.
Johansen, K. F. (1998). A history of ancient philosophy from the beginnings to Augustine. New York, NY: Routledge.
Kane, M. T. (2006). Validation. In R. L. Brennan (Ed.), Educational
measurement (4th ed., pp. 17 64). Westport, CT: Praeger.
Kane, M. T., & Mroch, A. A. (2010). Modeling group differences in OLS
and orthogonal regression: Implications for differential validity studies.
Applied Measurement in Education, 23, 215241. doi:10.1080/
08957347.2010.485990
Kaplan, R. M., & Saccuzzo, D. P. (2009). Psychological testing: Principles, applications, and issues (7th ed.). Belmont, CA: Wadsworth.
Koh, T.-h., Abbatiello, A., & McLoughlin, C. S. (1984). Cultural bias in
WISC subtest items: A response to Judge Gradys suggestion in relation
to the PASE case. School Psychology Review, 13, 89 94.
Koretz, D. (2003). Using multiple measures to address perverse incentives
and score inflation. Educational Measurement: Issues and Practice, 22,
18 26. doi:10.1111/j.1745-3992.2003.tb00124.x
Lane, S., Zumbo, B. D., Abedi, J., Benson, J., Dossey, J., Elliott, S. N., . . .
Willhoft, J. (2009). Prologue: An introduction to the evaluation of
NAEP. Applied Measurement in Education, 22, 309 316. doi:10.1080/
08957340903221436
Lee, J. (2002). Racial and ethnic achievement gap trends: Reversing the
progress toward equity? Educational Researcher, 31, 312. doi:10.3102/
0013189X031001003
Li, H., Lee, D., Pfeiffer, S. I., Kamata, A., Kumtepe, A. T., & Rosado, J.
(2009). Measurement invariance of the Gifted Rating ScalesSchool

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

EXPLORING TEST BIAS

Form across five cultural groups. School Psychology Quarterly, 24,
186 198. doi:10.1037/a0017382
Linn, R. L., & Drasgow, F. (1987). Implications of the Golden Rule
settlement for test construction. Educational Measurement: Issues and
Practice, 6, 1317. doi:10.1111/j.1745-3992.1987.tb00405.x
Lohman, D. F. (2005). Review of Naglieri and Ford (2003): Does the
Naglieri Nonverbal Ability Test identify equal proportions of highscoring White, Black, and Hispanic students? Gifted Child Quarterly,
49, 19 28. doi:10.1177/001698620504900103
Larry P. v. Riles, 495 F. Supp. 926 (N. D. Cal. 1979).
Maruyama, G. (2012). Assessing college readiness: Should we be satisfied
with ACT or other threshold scores? Educational Researcher, 41, 252
261. doi:10.3102/0013189X12455095
Mattern, K. D., Shaw, E. J., & Kobrin, J. L. (2011). An alternative
presentation of incremental validity: Discrepant SAT and HSGPA performance. Educational and Psychological Measurement, 71, 638 662.
doi:10.1177/0013164410383563
McDonald, R. P. (1999). Test theory: A unified treatment. Mahwah, NJ: Erlbaum.
Mensch, E., & Mensch, H. (1991). The IQ mythology: Class, race, gender,
and inequality. Carbondale, IL: Southern Illinois University Press.
Mercer, J. R. (1979). In defense of racially and culturally nondiscriminatory assessment. School Psychology Review, 8, 89 115.
Meredith, W. (1993). Measurement invariance, factor analysis and factorial
invariance. Psychometrika, 58, 525543. doi:10.1007/BF02294825
Millsap, R. E., & Everson, H. T. (1993). Methodology review: Statistical
approaches for assessing measurement bias. Applied Psychological Measurement, 17, 297334. doi:10.1177/014662169301700401
Millsap, R. E., & Kwok, O.-M. (2004). Evaluating the impact of partial
factorial invariance on selection in two populations. Psychological
Methods, 9, 93115. doi:10.1037/1082-989X.9.1.93
Morgan, P. L., Farkas, G., Hillemeier, M. M., & Maczuga, S. (2012). Are
minority children disproportionately represented in early intervention
and early childhood special education? Educational Researcher, 41,
339 351. doi:10.3102/0013189X12459678
Moss, G. (2008). Diversity study circles in teacher education practice: An
experiential learning project. Teaching and Teacher Education, 24,
216 224. doi:10.1016/j.tate.2006.10.010
National Center for Educational Statistics. (2009). NAEP 2008 trends in
academic progress (NCES Report 2009 479). Retrieved from http://
nces.ed.gov/nationsreportcard/pdf/main2008/2009479.pdf
Navas-Ara, M. J., & Gmez-Benito, J. (2002). Effects of ability scale
purification on the identification of dif. European Journal of Psychological Assessment, 18, 9 15. doi:10.1027//1015-5759.18.1.9
Neisser, U., Boodoo, G., Bouchard, T. J., Boykin, A. W., Brody, N., Ceci,
S. J., . . . Urbina, S. (1996). Intelligence: Knowns and unknowns.
American Psychologist, 51, 77101. doi:10.1037/0003-066X.51.2.77
Nichols, J. D. (2003). Prediction indicators for students failing the State of
Indiana high school graduation exam. Preventing School Failure, 47,
112120. doi:10.1080/10459880309604439
OBoyle, E. H., Jr., & McDaniel, M. A. (2009). Criticisms of employment
testing: A commentary. In R. P. Phelps (Ed.), Correcting fallacies about
educational and psychological testing (pp. 181197). Washington, DC:
American Psychological Association. doi:10.1037/11861-005
Olszewski-Kubilius, P., & Lee, S.-Y. (2011). Gender and other group differences
in performance on off-level tests: Changes in the 21st century. Gifted Child
Quarterly, 55, 5473. doi:10.1177/0016986210382574
Phelps, R. P. (2003). Kill the messenger: The war on standardized testing.
New Brunswick, NJ: Transaction Publishers.
Phillips, S. E. (2000). GI Forum v. Texas Education Agency: Psychometric
evidence. Applied Measurement in Education, 13, 343385. doi:
10.1207/S15324818AME1304_04
Phillips, S. E., & Camara, W. J. (2006). Legal and ethical issues. In R. L.
Brennan (Ed.), Educational measurement (4th ed., pp. 733755). Westport, CT: Praeger.

