Separating the Different Domains of Reading Intervention Programs

Providing a child with reading difficulties with the appropriate reading intervention as early as possible is critical to prevent future academic failure. As reading is composed of several sub-components (phonology, orthography, fluency, comprehension), choosing the appropriate intervention may be confusing. Here, we attempt to provide an up-to-date review of different reading intervention programs and their outcomes that currently are available for children 4 to 16 years of age. We also introduce the possible beneficial effect of including a component of executive-functions training to reading curricula to enhance the effects of reading intervention programs. These programs are separated by the sub-components of reading that each is designed to address, with discussion based on several leading models for reading acquisition. Our aim is to direct educators, professionals, and researchers to the most appropriate intervention according to either their domain of interest or the child’s needs.

The phonological processor is specialized for perceiving, remembering, interpreting, and producing the speech-sound system of a person's language (Moats, 2009), which enables the appropriate matching of the grapheme to the corresponding phoneme. A deficit in the phonological processor can result in an inability to correctly identify, recall, manipulate, and/or remember phonemes (Moats, 2009). The orthographic processor recalls the letters and letter sequences in the words to be written, and forms word templates in the reader's mental lexicon (see also "self-teaching" theory; Share 1995). A deficit in this domain will result in slower reading as well as poorer spelling and is tightly related to fluency difficulties (Breznitz, 2006). Fluent reading is defined as the overall timing and smoothness in reading, which is attributed to speed-of-processing abilities (Bowers & Wolf, 1993;Breznitz, 2006) and/or the level of proficiency with the phonological/orthographical routes (Seidenberg & McClelland, 1989). Impaired reading fluency may occur due to deficits in working memory (Swanson & Siegel, 2001), rapid automatic naming (RAN; Wolf, Bowers, & Biddle, 2000;Wolf & Katzir-Cohen, 2001), or timing (Bowers & Wolf, 1993) and/or an asynchrony between the auditoryphonologic and the visual-orthographic processing speeds (Breznitz, 2006;Breznitz & Misra, 2003). Reading comprehension is defined as the ability to extract and construct meaning through interaction with the written language (Snow, 2002). Impairment in this domain will harm reading comprehension on several levels: verbatim, gist, metacognitive questions, and conclusions (Nation, 2008). In 1974, LaBerge and Samuels pointed out the importance of automated technical reading (i.e., without investing attention resources), which involves fluent word recognition to enable the allocation of attention resources toward comprehension. The authors described the involvement of several key memory components in technical reading: visual, phonological, episodic, and semantic memory. Due to the characteristics of the attention system and based on the model for attention by LaBerge and Samuels, attention cannot be fully allocated both for decoding (visual memory, phonological memory, and episodic memory) and to comprehend the meaning (semantic memory), and therefore, to improve comprehension, technical reading should be automatic (LaBerge & Samuels, 1974). It is important to note that a major portion of the "technical" reading phase also relies on basic executive functions, including memory, attention, and speed of processing (i.e., automaticity).
The integration of the phonological, orthographical, and semantic components of reading is well described in the parallel-distributed processing (PDP) model (Seidenberg & McClelland, 1989). Word reading requires decoding that involves perception of the word in the visual modality and recoding of its sounds in the auditory-phonological system, and then, the semantic representation of the word from the mental lexicon is evoked (Breznitz, 2006;Seidenberg & McClelland, 1989). As a skilled reader acquires word reading skills, their decoding process becomes automatic and accurate (Van der Leij & Van Daal, 1999), resulting in few decoding errors. Support for this model has been provided from the neuroimaging field, illustrating simultaneous activation of brain regions related to phonology, orthography, and semantics (Lyon, Shaywitz, & Shaywitz, 2003). In the "Construction-Integration" model, contextual reading is organized into construction and integration phases. The construction phase focuses on decoding a single word, during which its semantic meaning is elaborated in a bottom-up process. The integration phase entails the integration of words into sentences, paragraphs, and stories, and is based on previous knowledge and context. It is conceivable that the integration phase processes information in a top-down manner, which relies on the skilled reader's ability to identify words that are variable in font/handwriting and words with different meanings in different contexts (Rayner & Pollatsek, 1989). Based on Kintsch's (1988) model, it can be concluded that readers with reading difficulties are more susceptible to errors in the construction phase than are skilled readers because of their phonological deficit (see also Snowling & Nation, 1997), whereas skilled readers are more prone to errors in the integration phase because of the more complex processes involved in top-down processing and the lack of deficits in the underlying reading mechanisms in these readers. For skilled readers, when reading becomes automatic, it also becomes effortless, and a trade-off occurs with highercontrol processes (Walczyk, 2000).
