Results
Proportions of Nonresponders Following Treatment
Dual-discrepancy approach. We also calculated proportions of students who continued to be unresponsive following implementation of the three treatments. The dual-discrepancy approach used to identify initial nonresponders in December was used again to identify nonresponders following the 13 weeks of additional treatment. As in December, students were identified as nonresponders if they scored more than 0.5 SD below average-performers' levels and slopes on the CBM measures. Accordingly, 81% of PALS students, 80% of Modified PALS students, and only 50% of Tutoring students remained unresponsive to treatment at the end of the study. Overall, 70% of the initial nonresponders were still unresponsive. This translates to 22% of the 174 at-risk students who were monitored during the first half of the study, or 12% of the total number of students participating in Standard PALS or PALS + Fluency (n = 323).
Performance-level-only and growth-rate-only approaches. To determine whether the procedures we used to identify nonresponders yielded similar proportions as alternative procedures used by other researchers, two additional approaches to identifying nonresponders (i.e., performance-level-only and growth-rate-only) were explored. Performance levels and growth rates on the Word Identification and Word Attack subtests and the near- and far-transfer fluency measures were examined (see Berninger et al., 1999; Brown & Felton, 1990; D. Fuchs, Fuchs, Thompson, Al Otaiba et al., 2001; Torgesen et al., 1999; Vellutino, 1996). Table 5 presents proportions of nonresponders identified using the dual discrepancy approach as well as the approaches used in previous studies.
Table 5: Proportions of Nonresponders at the End of the Study Identified by Dual-Discrepancy, Performance-Level-Only, and Growth-Rate Only Criteria
| Treatment | Performance Level Only | Growth Rate Only | ||||||||
| Dual Discrepancya | Percentile Rankb | Criterion Levelc | No Growthd | Limited Growthe | ||||||
| n | (%) | n | (%) | n | (%) | n | (%) | n | (%) | |
| PALS (n= 21) |
17 | (81) | 8 | (38) | 21 | (100) | 2 | (10) | 10 | (48) |
| Modified PALS (n= 15) |
12 | (80) | 8 | (53) | 15 | (100) | 1 | (7) | 7 | (47) |
| Tutoring (n= 20) |
10 | (50) | 9 | (45) | 20 | (100) | 1 | (5) | 5 | (25) |
| Total nonresponders (N= 56) |
39 | (70) | 25 | (45) | 56 | (100) | 4 | (7) | 22 | (39) |
| Total risk pool (N = 174) |
39 | (22) | 25 | (14) | 56 | (32) | 4 | (2) | 22 | (13) |
| Total PALS participants (N = 323) |
39 | (12) | 25 | (8) | 56 | (17) | 4 | (1) | 22 | (7) |
aDual discrepancy was determined by slopes and levels > .5 SD below average on the Dolch and/or Nonword Fluency measures.
bPercentile rank was determined by scores below the 30th percentile on the WRMT-R Word Identification and/or Word Attack subtests.
cCriterion level was determined by < 40 correct words per min on the Near-Transfer and/or Far-Transfer Fluency measures.
dNo growth was determined by a gain of 0 words (or less) on Word Identification and/or Word Attack.
eLimited growth was determined by a gain of 10 words or less on Word Identification and/or a gain of 5 words or less on Word Attack.
In terms of performance-level-only approaches, we first determined how many of the initial 56 nonresponders for whom we have complete data performed below the 30th percentile (as recommended by Torgesen, 2000) at the end of the study. Next, we calculated how many of the initial 56 nonresponders read less than 40 words correctly in 1 min (as suggested by Good, Kaminski, & Shinn, 1999) at study's end. In terms of growth rate, we figured how many students made no growth on the Word Identification and Word Attack subtests or limited growth (less than 10 words gained on the Word Identification subtest; less than 5 words gained on the Word Attack subtest).
As shown in Table 5, each approach to identifying nonresponders yielded a different proportion of students, sometimes strikingly so. Using percentile as a performance-level-only criterion (as per Torgesen, 2000) resulted in fewer nonresponders than the dual-discrepancy approach. However, students who scored above the 30th percentile on the WRMT-R subtests, but still met the dual-discrepancy criterion, had slopes that were .80 SD below the average-performers on the monitoring measures. This finding suggests that performance above the 30th percentile may mask very poor progress toward higher levels of reading. On the other hand, using a specific criterion level as a performance-level-only criterion (as per Good et al., 1999) resulted in many more nonresponders. Some of the students who had not yet reached 40 correct words per min were making mean gains of .08 SD above average, indicating that they were making similar word-reading gains to their average-performing peers. Thus, these students were probably not true "nonresponders." Similarly, the "no growth" (e.g., Berninger et al., 1999) and "limited growth" (e.g., Vellutino, 1996) criteria resulted in fewer nonresponders than the dual-discrepancy criteria, most likely overlooking many students who made some, but not sufficient growth.
Previous Page | Next Page
(Results: Effectiveness) | (Discussion)
