Examining the Word Level Skill and Reading Comprehension Profiles of Adolescents with and without Specific Learning Disabilities

Abstract

This study examined the heterogeneity of literacy profiles for adolescents with and without Specific Learning Disability (SLD). Student subgroups displaying common patterns of performance in word level skills and reading comprehension were identified through latent profile analysis. Results indicate most of the total sample demonstrated below average performance in one or both areas with word level skill difficulties being more common than difficulties in reading comprehension alone. Changes in reading performance by profile over time (Grade 6 to 8) were examined through a latent transition analysis revealing consistent patterns in the SLD sample and variable patterns in the typically developing sample. Resulting profiles were utilized to predict performance on an end-of-year broad reading comprehension measure indicating very little change in performance over time. Findings suggest large numbers of adolescents with concurrent word level and reading comprehension difficulties likely need sustained intervention in word level skills to support their reading comprehension.

As adolescent readers progress through secondary grades, they face increasingly difficult demands requiring them to read and apply knowledge from complex texts. This level of reading for understanding may be markedly difficult for students that possess weaknesses in one or more of the core skills necessary for reading comprehension. The Nation’s Report Card (2019) makes evident the existence of reading comprehension difficulties in adolescent students by reporting that only 34% of eighth-grade students scored at or above Proficient level on the 2019 National Assessment of Education Progress (NAEP) Reading Assessment.

Research has revealed that proficiency in multiple skills, including foundational text reading skills, as well as receptive and expressive language, are required for accurate reading comprehension. This is evident in the empirically validated framework of reading comprehension known as the Simple View of Reading (SVR; Gough & Tunmer, 1986), which describes reading comprehension as the product of decoding and language comprehension. The relationship between decoding and language comprehension can vary based on a few reasons, one of which being the number of years the individual has been reading (Florit & Cain, 2011). This specifically applies to adolescents because the strength of the relationship between decoding and reading comprehension decreases around the age of 10 (García & Cain, 2014). For novice, early readers, decoding has the largest influence on reading comprehension, but language comprehension becomes increasingly more important as students’ decoding skills develop and the texts they read become more complex (Language and Reading Research Consortium, 2015; Tilstra et al., 2009). This trend is likely because the relative importance of decoding and language comprehension changes based on students’ level of reading development and the complexity of the texts they are reading (Lonigan et al., 2018).

The Simple View of Reading underscores the importance of decoding and language comprehension; however, it does not thoroughly define all the component skills within these domains that influence comprehension. Better insight into the specific component skills required for reading comprehension, specifically in adolescent and older readers, is provided by the direct and inferential mediation (DIME) model of reading comprehension. The DIME model of reading comprehension (Cromley & Azevedo, 2007) extends the SVR by breaking comprehension into its component parts. The DIME model hypothesizes that the relationship between five components are responsible for reaching comprehension including background knowledge, vocabulary, inference, strategies, and word reading. The importance of word reading, vocabulary, and background knowledge are stressed because they directly affect reading comprehension and indirectly affect the reader’s ability to utilize comprehension strategies and draw inferences.

While the DIME model of adolescent reading does not specifically include a spelling construct, there is some agreement that orthography is related to skilled reading. Research has demonstrated a moderate to strong relation between decoding and spelling at the grapho-phonemic level (Robbins, et al., 2010). More complex grapho-phonemic patterns, like those encountered by readers in secondary grades, have higher correlations than those of less complex patterns, like those encountered by readers in elementary grades. When students read words, their memory retains word-specific knowledge that then contributes to their ability to spell words (Ehri, 2000). Spelling has been shown to have a small direct effect on reading comprehension that increases as students’ progress from sixth grade to tenth grade (Reed et al., 2016). Additionally, it has been found that spelling mediates the relationship between vocabulary knowledge and reading comprehension for students in this age group.

Research has demonstrated that improving the reading comprehension skills of adolescents, especially on standardized measures of reading comprehension, is quite difficult (Scammacca et al., 2015; Solis et al., 2014; Wanzek et al., 2013). Reading comprehension ability can break down for a multitude of reasons (Snow, 2002), but when focusing on the SVR to determine the origins of reading comprehension difficulties, the most obvious origins of such difficulties would include problems connected to language comprehension and/or word reading problems. Within the context of the DIME model of reading comprehension, all five components made significant contributions to reading comprehension for older students including word reading which had a small, significant direct effect (Cromley & Azevedo, 2007). In adolescents, extant research has shown that the majority of students with reading comprehension difficulties commonly have additional difficulties in more basic word level reading skills (Cirino et al., 2013).

The Heterogeneous Nature of Reading Skills Among Adolescents with Reading Difficulties

Studies investigating the component reading skills of adolescents with reading difficulties suggest a heterogeneity of profiles (Brasseur-Hock et al., 2011; Cirino et al., 2013; Clemens et al., 2017). Brasseur-Hock et al. (2011) investigated reading skill profiles of adolescent struggling readers (n=319) on three standardized measures of reading comprehension and eight standardized measures of component skills including vocabulary, listening comprehension, word- and text-level reading accuracy and fluency. Importantly, five profiles of component skills emerged among below average comprehenders distinguished by their specific strengths and weaknesses. Findings revealed subgroups of students with specific weaknesses in listening comprehension, reading comprehension, fluency, and students with moderate and severe levels of global weaknesses. Cirino et al. (2013) investigated the pattern of overlap between middle school typical readers (n=723) and those with reading difficulties (n=1,025) on measures of decoding, fluency, and comprehension. They found that the majority of sixth to eighth grade students with reading comprehension difficulties also had difficulties in decoding or fluency. Their findings also revealed a varying relation between reading components for students with reading difficulties.

Clemens and colleagues (2017) investigated the prevalence of reading fluency and vocabulary difficulties among 180 adolescents in Grades 6 though 8 with low reading comprehension. Findings revealed four distinct score profiles with 96% of the sample demonstrating difficulties in one or both areas. The largest group, which was comprised of over half the sample (n=103, 57%), included the low fluency and low vocabulary profile. Two mixed profiles included less than a quarter of the sample: low fluency and average vocabulary (n=41, 23%); average fluency and low vocabulary (n=28, 16%). Lastly, a very small portion of the sample, only 4% (n=8), fit the profile for average fluency and average vocabulary. Altogether, these studies suggest that the majority of adolescent readers with reading comprehension difficulties, regardless of disability status, display areas of weakness in component skills including word reading and fluency.

