xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam a22 u 4500
controlfield tag 008 c20039999azu 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E11-00307
Educational policy analysis archives.
n Vol. 11, no. 9 (February 28, 2003).
Tempe, Ariz. :
b Arizona State University ;
Tampa, Fla. :
University of South Florida.
c February 28, 2003
Creating a system of accountability : the impact of instructional assessment on elementary children's achievement test scores / Samuel J. Meisels, Sally Atkins-Burnett, Yange Xue, Julie Nicholson, Donna DiPrima Bickel [and] Seung-Hee Son.
Arizona State University.
University of South Florida.
t Education Policy Analysis Archives (EPAA)
xml version 1.0 encoding UTF-8 standalone no
mods:mods xmlns:mods http:www.loc.govmodsv3 xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govmodsv3mods-3-1.xsd
mods:relatedItem type host
mods:identifier issn 1068-2341mods:part
mods:detail volume mods:number 11issue 9series Year mods:caption 20032003Month February2Day 2828mods:originInfo mods:dateIssued iso8601 2003-02-28
1 of 18 Education Policy Analysis Archives Volume 11 Number 9February 28, 2003ISSN 1068-2341 A peer-reviewed scholarly journal Editor: Gene V Glass College of Education Arizona State University Copyright 2003, the EDUCATION POLICY ANALYSIS ARCHIVES .Permission is hereby granted to copy any article if EPAA is credited and copies are not sold. EPAA is a project of the Education Policy Studies Laboratory. Articles appearing in EPAA are abstracted in the Current Index to Journals in Education by the ERIC Clearinghouse on Assessment and Evaluation and are permanently archived in Resources in Education .Creating a System of Accountability: The Impact of Instructional Assessment on Elementary Children's Achievement Test Scores Samuel J. Meisels Erikson Institute Sally Atkins-Burnett University of Michigan Yange Xue Julie Nicholson Palo Alto, CA Donna DiPrima Bickel University of Pittsburgh Seung-Hee Son University of Michigan
2 of 18Citation: Meisels, S. J., Atkins-Burnett, S., Xue, Y., Nicholson, J., Bickel, D. D., and Son, S-H. (2003, February 28). Creating a syste m of accountability: The impact of instructional assessment on elementary ch ildren's achievement test scores, Education Policy Analysis Archives, 11 (9). Retrieved [Date] from http://epaa.asu.edu/epaa/v11n9/.Abstract This study examined the trajectory of change in sco res on the Iowa Tests of Basic Skills (ITBS) of low-income, urban, third and fourth graders who had been enrolled in classrooms where the Work Sampling System (WSS), a curriculum-embedded performance assessment was used for at least three years. The ITBS scores of children expo sed to WSS were compared with those of students in a group of non-W SS contrast schools that were matched by race, income, mobility, school size, and number of parents in the home and to a comparison group of al l other students in the school district.Results indicated that students who were in WSS cla ssrooms displayed growth in reading from one year to the next that fa r exceeded the demographically matched contrast group as well as t he average change shown by all other students in the district. Childr en in WSS classrooms made greater gains in math than children in the oth er two groups, although the results were only marginally significa nt when compared with gains by the matched contrast group. The discu ssion concerns the complementarity of performance-based and normative tests in systems of accountability and the potential value of using a c urriculum-embedded assessment to enhance teaching, improve learning, a nd increase scores on conventional accountability examinations. More group-administered achievement testing is taki ng place in states and local school districts than ever before (Achieve, Inc., 2002). A ccording to a survey published in Education Week (Olson, 2001), every state has adopted mandatory t ests at one or more grades in elementary, middle, and high school, and 49 states have linked their academic standards to these tests. Moreover, legislation man dating annual testing in reading and mathematics for all children in grades 3 Â– 8 was re cently enacted by Congress. The tests in use nationwide are standards-based, bu t primarily norm-referenced; only 10 states report use of supplementary criterion-refere nced tests (Olson, 2001). The principal purpose of these tests is to ascertain the current status of student achievement, rather than to identify students in need of intervention or to det ermine appropriate instructional strategies. Only seven states provide extra funding for low-per forming schools and just nine states allocate funds for remediation of failing students. By the close of the year 2000, 18 states had made graduation contingent on student test perf ormance and an additional five states were about to begin administering exit exams. As ma ny as 27 states could withhold diplomas to students who fail to pass state account ability examinations by 2003 (Olson, 2001).In addition to the increased prevalence of such tes ts and the escalation of consequences or Â“stakesÂ” associated with them, the most notable cha nge in the assessments themselves
3 of 18concerns their alignment with curriculum standards. Never before in U. S. history have we witnessed such an explosion of attention to standar ds. The minimum competency testing movement of the 1980s that was inaugurated by the Nation At Risk report (National Commission on Excellence in Education, 1983) has be en overtaken by a focus on high standards of achievement inspired by the nation's e mbrace of the national educational goals of the 1990s.Ironically, one result of the standards-based refor m movement has been a heightened emphasis on high-stakes, group-administered, decont extualized testing practices. This is due to the linkage between standards-based reform and t he national political drive for accountability (Kohn, 2000). As Popham (2000) point s out, these tests lend themselves to the Â“score-boosting gameÂ” in which educators devote most of their energy to raising students' scores on conventional achievement tests. Rather than trying to improve student performance by enhancing instruction, this approach views high stakes testing as a policy tool to leverage learning (Firestone & Mayorowetz, 2000). In short, funding, availability of other resources, and state and local prestige are a ll devoted to improving student test scores by changing the curriculum to match more closely th e items or content standards of the assessments. This practice of Â“curriculum alignment Â” has the consequence of increasing test scores for