Instruction in strategies versus direct instruction for students with learning disabilities

The empirical evidence gathered from three decades of reading intervention research with students with learning disabilities was synthesized through ^{Swanson's (1999}) meta-analysis, which clarified the possibility of improvement in reading disabilities in both the subdomains of word recognition (WC) and reading comprehension (RC). We compared, in terms of effect size, studies that focused directly on reading the word - decoding printed material - through direct, fairly structured, and explicit instruction - those that indirectly influenced reading through instruction in strategies - or instruction in metacognitive skills -, and those that focused on a combination of both. Included were 92 studies that involved treatment in reading and had measures of word recognition and reading comprehension as dependent variables. On average, the WC intervention studies had more treatment sessions than the RC studies (M = 38.60, SD = 42.12 vs. M = 22.78, SD = 23.30, respectively). Conversely, the time for each instruction session in RC was longer (minutes: M = 38.94, SD = 29.20 vs. M = 43.14, SD = 26.45).

The initial analysis of effect sizes of the studies classified by intervention domain (WC and RC), done using Cohen's d, showed a large effect size - 0.72, with the threshold to be considered a large effect being 0.80 - for a prototype intervention study in RC. The effect size is largest in studies where comprehension was assessed using researcher-created measures (0.81) and smallest when normed measures were used (0.45). The prototypical intervention in WC, on the other hand, produced a moderate effect size (0.59). When effect sizes were analyzed according to the instructional model, it was found that the combined model produced larger effect sizes for RC, and the direct instruction model showed larger effect sizes for WC.

Another set of analyses allowed us to identify only a few instructional components that effectively contributed to treatment effectiveness, regardless of the overall instructional model: (1) teacher-directed questioning - dialectical teaching, in which the teacher asks students questions or directs them to ask, thus engaging everyone in a dialogue -, (2) difficulty control of task processing demand - tasks sequenced from easy to difficult, and teacher demonstrating the task and offering assistance -, (3) elaboration - when the teacher offers additional information and explanations -, (4) demonstration - when the teacher offers a model of the steps to follow in front of the task -, (5) small group instruction, and finally, (6) strategy suggestions by the teacher. Hierarchical regression analysis also showed that the last two variables were the ones that contributed significantly to the variance in RC scores.

Instruction in strategies for students with learning disabilities

^{Sencibaugh's (2007}) meta-analysis addresses intervention studies with metacognitive instructional strategies targeting students with learning disabilities. Among the studies - collected between 1985 and 2005 - 15 were included, which met the criteria: experimental design, samples of students from 1st grade to 12th grade, and who received the treatment to increase RC ability. The duration of the interventions ranged from 1 day to 12 weeks, with session time ranging from 25 to 95 minutes.

The variety of strategic instructions was categorized into two types: strategy dependent on visual resources (visual strategies) and strategy dependent on auditory or language resources (auditory strategies). The first involves using pictures or visual skills — such as visual attention therapy, text illustrations, and graphic semantic organizers. The second strategy, termed "auditory," encompasses the use of language in pre-reading activities or post-reading exercises to aid comprehension-such as main idea strategy, summarization, paragraph rephrasing, story retelling, and strategies based on text structures. The meta-analysis found a large effect size for visual strategies (ES = 0.94) and an even larger effect size for auditory strategies (ES = 1.18). Among these, questioning strategies involving self-instruction, paragraph reformulations, and text structure-based strategies produced the most significant results.

Instruction in reading strategies applied by the teacher in the classroom

The meta-analysis by ^{Okkinga et al. (2018}) also sought to verify the effectiveness of interventions in reading strategies, specifically in classroom settings, thus in a situation where the teacher himself applies the intervention, unlike previous meta-analyses — in which the researchers themselves instruct small groups. In this case, the authors refer to previous meta-analyses that, by including studies conducted with the whole class, highlighted the moderating effect of who applies the intervention, with teachers being less successful than the researchers themselves in ensuring progress in students' comprehension skills. Possible explanations for this include the teachers' difficulty in understanding the strategies to be applied, the didactic complexity of combining teaching the strategy, demonstrating how it works, and monitoring the groups when the strategy provides for small groups of students interacting with each other.

Fifty-two studies were included in the meta-analysis that, among other inclusion criteria, met the requirement of comparing groups - the experimental group, which underwent the intervention, and the control group, which did not undergo the intervention. The outcome variables of the intervention studies were measures of RC and measures of strategic skills. The reading strategies coded, and whose presence or absence was verified in the studies, were subdivided into "before reading" (e.g., setting the goal to be achieved in reading, predicting the content, and activating prior knowledge), "during reading" - questioning, paraphrasing, making inferences, underlining - and "after reading." As for the duration of the interventions, the mean in hours was 47 hours (standard deviation 55), and the mean in weeks was 17 weeks (standard deviation 11).