The existence of heterogeneous profiles is also evident in adolescents with late emerging reading difficulties (Catts et al., 2012; Etmanskie et al., 2016; Leach et al., 2003). The identification of this subgroup of students was first introduced in Chall’s (1983) theory of reading through discussion of the “fourth grade slump”. Students in this subgroup may exhibit adequate or better progress in beginning reading but begin to fall behind when advanced texts require them to integrate their vocabulary and critical thinking skills with their basic reading skills. Essentially, as reading material in Grade 4 becomes more complex and reading instruction is not continued past the primary grades, some skilled readers can suddenly present with reading problems (Etmanskie et al., 2016).

Leach and colleagues (2003) classified children with late-emerging reading difficulties into three reading subtypes: children with primary problems in word recognition, children with primary problems in comprehension, and children with both word recognition and comprehension difficulties. Children in any of these groups may have different underlying deficits. Through their longitudinal study, Leach et al. (2003) found that more than two thirds of the late emergers also had at least some difficulty with word identification and decoding, providing support that many children who exhibit late-emerging reading difficulties have difficulty at the word level. Another longitudinal study by Catts et al. (2012) produced conflicting results suggesting that children with late-emerging reading difficulties may be more likely to have difficulties with reading comprehension than with word recognition. To further investigate this subgroup of students, Etmanskie et al. (2016) conducted a longitudinal study to explore the permanence or lack of late-emerging reading difficulties. Their findings in the prevalence of late-emerging reading difficulties were that 0.2% of children in their sample had poor word reading skills and 3.3% of children had poor reading comprehension skills for the first time in Grade 4 which were consistent with those of Catts and colleagues in that difficulties with reading comprehension were more common than difficulties with word recognition.

Current Study

This study builds upon extant research that has described the heterogeneous nature of reading skill profiles among adolescents, specifically those with reading difficulties. This study will contribute to the growing body of work showing that adolescents with reading comprehension difficulties may experience difficulties in component skills that contribute to reading comprehension (Biancarosa & Snow, 2004). Previous research on adolescent literacy profiles utilized smaller sample sizes and rarely compared typically developing students to those with reading difficulties. The current investigation draws upon a much larger sample of students while also separating groups of students with and without formal diagnosis of a specific learning disability (SLD) for analysis. Additionally, this study utilizes resulting student skill profiles to predict performance on a broad measure of reading comprehension in two subsequent years which has not previously been included in analyses in this area of research. Latent profiles were extracted from a large state-level dataset, utilizing Latent Profile Analysis (LPA) in order to determine the number of unique profiles that emerge based on measures of word level skills and reading comprehension for sixth-grade students with and without SLD. Next, the component reading skill profiles were examined for changes in performance over time through a Latent Transition Analysis (LTA) before being used to predict students’ scores on a broad measure of reading comprehension in grades seven and eight for each group of students. The research questions for this study include:

1. How many reading component skill profiles emerge for students with SLD and those without SLD?

2. How does reading performance change or stay the same for the latent profiles across all three middle school years?

3. How do the latent transition profiles predict performance on a broad measure of reading comprehension in subsequent grades one year and two years later?

Method

Participants

This study used data from Florida’s Progress Monitoring and Reporting Network (PMRN), a statewide educational database of standardized assessment data for all Florida public school children in kindergarten through twelfth grade. Sixth grade reading data came from 2011–2012 school year, and seventh and eighth grade reading data were from the subsequent 2012–2013 and 2013–2014 school years, respectively. In sixth grade, the students with SLD had an average age of 11.32 (SD = .62) and typically developing (TD) students had an average age of 10.71 (SD = .46). Data on specific learning disability and other exceptionality status were drawn from the PMRN and included students classified under the Individuals with Disabilities Education Act (IDEA). According to IDEA, a child has SLD if, “a child does not achieve adequately for the child’s age or to meet State-approved grade-level standards in one or more of the following areas, when provided with learning experiences and instruction appropriate for the child’s age or State-approved grade–level standards: oral expression, listening comprehension, written expression, basic reading skills, reading fluency skills, reading comprehension, mathematics calculation, mathematics problem solving” (US Department of Education, 2014). Overall, IDEA provides free appropriate public education, evaluation and individualized education plans for children who qualify for an exceptionality under 14 categories: (1) autism, (2) deaf-blindness, (3) deafness, (4) emotional disturbance, (5) hearing impairment, (6) intellectual disability, (7) multiple disabilities, (8) orthopedic impairment, (9) other health impairment, (10) specific learning disability, (11) speech or language impairment, (12) traumatic brain injury, (13) visual impairment, and (14) developmental delay (Dragoo & Lomax, 2020). For our operationalization of students with SLD, only those students who were classified under the IDEA category of SLD (and no other IDEA category) as reported in the PMRN were included. Only children with complete data availability were included in the analyses.

As a first step, the overall sample underwent a series of status checks for SLD status, SES status, and school attendance. Given the current study’s interest in determining the differences in the component reading skills of both SLD and TD students, any child who did not retain a designation of SLD or TD for all three school years (2011–2014) was dropped from the overall sample (n = 141). Similarly, due to the importance of SES in achievement outcomes and the inclusion of SES as a covariate in the models, any child who did not retain the same designation as eligible or not eligible for free or reduced lunch for all three years was dropped from the overall sample (n = 6,335). Finally, due to the nesting of children within schools and the need to account for school effects, any child who did not attend the same school for all three years was dropped from the overall sample (n = 8,552).

The final sample at Time 1 included 23,800 students in 6th grade during the 2011–2012 school year who were attending schools across the state of Florida. Prior to running analyses, the component reading skill scores were mean-centered for the whole sample. Then, the sample was divided into two subsamples: SLD and TD students. Specifically, there were 2,355 SLD students across 399 schools and 21,445 TD students across 504 schools. Times 2 and 3 included the same students with the same designation (SLD or TD) who were in 7th grade during the 2012–2013 school year and 8th grade during the 2013–2014 school year, respectively. For TD students, 53.99% qualified for free or reduced-price lunch status, 47.92% were male, and the racial/ethnic composition included 49.82% White, 27.69% Black, 16.11% Hispanic, and 6.37% Other. For students with SLD, 60.00% qualified for free or reduced-price lunch status, 49.33% were male, and the racial/ethnic composition included 47.28% White, 31.39% Black, 15.21% Hispanic, and 5.99% Other.