some students, but often leaving general knowledge and mastery of curriculum domains virtually untouched (Amrein & Berliner, 200 2; Corbett & Wilson, 1991; Haney, 2000; McNeil, 2001). High-stakes testing has been s hown to impair the validity of the tests themselves as the test-taking experience becomes le ss a sampling of students' adaptive skills and higher-order thinking and more of an exercise i n rote memory and mastery of basic skills (Orfield & Kornhaber, 2001). In a study of 1 8 states with high-stakes testing programs, the learning levels of the students in all but one of the states were at the same level as before the testing policies were implemented (Amrein & Ber liner, 2002). The alternative to teaching to the test, or measure ment-driven instruction, is to transform instruction, but to do so in such a way that the st andards that are intended to serve as the basis of the tests inform instructional decisions a nd are incorporated into new forms of assessment. Curriculum-embedded performance assessm ents (see Baron & Wolf, 1996; Darling-Hammond, 1992) represent an instructional-d riven measurement in which students' actual classroom performance is evaluated in terms of standards-infused criteria. These criteria in turn suggest next steps in curriculum d evelopment which are consistent with advancing progress toward attainment of the defined standard. It is reasonable to assume that as students' learning improves, so will their score s on accountability examinations. Unfortunately, the research literature contains few studies of the impact of curriculum-embedded performance assessments on grou p-administered achievement test scores (Borko, Flory, & Cumbo, 1993; Falk & Darling -Hammond, 1993). Rather, most studies of the impact of testing provide data about conventional results-driven accountability testsÂ–Â–tests that rely on public reporting of perfo rmance data and utilization of these data for reward or sanction. This approach to assessment (wh ich can also be called Â“norm-referenced accountabilityÂ”) emphasizes the use of test data fo r instrumental purposesÂ–Â–purposes external to the classroomÂ–Â–rather than direct appli cation of the data to improve educational practice.In this paper we describe an alternative to typical conceptions of accountability. Instead of relying on either norm-referenced or performance-ba sed assessments in isolation, we suggest a complementary approach that incorporates both typ es of assessment. In short, consistent with recent federal and state initiatives, we sugge st the addition of performance assessments
4 of 18to conventional norm-referenced testing.This paper investigates whether students who are en rolled in classrooms in which a curriculum-embedded performance assessment is in us e will show greater gains on a conventional test used for accountability than stud ents who have not had exposure to the performance assessment. The research question this study poses is the following: Can ongoing, focused instructional assessments influenc e performance on group-administered achievement tests? A corollary to this question con cerns whether instructional and high-stakes assessments can be linked to create an accountability system that relies on both classroomand test-based information about student achievement.MethodProceduresThis study is part of a larger investigation that e valuated the validity of a curriculum-embedded performance assessmentÂ–Â–the Wor k Sampling System (WSS; Meisels, Jablon, Marsden, Dichtelmiller, & Dorfman, 1994, 2001)Â–Â–and its influences on teacher practices and children's achievement in the Pittsburgh Public Schools (PPS). The overall study included data from teachers, parents, and children. This paper focuses on the impact of WSS on the trajectory of children's chang e in scores on a group-administered achievement test (the Iowa Tests of Basic Skills [I TBS]) from grade 3 to grade 4. Related studies focus on other aspects of the validity of W SS (Meisels, Bickel, Nicholson, Xue, & Atkins-Burnett, 2001), parental reactions to WSS (M eisels, Xue, Bickel, Nicholson, & Atkins-Burnett, 2001), and teachers' views of the c onsequences of using WSS (Nicholson, 2000).The Work Sampling System (Meisels, 1997; Meisels et al., 1994; 2001) is a curriculum-embedded performance assessment designed for children from preschool through grade 5. WSS is comprised of developmental guidelines and checklists, portfolios, and summary reports. It uses teachers' perceptions of their students in actual classroom situations as the data of assessment while simultan eously informing, expanding, and structuring those perceptions. It involves students and parents in the learning and assessment process and it makes possible a systematic document ation of what children are learning and how teachers are teaching. This approach to perform ance assessment allows teachers the opportunity to learn about children's processes of learning by documenting children's interactions with materials, adults, and peers in t he classroom environment and using this documentation as the basis for evaluating children' s achievements and planning future educational interventions through comparisons with standards-based guidelines. Evidence of the reliability and validity of Work Sampling is av ailable in Meisels, Bickel, Nicholson, Xue, and Atkins-Burnett (2001) and Meisels, Liaw, D orfman, and Nelson (1995). Further descriptions of WSS are found in Meisels (1996, 199 7) and Meisels, Dorfman, and Steele (1995).The Iowa Test of Basic Skills (ITBS; University of Iowa and Riverside Publishing, 1994) is a group-administered achievement test designed to m onitor year-to-year achievement differences in students from K Â– Grade 12. Norming for the 1993 edition (Form K) was completed on 136,934 individuals. KR-20 internal co nsistency ratings are reported at >.84 for all reading and mathematics subtests. The third and fourth graders in this study were administered the reading comprehension and the voca bulary subtests of the ITBS Survey