581

Popham, W. J. (1997). Consequential validity: Right concernwrong concept. Educational Measurement: Issues and Practice, 16, 9 13. doi:
10.1111/j.1745-3992.1997.tb00586.x
Posselt, J. R., Jaquette, O., Bielby, R., & Bastedo, M. N. (2012). Access
without equity: Longitudinal analyses of institutional stratification by
race and ethnicity, 19722004. American Educational Research Journal, 49, 1074 1111. doi:10.3102/0002831212439456
Reise, S. P., Widaman, K. F., & Pugh, R. H. (1993). Confirmatory factor
analysis and item response theory: Two approaches for exploring measurement invariance. Psychological Bulletin, 114, 552566. doi:
10.1037/0033-2909.114.3.552
Reschly, D. J. (1980). Psychological evidence in the Larry P. opinion: A
case of right problemwrong solution? School Psychology Review, 9,
123135.
Reschly, D. J., & Sabers, D. L. (1979). Analysis of test bias in four groups
with the regression definition. Journal of Educational Measurement, 16,
19. doi:10.1111/j.1745-3984.1979.tb00080.x
Reynolds, C. R. (1980). An examination of bias in a preschool battery
across race and sex. Journal of Educational Measurement, 17, 137146.
doi:10.1111/j.1745-3984.1980.tb00822.x
Reynolds, C. R. (2000). Why is psychometric research on bias in mental
testing so often ignored? Psychology, Public Policy, and Law, 6, 144
150. doi:10.1037/1076-8971.6.1.144
Reynolds, C. R., & Lowe, P. A. (2009). The problem of bias in psychological assessment. In C. R. Reynolds & T. B. Gutkin (Eds.), The
handbook of school psychology (pp. 332374). New York, NY: Wiley.
Richert, E. S. (2003). Excellence with justice in identification and programming. In N. Colangelo & G. A. Davis (Eds.), Handbook of gifted
education (3rd ed., pp. 146 158). Boston, MA: Allyn & Bacon.
Roid, G. H. (2003). Stanford-Binet Intelligence Scales, fifth edition, technical manual. Itasca, IL: Riverside Publishing.
Rushton, J. P., & Jensen, A. R. (2005). Thirty years of research on race
differences in cognitive ability. Psychology, Public Policy, and Law, 11,
235294. doi:10.1037/1076-8971.11.2.235
Salend, S. J., Garrick Duhaney, L. M., & Montgomery, W. (2002). A
comprehensive approach to identifying and addressing issues of disproportionate representation. Remedial and Special Education, 23, 289
299. doi:10.1177/07419325020230050401
Santor, D. A., Ramsay, J. O., & Zuroff, D. C. (1994). Nonparametric item
analyses of the Beck Depression Inventory: Evaluating gender item bias
and response option weights. Psychological Assessment, 6, 255270.
doi:10.1037/1040-3590.6.3.255
Schafer, W. D. (2000). GI Forum v. Texas Education Agency: Observations
for states. Applied Measurement in Education, 13, 411 418. doi:
10.1207/S15324818AME1304_07
Schmitt, A. P., & Dorans, N. J. (1990). Differential item functioning for
minority examinees on the SAT. Journal of Educational Measurement,
27, 67 81. doi:10.1111/j.1745-3984.1990.tb00735.x
Schmitt, N., & Kuljanin, G. (2008). Measurement invariance: Review of
practice and implications. Human Resource Management Review, 18,
210 222. doi:10.1016/j.hrmr.2008.03.003
Sireci, S. G. (2005). The most frequently unasked questions about testing.
In R. P. Phelps (Ed.), Defending standardized testing (pp. 111121).
Mahwah, NJ: Erlbaum.
Smith, R. A. (2003). Race, poverty, & special education: Apprenticeships
for prison work. Poverty & Race, 12, 1 4.
Swaminathan, H., & Rogers, H. J. (1990). Detecting differential item
functioning using logistic regression procedures. Journal of Educational
Measurement, 27, 361370. doi:10.1111/j.1745-3984.1990.tb00754.x
Terman, L. M. (1928). The influence of nature and nurture upon intelligence scores: An evaluation of the evidence in Part I of the 1928
Yearbook of the National Society for the Study of Education. Journal of
Educational Psychology, 19, 362373. doi:10.1037/h0071466