Chall describes reading as a multi-level process, where each component is integral to the development of the next (Chall, 1983). The first level of the Chall model is "learning to read," whereas the second level is "reading to learn." The first level is divided into several stages: (a) pre-reading stage (0-6 years old), (b) predecoding stage (first and second grades), and (c) fluent-reading stage (second and third grades). The second level is composed of reading to learn (extracting knowledge from the text; fourth-sixth grades), reading for different perspectives (high school grades), and building new knowledge (high school and college). The Chall model emphasizes that phonology is the most basic component, followed by orthography (or holistic word reading), and then an automatic mastery of both of them resulting in fluent reading. Only when the child reads fluently can he or she move to the comprehension level.
For this review, we attempt to provide researchers, clinicians, teachers, and parents with an up-to-date overview of the research-based reading intervention programs currently available for children ages 4 to 16 years. Using a PubMed search (U.S. National Library of Medicine, 2015), we have included only research-based articles describing reading intervention programs for this age range (key words were "reading intervention" and "children"). The biomedical perspective of the programs reviewed broadens the scope of existing reading program reviews created by the U.S. Department of Education's Institute of Education Sciences (IES). Founded in 2002 to promote informed educational decision making, the IES What Works Clearinghouse website provides educators, legislators, and community members with a central source for research-based educational programs, including reading intervention programs, from an educational perspective (U.S. Department of Education, 2005). Additional reviews of reading intervention programs are also available through national and state Response to Intervention (RTI) search engines, such as the American Institutes for Research (2015) Center for Response to Intervention. However, these reviews do not provide a review on only research-based interventions and the actual quantitative information reflecting the level of improvement, and they do not separate the interventions according to the different aspect of reading they are treating (i.e., phonology, fluency, comprehension, multi-component). Our review aims to direct educators, professionals, and researchers to the most appropriate intervention according to either their domain of interest or the child's needs.
An attempt was made to divide the presented intervention programs into one of the three main reading sub-components (Table 1-phonology,Table 2-fluency, Table 3-comprehension). As intervention programs that train orthography use context to train this domain, we included these programs under "fluency." We also included programs that involve several components (Table 4-multiple components) and the outcomes that are available for each of these programs in the following domains: phonology, orthography, fluency, comprehension, and other general domains if reported (spelling, motivation, knowledge, and others). Results in the tables report the effect measured immediately after intervention. We then attempted to conclude as to a preferable trained reading domain that results in many improved reading domains.
Phonological processing interventions, such as Lindamood-Bell, code-based classroom instruction, and Phono-Graphix, result in improvements in phonology, orthography, and comprehension, whereas reading fluency (contextual accuracy and speed) was not affected (Table 1). The lack of the "timing" component (i.e., the encouragement of faster contextual reading) that is absent in the listed interventions may contribute to the lack of changes in the fluency domain. However, the ages of the participants listed in some of the phonology-based interventions that did not show an improvement in fluency were less than 7 years (i.e., Graphogame and the classroom instructions). Also, in the intervention program that did have an effect on reading fluency (i.e., Lindamood-Bell), the intervention lasted 2.5 years as opposed to 12 to 16 weeks for the other two interventions, which may have had an effect on the development of fluency skills.
Interestingly, the number of research-based intervention programs is limited. To date, we found that only the Reading Acceleration Program (RAP), which is a computer-based intervention, has a published effect on orthography, fluency, and comprehension (Table 2). The effect was found in a variety of orthographies (i.e., German, Hebrew, and English) and different age groups (6-12 years), with reports in adults as well . The effect of this intervention was found after 3.5 to 8 weeks of training.
