ABSTRACT As part of a grant to recruit and retain multicultural and economically disadvantaged nursing students, funding was approved for a study analyzing multiple-choice nursing course exams for linguistic and cultural bias. Nineteen exams, for a total of 673 multiple-choice test items, were analyzed for this study. The categories of analysis used to identify flaws in the test items were: testwise flaws, irrelevant difficulty in stem, irrelevant difficulty in option, linguistic/structural bias, and cultural bias. Twenty-eight types of flaws occurred at least 10 times in the reviewed exams. Flaws from each of the categories are presented and discussed; specific examples are analyzed and revised.
In recent years an increasing number of culturally diverse students, including non-native speakers of English, have enrolled in nursing programs throughout the United States (1-11). Once they enter a nursing program, many of these students have difficulty succeeding academically (9,12,13). Language difficulties are often cited as a factor that contributes to student attrition (7,14).
Multiple-choice tests comprise an important means of assessing the mastery of content in nursing courses. In many courses, these tests also constitute a large percentage of the final grade. ESL (English as a second language) students, or non-native speakers of English, often do not perform well on multiple-choice exams (15). Commitment to diversity in the nursing profession, and a concomitant commitment to improving the success rate of multicultural and international students in nursing programs, should include "the reduction of biased items in nursing examinations, with the goal of making the evaluation process fair for all students" (15, p. 35).
What is bias in tests and how can it be determined? E. Julian, psychometrician at the National Council of State Boards of Nursing, states that "an item is ... free of bias if the probability of responding correctly, given total score, is the same for all sub-populations" (15, p. 36). The term given total score refers to statistical procedures that assess for the overall ability of groups of majority and minority examinees. If it is determined that "students of equal ability but [from] different ethnic groups do not have the same probability of answering the item correctly, the item may be biased" (15, p. 36).
Item bias occurs at two different levels: linguistic/structural and cultural. Linguistic bias refers to overly complex items that are not easily understood (16). Klisch defines structural bias as items that are "long, unclear, or contain awkward or misused grammatical construction[s]" (15, p. 36). Such items "contribute noise, rather than information, to the testing process" (E. Julian, 15, p. 36) and can interfere with the reader's understanding of the writer's intent.
Cultural bias refers to content in test items that is not equally available to all cultural groups. Mohan (17) defines cultural bias as the presence of culturally specific information, such that special cultural knowledge is necessary to answer certain test items correctly. Klisch defines a culturally biased item as one that "contains references to a particular culture and which, given total score, is answered incorrectly more frequently by examinees from another culture." She claims, however, that some culturally biased items should be retained if it is determined that the information is "essential to safe and effective nursing practice" (15, p. 37).
One common source of cultural bias in test items is humor. Haladyna (18) cautions against humor because it detracts from the purpose of the test and does little good. According to Klisch, "Many ESL students may not understand the attempted humor and could become confused further as to the meaning of the item" (15, p. 37). Furthermore, humor that depends on shared cultural knowledge may be culturally biased.
Background of Study The nursing program at the College of St. …