Very small effect sizes were found on RC when measured using standardized tests, immediately after the intervention (Cohen's d = 0.186) and sometime after the intervention ended (Cohen's d = 0.167). When comprehension was measured through researcher-developed tests, the effect size was larger, but still small at the immediate assessment (Cohen's d = 0.431), and large at the late assessment through researcher-created tests (Cohen's d = 0.947). Regarding moderating effects in terms of intervention characteristics, setting the reading objectives was shown to have an important contribution to the effect sizes on the outcome variables. Regarding moderating effects in terms of student characteristics, effect sizes of interventions were found to be larger on the strategic skills of students with difficulties - compared to those with typical development. Grade also had a bearing on the effect sizes of the interventions, with those being larger for students between the 6th and 8th grades.

Instruction in writing activities to foster reading comprehension

^{Graham and Hebert (2011}) examined the effectiveness of the writing activity as a tool in improving students' reading, starting from the theoretical perspectives that explain the possible impact of writing on reading. Writing about the text facilitates its comprehension as it provokes the explicitness and integration of the most important ideas from an active position in front of the text, which leads to the selection of information and the transformation and manipulation of the language of the text into one's ideas. The fact that writing characterizes a visible and permanent record facilitates revision, criticism, and the connections that make understanding deeper and deeper. Additionally, there is the theoretical principle that, although distinct, reading and writing activities share certain skills and processes, so that instruction in the process of writing could facilitate comprehension of read texts — for example, by being taught to write complex sentences, the student would more easily understand these types of sentences when read. Finally, it is believed that the process of composing a text to be read by others provides the student with insights that facilitate the understanding of others' texts (i.e., to write he needs to make premises explicit, obeying a logic, which may leave him more attentive, searching for these same questions during the reading of a text).

The meta-analysis included experimental or quasi-experimental studies involving treatments in which elementary school students — starting in 4th grade — and high school students wrote about what they read, were taught to write, or underwent an increase in the amount of writing tasks. The studies, which were to have a measure of RC as the dependent variable, were categorized according to the type of writing activity: responding in writing to questions about the text or devising written questions about the text; taking written notes; making summaries; making interpretive, analytical, or explanatory written comments; and writing short answers about the text. As a result, the meta-analysis indicated that writing about the material read improves comprehension of the text, which proved true for both general students and students who were not very skilled at reading and writing. The effect size was 0.37 in studies using standardized reading tests, and 0.50 in studies using RC measures developed by the researchers. The beneficial effect of writing activities was confirmed for both expository and narrative texts. Also, four activities were shown to be effective: responding in writing elaborately, writing summaries, taking notes, and answering and/or asking questions. Finally, it was seen that increasing the time that typically developed, 1st through 6th-grade students engaged in writing activities had a positive effect (effect size = 0.35).

The same authors, ^{Hebert, Gillespie, and Graham (2012}), conducted a subsequent meta-analysis in which they sought to compare the effects of different writing activities on comprehensive reading. The study synthesized the results of four comparisons: Summary versus Developing and Answering Questions; Answering Questions versus Taking Notes; Summary versus Taking Notes; and Answering Questions versus Extended Writing. On the one hand, the mean scores did not indicate that one type of writing activity favored RC more than another. Nevertheless, the effectiveness of writing varied according to the test that assessed reading comprehension: in the Summary versus Answering Questions comparison, the former activity was more effective in assessing by free recall, and in the Extended Writing versus Answering Questions comparison, the former type of writing activity was more effective in assessing by making elaborate responses about the text.

As for the implications for educational practice, based on this second meta-analysis (^{HEBERT; GILLESPIE; GRAHAM, 2012}), the researchers concluded that the study did not produce evidence to recommend that teachers use one writing activity over another. Rather, the recommendation was that they combine the activities considered in their classrooms.

Instruction in expository text structures for students at different reading levels (with difficulty, without difficulty, and at risk)

The expository texts used in school to present the most different curricular contents admit a classification. According to the information passed on and the didactic objectives, one can find texts of descriptive, grouping, causal, clarifying or comparative structure. By recognizing the type of text, students can identify some indicators in order to anticipate the information and facilitate interpretation because there are expressions typically used in the different structures: a causal text, for example, may contain keywords or expressions such as "The reason why [...]; due to [...]; for the fact that [...]" (^{SOLÉ, 1998}).

^{Hebert et al. (2016}) conducted a meta-analysis on expository text structure instruction's effect on student comprehension. They even pointed out that many researchers consider two factors possibly influential in the effectiveness of text structure instruction: the characteristics of the students who might benefit from the such instruction-their reading level or their grade-and the approaches and strategies used in instruction.

