Today, schools actually have more than two dozen individ-ually administered early screening assessments to choosefrom that are appropriate for kindergarten through thirdgrade. These assessments cover a variety of reading skills—phonemic awareness, phonics, fluency, vocabulary, andreading comprehension—as well as a variety ofassessment purposes: screening, diagnosis, andprogress monitoring. (There are also outcomesassessments that educators should be aware of ifthey plan to conduct studies of the effectivenessof their reading programs.) As an introductionto the kind of information that teachers canglean from early reading assessments, examplesfrom TPRI and Dynamic Indicators of BasicEarly Literacy Skills (DIBELS) are includedhere. The figure (immediate right) shows someitems from the TPRI Screening Section for chil-dren at the beginning of the first grade. The fig-ure on page 15 shows how a kindergarten stu-dent’s progress is monitored using DIBELS.

Fortunately, a team of researchers has re-viewed the current crop of assessments, identi-fied which have sufficient reliability and validity,and developed a Web site for educators thatclearly indicates which assessments are appropriate for dif-fering grades, skills, and purposes (available athttp://idea.uoregon.edu/assessment/).

Most currently available assessments identify childrenusing national norms. So, for example, schools can decide tointervene with all children who score in the bottom 10 to 20percent nationally. Of course, just how many students thiswill be varies greatly by school. A few assessments have es-tablished benchmarks, or cut scores, that represent evi-dence-based thresholds indicating the likelihood of readingsuccess (or failure), and recommend that schools intervenewith all students who fall below the benchmark.

A key issue that has arisen during the instrument-devel-opment research is creating accurate instruments that arenot too long. Assessment developers have been grapplingwith the fact that longer assessments provide more detaileddata, but shorter assessments are more practical for the

classroom. This led to differentiating between screeningand diagnostic assessments. Typically, screening instru-ments tend to be short, taking as little as five to 10 minutesper child, and they identify which students are at risk orbehind, as well as some information on which skills the

Early intervention works. Because it isalso expensive, it’s important to be

able to identify the kids who are most atrisk of reading failure. Thanks to a newgeneration of screening assessments, wecan identify these students as early askindergarten—and then invest in inter-

ventions for them. The new assessmentsare brief, trustworthy, and easy to adminis-ter. They can be administered to allkindergartners through third-graders a fewtimes a year, allowing teachers to identifywhich students need extra help. They takeonly five to ten minutes per child to ad-

minister and can typically be given byclassroom, reading, or special educationteachers or aides. Once identified, thesestudents can receive the assistance theyneed, and the downward spiral that resultsfrom weak early reading skills can beaverted.

Xi AMERICAN EDUCATOR FALL 2004

Early Screening Is at the Heart of Prevention

Selecting Assessments for Your School

Shown above are assessments and benchmarks from the Screen-ing Section of the TPRI, the first early reading assessment to beused throughout a state. These two Screenings are used at thebeginning of first grade along with a 10-item screening ofchildren’s knowledge of letter names and sounds. As you cansee, the TPRI provides empirically derived criteria to indicateif students have developed adequate knowledge and skills.When students do not meet those criteria, the teacher movesdirectly into more in-depth assessments from the Inventory Section of the TPRI. Therefore, the amount of assessment isindividualized: Students who are “developed” on the Screen-ings will be done in just five minutes; students lacking skillswill continue through the Inventory assessments until the skillsthat need to be developed are identified—a process that cantake an experienced teacher anywhere from 10 to 25 minutes.(For more information on the TPRI, see www.tpri.org.)

students are lacking. Diagnostic instruments—used onlyfor the smaller group of students deemed at risk in thescreening—tend to be longer, taking roughly 20 to 45 min-utes per child, and they offer a much more thorough lookat students’ strengths and weaknesses. (These time esti-mates are for teachers who are experienced in using theseassessments. More time will be needed while teachers be-come accustomed to using those tools.) Sometimes assess-

ments have both screening and diagnostic components. Forexample, the TPRI has “Screening” and “Inventory” sec-tions. When children don’t meet criteria in the ScreeningSection, the teacher can immediately switch to a more in-depth assessment from the Inventory Section to pinpointthe knowledge and skills that the child still needs to de-velop. Teachers can also use Inventory data to match in-struction with specific student needs.