Procedure and Measures

All reading-related assessments were administered as a part of regular school attendance during the fall semester of each school year, with the exception of The Florida Comprehensive Assessment Test (FCAT), which was administered during the spring semester of each school year. The measures used to create the latent profiles came from the Florida Assessment for Instruction in Reading (FAIR, Florida Department of Education, 2009) and include the Reading Comprehension subtest, the Word Analysis subtest, and the Maze subtest. All three subtests are administered on a computer and require students to wear headphones and have a live Internet connection. Each task includes directions and practice items with feedback. The combination of all three FAIR assessments in Grades 3 −10 has been shown to predict FCAT reading performance (Florida Department of Education).

Socioeconomic Status (SES)

SES was operationalized as student-level free or reduced-price lunch rate and was drawn from the PMRN. Free or reduced-price lunch rate is measured as the percentage of students eligible for subsidized lunch. Free lunch is offered to students from households with an income at or below 130 percent of the poverty income threshold, and reduced-price lunch is available for students from households with an income in the range of 130–185 percent of the poverty threshold (Kena et al., 2015).

FAIR Reading Comprehension subtest.

The FAIR Reading Comprehension (RC) subtest is a computer-adaptive assessment comprised of one to three literary and informational passages, which are read silently and followed by seven to nine multiple- choice questions. Raw scores were used in the current study. Generic estimates of reliability from item response theory are .92 in Grades 5–12.

FAIR Maze Task.

The Maze is a timed, computer-administered cloze-format test of text reading efficiency. Students are instructed to silently read a passage and choose which one of three words best completes the cloze items embedded within each passage. The score on the Maze based on the average number of Maze items correctly in 3 minutes on two passages. Raw scores were used in the current study. Corrected parallel-form reliability ranges from .77 to .90 in Grades 7–12.

FAIR Word Analysis Task.

The Word Analysis Task (WA) is a computer-adaptive test of spelling that assesses students’ knowledge of the phonological, orthographic, and morphological information necessary to accurately identify words in text. Students begin by spelling five words at their grade level. Then, the system adapts based on their ability and they receive harder or easier words, with an average of approximately 12 words spelled and a maximum of 30 words total. Raw scores were used in the current study. Based on generic IRT, estimates of reliability for the Word Analysis task range from .90 in Grades 4–6 to .95 in 8th grade (Foorman & Petscher, 2010).

FCAT Reading Comprehension Subtest.

The Florida Comprehensive Assessment Test 2.0 (FCAT) is a high-stakes standardized assessment that is administered every spring to Florida public school students in grades 3 through 12 to test for reading comprehension skills that students are expected to master at each grade level (Tannenbaum et al., 2006). The FCAT has shown high reliability, ranging from .86–.91 (Tannenbaum et al., 2006; Florida Department of Education, 2002). In the present study, Developmental Scale Scores were used which is a vertically scaled score that allows for the development of the same child across grade levels to be measured. For grades 3–12, alpha reliability coefficients exceed .90, ranging from .92 to .95 (Carlson et al., 2010).

Data Analyses

Latent Profile Analysis (LPA) operates under the assumption that the observed sample is a combination of individuals from different latent profiles, with individuals who have similar observed scores on a set of selected measures assumed to come from the same probability distributions and thus, belonging to the same profile (Vermunt & Magidson, 2002). LPA is conducted using a systematic model comparison approach to determine the best model based on a balance of parsimony, model fit, and the interpretability of the profiles for a series of models that increase by one profile (i.e., subgroup) at a time. When an increase in model fit is not statistically significant, it is an indication that that the more parsimonious model, with one fewer profile, should be chosen. Starting with a two-profile solution, the three FAIR measures (RC, WA, and Maze) were used to systematically test solutions with increasing numbers of profiles. The fit statistics used to evaluate the models were the log-likelihood ratio test, entropy, Akaike information criterion (AIC), Bayesian information criterion (BIC), sample-size-adjusted BIC (SABIC), and the Lo-Mendell-Rubin adjusted likelihood ratio test (LMR-LRT) and its accompanying p-value. For the log likelihood ratio test and entropy statistics, higher values signify a better fit. For AIC, BIC, and SABIC, smaller values are indicative of better fit. For the LMR-LRT test, a higher value and a statistically significant p-value (p < .05) indicate a better fit

Given the longitudinal nature of the data, instead of assuming measurement invariance like a traditional LPA approach, Latent Transition Analysis (LTA; Velicer et al., 1996) was used to examine the developmental trajectories of latent profiles across three time points (Grades 6, 7, and 8) and test for measurement invariance directly. Conducting an LTA allowed for examination of how the presence of latent profiles and their associations varied over time (Ryoo et al., 2018), providing a picture of how students performed across the middle school years and enabling the consistent interpretation of latent status characteristics.

All analyses were conducted in Mplus version 7.4 (Muthén & Muthén, 1998–2012) using Full Information Maximum Likelihood (FIML) and an integration algorithm. LPAs were conducted based on the three FAIR reading subtests (RC, WA, and Maze) for each of the three time points (Time 1 = 6th grade, Time 2 = 7th grade, Time 3 = 8th grade), and the same three measures were used for the LTAs. Because children were nested within schools, the multilevel aspect of the data was incorporated into the models implicitly through the estimation of cluster-robust standard errors (McNeish et al., 2017), allowing the focus of the present analysis to stay on student-driven differences rather than the role played by schools. Covariates included age, and sex, and socioeconomic status (SES). In order to use the age, sex and SES covariates as control variables, they were included when enumerating the latent statuses (Li & Hser, 2011). To compare relative fit among LTA models, -LL, AIC, BIC, SABIC, and entropy were used. Consistent with other longitudinal data analyses, it was necessary to diagnose and explore cross-sectional data first at each time point. Therefore, a series of LPA models were conducted separately for SLD and TD children to determine the most likely number of latent profiles that were present in each group for each school year. The results from the LPAs at each time point are presented separately for TD students and students with SLD in Table 1. Once the most likely number of profiles was determined for each group, each candidate number of profiles was tested as an LTA that included all three time points (6^th, 7^th, and 8^th grade) to determine which model fit the longitudinal data best while providing the most theoretically informative profile divisions.