A variety of intervention programs are focused on comprehension (Table 3). These programs vary from listening comprehension programs that are computer-based (van Kleeck, Vander Woude, & Hammett, 2006) to a teacherbased setting (Garner & Bochna, 2004). Some programs, such as Concept-Oriented Reading Instruction (CORI), improved comprehension and other domains not related to reading (e.g., reading motivation, strategies, knowledge, searching, and more). Only the Guided Reading (GR) intervention had an effect on the phonology, orthography, and comprehension domains related to reading, but without a direct effect on fluency. Interestingly, GR was the shortest intervention compared with the other interventions listed in the table (i.e., only 6 weeks), but the participants were the oldest (9-14 years of age) Multi-component interventions are both computer based as well as teacher based, with a major difference in the training time frame (4 weeks to 3 school years; Table 4). The only intervention that showed an improved reading on all four examined domains of reading (phonology, orthography, fluency, and comprehension) was the Multi-component Tier 2 intervention for 9-to-12-year-old children, which lasts from a couple of weeks to a couple of years. Another program for this age group is computer based for 6 weeks (i.e., video instructions; Xin & Rieth, 2001), which only showed an effect on orthography and comprehension.

Discussion
The purpose of this current-literature review was to provide an overview of research-based phonological, fluency (includes orthography), comprehension, and multi-component reading intervention programs. The intervention programs we have reviewed include those for children aged 4 to 16 years, both computer-based and teacher-based in each of the following reading domains: phonemic awareness, phonology/decoding, orthography, fluency, and comprehension. Other outcome measures, if given in the original studies, are also provided. Due to the different measures in each study, we are unable to provide a meta-analysis that compares the main effects of the intervention programs to identify the most effective intervention. However, based on the literature review, greater gains in all levels of reading measures and comprehension are achieved when using techniques that specifically address reading fluency (such as the RAP; Horowitz- . We base our conclusions on the reports of increase in orthography, fluency, and comprehension abilities (i.e., all components involved in contextual reading; see Horowitz- Horowitz-Kraus, Cicchino, Amiel, Holland, & Breznitz, 2014) following a fluency training.
Reading fluency recently has been acknowledged as a composite of key cognitive processes (phonological, orthographical, and semantic processes, as well as more basic higher-order abilities such as executive functions [attention, working memory, and speed of processing]; Benjamin & Gaab, 2012;Berninger, Abbott, Billingsley, & Nagy, 2001;Wolf & Katzir-Cohen, 2001). Only a synergy in the activation of all of these components results in fluent reading (Breznitz, 2006). Recent neuroimaging studies also support this concept; fluent-reading training resulted in increased activation in phonological, orthographical, and semantic regions, as well as executive functions-related brain regions (Horowitz- Horowitz-Kraus et al., 2014;Horowitz-Kraus & Holland, 2015). We suggest that by encouraging fluent reading, more information units can be processed in a given time, and reading fluency training "releases" the speed of the cognitive bottleneck resulting from the slow speed of processing and deficits in working memory demonstrated by children with reading difficulties Length of sessions GG Rime mean raw score increased from 6.8 at T1 to 10.0 at T2 for Rhyme Oddity task, GG Phoneme increased from 10.0 to 11.2, and control group increased from 8.9 to 9.5 Phonology: GG Rime group read seven more non-words on the TOWRE Phonemic Decoding Efficiency test, GG Phoneme group read five more non-words, and control group read four more non-words at T2 than at T1 Orthography: GG Rime group read 14 more words on the BAS II, GG Phoneme read 11 more words, and control group read 10 more words at T2 than at T1 GG Rime group read 10 more words on the TOWRE Sight Word Efficiency subtest, GG Phoneme read nine more words, and control group read seven more words at T2 than at T1 Spelling: GG Rime group correctly spelled seven more words on the BAS II, GG Phoneme spelled five more words, and control group spelled four more words at T2 than at T1 Code-based classroom instruction and a supplemental maintenance intervention.    (Breznitz, 1997a(Breznitz, , 1997b Computer 6-7 (typical readers)    Gains noted in phonological skills and remained 2 years post-training In the 1-and 2-year follow-up, there were no significant effects on word reading at any age/grade Additional research is suggested to look at interventions that will provide a carryover from phonological skills to word reading    Children who read twice showed an increase of 22% of instructed words and 12% of non-instructed words after intervention. Children who read 4 times showed an increase of 24% of instructed words and 15% of non-instructed words after intervention.