This meta-analysis examined 45 studies conducted with school-aged participants in grades 1 through 12. As a result, teaching students about the structures of expository texts improves RC in all comparison groups used in the studies that comprised this paper. A relevant finding is that instruction in more text structures resulted in larger effects: there was a 0.13 standard deviation increase in effect size for each new structure taught after the first. A qualitative examination of the text types and structures taught, the order in which they were taught, and groupings of text structures revealed little variation in the approach to teaching multiple text structures. Thus, teaching more structures and comparing/contrasting would also lead to greater gains. Thus, it seems logical that knowledge in more text structures, not just one, would improve comprehension.

Another important finding of the meta-regression model was that writing appeared to predict intervention effectiveness significantly. In this study, writing was considered to include note taking, sentence writing — for instance, short answers to questions about the text — and writing paragraph-sized answers about the text. Studies that included writing had an increased effect size — mean of 0.38 standard deviation units.

Individual instruction for students at risk: components of decoding and comprehension

^{Elbaum et al. (2000}) conducted a meta-analysis of empirical findings related to individual instruction provided by adults to early-grade children at risk for reading failure. Twenty-nine studies were pooled, conducted between 1976 and 1998, with interventions lasting between eight and 90 weeks, and total instruction time between eight and 150 hours. The intervention types, corresponding to the independent variables, were coded as: WR (word reading); RC; mixed (WR and RC); phonemic awareness; visual perceptual skills; unspecified. The outcome variables were: RC; spelling; decoding, oral reading of words, oral reading of passages, subtest-based task compositions, and writing; listening comprehension; written vocabulary; phonemic awareness.

Reported effect sizes were largest in the face of RC (2.41) and mixed (0.50) instruction. Regarding the outcome variables, written vocabulary (0.94) and listening comprehension (0.68) had the largest effect sizes. Regarding the duration of the intervention, the effect sizes were larger for interventions carried out in up to 20 weeks, with an average of 63 instructional hours (0.65), than those with more than 20 weeks, with an average of 61 hours (0.37). Interventions delivered by college students had the largest effect size (1.65) compared to those delivered by teachers (0.36), community volunteers (0.26), or paraprofessionals (0.68).

In summary, it was revealed that students who received individual instruction demonstrated performance of one standard deviation higher than that of the control group. The focus of the intervention on RC had the greatest effects for older students, while instruction with a focus on phonemic awareness or of the mixed type had moderate effects. The meta-analysis also showed that trained volunteers and college students could provide significant help to at-risk students, indicating the possibility of cost savings in offering instruction to children. Importantly, however, this instruction should be supplemental to classroom instruction provided by teachers, not a substitute for it. The findings on small group interventions have also been positive: when highly qualified teachers implement a well-developed intervention, students experience the same academic benefits regardless of whether they are taught individually or in smaller groups.

Reading instruction and its long-term effects: components of decoding and comprehension

The meta-analysis ^{Suggate (2016}) developed included experimental and quasi-experimental studies of reading interventions, focusing on phonemic awareness, phonics, fluency, and comprehension approaches. Unlike the previous studies, Suggate dealt with studies that included long-term post-intervention follow-up. In addition to coding the type of intervention and the risk status for the sample's reading difficulty, the intervention's methodological characteristics were coded to clarify why the effect size changes from post-test to follow-up.

Regarding the approaches of the interventions, four types were coded. The phonemic awareness intervention specifically addresses the perception of sounds at the phoneme level - boi as b/o/i. The phonemic intervention focuses on associations between phoneme and spelling, incorporating letters or text. Fluency intervention works on the ability to read with speed and accuracy. The comprehension intervention generally involved metacognitive strategies, such as reflecting, summarizing, and predicting.

The main results of this meta-analysis, which included 71 intervention studies using a control group, are presented below: Regular readers were shown to lose their advantages over the control group at follow-up (ES = 0.13) and researcher-administered interventions resulted in larger effect sizes (ES = 0.34) at post-test. Long-term effect sizes were larger in the higher grades, 3rd through 6th grade (ES = 0.43). Phonemic awareness interventions result in larger effect gains at follow-up, while phonics and fluency interventions lost more effect at follow-up. In addition, fluency interventions did not result in good transfer to reading comprehension, with benefits more directed toward decoding and reading skills at follow-up. The largest effect sizes at follow-up resulted from interventions with comprehension components that were worked on with the older children.

Considering these meta-analyses presented, it is important to compare their evidence and reflect on the implications for educational practice. In fact, as mentioned before, the transposition of scientific results to the teaching activities in the classroom requires mediation and a new elaboration. In order to clarify this issue, we will discuss the main indications of the meta-analyses below, seeking a reflective synthesis that integrates the main findings.