Table 1.

Latent profile analysis model fit indices for TD students and students with SLD

Typically Developing Students
Grade	Profiles	DF	-LL	AIC	BIC	SABIC	Entropy	LMR-LRT	p-value
6	2	19	−276025.68	552089.35	552238.58	552178.20	0.53	4004.14	<.0001
6	3	29	−275575.97	551209.93	551437.70	551345.54	0.52	890.38	.0001
6	4	39	−275354.25	550786.51	551092.82	550968.88	0.53	438.97	.0128
6	5	49	−274714.00	549525.99	549910.85	549755.13	0.65	415.00	.0792
7	2	19	−256843.15	513724.30	513872.08	513811.70	0.54	5246.51	<.0001
7	3	29	−255948.83	511955.66	512181.22	512089.06	0.57	1770.54	<.0001
7	4	39	−255683.73	511445.45	511748.80	511624.86	0.53	524.84	.0044
7	5	49	−255488.91	511075.82	511456.95	511301.23	0.58	385.69	0.3900
8	2	19	−219127.44	438292.88	438437.31	438376.93	0.50	2357.01	<.0001
8	3	29	−218828.62	437715.25	437935.70	437843.54	0.46	591.47	<.0001
8	4	39	−216041.60	432161.20	432457.67	432333.73	0.61	735.24	<.0001
8	5	49	−218553.39	437204.78	437577.27	437421.55	0.54	253.29	.0957
Students with Specific Learning Disability
Grade	Profiles	DF	-LL	AIC	BIC	SABIC	Entropy	LMR-LRT	p-value
6	2	19	−31280.04	62598.08	62706.35	62645.98	0.52	584.78	<.0001
6	3	29	−31227.89	62513.78	62679.03	62586.90	0.57	102.96	.0093
6	4	39	−31198.88	62475.75	62697.99	62574.08	0.51	57.28	.4276
7	2	19	−30875.59	61789.18	61896.98	61836.62	0.50	515.79	<.0001
7	3	29	−30747.24	61552.48	61717.02	61624.88	0.60	253.40	.0006
7	4	39	−30697.53	61473.05	61694.33	61570.42	0.61	98.15	.0159
7	5	49	−30787.51	61673.01	61951.03	61795.35	0.77	−80.02	.9745
8	2	19	−29113.18	58264.35	58370.46	58310.10	0.53	242.13	<.0001
8	3	29	−29058.43	58174.86	58336.82	58244.69	0.60	108.06	.1063
8	4	39	−28525.24	57128.48	57346.28	57222.38	0.68	173.00	.4502

Note. TD = typically-developing, SLD = Specific Learning Disability; -LL = -Log likelihood; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion; SABIC = Sample-size-adjusted Bayesian Information Criterion. The best-fitting models for each grade are bolded.

Then, for each candidate LTA model, measurement invariance was tested to determine whether item response probabilities should be constrained or freely estimated. The reason for this comparison was to determine whether item response probabilities caused ambiguity when latent statuses were defined because their characteristics were attributable to not only observed items but also to measurement model variance. When longitudinal measurement invariance is met (because an invariant model fits better than a freely-estimated model), it means that latent status characteristics are attributable to observed items over time and not to measurement model variability (Ryoo et al., 2018). Importantly, even the freely-estimated model was restrained to be configurally invariant, meaning that all latent statuses were constrained to have the same factor structure (i.e., 3 profiles for each grade; Asparouhov & Muthén, 2014).

Once the final LTA model was estimated for TD students and students with SLD separately, information related to latent profile membership and transitions was extracted for use in a subsequent model for each group to estimate the relationship between latent profile membership and a distal measure of broad reading performance: average FCAT reading scores in 7^th and 8^th grade using the ML three-step approach (Vermunt & Magidson, 2013). The ML three-step approach was developed based on an error-in-variable schema, meaning that it accounts for the fact that measurements used to define latent profiles are subject to measurement error. Accordingly, this approach aims to account for this measurement error and estimate the true underlying latent profiles by using a three-step process. The steps involve (1) identifying the best-fitting unconditional model and saving the posterior probabilities and modal profile assignment for that model, (2) computing estimated conditional probabilities for modal profile assignment given true latent profile membership, and (3) specifying a new analytic model (that includes FCAT reading scores in 7^th grade and 8^th grade) with fixed parameters representing the classification error (Vermunt & Magidson, 2013).

Results

Descriptive statistics for all achievement measures are presented separately for SLD and TD students in Table S1, and the bivariate correlations between the reading assessments, age, and SES are presented in Table S2 in the supplemental materials.

LTA of Typically Developing (TD) Children

After conducting a series of LPA models for each time point, measurement invariance was tested for each of the potential LTA profile solutions. The fit indices for all LTA models tested are presented in Table 2 with TD students and students with SLD reported separately. The final LTA model chosen for TD children was a freely-estimated model with 3 latent profiles for each middle school year, with freely estimated means, variances, latent status prevalences, and latent status transitions. Although fit indices indicated that the four-profile model provided a better fit for all three grades, an examination of the profile means revealed that the third and fourth profiles from the 4-profile model were barely discernible in 7^th and 8^th grade. As such, the three-profile model was chosen, which showed a unique pattern of performance for each profile on the three FAIR reading measures over time (depicted in Figure 1).

Table 2.

Fit indices for the LTA models tested for typically developing children and children with Specific Learning Disability

Invariance	Group	Profiles	N	Clusters	DF	-LL	AIC	BIC	SABIC	Entropy
Free	TD	3	21445	504	122	−728955.62	1458155.25	1459127.98	1458740.27	.740
Free	TD	4	21445	504	162	−728256.44	1456836.90	1458128.55	1457613.72	.695
Invariant	TD	3	21445	504	86	−731490.89	1463153.78	1463839.48	1463566.18	.745
Invariant	TD	4	21445	504	114	−730617.54	1461463.08	1462372.03	1462009.74	.644
Free	SLD	2	2355	399	86	−88381.28	176934.55	177430.28	177157.04	.811
Free	SLD	3	2355	399	122	−88144.06	176532.13	177235.37	76847.75	.813
Free	SLD	4	2355	399	162	−87994.83	176313.65	177247.47	176732.76	.819
Invariant	SLD	2*	2355	399	63	−88753.23	177632.45	177995.61	177795.44	.764
Invariant	SLD	3	2355	399	86	−88594.70	177361.39	177857.12	177583.88	.771
Invariant	SLD	4*	2355	399	95	−88514.10	177218.20	177765.81	177463.98	.767

Note. TD = typically-developing, SLD = Specific Learning Disability; -LL = -Log likelihood; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion; SABIC = Sample-size-adjusted Bayesian Information Criterion. Model fit indices for the 5-profile invariant LTA for TD children were not reported because the model did not converge. The 4-profile invariant LTA was not fully invariant because it was necessary to allow 6th grade FAIR RC scores to vary between classes so the model would converge.