5 of 18Battery (Levels 7 11). The Reading total score in cludes both the comprehension and vocabulary scores. The Mathematics total score incl udes items tapping computation, estimation, calculation, problem solving, and data interpretation (Levels 7 11). Developmental Standard Scores (DSS) were calculated using the raw score/DSS conversion tables provided by ITBS (Hoover et al., 1993). The DSS scales allow a comparison of both status and growth even when children take different level tests. The mathematics computation score is combined with the total score to determine the DSS for mathematics plus computation.DesignThis report describes the results of a natural expe riment in which two groups of demographically matched students who were administe red the ITBS were compared with one another and with all other students in their gr ade levels in the PPS who were also administered the ITBS. One of the target groups was composed of students who had been exposed to WSS for three years prior to being admin istered the ITBS; all other students had no experience with WSS. The ITBS was administered t o all children by their teachers in the spring of 1997 (end of third grade) and the spring of 1998 (end of fourth grade). A total of 2708 students received the ITBS reading assessments in both 1997 and 1998 and 2564 students took the ITBS math assessments in both yea rs. Data were coded and reported by school district personnel.A longitudinal design was selected because the scho ols using WSS were among the lowest-performing schools on the ITBS in the distri ct. It was evident to us that comparisons of absolute scores at the end of third or fourth gr ade would only confirm that the children in these low-scoring schools were still low scoring, d espite potential improvements in comparison with children from the same schools but from different age cohorts. The longitudinal design focuses on the trajectory of ch ange from third to fourth grade as a way of capturing growth over time within students. This intra-individual use of normative d ata enabled us to examine the relative change in studen t achievement without regard for comparisons of absolute differences in student scor es. WSS was adopted as part of the District's restructu ring effort, but it was not the sole innovation taking place in the district. New readin g, mathematics, and social studies curricula were introduced at the same time as WSS, however, not all schools and classrooms in the district implemented all of these practices. Since information was unavailable about which schools implemented which instructional innov ations, it is not possible to test alternative explanations for our results by isolati ng WSS versus any of these interventions. SampleAt the time this study took place WSS had been used in a sample of the Pittsburgh Schools for three years. All teachers in the WSS schools we re voluntary participants. Selection criteria for participation of teachers in this stud y included use of WSS for at least two years and a determination of the fidelity of implementati on of participating teachers. This was ensured by a review by external examiners conducted in the spring of 1996 of portfolios and of the teachers' 1996Â–97 WSS materials by the resea rch staff. These selection criteria limit the generalizability of our results but provide a t est of the full implementation of WSS. The longitudinal data consist of 96 third grade stu dents in the WSS schools, 116 students in
6 of 18 the non-WSS comparison schools, and 2922 students e nrolled in all other PPS Grade 3 and 4 classrooms in 1996Â–98. For students in WSS school s, 71% were African-American and 90% received free or reduced lunch (see Table 1). T here were more girls (58%) than boys in the sample. To form the comparison group, classroom s were chosen that matched those in the WSS schools as closely as possible on race, inc ome, mobility, school size, and number of parents in the home. In other PPS schools, 70% o f the students were African-American and 87% received free or reduced lunch (see Table 1 ).Table 1 Demographic characteristics of WSS schools, compari son schools, and other PPSGroupSchool Size (N) F/R Lunch (%) Other Parents in Household (%) Mobility (%) African-American (%) WSS400.490.070.69.870.6 Comparison298.889.874.09.475.2 PPS Other311.287.070.89.471.5 WSS = Work Sampling SystemPPS = Pittsburgh Public SchoolsAnalysisComparisons of mean change in reading and math scor es on the ITBS from Grades 3 to 4, as well as regression analyses, were conducted in orde r to study the average change in test scores from one year to the next among the three gr oups. As noted, analysis of previous school district results showed that the WSS schools scored at or near the bottom of the district's ITBS test scores. By controlling for ini tial ability we were able to study the trajectory of change in students' ITBS scores from one year to the next in order to determine if differential change on the ITBS took place despi te the low levels of initial competence on this assessment shown by students in the WSS school s. Our analytic approach began with a comparison of th e change in ITBS scores of the WSS third graders in 1997 1998 as compared with a com parison group matched on key demographic characteristics and all other PPS third graders in that cohort. Differences between the 1997 and 1998 ITBS Developmental Standa rd Scores (DSS) were computed to create DSS change scores, which we used as one indi cator of academic growth. In order to examine whether WSS had a differential impact on hi gh or low achievers, the means of the 1997 DSS scores for the PPS district were used to d ivide all the PPS students into above average and below average groups. The gains for hig h and low achieving students within each group (WSS, comparison, and all other PPS) wer e then compared. Next, three-step hierarchical regressions were perf ormed to examine the relative effect of participating in WSS on the students' change in per formance from third to fourth grade. Because change scores may be particularly sensitive to problems with floor and ceiling, we used a covariance model with the Grade 4 score as t he outcome, controlling for initial ability
7 of 18and the level of form administered. In the first st ep, children's 1997 DSSs were entered to control for differences in initial ability levels. Typically, fourth graders take a Level 10 form, however, PPS allows for administration of out-of-le vel tests. Although the publishers equated the forms, and the level of administration should have had minimal effect on the DSS, we entered the level form into the regression to ensure that results were not biased due to differences in forms or variations in administra tion. Children taking a below-grade form might be more apt to reach the ceiling of the measu re and receive an inflated estimated ability. Only three children in math and two childr en in reading took a form above the grade level; we excluded these cases from analyses. For t he remaining students we created a dummy variable that was entered in the second step of the regression to indicate if the student received a below-grade level form for the G rade 4 administration. (The percentage of students in