(continued) Treatment group correctly answered 3.59 more questions post-test than pre-test. They correctly answered on average 8.03 questions post-test (max score = 15). There was no interaction between instructional condition and testing time. Sentence Cloze: Treatment group correctly completed 9.43 more sentences post-test than pre-test. They completed on average 14.46 sentences post-test (max score = 30). There was no interaction between instructional condition and testing time.
Note. All three measures were teacher-made tests. Test-retest interrelated correlation coefficients were all within the coefficient of stability acceptable for a teacher-made test.
The definition method: Learning novel words by being told their meanings.
The context method: Teaching of a strategy for deriving meanings from written context (clue words).
Trains vocabulary and comprehension (Nash & Snowling, 2006)    Note   (Breznitz & Share, 1992). However, until inferential statistics using experimental research methodologies come into play concerning functional magnetic resonance imaging (fMRI), we must be careful not to overgeneralize correlational-based findings (Goswami, 2008). As noted, it is now acknowledged that in addition to the linguistic domains that reading relies on (phonology, orthography, and semantics), fluent reading will be achieved by also relying on intact executive functions (Brosnan et al., 2002). A deficit in several domains of executive functions is common in individuals with reading difficulties and, specifically, in those with dyslexia. These challenges included difficulties in attention (Facoetti, Paganoni, Turatto, Marzola, & Mascetti, 2000;Shaywitz & Shaywitz, 2008), inhibition (Brosnan et al., 2002), working memory (Ackerman & Dykman, 1993;Helland & Asbjornsen, 2004;Swanson & Ashbaker, 2000), shifting , and self-monitoring (a deficit in error monitoring in nonlinguistic and linguistic domains; Horowitz- Kraus & Breznitz, 2008, 2009). However, although executivefunctions training has been shown to improve school readiness (Diamond & Lee, 2011), the number of studies looking at the effect of executive-functions training specifically on reading ability is limited (Franceschini et al., 2013;Horowitz-Kraus & Breznitz, 2009). Breznitz and colleagues (2013) demonstrated that an executive functions-based fluent-reading program that trains reading by forcing the reader to allocate his or her visual attention toward the written material in a speeded manner (i.e., the RAP) had a greater effect on reading fluency and comprehension as compared with the same program without the speeded manipulation. We suggest that a short executive-functions "warm-up" training prior to a reading session may be beneficial for better reading outcomes for children with reading difficulties. However, future research should verify this point with in-depth study.
Interestingly, in addition to the observed improvement in the fluency and multi-component domains, it seems that training with the comprehension-based reading intervention GR also shows an improvement in all reading components. GR is a research-based instructional strategy that has proven to be effective in all sub-domains of literacy. This smallgroup, homogeneous teaching approach provides students with intentional and intensive literacy support that is differentiated to meet their unique learning needs (Fountas & Pinnell, 2012). Daily GR lessons use a balanced literacy approach and include explicit instruction in fluency, oral language, and vocabulary development; direct instruction in phonemic awareness and phonics; and opportunities to write about reading (Pinnell & Fountas, 2010). We relate this improvement to the fluency domain that is trained using the GR program (Pinnell & Fountas, 1998). Regular formative reading assessments that measure rate and accuracy, both components of reading fluency, are also recommended components of GR (Pinnell & Fountas, 1998). In-depth research into the speed elements involved in GR, compared with the effect of a fluency-based intervention program (such as the RAP), is warranted and can verify whether the fluency component drives the positive effect of GR or whether it is due to the inclusion of other reading domains as well.
