探究英语学习者英语水平与ACT考试成绩之间的关系附录：ACT写作

Joann L. Moore, PhD Two recent research reports investigated relationships between English language proficiencyandtheperformance of English learners (ELs) taking the ACT®test:ExploringRelationshipsBetween EnglishProficiency and ACTTestPerformance of EnglishLearners: State 1(Moore &Schnieders,2026) andExploringRelationshipsBetween EnglishProficiency and ACTTestPerformance of EnglishLearners: State 2(Mooreet al.,2026). Thisissue briefprovides the ACT began offering testing supports to ELs in the fall of 2017, including (a) extended time (oneand one-half time, single day), (b) translated test instructions that ACT provides, and (c) anapproved word-to-word bilingual dictionary without definitions.These supports are intended toremove construct-irrelevant variance and allow ELs to more accurately demonstrate theirknowledge and skills. It should be noted that there is aprocess for requesting supportswherestudents or school officials initiate the request, school officials provideACT withdocumentation TwoU.S.states agreed to participate in this study. Both states are part of theWIDAconsortiumand havehad statewide contracts to administerthe ACT to virtually allGrade 11 students sincebeforethe 2017–2018 school year.WIDAis a consortium of states led by the University ofWisconsin–Madison anddevelops resources for supporting ELs and their teachers, includingACCESS, a set of summative English proficiency assessments administered annually toallK–12 students in more than 40 states. The ACCESS assessments include four domains— Each state provideddatafiles containing demographic information about their students,includingEL status, and each state granted permission for WIDA to share student-levelACCESSdatawith ACT.State 2 also providedGrade 11 high school transcript data, whichallowed us to conduct additional analyses usingGrade 11 GPA, courses taken, and credit hours the 2018–2019 through 2022–2023 school years. State 1is asmallersamplewithapproximately 2% ELs in the study sample (N= 642 ELs and 35,234 non-ELs), and State 2is alargersamplewith approximately 10% ELs in the study sample (N= 15,788 ELs and 143,461non-ELs). Approximately 7% ofGrade 11 students (11% across K–12) in public schools in the Table 1andTable 2contain the demographic characteristics of the data samples for State 1and State 2, respectively. In both states, ELs were predominantly Hispanic/Latino (79% of allELs in State 1 and 87% of all ELs in State 2).Non-ELs in State 1 were predominantlywhite(81%), whereas non-ELs in State 2 were a more diverse group, with 38% Hispanic/Latino, 35%white, 11% Black/African American, and smaller percentages of students from otherracial/ethnic groups.ELsin both stateswere also more likely to be eligible for free/reduced-price The ACT writing test is optional for students who take thenational ACT, but states thatadminister the ACT statewide may choose to require the writing test for all their students, whichwas the case for the two states participating in this study. The writing test is a singlepromptdescribing three perspectives on a complex issue, and students are instructed to write an essayin 40 minutes analyzing the issue and developing their own perspective on the issue. The essayis scored bytwotrained raters, oneof which may be ACT’s CRASE+®automated essay scoringengine,on a1-to-6 scale in four domains(ideas and analysis, development and support, Table 3contains the percentages of students who took the ACT writing testby state and group.Approximately 99% ofthestudents in each state took the ACT writing test.ELs in both stateswere less likely to have taken the writing testthannon-ELs, and ELs who took the ACT with Table 4containsthe average ACT writing scores and standard deviationsof studentsby stateand group, andTable 5containsthescore distributions (percentages of students at each ACTwriting test score point)by state and group.In both states,ELs had lower scores than non-ELs,and ELs who took the ACT with supports had lower writing scores than ELswho took the testwithout supports. Thescore distributionsshow that ELsin both stateswere much more likely tohave scores at or below a6, and non-ELs were more likely to have scores between6 and 8.In Average ACCESS scores and distributions of ACCESS levels by state and group can be foundin the full reports. In both states, ELs who took the ACT with supports tended to have slightlylower ACCESS scores than ELs who took the ACT without supports, but thedifferences weregenerally small, indicating that the two groups had similar levels of English proficiency on Table 6contains correlations between ACT writing scores and ACCESS scores by state andgroup.Acrossall theACCESS scores and composites, State 1 had higher correlationswith ACTwriting scores than State 2.For both states, higher correlations were also found betweenlanguage proficiency and ACT writing scores for ELs who took the ACT with supportsthan forELs who took the ACT without supports.For both states, the highest