below-grade level forms was similar in the WSS and all other PPS groupsÂ–Â–10 and 11% respectively. Only 6% of student s in the comparison schools took below-levels forms.) Finally, students' group membe rship (comparison and all other PPS with WSS as the referent group) was entered into th e regression model. Since dependent variables (1998 DSS scores) and the children's init ial scores (1997 DSS scores) were standardized. The regression coefficients can be in terpreted as effect sizes. Missing Data In order to study the impact of missing data on our conclusions, the means of the students with missing subtests were compared by group member ship. In 1997, 21.9% of the WSS students were missing the reading total score; 15.6 % were missing the 1998 reading total score. The comparison group was missing 13.8% of it s reading total scores in 1997 and 17.2% in 1998. Among the other PPS students, 3.7% w ere missing the 1997 reading total score and 11.2% were missing the 1998 reading total score. Due to a missing score in either of the years, 30.2% of the WSS students, 17.2% of t he comparison students, and 13.4% of other PPS students were excluded from the reading c hange score analysis. In mathematics, 7.3% of the WSS group was missing the 1997 math tot al score compared with 13.8% of the comparison group and 5.7% of the other PPS. For the 1998 math total score 18.8% of the WSS group, 17.2% of the comparison group, and 16.1% of the other PPS students had missing data. Thus, 18.8% of the WSS group, 22.4% o f the comparison group, and 19% of the other PPS group were not included in the analys is of the mathematics change score. For the WSS group, we compared students' missing sc ores for one of those years with students who had scores for both years in terms of gender, ethnicity, age, SES, and Woodcock Johnson Psychoeducational Battery-Revised (WJ-R) broad reading, broad math, and broad writing scores in third grade. (The WJ-R was available only for the WSS group.) The results showed no significant differences acros s these variables except that fewer students who were missing at least one mathematics score received free or reduced lunch (58% vs. 92%, p < .01). There were no gender or race differences b etween missing and non-missing students for the comparison group.The means for the ITBS subtests that were taken by the missing students were compared to the means of the other students in that group. For the WSS group, the means on the mathematics and reading total scores were not signi ficantly different between the missing and non-missing groups. However, the 1997 reading D SS score was significantly lower for the comparison group students who were missing the reading subtests in one of those years (171 vs. 197, p < .05), but there were no significant differences in mathematics total score between the missing and non-missing students for th e comparison group. For the other PPS
8 of 18students, the reading total scores in both 1997 and 1998 were significantly lower among the missing group (173 vs. 181 in 1997, p < .001; 186 vs. 197 in 1998, p < .01), but the means on the mathematics total score were not significant ly different for the missing and non-missing groups. Thus, the effects of the missin g data on our findings were relatively limited.ResultsIn this analysis we compare ITBS scores of third an d fourth grade WSS and non-WSS students (both the matched comparison group and all other PPS students) in order to determine if differential achievement on this outco me is associated with participation in WSS classrooms. We begin by comparing the change in mean DSS in reading and math among the WSS students, the comparison group, and t he remainder of the PPS students (see Figure 1). The mean change scores of the WSS group (27 and 20 points for reading and math, respectively) are substantially greater than those of the other groups (0 and 6 for the comparison and 15 and 17 points for all other PPS i n reading and math, respectively). The differences in the groups are particularly strong i n reading, with a mean change on the reading total DSS that is more than 11 points great er for the WSS students when compared with the other PPS students (t = 4.33; p < .001). The moderately large effect size of .68 indicates meaningful as well as significant differe nces in reading change scores (Cohen, 1988). When considered in relation to the compariso n group, the discrepancy is even greater. The WSS group mean change score is more than 27 poi nts higher than the comparison group's mean change score (t = 8.86; p < .001). The unusually large effect size ( d = 1.60) indicates strong differences in the sample in measu red change in reading. In math, results follow a similar pattern, although they are not as dramatic. The mean change score of the WSS students is greater than that of t he other PPS students by more than three
9 of 18 points, a marginally significant finding (t = 1.89; p = .059). The examination of the effect size ( d = .20) suggests a small but nontrivial effect (Coh en, 1988). The mean change of DSS math score in WSS students is almost 14 points high er than that of the comparison group (t = 4.88; p < .001), indicating a large effect ( d = .76). To investigate whether WSS had a differential effec t on high and low achieving students, we used a segmentation analysis. The mean 1997 DSS rat ings of the entire sample were used to divide all students into above and below average gr oups (see Table 2). Before comparing differential changes of score in these groups, we e xamined the initial scores of the 1997 DSS. As expected, the initial scores of the WSS stu dents were lower than the comparable group (above and below average) in the other sample s (comparison and other PPS), although the mean differences were relatively small (2-4 poi nt differences) and not highly significant, with the exception of the above average group of co mparison students in reading. The mean of the 1997 reading DSS of the above average studen ts differed by more than one SD between WSS and the comparison group and the above average performers represented a greater percentage of the comparison sample than wa s the case in the WSS sample.Table 2 Mean of 1997 DSS in above and below average perform ersSubtestWSS Comparison Other PPSWSS vs. Comparison WSS vs. Other PPS MSDnMSDnMSDn t t Reading total Above Average191.4011.0510 205.3617.1577 199.9115.9 01295 -3.49** -1.69 Below Average160.6913.1765 164.8113.8731 164.1212.1 31519 -1.41 -2.22* Math total Above Average197.5814.5624 202.3115.4884 199.3013.3 11534 -1.34 -.63 Below Average163.579.9565 168.619.9923 167.199.4213 77 -2.09* -3.17**Note. t = t-score, DSS = Developmental Standard Score p <.05, ** p < .01 With one exception, above and below average WSS stu dents in reading and math made gains that were greater than the comparison group a nd the other PPS students (see Table 3). With small differences in initial ability in all bu t one area, the effect sizes of the differences in change scores were all moderate to high in readi ng. In mathematics the results were more equivocal. Among below average math achievers, WSS had significantly higher change scores than other PPS students (t = 2.14; p < .05; d = .29). The change scores of low performing PPS students in math were greater than t he change scores of either the WSS or the comparison group of low performers. However, th ese results were not significant and the effect size was negligible ( d = .05).Table 3 Mean of DSS change in above and below average perfo rmers (SD)