^*The variances could not be estimated for these models in order to achieve convergence.

Figure 1.

3-Profile LTA Results for TD Students (Left) and Students with SLD (Right) in 6^th-8^th Grade

The overall means and variances (in parentheses) for the component reading skills were −10.21 (7684.68), 13.90 (8333.07), and 1.26 (79.75) for 6^th grade RC, WA, and Maze, respectively. For 7^th grade, the overall means (and variances) for RC, WA, and Maze were 11.42 (8391.61), 11.83 (5960.04), and 1.64 (52.49), respectively. For 8^th grade, the overall means (and variances) for RC, WA, and Maze were 12.39 (8938.28), 13.49 (7930.36), and 1.46 (117.65), respectively. The means for all three reading components were similar over time, with RC means steadily increasing from 6^th-8^th grade, and WA and Maze performance means following less consistent patterns. The latent prevalences, means and variances for each reading measure by latent profile are presented in Table 3.

Table 3.

Unstandardized Latent Indicator Prevalences, Means, and Standard Errors by Profile from the 3-Profile LTA

Latent Profile	Grade	Prevalence (%)	N	Comprehension (SE)	Word Level Skills (SE)	Maze (SE)
Typically Developing Students
Profile 1: Persistently Low Performers with Declining Comprehension
	6	53	11,346	−33.77 (2.24)	−18.01 (1.78)	−3.13 (0.26)
	7	53	11,385	−27.28 (2.47)	−10.76 (1.43)	−3.74 (0.27)
	8	7.4	1,595	−70.31 (7.43)	−22.58 (6.65)	−5.36 (0.69)
Profile 2: Improvers in Comprehension and Word Level Skills
	6	29	6,224	62.61 (4.05)	30.85 (2.12)	5.52 (0.42)
	7	7	1,439	55.23 (6.75)	76.46 (6.94)	17.08 (1.37)
	8	46	9,880	72.29 (3.02)	54.78 (2.31)	8.97 (0.45)
Profile 3: Decliners in Comprehension and Word Level Skills
	6	18	3,875	59.06 (5.25)	83.07 (4.97)	7.43 (0.50)
	7	40	8,620	69.51 (3.32)	40.78 (2.59)	8.10 (0.60)
	8	46	9,970	−14.85 (3.80)	−6.83 (1.88)	−2.66 (0.31)
Students with Specific Learning Disability
Profile 1: Consistently Low Performers with Improving Word Level Skills
	6	16.10	379	−118.70 (7.81)	−169.40 (23.04)	−14.62 (1.16)
	7	14.32	337	−117.91 (6.46)	−157.54 (15.39)	−19.75 (1.17)
	8	16.54	390	−140.83 (8.35)	−144.61 (13.03)	−19.88 (1.07)
Profile 2: Consistently Average Performers
	6	54.57	1285	−75.62 (5.24)	−97.25 (7.49)	−9.52 (0.51)
	7	56.01	1319	−78.28 (5.90)	−73.06 (5.30)	−10.69 (0.45)
	8	54.79	1290	−68.16 (7.30)	−87.26 (7.33)	−10.33 (0.62)
Profile 3: Consistently High Performers with a Word Level Skill Deficit
	6	16.54	390	−6.80 (6.77)	−46.43 (4.95)	−2.57 (0.78)
	7	29.68	699	−2.62 (7.80)	−36.55 (4.86)	−2.70 (0.98)
	8	28.67	675	6.11 (8.63)	−36.43 (6.66)	−1.42 (1.22)

Note. Comprehension = Florida Assessments for Instruction in Reading, Reading Comprehension subtest, Word Level Skills = Florida Assessments for Instruction in Reading Word Analysis subtest, Maze = Florida Assessments for Instruction in Reading Maze subtest.

Based on their latent means, the three profiles were named Persistently Low Performers with Declining Comprehension (Profile 1), Improvers in Comprehension and Word Level Skills (Profile 2), and Decliners in Comprehension and Word Level Skills (Profile 3). Across all profiles and grades, TD children showed consistent and average Maze performance, indicating that Maze may not be a good component skill for differentiating latent profiles of TD readers. The three profiles are described in detail below.

Profile 1: Persistently Low Performers with Declining Comprehension.

This group started as the largest profile, with identical prevalences in 6^th and 7^th grade (53%) but dropped to just 7.4% of the sample, and thus the smallest group, by 8^th grade. Students in this group were characterized by consistently low performance on all three reading measures, with WA and Maze performance that stayed about the same and declining RC performance that reached its lowest point in 8^th grade. The transition probabilities for this profile suggested that all students who started in this group transitioned to a different group at both transition points, with 98.6% transitioning to Profile 3 with just 1.3% transitioning to Profile 2 in 6^th-7^th grade and 97% transitioning to Profile 2 with just 3% transitioning to Profile 3 in 7^th-8^th grade.

Profile 2: Improvers in Comprehension and Word Level Skills.

This group demonstrated inconsistent prevalences across the three school years, starting at 29% in 6^th grade, declining to 7% in 7^th grade, and increasing to 46% (the largest profile) in 8^th grade. This group was characterized by high word level skills and especially high comprehension skills. They began 6^th grade as the second highest performing profile with especially high comprehension performance, showed a large jump in word level skills performance in 7^th grade, and then another large jump in comprehension performance in 8^th grade, ending middle school with the highest scores in all reading domains. The transition probabilities for this profile suggested that all students who started in this group transitioned to a different group in the 6^th-7^th grade transition, with most students (96.2%) transitioning to the Profile 3 and 3.8% transitioning to Profile 1. In contrast, the 7^th-8^th grade transition probabilities for this profile suggested that a majority (98.6%) of students would remain in this profile, with just 1% transitioning into each of the other two profiles.