In addition to the improvement of all reading domains following fluency training, our review demonstrates that multicomponent reading intervention programs that remediate different combinations of sub-components of reading result in a massive improvement in several domains of reading (see Table 4). Some examples for such programs are reading recovery, multi-components, treatment mastery, word building, Auditory Discrimination in Depth (ADD), and Fast ForWord (FFW). The reason for the positive effect of the multi-component intervention programs is obviously the treatment provided in each of the examined reading domains. However, probably due to this reason, the intervention time ranges from a couple of weeks to several years. These programs may use some of the training time to train nonimpaired domains that are part of the program curricula instead of devoting time specifically to the reading domains in which each individual child shows the greatest impairments. As mentioned, the most effective program (Multicomponent Tier 2, by O'Connor, Harty, & Fulmer, 2005) lasts about 3 years. This long intervention period should be considered if a limited time is available for intervention or if a short-term intervention is needed. Again, due to our inability to compare standard scores across studies, we cannot compare the effect of the RAP fluency training program with that of the Multi-component Tier 2 program. Although in the clinical arena the use of non-standardized tests for the use of intervention planning and instruction is acceptable, for the purpose of our review, it is challenging to objectively define the effect of interventions between the different studies. A future study should verify this point in depth to examine which program will have a greater effect on all reading domains. Also, educators and parents should pay particular attention to each child's needs, according to the individual's reading and cognitive ability relative to the expected agematched scores, and invest their time to remediate the specific weaknesses of each child.
The effect of reading intervention administered by computer versus an intervention delivered by a tutor or teacher is yet to be clearly understood. Obviously there are advantages for a personal connection and the established relationship that a teacher has that can affect the child's overall performance. However, when the human interaction is limited, as often is the case in schools with large class sizes, computer programs have some advantages, such as the lack of judgmental feedback, the potential positive effect of working in a group with peers, and the capacity to provide manipulations that may be impossible without using a computer. Other advantages of computer-based instruction programs are the ability to use digitized speech (Foster, Erickson, & Foster, 1994) and the tailored instruction the child receives combined with an individualized feedback, which may increase motivation (Speziale & La-France, 1992). Mioduser, Tur-Kaspa, and Leitner (2000) showed that a computer-based versus a teacher-based reading instruction for 5-to-6-yearold children at risk of developing reading difficulties resulted in significantly higher scores in phonological awareness, word recognition, and letter-naming skills in children using the computer-based instruction versus their teacher-instructed peers. These results highlight the relative benefit that a computer-based instruction may have on reading acquisition, such as learning the letter-sound relationship using the special auditory/visual characteristics of the computer environment. Future studies to examine the effect of the instruction and the same curricula administered by computer versus a teacher, even in typical developing children, may verify which condition is better than the other. Alternatively, these two types of instruction/intervention modalities could be complementary to one another, a point that should be examined in depth.
Reading intervention is a labor-intensive process. A large number of studies recommend intensive phonological-based intervention programs, even for a couple of years (e.g., Alphabetic Phonics, Orton-Gillingham). Educators, reading specialists, and parents are seeking the most effective, yet efficient, intervention for their student/child to not lengthen the gap in reading achievements. Effective reading intervention, as explained by Nicolson and colleagues (1999) needs to be systematic and comprehensive, which may be costly, but also cost-effective. Despite the desire to achieve an effect as soon as possible, especially due to the overall concern regarding the quality of the American education system and the economic consequences of long interventions (National Commission on Excellence in Education, 1983; Task Force on Education for Economic Growth, 1983), it is important to remember that reading is a developmental, explicitly acquired ability. Each domain in it relies on an intact acquisition of the previous one. Knowing which reading domain should come next may guide the choice to a specific intervention. The Frith (1985) reading acquisition model describes how children transition from stage to stage in a Piaget manner until reading is completely mastered; from contextual to logographic reading, to partial and full decoding, and then to the orthographical stage of automatic word recognition. Interestingly, both the Frith model and the Chall (1983) developmental model describe automatic word recognition, which leads to fluent reading, as the stage at which children master reading and can devote their attention to comprehension (see also LaBerge & Samuels, 1974).
One limitation of the current review is that it describes the results as reported by other researchers, who used different tests to measure the effectiveness of the examined intervention. Hence, we are unable to report the results of a metaanalysis among the different domains. Such meta-analysis would have to take into account the variety of tests that were used to assess reading improvement, most of which are very different from one study to the other, as well as the differences in training intensity and the involvement of a tutor (or alternatively, a computerized program), and the different ages of the children who trained on the program. However, such meta-analysis has the potential to provide indepth and quantified information regarding the reading intervention program or programs that affect the most components of reading.