10 of 18 SubtestWSS Comparison Other PPS WSS vs Comparison WSS vs Other PPS MSDnMSDnMSDn td td Reading total High26.5620.449 -4.0117.4272 12.2916.781193 4.87*** 1.77 2.54*0.82 Low26.6721.6758 10.5012.7124 17.9315.441337 3.41*** 1.02 3.04**0.55 t (high vs. low).02 3.76*** 8.76*** d (high vs. low).00 .83 .35 Math total High8.3223.3119 2.776.2670 12.9123.311138 .94.24 -. 25.05 Low24.2517.1859 18.8515.4920 20.0416.791230 1.21.32 2.14*.29 t (high vs. low)3.21** 3.94*** 10.31*** d (high vs. low).80 .83 .39 Note. t = t score, d = standardized estimates of effect size *p < .05, ** p < .01, *** p < .001 The publishers of the ITBS contend that children wi th above average ability will show greater gains than other children from year to year However, in each of the three samples, lower performing students made greater gains than h igher performers. The difference in the mean reading change score between high and low perf orming students was greater in the comparison group (14.51) and the PPS group (5.64) t han in the WSS group (0.12). The WSS group did not show substantial differences in readi ng score change between high and low performers (t = .02; p = N.S), although there were significant difference s between high and low performing students in the comparison group (t = 3.76; p < .001; d = .83) and the other PPS group (t = 8.76; p < .001; d = .35). This suggests that high and low performing students profit equally from WSS in reading. In mathematics, the other PPS group demonstrated the least difference in change scores between high and low performing students (7.13) compared with the WSS group (15.94) and the comparison group (16.08). The difference in change scores between high and low performing students was significant in WSS (t = 3.21; p < .01; d = .80) and the comparison group (t = 3.94; p <. 001; d = .83), as well as the other PPS groups (t = 10.31; p <. 001, d = .39). To examine whether group membership accounts for di fferences in achievement growth after controlling for differences in both initial a chievement (i.e., 1997 DSS) and the level form taken in 1998, three step regressions were per formed .The initial achievement and the 1998 level form (below-grade level form = 1, on-gra de level form = 0) were entered in the first and the second step, respectively. Then, grou p membership was entered in the third step. All three models were statistically significa nt (see Table 4). Initial ability, level of form administered, and membership in WSS all predicted s tudents' fourth grade reading DSS (see Table 4). Children who took on-grade level forms ha d fourth grade reading DSS more than one-half SD higher than those who took below-grade level forms, even after controlling for initial ability ( p < .001). The children who were in WSS schools had higher fourth grade scores in reading by .17 SDs when compared with the other PPS students, after controlling for the effect of initial achievement and the level form taken in 1998 ( p < .05). The average fourth grade score for the WSS group was .60 SDs gr eater than the average score for the comparison group after controlling for initial achi evement and 1998 level form ( p < .001)Table 4
11 of 18 Covariance models of Grade 4 achievement Grade 4 DSS Reading Total (N=2,772) Grade 4 DSS Mathematics Total (N=2828) VariableStep 1Step 2Step 3Step 1Step 2Step 3 Initial abilitya.74 ***.67***.68***.56***.56***.56*** 1998 Level Form (below-grade level) -.55***-.54*** -.10*-.10* Other PPS group -.17* -.06 Comparison group -.60*** -.18 R2.55***.58***.58***.32***.32***.32*** R2 change .03***.01*** .00*.00Note. All regression coefficients are shown in effect si ze using standardized scores. DSS = Developmental Standard Score* p < .05, *** p < .001aInitial ability is the third grade DSS on the respe ctive measures (reading and mathematics) The covariance model for the mathematics total four th grade DSS showed a somewhat different pattern. Group membership was not a signi ficant predictor of the fourth grade mathematics DSS after controlling for initial abili ty and level form taken. Students who had higher mathematics DSS on the 1997 ITBS ( p < .001) and took an on-grade level form in 1998 ( p < .05) had higher fourth grade scores than other s tudents. However, the variance explained by this model was unusually low for a cov ariance model that uses the identical assessment to control for initial ability (R2 = .32).DiscussionThis paper examined whether students enrolled in cl assrooms that use a curriculum-embedded performance assessment will sho w greater gains on a conventional test of academic accountability than students witho ut exposure to such a performance assessment. Our results indicate that students who were in WSS classrooms display growth in reading from one year to the next that far outst rips a demographically matched comparison group and that also exceeds the average change shown by all other students in the district. Further, by examining the results of above and below average students separately, we were able to demonstrate that the im pact of the curriculum-embedded performance assessment is not limited solely to tho se who start with either low or high skills. Rather, the impact appears to be across the board with high and low performing students making comparable gains in reading. This a nalysis effectively dismisses the objection that these results are attributable to re gression to the mean, since students with higher scores in both groups performed better than low scoring students in either group. The three-step regressions further demonstrate that par ticipation in WSS classrooms accounts for these differences even after taking into account in itial achievement ranking and level of test form administered.