Profile 3: Decliners in Comprehension and Word Level Skills.

This group started with a low prevalence (18%) in 6^th grade but increased to much higher prevalences in 7^th and 8^th grade (40% and 46%). They started off as the highest achieving profile, with especially high WA performance but showed inconsistent RC and WA performance across the three middle school years. In 7^th grade, there was a large decline in WA performance, followed by a large decline in RC performance in 8^th grade, so by the end of middle school, RC, WA, and Maze performance were all at the same level and below average. The transition probabilities for this profile suggested that just only 3.8% of students who started in this group stayed in this group in the 6^th-7^th grade transition, with moderate proportions (38.4% and 57.8%) transitioning to Profile 1 and 2, respectively. In contrast, the 7^th-8^th grade transition probabilities for this profile suggested that most of these students (88.9%) transition to Profile 1 and the remaining 11.1% would remain in the same profile.

LTA for Students with Specific Learning Disability (SLD)

After conducting a series of LPAs for each time point for children with SLD, a differing number of profiles were found to provide the best fit for each of the three grades: 3 profiles in 6^th grade, 4 profiles in 7^th grade, and 2 profiles in 8^th grade. As such, 2-, 3- and 4-profile LTA models were tested. The LTA models tested included the covariate effects of age, sex and SES in the creation of latent profiles and latent transition probabilities. For each number of profiles, measurement invariance was tested, comparing a freely-estimated model (with freely-estimated means, variances, prevalences, and transition probabilities) to a measurement invariant model. All fit indices for the LTA models tested are presented in Table 2. The freely-estimated LTA that included 4 latent profiles demonstrated the best fit according to most indices (-LL, AIC, SABIC, and entropy), but the 3-profile model showed better fit according to the BIC. After plotting and comparing the two models, the 3-profile model (depicted in Figure 1) was chosen because it showed performance patterns that were more interpretable and theoretically stable than the 4-profile model.

The overall means and variances (in parentheses) for the component reading skills were −62.62 (4516.97), −94.94 (8107.67), and −8.31 (58.32) for 6^th grade RC, WA, and Maze performance, respectively. For 7^th grade, the overall means (and variances) for reading RC, WA, and Maze performance were −62.93 (5822.81), −75.98 (6445.73), and −9.79 (73.82), respectively. For 8^th grade, the overall means (and variances) for RC, WA, and Maze performance were −61.47 (8526.08), −84.85 (9274.71), and −9.64 (93.52), respectively. The means for RC and Maze performance were similar between the time points, but WA performance had mean differences ranging from 10 to 19 points across the three grades. Notably, all means were much lower for children with SLD than those of TD children, lending support to the choice to separate the two groups.

The latent prevalences, means, and variances for each reading measure by latent profile are presented in Table 3. The three profiles that emerged for students with SLD were characterized based on their means as Consistently Low Performers with Improving Word Level Skills (Profile 1), Consistently Average Performers (Profile 2), Consistently High Performers with a Word Level Skill Deficit (Profile 3). Overall, the latent transition probabilities suggested that children were likely to stay in the same profile, both in the transition from 6th to 7th grade and from 7th to 8th grade.

Profile 1: Consistently Low Performers with Improving Word Level Skills.

This profile had similar prevalences in all grades (16.1% in 6^th grade, 14.3% in 7^th grade, and 16.5% in 8^th grade) and is characterized by very low 6^th grade RC and even lower WA performance. Even though Maze performance was the highest skill for this group, they still performed the lowest, and below the overall average, in Maze. WA performance improved each year, and RC dropped in 8^th grade. The transition probabilities for this profile suggested that a large proportion (76.7%) would stay in this profile in the transition from 6^th to 7^th grade and the transition from 7^th to 8^th grade (90.7%), with a smaller proportion of students (21.5% and 9.3%) transitioning to Profile 2 in the 6^th-7^th and 7^th-8^th grade transition, respectively. An even smaller (1.8%) proportion of students were likely to transition to Profile 3 in the 6^th-7^th grade transition, but no students would transition to the same group in the 7^th-8^th grade transition.

Profile 2: Consistently Average Performers.

This profile was the largest and had very similar prevalences in all grades (54.6% in 6^th grade, 56% in 7^th grade, and 54.8% in 8^th grade). They were characterized by consistent average performance on Maze and WA with some improvement in RC by 8^th grade. For both transitions, the transition probabilities for this profile indicated that a vast majority of students (98% and 95.6%, respectively) stayed in the Profile 2, with a very small proportion of students (1.8%) likely to transition to Profile 3 in the 6^th-7^th grade transition or to Profile 1 in the 7^th-8^th grade transition (4.4%).

Profile 3: Consistently High Performers with a Word Level Skill Deficit.

This profile had prevalences of 29–30% in all grades and was characterized by high performance on all component skills, with above average Maze performance and RC that was higher than Maze by 8^th grade. Although their performance was higher than the other two profiles, they had relatively lower WA performance compared to the other reading skill domains. For both transitions, the transition probabilities for this profile indicated that a vast majority of students (98.2% and 96.8%, respectively) stayed in Profile 3, with just 1% and 1.3% transitioning to Profile 1 and 2 in 6^th-7^th grade and 1.5% and 1.7% transitioning to Profile 1 and 2 in 7^th-8^th grade.

Covariates.

The role played by the control variables (age, sex, and SES) in the assignment to and transitions between profiles are discussed in Supplemental Materials and reported in supplemental tables S3 and S4.

Distal Outcomes: FCAT Reading in 7^th and 8^th Grade

Overall, compared to students with SLD, TD students demonstrated higher FCAT reading performance in both 7^th and 8^th grade. The 3-step ML approach revealed that, for both groups of students, the pattern of FCAT reading scores was consistent across 7^th and 8^th grade, with the same profiles scoring the highest, second highest or lowest on FCAT reading in both grades with p-values < .001. For TD students, Profile 1 demonstrated the lowest FCAT scores with a mean score of 210.05 (SE = 0.58) in 7^th grade and 214.50 (SE = 0.59) in 8^th grade. Next was Profile 3 with a mean score of 223.86 (SE = 0.51) in 7^th grade and 231.45 (SE = 0.52) in 8^th grade. The highest scores were found for Profile 2 with a mean score of 247.02 (SE = 0.54) in 7^th grade and 254.44 (SE = 0.53) in 8^th grade. For students with SLD, Profile 1 demonstrated the lowest FCAT reading scores with a mean score of 151.64 (SE = 7.75) in 7^th grade and 184.61 (SE =10.61) in 8^th grade. Next was Profile 2 with a mean score of 213.29 (SE = 1.16) in 7^th grade and 220.66 (SE =1.22) in 8^th grade. Highest scores were found for Profile 3 with a mean score of 233.69 (SE = 1.80) in 7^th grade and 241.48 (SE =1.61) in 8^th grade.

