Enhancing Test Items: Students' Insights on Modifying Items

1. Students’ Perceptions of Item Modifications: Using Cognitive Labs and QuestionnairesAndrew RoachPaper presented as part of “Design and Evaluating Modified Items for Students with Disabilities: Research Results” Coordinated SessionNCME 2009 Annual ConventionSan Diego, CA Special thanks to Cori Wixson, Tanya Talapatra, and Tamika LaSalle for their assistance in coding the think-aloud videos.

2. Objectives • To discuss the rationale for including students’ perceptions in research on test item development and modification. • To discuss possible research strategies for collecting student responses. • To present data from post-assessment questionnaires and cognitive lab studies.

3. Why collect student response data? Support from the Test Standards • “Questioning test takers about their performance strategies can yield evidence that enriches the definition of a construct…” (p. 12). • “Process studies involving examinees from different subgroups can assist in determining the extent to which capabilities irrelevant or ancillary to the construct may be differentially influencing (student) performance” (p. 12). • “Educational tests…may be advocated on the grounds that their use will improve student motivation….Where such claims are central to the rationale of testing, the direct examination of testing consequences necessarily assumes even greater importance” (p. 17).

5. Why collect student response data? • Item enhancements or modifications could be conceptualized as a form of educational intervention. • …Student perceptions are essential evidence about the acceptability of these assessment strategies. • Acceptability refers to an individual’s perceptions regarding the appropriateness, fairness, and reasonableness of an intervention (Kazdin, 1981).

6. Using Student Response Data: Applications to Test Item Modifications Cognitive Lab Study Post-Test Survey Test Development or Item Enhancement/ Modifications Additional modifications or enhancements based on results Field Test

7. Study #1: CAAVES Cognitive Lab--An Initial Application of Think-Aloud Methodology • Purpose: To evaluate the influence of test item modifications on students’ problem-solving and test-taking behaviors. • Our study involved three components: 1. Students completed a series of 16 assessment items (8 reading; 8 mathematics). 2. Students were asked to think aloud as they completed or solved these items. 3. We also asked follow-up questions about students’ perceptions of the assessment items.

8. Distribution of item modifications X = Item modifications used.

9. Sample Size by Sub-group

10. Method • We explained the think-aloud procedures, had the students restate their understanding of the process, and modeled thinking aloud on a practice item. • We used a script adapted from a study conducted by Johnstone, Bottsford-Miller, and Thompson (2006). • Students were prompted only when they were silent for 10 consecutive seconds. • If students verbalized infrequently, we reminded them to “keep thinking aloud” or “keep talking.” Otherwise we generally did not give encouragement or support.

11. Results: Data on Reading Items

12. Results: Data on Math Items

13. Results: Use of Visuals • Visuals in Reading Passages/Items • Most SWDs (67%) saw the visuals as being helpful and providing support on reading questions and passages. • 100% of the students without disabilities indicated the pictures made no difference. • Visuals/Graphs in Mathematics Items • Students with (50%) and without disabilities (67%) generally saw the visuals and graphs as being helpful and providing support. • …However, 33% of SWDs indicated that the visuals/graphs were distracting or made items harder.

14. Students’ Comments:Use of Visuals • The one talking about the $100 bills…well it showed me--and I was understanding—how it goes with what it was talking about, and I looked at it and it helped me even more.” -- Student with disability (eligible for AA-MAS)" • “When people do math, they're working on a sheet and what's the point of looking at a picture. It doesn't really help you. For example, on (questions) #1 and #2, those two pictures were really messing me up.” --Student with disability (not eligible for AA-MAS)

15. Results: Removing Answer Choices • Reading • SWDs (with one exception) perceived no difference in difficulty between items having 3 or 4 possible answers. • Conversely, 67% of the students without disabilities identified the 3-answer modification as making the reading items easier. • Mathematics • Students without disabilities (67%) and non-eligible SWDs (67%) generally indicated removing answer choice made the math items easier. • Some non-eligible students appeared to use the possible answer choices to help solve math items, but it was not clear that they used this same strategy in reading. • “If you didn't get the answer right the first time, you know you only had 3 choices to go back and look at, instead of 4.”—Student without disability

16. Results: Format of Analogies • Most students (including 2/3 of SWDs) found the traditional format for the analogy easier (i.e.,“meteor:space::dolphin:_______”). Some students indicated they had been taught analogies using this format and it was familiar to them. • This was supported by the results as SWDs correctly answered all the traditional analogy items. SWDs missed items with a modified analogy format (i.e., “meteor is to space as dolphin is to ___”) 40% of the time.

17. Study #2: Post-Test Survey • Original and modified versions of the 39 item tests were field tested experimentally using DEA’s online test delivery system. • A large sample of students (N = 755) in grade eight from the four states (AZ, HI, ID, and IN) participated in the study. • Sample was comprised of three groups: SWOD (n = 269), SWD-NE (n = 236), and SWD-E (n = 250). • Students received 13 items in each of three conditions: Original, Modified, and Modified with Reading Support. • After the test, students were presented with a follow-up survey that contained seven questions about their perceptions of particular item modifications.