12 of 18The pattern of change is similar though not as stro ng in mathematics, although it does not appear to benefit high and low performing students equally. Students who were in the WSS group had higher mathematics scores after controlli ng for initial ability and level of form administered, but these findings were not statistic ally significant with this size sample (the effect size with the other PPS group was .18). In a ddition, the segmentation analysis indicated that the mean change score was lower for high performing students in all groups than for below-average students.Examination of the construction of the DSS on this edition of the ITBS provides a clue to potential reasons for this finding. In this version of the ITBS, the mathematics DSS appears to be heavily weighted by number concepts and opera tions and, in particular, by computation (procedural knowledge of operations). I n contrast, national standards in effect at the time (NCTM, 1989) and WSS (which is based on these and other standards) address multiple strands of mathematical thinking, includin g geometry, measurement and spatial sense, data analysis, statistics and probability, p atterns, functions, and algebra, as well as number concepts and operations. The findings in mat hematics may be a reflection of a mismatch between the broad standards WSS reflects a nd the test specifications of this edition of the ITBS. In addition, the covariance mo del for mathematics suggests that there is a problem with this measure of mathematics skills. Only 32% of the variance in the fourth grade scores could be explained by the third grade scores. The findings on the reading assessment are robust a nd pervasive and make an important contribution to discussions of accountability. Whet her looking at high or low performing students, examining gain scores or using a covarian ce model, students enrolled in WSS classrooms made greater gains in reading than stude nts who did not have this exposure. These findings, as well as those for math, though n ot definitive because of our inability to disentangle the impact of WSS from the other innova tions co-occurring in the district, suggest a new way to approach accountability testin g. For too long, accountability examinations have been assumed to be of a particula r kind with an unambiguous focus: normative assessments intended to rank students num erically and compare them to the performance of a specified group. As noted at the o utset of this paper, it is likely that the U.S. is spending more money on tests at this time t han at any previous point in its history. However, children are not faring well on these asse ssments. Media reports of large numbers of failures in numerous states are interpreted eith er in terms of students' lack of skills or teachers' inability to align curricula to the stand ards that are used to design high stakes tests (Manzo, 2001).However, another explanation is also possible, and this perspective provides a link to our second research question: Is it possible to design an accountability system that relies on both classroomand test-based information about student achievement? The alternative view of the accountability debate is that we not only need high standards and tests that reflect these standards, we also need curricula that will enable students to be successful on these assessmentsÂ–Â–but that are not simply instances of m easurement-driven instruction.. By implementing an instructional assessment such as Wo rk Sampling, teachers obtain information about their students on a continuous ba sis across multiple curriculum domains and from several assessment sources. They compare s tudent performance with standards-based guidelines. They collect multiple s ources of information from checklists, portfolios, and student and parent reports. They en gage in curriculum analysis in order to evaluate artifacts included in portfolios. And they participate in processes of planning, review, and analysis with their colleagues. Student s also have a meaningful role in the
13 of 18assessment process and thus become active participa nts in the evaluation process by becoming more familiar with the standards and how t o progress towards those standards. This appears to enhance teaching and improve learni ng. Perhaps the most important lesson that can be garne red from this study is that accountability should not be viewed as a test, but as a system When well-constructed, normative assessments of accountability are linked to well-de signed, curriculum-embedded instructional assessments, children perform better on accountability exams, but they do this not because instruction has been narrowed to the sp ecific content of the test. They do better on the high stakes tests because instruction can be targeted to the skills and needs of the learner using standards-based information the teach er gains from ongoing assessment and shares with the learner. Â“Will this be on the test? Â” ceases to be the question that drives learning. Instead, Â“What should I learn next?Â” beco mes the focus. When accountability is seen as a system that incorp orates both instructional assessment and on-demand tests, both teaching and learning can be affected positively. Moreover, this methodology provides policy makers with clear docum entation not only of summative accomplishments, but also of the process of teachin g and learning. The approach described in this study places emphasis where it belongs: on teaching and learning, rather than on testing. And it does so without sacrificing either the student or the teacher on the altar of accountability.AcknowledgementWe wish to thank Jack Garrow for assisting us in ob taining and interpreting school district data. We are also deeply grateful to the principals teachers, parents, and children who participated in this study, and to the staff and ad ministrators of the Pittsburgh Public Schools. This study was supported by a grant from t he School Restructuring Evaluation Project, University of Pittsburgh, the Heinz Endowm ents, and the Grable and Mellon Foundations. The views expressed in this paper are those of the authors and do not necessarily represent the positions of these organi zations.ReferencesAchieve, Inc. (2002). Achieve + McRel Standards Database http://frodo.mindseye.com/achieve/achieved.nsf Amrein, A. L. & Berliner, D. C. (2002). High-stakes testing, uncertainty, and student learning. Education Policy Analysis Archives 10 (18). Retrieved 4/1/02 from http://epaa.asu.edu/epaa/v10n18/ Baron, J. B. & Wolf, D. P. (Eds.), Performance-based student assessment: Challenges an d possibilities (95th Yearbook of the Society for the Study of Education pp. 166-191). Chicago, IL: Society for the Study of Education. Borko, H., Flory, M., & Cumbo, K. (October, 1993). Teachers' ideas and practices about assessment and instruction. A case study of the eff ects of alternative assessment in instruction, student learning, and accountability p ractice CSE Technical Report 366. Los Angeles: Center for Research on Evaluation, Sta ndards, and Student Testing (CRESST).