Discussion

Persistent difficulties in word level skills may be a reason why many adolescents, regardless of disability status, have reading comprehension difficulties. As described by numerous models of reading comprehension (Cromley & Azevedo, 2007; Hoover & Gough, 1990), and empirical studies of comprehension in adolescents (Foorman & Petscher, 2010; García & Kain, 2014) word level skills are critical to supporting the processes necessary to construct meaning from text. The study examined the word level skill and reading comprehension profiles of adolescents with and without SLD. This investigation allowed for the examination of profile heterogeneity within student groups, which has been established in extant research (Brassuer-Hock et al., 2011; Cirino et al., 2013; Clemens et al., 2017), as well as a comparison across student groups which has been less commonly examined (Cirino et al.). Further, this study investigated these groups longitudinally to better understand group stability across grade bands. The current study further expands on previous work in this area by including an increased sample size of both TD (n= 21,445) and SLD students (n=2,355) as compared to Brassuer-Hock et al. (below-average comprehenders n=319), Clemens et al. (struggling readers n=180), and Cirino et al. (typical readers n=723, struggling readers n=1,025). As such, the current study better informs the design and focus of assessment and intervention practices for adolescent students that may have reading comprehension difficulties.

Reading Component Skill Profiles

The first research question investigated the number of reading component skill profiles that emerge for TD students and those identified with SLD. Results indicate that three unique skill profiles emerge for both student samples with varying performance for word level skills (Word Analysis subtest), the cloze (Maze subtest), and reading comprehension (RC) subtest as measured by the state level reading assessment. Data in this study suggest heterogeneity in both groups of students, those with typically developing literacy skills and those with SLD. Both samples contained profiles of higher performers and lower performers but with more nuance than just low, medium, and high scores; SLD student means were well below the TD student means, with greater than a 100-point discrepancy between the two groups on some measures. The greatest discrepancy across groups is seen in the word level skills performance. The score differential between the highest performing TD profile and the lowest performing SLD profile was over 200 points at all timepoints. This great variability in performance highlights the challenge that schools and teachers have when planning and implementing reading instruction during the middle school years. The pronounced difference in word level skill performance across the middle school years underscores the importance of assessing word level component skills including spelling. Administering a diagnostic spelling inventory annually, beyond just in the primary grades, can provide information about a student’s decoding and encoding skills. This information informs the type and intensity of word level instruction (e.g., phonological, orthographic, morphological) that will meet varied student needs. Results indicate that almost all of the students in the full sample, including both students with and without SLD, had scores below the mean in either one or both skill areas, i.e., word level skills and reading comprehension, during at least one timepoint in the middle school years. These findings are similar to those of Clemens and colleagues (2017) who found that a majority of their adolescent sample (96%) displayed below average performance in one or both skill areas assessed. Nearly half of the current study’s total sample (58%), regardless of disability status, displayed profiles with concurrent difficulties in word level skills and reading comprehension; this is most pronounced in Time 3 (Grade 8). Most skill profiles presented a high incidence of word level skill difficulties. In total, all three profiles in the SLD sample and two of the three profiles in the typically developing sample (54%) demonstrated below average word level skills either with or without reading comprehension difficulties at one of the timepoints. This is similar to findings from both Leach et al. (2003) and Cirino et al. (2013) in that most students who exhibit reading comprehension difficulties also have difficulties at the word level.

Given that students with reading disabilities, including SLD and dyslexia, can present with a combination of symptoms including poor spelling and difficulty comprehending what was read (Hulme & Snowling, 2016) it was expected that heterogeneity within the skills assessed in this study would emerge. Unexpectedly, in this analysis all profiles of students in the SLD sample performed below average across all measures examined at multiple timepoints. Previous studies of students with reading difficulties and disabilities (Brasseur-Hock et al., 2011; Cirino et al., 2012; Clemens et al., 2017; Solis et al., 2014) found one or multiple classes or profiles of students displaying performance on a number of skills in the average or even above average range. Additionally, an overlap among different types of difficulties within student samples was expected along with a gap in observed performance between the typically developing and SLD sample (Solis et al., 2014). Instead, the analysis revealed an overlap among difficulties across profiles with TD Profile 1 completely overlapping with SLD Profile 3 on the reading comprehension and cloze measures.

These findings underscore the need for schools to attend to the teaching of reading and writing alongside typical Language Arts (LA) instruction because a majority of adolescents with disabilities enter middle school with established reading difficulties (National Center for Education Statistics, 2021). Extant research has documented that middle school students with reading difficulties, regardless of disability status, not only experience challenges comprehending complex grade-level texts but continue to present with difficulties in foundational reading skills (e.g., word reading or fluency) throughout adolescence (Cirino et al., 2013). Spelling knowledge is distinctly related to text reading efficiency and reading comprehension at the sentence and text level, even for adolescents, but a lack of spelling growth in secondary classrooms suggests a lack of spelling instruction (Foorman & Petscher, 2010). These foundational reading gaps, while not typically prioritized in middle school, must be addressed with specialized instruction and supports if students are expected to show improvement in their understanding of grade-level complex texts.

Longitudinal Transition of Latent Profiles

The findings from the current study support extant research that discovered a high prevalence of basic skill difficulties among adolescent readers (Brassuer-Hock et al., 2011; Cirino et al., 2013; Clemens et al., 2017) but extend this work by investigating how reading performance by profile changes over time. The rank order among the profiles and the pattern of performance among the component reading skills remained the same across the middle school years for students with SLD, while TD students demonstrated much more fluctuation, with changes in rank order and performance patterns at each transition point (from 6^th to 7^th grade and 7^th to 8^th grade). The most struggling SLD students were more stagnant and showing less improvement over the middle school years. This is consistent with research suggesting that students with reading difficulties, particularly word level reading difficulties, require sustained targeted instruction over multiple years to demonstrate improvement in performance (Solis et al., 2014).