A third type of assessment is forprogress monitoring. These instrumentstypically come in short, multiple forms sothat students’ skills can be assessed everytwo weeks (or even more frequently) toquickly determine if an intervention issufficiently effective. If not, the interven-tion can be altered (by changing the in-structional content, methods, and/or in-tensity), the child may be given a diag-nostic assessment, or the child may be re-ferred for special education. For example,DIBELS is a widely used screening andprogress-monitoring assessment. DIBELSmeasures take just a few minutes each andusually come in 20 alternate forms for fre-quent checkups. The figure (left) showshow a kindergarten student’s progress is

monitored using DIBELS.While screening instruments are used with all students,

diagnostic instruments are only necessary for studentswhose screenings reveal serious skill deficits and/or whoseprogress monitoring indicates that they are not respondingto the intervention. Ideally, all K-3 students should bescreened three times per year starting in mid-kindergarten;diagnostic and progress monitoring assessments can bedone as needed, with progress monitoring of children in anintervention being quite frequent to make sure that inter-ventions are as effective as possible.

FALL 2004 AMERICAN FEDERATION OF TEACHERS Xii

Source for the DIBELS figure: Good, R.H., Gruba, J., and Kaminski, R.A. (2002). UsingDynamic Indicators of Basic Early Literacy Skills (DIBELS) in an Outcomes-Driven Model. In AThomas and J. Grimes (Eds.), Best Practices in School Psychology IV (pp, 679-700).Washington, D.C.: National Association of School Psychologists.

How Do They Work?The key to our new ability to predictwhich children are likely to have prob-lems in learning to read is the researchfinding that almost all struggling read-ers have problems with phonemicawareness—identifying and being able

to manipulate the sounds in words(Torgesen, 1998). Not surprisingly,given their troubles with the phonolog-ical features of language, these childrenalso have difficulty grasping the alpha-betic principle and are slow to build upa “sight vocabulary,” meaning words

that they can read automatically with-out sounding them out. Building onthese highly consistent findings, re-searchers have found that by midwaythrough kindergarten (assuming pre-reading skills are being taught), knowl-edge of letter names predicts future

This chart shows one kindergarten student’s improvementbased on a progress monitoring assessment from January toJune. Using alternate forms of the DIBELS measure ofPhoneme Segmentation Fluency (PSF), the teacher screenedthis child three times in January to be sure that he reallyneeded an intervention. As noted by the horizontal line in thechart, all students should score at or above 40 on PSF by theend of kindergarten. Drawing a line from the student’s initialscores to that benchmark creates clear goals for the rest of theyear and allows the teacher to judge the success of his interven-tion. Initially, the intervention was not sufficient. The teachermade his intervention more intense by providing additionalmodeling, examples, and practice—and the student reachedthe PSF benchmark by the end of May. (To learn more aboutDIBELS, visit http://dibels.uoregon.edu/ andwww.dibelsassessment.com.)

reading ability. And by first grade, let-ter-sound knowledge is highly predic-tive. (For more on this topic, see“Catch Them Before They Fall: Identi-fication and Assessment To PreventReading Failure in Young Children”from the Spring/Summer 1998 issue ofAmerican Educator; it is available onlineat www.aft.org/pubs-reports/american_educator/spring_sum98/torgesen.pdf.)

How Accurate Are They?Just how accurate are these early assess-ments? Accuracy varies by instrument.Rather than reviewing several assess-ments, let’s look at the average predictivepower of assessing kindergartners’ letteridentification skills (Snow et al., 1998).A meta-analysis of 20 studies that mea-sured 11 different possible predictors ofreading difficulties (including receptivevocabulary, expressive language, con-cepts of print, and verbal memory ofstories or sentences) found that letteridentification was the strongest single in-dicator of future reading. The mean cor-relation between letter identification inkindergarten and reading scores ingrades one through three was .52. Infact, letter identification was almost asgood a predictor by itself as an entirereading-readiness test (which includes awhole host of reading skills). But whatdoes a moderately strong correlation likethis mean when it comes to designatingchildren at risk or not? Another study(Snow et al., 1998) used 1,000 kinder-gartners’ letter identification skills tofind out. The researchers consideredtheir predictions accurate if the childrenwho were designated at risk in kinder-garten were then in the bottom 20 per-cent on teachers’ ratings in first grade.

To begin with, the researchers tested astrict letter-identification cutoff; theydesignated students at risk only if theyfell in the bottom 10 percent. Accordingto the first-grade teachers’ ratings, thisstrict cutoff correctly identified 83.2 per-cent of children. Since there were 1,000children in the study and the bottom 10percent were designated at risk, 100 chil-dren were so designated. Of these, 63were correctly identified (meaning theywere in the bottom 20 percent accordingto teachers’ ratings in first grade), but 37were false alarms (meaning they werenot in the bottom 20 percent). Of the900 children designated not at risk, 769were correctly identified, but 131 were

misidentified (meaning they were in thebottom 20 percent in first grade).