14 of 18Cohen, J. (1988). Statistical power analysis for the behavioral scien ces New York: Wiley Corbett, H. D. & Wilson, B. L. (1991). Testing, reform, and rebellion Norwood, NJ: Ablex. Darling-Hammond, L. (1992). Standards of practice for learner-centered schools New York: Columbia University, Teachers College, Nation al Center for Restructuring Education, Schools, and Teaching. Falk, B., & Darling-Hammond, L. (March, 1993). The primary language record at P.S. 261: How assessment transforms teaching and learning New York: National Center for Restructuring Education, Schools, and Teaching. Hoover, H. D., Hieronymus, A. N., Frisbie, D. A., D unbar, S. B., Oberley, K. R., Cantor, N. K., Bray, J. C., Lewis, J. C., & Qualls-Payne, A. L (1993), Iowa Tests of Basic Skills Norms and scores conversions Form K Survey Battery Chicago: Riverside. Jennings, J. (1998). Why national standards and tests? Politics and the quest for better schools. Thousand Oaks, CA: Sage Publications, Inc. Kohn, A. (2000). The case against standardized testing: Raising the scores, ruining the schools Portsmouth, NH: Heinemann. McNeil, L. M. (2000). Contradictions of school reform: Educational costs of standardized testing New York: Routledge. Manzo, K. K. (2001, June 20). More than half of Cal ifornia 9th graders flunk exit exam. Education Week p. 19. Meisels, S. (1996). Performance in context: Assessi ng children's achievement at the outset of school. In A. Sameroff and M. Haith (Eds.), The five to seven year shift: The age of reason and responsibility (pp. 407-431). Chicago: The University of Chicago Press. Meisels, S. (1997). Using Work Sampling in authenti c assessments. Educational Leadership, 54 (4), 60-65. Meisels, S., Dorfman, A., & Steele, D. (1994). Equi ty and excellence in group-administered and performance-based assessments. In M. Nettles an d A. Nettles (Eds.), Equity in educational assessment and testing (pp. 195-211). Boston: Kluwer. Meisels, S.J., Jablon, J., Marsden, D.B., Dichtelmi ller, M.L., & Dorfman, A. (1994). The Work Sampling System (3d ed.). Ann Arbor, MI: Rebus Inc. Meisels, S.J., Jablon, J., Marsden, D.B., Dichtelmi ller, M.L., & Dorfman, A. (2001). The Work Sampling System (4th ed.). Ann Arbor, MI: Rebus Inc. Meisels, S. J., Bickel, D. D., Nicholson, J., Xue, Y., & Atkins-Burnett, S. (2001). Trusting teachers' judgments: A validity study of a curricul um-embedded performance assessment in Kindergarten-Grade 3. American Educational Research Journal 38 (1), 73 Â– 95. Meisels, S.J., Liaw, F-r, Dorfman, A., Nelson, R. F (1995). The Work Sampling System: Reliability and validity of a performance assessmen t for young children. Early
15 of 18Childhood Research Quarterly, 10 277-296. Meisels, S. J., Xue, Y., Bickel, D. D., Nicholson, J., & Atkins-Burnett, S. (2001) Parental reactions to authentic performance assessment. Educational Assessment 7 (1), 61-85. National Council of Teachers of Mathematics (1989). Curriculum and evaluation standards for school mathematics, grades K-6 Reston, VA: Author. Nicholson, J. M. (2000). Examining aspects of consequential validity in a curriculum-embedded performance assessment Doctoral dissertation, Ann Arbor, MI: University of Michigan. National Commission on Excellence in Education. (19 83). A nation at risk: The imperative for educational reform: A report to the nation and Secretary of Education, United States Department of Education. Washington, D.C.: The Commission [Supt. Of Docs., U.S. G.P.O. distributor]. Orfield, G. & Kornhaber, M. L. (Eds.) (2001). Raising standards or raising barriers? Inequality and high-stakes testing in public educat ion New York: The Century Foundation Press. Popham, W. J. (2000). Testing! Testing! What every parent should know abo ut school tests Boston: Allyn & Bacon. Olson, L. (2001, January 11). Finding the right mix Education Week, Quality Counts Special Issue p. 12 20. University of Iowa and Riverside Publishing Co. (19 94). Riverside 2000 Integrated Assessment Program: Technical Summary I Chicago: The Riverside Publishing Company.About the AuthorsSamuel J. MeiselsErikson Institute420 N. WabashChicago, IL 60614Email: firstname.lastname@example.org Samuel J. Meisels is president of the Erikson Insititute. An emeritu s professor of education at the University of Michigan, Dr. Meisels's resear ch concerns the development of alternative assessments from birth through grade 5, policy studies of early intervention, and research regarding the developmental consequences o f high risk birth. He has co-authored The Handbook of Early Intervention the Â“Work Sampling System,Â” and the Â“Early Screening Inventory-Revised,Â” among many other publ ications. Sally Atkins-Burnett is an assistant professor at the University of Tol edo. She earned her Ph.D. from the University of Michigan. She has been a major contributor to the NCES-sponsored Early Childhood Longitudinal Study a nd has also worked on the development of instruments used in the Study of Ins tructional Improvement, a national longitudinal study of the impact of school improvem ent efforts on instruction and student