582

WARNE, YOON, AND PRICE

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Ward, C. A. (2000). GI Forum v. Texas Education Agency: Implications for

state assessment programs. Applied Measurement in Education, 13,
419 426. doi:10.1207/S15324818AME1304_08_1
Warne, R. T. (2011). An investigation of measurement invariance across
genders on the Overexcitability QuestionnaireTwo. Journal of Advanced Academics, 22, 578 593. doi:10.1177/1932202X11414821
Woodford, F. P. (1967). Sounder thinking through clearer writing. Science,
156, 743745. doi:10.1126/science.156.3776.743
Yoakum, C. S., & Yerkes, R. M. (1920). Army mental tests. New York,
NY: Henry Holt and Company. doi:10.1037/11054-000
Young, J. W., Kane, M., Monfils, L., Li, C., & Ezzo, C. (2013, May).
Investigating differential validity and differential prediction using orthogonal regression analysis. Paper presented at annual meeting of the

American Educational Research Association, San Francisco,

CA.
Young, J. W. (2004). Differential validity and prediction: Race and sex
differences in college admissions testing. In R. Zwick (Ed.), Rethinking
the SAT: The future of standardized testing in university admissions (pp.
289 302). New York, NY: Rouledge-Falmer. doi:10.4324/
9780203463932_Differential_Validity_and_Prediction
Zwick, R. (2006). Higher education admissions testing. In R. L. Brennan
(Ed.), Educational measurement (4th ed., pp. 647 679). Westport, CT:
Praeger.
Zwick, R. (2007). College admission testing. Retrieved from http://www
.nacacnet.org/research/PublicationsResources/Marketplace/Documents/
TestingWhitePaper.pdf

Members of Underrepresented Groups:

Reviewers for Journal Manuscripts Wanted
If you are interested in reviewing manuscripts for APA journals, the APA Publications
and Communications Board would like to invite your participation. Manuscript reviewers
are vital to the publications process. As a reviewer, you would gain valuable experience
in publishing. The P&C Board is particularly interested in encouraging members of
underrepresented groups to participate more in this process.
If you are interested in reviewing manuscripts, please write APA Journals at
Reviewers@apa.org. Please note the following important points:
To be selected as a reviewer, you must have published articles in peer-reviewed
journals. The experience of publishing provides a reviewer with the basis for preparing
a thorough, objective review.
To be selected, it is critical to be a regular reader of the five to six empirical journals
that are most central to the area or journal for which you would like to review. Current
knowledge of recently published research provides a reviewer with the knowledge base
to evaluate a new submission within the context of existing research.
To select the appropriate reviewers for each manuscript, the editor needs detailed
information. Please include with your letter your vita. In the letter, please identify
which APA journal(s) you are interested in, and describe your area of expertise. Be as
specific as possible. For example, social psychology is not sufficientyou would
need to specify social cognition or attitude change as well.
Reviewing a manuscript takes time (1 4 hours per manuscript reviewed). If you are
selected to review a manuscript, be prepared to invest the necessary time to evaluate
the manuscript thoroughly.
APA now has an online video course that provides guidance in reviewing manuscripts. To
learn more about the course and to access the video, visit http://www.apa.org/pubs/
authors/review-manuscript-ce-video.aspx.