18. Results: Relative Difficulty of Items • Most students reported the test had about the same difficulty all the way through (61% for reading; 46% for mathematics). • Some students reported the test was easier toward the beginning (19% for reading; 29% for mathematics), despite the fact that some students received the Modified or Modified with Reading Support conditions first. • Actual field test results showed decreases in student performance for each successive part across groups for both content areas, independent of the order of conditions (i.e., Original, Modified, or Modified with Reading Support).

19. Results: Relative Difficulty of Items • Fewer students in the SWD-E group reported the reading test was the same difficulty throughout (49% versus 71% of SWODs). • Fewer SWDs reported the mathematics test was the same difficulty all the way through (42% and 41% of students in the SWD-NE and SWD-E groups, respectively, compared to 54% of SWODs).

20. Results: Adding Visuals to Items • Reading Items: 62% of students in the SWD–E group reported the visuals provided helpful clues compared to 50% of students in SWD-NE and 44% of students in SWOD (44%) groups. • Mathematics Items: 58% of students in the SWD-E group reported visuals gave helpful clues, compared to 37% of students in the SWOD group and 44% of students in the SWD-NE group.

21. Results: Using bold font for key terms • We expected this modification to be most strongly endorsed by the SWD-E group, but fewer students in the eligible group reported bold type as helpful for vocabulary items (73%) compared to SWD-NEs (81%) and SWODs (84%). • Actual performance data indicated that for the 17 items with key vocabulary terms in bold type, difficulty was lower for the Modified condition than for the Original condition.

22. Results: Reading Support • More students in the SWD-E group reported reading support made the items easier (67% on the reading test, 68% on the mathematics test) compared to students in the SWOD group (41% for reading; 40% for mathematics). • Field test results in both content areas, however, indicated only small differences in student performance between the Modified condition and the Modified with Reading Support condition (effect sizes of .07 for reading and .05 for mathematics items).

23. Study #3: CMAADI Cognitive Lab Study • A replication and extension of the CAAVES study. • 60 students with and without disabilities in grades 5-8 and 10. • We collected data on… • Students’ mental effort/mental ease for each item. How hard did you have to work to answer the reading/math item above? 1 2 3 4 5 6 7 Not very hard Very hard • Students’ instructional experiences. • Students’ oral reading fluency.

25. Test Construction • 12 items per grade level from Arizona’s item pool were selected for inclusion in the study by the state’s item modification/writing team members. • Items were used to create two versions of each test, Forms A and B. • Each test version included 6 items in their original forms and 6 items that had been modified. • Using two versions of the test allowed us to make comparisons between behavior and responses on original (O) and modified (M) versions of each item at each of six grade levels (4th through 8th and 10th grades).

26. Sample: Demographics

28. Sample: Oral Reading Fluency

29. Reading Content Coverage by Grade

30. Math Content Coverage by Grade

31. Reading Test Outcome Measures by Condition & Group

32. Mathematics Test Outcome Measures by Condition & Group

33. Reading Mean total test scores by condition and group

34. Reading mean cognitive ease score by condition and group

35. Math mean total test scores by condition and group

36. Math mean cognitive ease score by condition and group

37. Students’ Perceptions of Tests and Items

38. Cognitive efficiency plot for grade 7 reading items #4-6

39. Take Away Ideas • Verbalizing about “automatized” procedures and skills (i.e. low DOK levels or p-values) is difficult, but many of our test items are at lower levels of cognitive complexity. • Follow-up questions may provided valuable information that make think-aloud data easier to understand and interpret (Branch, 2000; Fonteyn, Kuipers, & Grobe, 1993; Johnstone, Bottsford-Miller, & Thompson, 2006).

40. Take Away Ideas • SWDs often appeared unfamiliar with some concepts (e.g., percentages). In these cases, item modifications are unlikely to provide necessary support or facilitate access. • Reading fluency may be an issue for SWDs. In some cases, SWD’s slower rates of reading resulted in testing sessions that were almost twice as long as their peers. How could (or should) technology be used to address this barrier?

41. Take Away Ideas • To understand the students’ cognitive processing and problem solving behavior, researchers must understand: • The instructional/assessment task; • Individual participants’ knowledge about the task; and • How prior knowledge may affect processing and problem solving during the task (Pressley & Afflerbach, 1995).

42. References Branch, J. L. (2000). Investigating the information-seeking processes of adolescents: The value of using think-alouds and think-afters. Library and Information Science Research, 22(4), 371–392. Ericsson, K. A., & Simon, H. A. (1993). Protocol analysis: Verbal reports as data (Revised edition). Cambridge, MA: MIT Press. Johnstone, C. J., Bottsford-Miller, N. A., & Thompson, S. J. (2006). Using the think aloud method (cognitive labs) to evaulate test design for students with disabilities and English language learners (Technical Report 44). Minneapolis, MN: University of Minnesota, National Center on Educational Outcomes. Retrieved from the World Wide Web: http://education.umn.edu/NCEO/OnlinePubs/Tech44/. Pressley, M., & Afflerbach, P. (1995). Verbal protocols of reading: The nature of constructively responsive reading. Hillsdale, NJ: Lawrence Earlbaum.