16 of 18 performance. Her research interests include assessm ent in early childhood, social development, literacy instruction, and children wit h special needs. Yange Xue is a Research Scientist at the National Center for Children and Families, Teachers College, Columbia University, New York, NY She holds a doctorate in early childhood education from the University of Michigan Her areas of specialization include early development and education and quantitative me thodology. Julie Nicholson received her PhD in Early Childhood Education from the University of Michigan. Her research interests include emergent l iteracy, the use of technology in early childhood classrooms, and the positive and negative consequences associated with different methods of early childhood assessment.Donna DiPrima Bickel is a Resident Fellow at the University of Pittsbu rgh's, Institute For Learning (IFL). Since 1999, Dr. Bickel has coordina ted the IFL program in Content-Focused Coaching in elementary literacy. She has co-develop ed the professional development video and print materials used in preparing school distri ct coaches to work with teachers using this model of professional development.Seung-Hee Son is a doctoral student in Early Childhood Education at the University of Michigan. Her research interests include parenting and preschool influences on early language and literacy development, and policy issue s in child development.Copyright 2003 by the Education Policy Analysis ArchivesThe World Wide Web address for the Education Policy Analysis Archives is epaa.asu.edu Editor: Gene V Glass, Arizona State UniversityProduction Assistant: Chris Murrell, Arizona State University General questions about appropriateness of topics o r particular articles may be addressed to the Editor, Gene V Glass, email@example.com or reach him at College of Education, Arizona State University, Tempe, AZ 8 5287-2411. The Commentary Editor is Casey D. Cobb: firstname.lastname@example.org .EPAA Editorial Board Michael W. Apple University of Wisconsin David C. Berliner Arizona State University Greg Camilli Rutgers University Linda Darling-Hammond Stanford University Sherman Dorn University of South Florida Mark E. Fetler California Commission on Teacher Credentialing Gustavo E. Fischman California State UniveristyÂ–Los Angeles Richard Garlikov Birmingham, Alabama Thomas F. Green Syracuse University Aimee Howley Ohio University
17 of 18 Craig B. Howley Appalachia Educational Laboratory William Hunter University of Ontario Institute ofTechnology Patricia Fey Jarvis Seattle, Washington Daniel Kalls Ume University Benjamin Levin University of Manitoba Thomas Mauhs-Pugh Green Mountain College Les McLean University of Toronto Heinrich Mintrop University of California, Los Angeles Michele Moses Arizona State University Gary Orfield Harvard University Anthony G. Rud Jr. Purdue University Jay Paredes Scribner University of Missouri Michael Scriven University of Auckland Lorrie A. Shepard University of Colorado, Boulder Robert E. Stake University of IllinoisÂ—UC Kevin Welner University of Colorado, Boulder Terrence G. Wiley Arizona State University John Willinsky University of British ColumbiaEPAA Spanish Language Editorial BoardAssociate Editor for Spanish Language Roberto Rodrguez Gmez Universidad Nacional Autnoma de Mxico email@example.com Adrin Acosta (Mxico) Universidad de Guadalajaraadrianacosta@compuserve.com J. Flix Angulo Rasco (Spain) Universidad de Cdizfelix.firstname.lastname@example.org Teresa Bracho (Mxico) Centro de Investigacin y DocenciaEconmica-CIDEbracho dis1.cide.mx Alejandro Canales (Mxico) Universidad Nacional Autnoma deMxicocanalesa@servidor.unam.mx Ursula Casanova (U.S.A.) Arizona State Universitycasanova@asu.edu Jos Contreras Domingo Universitat de Barcelona Jose.Contreras@doe.d5.ub.es Erwin Epstein (U.S.A.) Loyola University of ChicagoEepstein@luc.edu Josu Gonzlez (U.S.A.) Arizona State Universityjosue@asu.edu Rollin Kent (Mxico)Universidad Autnoma de Puebla email@example.com Mara Beatriz Luce (Brazil)Universidad Federal de Rio Grande do Sul-UFRGSlucemb@orion.ufrgs.br
18 of 18 Javier Mendoza Rojas (Mxico)Universidad Nacional Autnoma deMxicojaviermr@servidor.unam.mxMarcela Mollis (Argentina)Universidad de Buenos Airesmmollis@filo.uba.ar Humberto Muoz Garca (Mxico) Universidad Nacional Autnoma deMxicohumberto@servidor.unam.mxAngel Ignacio Prez Gmez (Spain)Universidad de Mlagaaiperez@uma.es DanielSchugurensky (Argentina-Canad)OISE/UT, Canadadschugurensky@oise.utoronto.ca Simon Schwartzman (Brazil)American Institutes for ResesarchÂ–Brazil(AIRBrasil) firstname.lastname@example.org Jurjo Torres Santom (Spain)Universidad de A Coruajurjo@udc.es Carlos Alberto Torres (U.S.A.)University of California, Los Angelestorres@gseisucla.edu