Believing that too many childrenwho did end up having reading difficul-ties were missed with the strict cutoff,the researchers also examined a more le-nient letter-identification cutoff. In thissecond analysis, they designated thebottom 25 percent of kindergartners atrisk. Of these 250 children, 118 werecorrectly identified, but 132 were falsealarms. Of the 750 children designatednot at risk, 677 were correctly identi-fied, but 73 were not. Overall, the morelenient cutoff meant that the overall ac-curacy of the prediction was reducedslightly (79.5 percent of children werecorrectly identified)—but the percent-age of struggling readers who weremissed dropped from 15 to 11.

Obviously, educators have to make aconscious choice when they decidewhat percentage of children to inter-vene with. Intervening with the bottom10 percent means that many at-riskchildren will not be appropriatelyserved. And intervening with the bot-tom 25 percent means that many not-at-risk children will be served.

No assessment can completely over-come these potential errors in identify-ing at-risk children. Even with the bestassessment, some children who will havereading problems are not identified andsome who will not are. But there arestrategies to greatly reduce the errors inidentification. To minimize under-iden-tification, schools are encouraged toscreen all children—three times peryear—starting with mid-K. (Assess-ments at the very beginning of kinder-garten tend to be unreliable because stu-dents may lack skills simply becausethey haven’t been taught, not becausethey will have trouble with the conceptsonce they have been presented in theregular classroom setting.) To minimizeover-identification, assessments oftencome with multiple forms so that teach-ers can confirm the results (and be surethat the child was not just having a badday) before the intervention begins.Given the importance of addressing skilldeficits, over-identification of childrenmay be the best policy. For not-at-riskstudents, the intervention will simplyreinforce their skills, acting like an “in-surance policy” against future problemswith reading. And, with adequateprogress monitoring, such students willtest out of the intervention quickly.

Fortunately, predictions of whichstudents are at risk for reading failurebecome even more accurate by the endof first grade. This is what one wouldexpect given that, starting at the end offirst grade, students’ word-reading abil-ity can be assessed directly instead ofindirectly through such pre-readingskills as letter naming and phonemesegmentation. While it is clearly truethat early word reading ability is astrong predictor of later word readingability, very brief measures of oral read-ing fluency are also a strong predictor,and thus a good screening measure, fordifficulties in reading comprehension.In fact, Fuchs, Fuchs, Hosp, and Jenk-ins (2001) reported evidence that a verybrief measure of oral reading fluencywas a better predictor of performanceon a reading comprehension outcomemeasure than was a brief measure ofreading comprehension itself. In thisstudy, with middle and junior highschool students with reading disabili-ties, the correlation between oral read-ing fluency and the reading compre-hension measure was a nearly perfect.91. More recently, researchers comparingthird graders’ performance on the Dy-namic Indicators of Basic Early LiteracySkills measure of Oral Reading Fluencyto their scores on state assessments ofreading comprehension have found cor-relations of .70 with the Florida Com-prehensive Assessment Test (Buck andTorgesen, 2003) and .73 with theNorth Carolina end-of-grade assess-ment (Barger, 2003).

—EDITORS

ReferencesBarger, J. (2003). Comparing the DIBELS OralReading Fluency indicator and the North Carolinaend-of-grade reading assessment. Asheville, N.C.:North Carolina Teacher Academy.Buck, J. and Torgesen, J. (2003). The RelationshipBetween Performance on a Measure of Oral ReadingFluency and Performance on the Florida Compre-hensive Assessment Test. Tallahassee, Fla.: FloridaCenter for Reading Research.Fuchs, L.S., Fuchs, D., Hosp, M.K., and Jenkins,J.R. (2001). Oral reading fluency as an indicator ofreading competence: A theoretical, empirical, andhistorical analysis. Scientific Studies of Reading, 5,239-256.Snow, C. E., Burns, M. S., and Griffin, P., Eds.(1998). “Predictors of Success and Failure in Read-ing” in Preventing Reading Difficulties in YoungChildren. Washington, D.C.: National AcademyPress.Torgesen, J. K. (1998). “Catch Them Before TheyFall: Identification and Assessment to Prevent Read-ing Failure in Young Children, American Educator,22 (1&2), 32-39.

XiiiAMERICAN EDUCATOR FALL 2004