For any given middle school year and any given profile, students with SLD showed a pronounced deficit in word level skills compared to their comprehension and cloze performance, which appeared as a consistent “V” shape across all the latent statuses depicted in their LTA figures. In contrast, TD students’ performance patterns on the component reading skills ranged anywhere from a diagonal line, in which cloze performance was better than word level skill performance, which was better than comprehension performance (i.e., Profile 1 in 6^th grade), to a mostly straight line, in which performance on all three reading skills was relatively equal (i.e., Profile 3 in 8^th grade). TD students even showed an “inverted V shape”, capturing the opposite performance pattern to students with SLD, wherein scores on word level skills were highest relative to the other reading domains.

Variability in performance across profiles over time demonstrates the diverse needs of students, even among the highest performing student profiles, regardless of disability status. This further emphasizes the need for screening of reading comprehension and related component skills in adolescence to inform instructional practice in the classroom. Students with reading comprehension difficulties but higher word level skills will benefit from explicit instruction of vocabulary and reading comprehension strategy instruction implemented in both LA and content-area classrooms (Capin et al., 2022). Students with documented evidence of reading difficulties, particularly in word level skills, will require intensive interventions targeting different aspects of reading. This intensive instruction, taking place in a small group, would include a focus on fluency, multisyllabic word reading, or foundational word reading skills (e.g., phonics, word recognition) depending on students’ unique learning needs (Capin et al.).

Prediction of Broad Reading Comprehension

The third research question investigated how the component reading skill profiles predict performance on a broad measure of reading comprehension in subsequent grades one year and two years later. Correlations indicate that the FAIR reading comprehension subtest strongly predicts FCAT performance for students in grades 4–10 (Foorman et al, 2013). Utilizing this reliable predictor along with student performance on the FAIR Maze (cloze) and word analysis (WA) measures to predict FCAT performance, results indicate that student performance on the FCAT broad measure of reading comprehension remain stable over time. Results indicate there is very little change in overall performance over time, when looking from grade 7 to 8. This overall pattern of stagnant reading comprehension growth mirrors the national trend for U.S. adolescents (Ellerman & Oslund, 2019).

Implications for Practice

The results of this study have implications for reading achievement assessment and intervention design and delivery for adolescents both with and without a specific learning disability (SLD). Like elementary age students, adolescents should be screened with reliable and valid reading assessments that include not only more complex skills like comprehension but also word-level component skills, like spelling, which moderate reading comprehension performance (Reed et al., 2016). Given there is no “one size fits all” approach to effective reading instruction, these assessments are key in determining the type and intensity of instruction required to meet students’ unique learning needs. As shown by the current findings, adolescents may demonstrate average reading comprehension while simultaneously demonstrating below average word level skills. While instruction in spelling, or orthographic knowledge, is documented as receiving very little attention in grades 8–12 (Foorman & Petscher, 2010), these results support the provision of specific, targeted instruction in word level skills for adolescent students demonstrating weakness in these areas in order to further support their reading comprehension development.

Improving reading comprehension in adolescents will require a sustained focus on developing long-term solutions as opposed to short-term gains on measures utilizing low-level comprehension (Elleman & Oslund, 2019). This focus should include developing word level skills in conjunction with developing background knowledge, vocabulary, inference, and comprehension monitoring skills which represent the additional components of the DIME model of reading comprehension. This type of instruction will benefit the majority of older students that require interventions that address several components of reading (Cirino et al., 2012). Extant research in the form of longitudinal studies with adolescents with poor reading comprehension and reading disabilities suggests that one year of supplementary reading intervention may not be sufficient to address reading difficulties in students with deficits in word reading skills, vocabulary, and background knowledge (Solis et al., 2014). Solis and colleagues (2014) posit that for adolescents, continued remediation beyond just one year that includes intensive interventions utilizing content area texts to bolster background knowledge and content learning are essential to prevent these students from falling further behind their normative peers. Additionally, adolescents should receive interventions at an appropriate intensity based on their current reading achievement scores (Fuchs et al., 2010).

Limitations

It is important to acknowledge that reading comprehension involves several components and processes beyond the skills assessed by the Word Analysis and Maze tasks used in the present analysis. The adolescents in this sample with reading comprehension difficulties may also experience difficulty in other skill areas that were unable to be included in this analysis. This includes higher order processes such as inference making, comprehension monitoring, and story structure knowledge (Cain & Oakhill, 1999) as well as working memory (Cain et al., 2004).

Another limitation is that the SLD sample includes any student identified with a disorder in understanding or using spoken or written language. This disorder may manifest in an imperfect ability to listen, think, speak, read, write, spell, or do mathematical calculations and represents about 33% of all students in special education (National Center for Education Statistics, 2021). The PMRN database did not contain the level of detail to isolate a sample of students with SLD that have an impairment in reading only, meaning that the sample may include students with SLD that do not have documented impairments in reading specifically. It is worth mentioning that the National Center for Education Statistics estimates that over 80% of students with SLD have difficulties in word-level reading and all the students in our SLD sample demonstrated below average performance in both word level and reading comprehension skills. While all the students in the SLD sample may not have documented impaired reading, their assessment and subsequent intervention should target each students’ specific areas of weakness regardless of the type of SLD.

There were also some limitations introduced by the methods employed in the current study. For instance, despite providing more nuanced information about the reading component skills of SLD and TD readers, extracting three profiles from three FAIR reading measures could represent over-fitting, a possibility that is difficult to judge in mixture models (Loken & Molenaar, 2008). However, the use of three time points for each reading measure (from 6^th, 7^th, and 8^th grade) is likely to have made the results more robust than estimating the latent profiles based on a single time point. Given this study’s interest in examining how profiles differed qualitatively between SLD and TD students and the fact that an SLD designation is imposed on students by their schools, which translates to a child with an SLD designation being treated differently within the school environment in terms of accommodations, general labeling, and in other ways, separate models were conducted for TD and SLD children. Notably, this choice imposed two empirical limitations to the analysis: one, SLD and TD students could not be allocated to the same profiles and two, joint school clustering was not allowed.

Conclusion

This study examined the word level skill and reading comprehension performance of over 2,000 adolescents with SLD and 20,000 without SLD. Within both subgroups, three distinct profiles of student skills were identified and then examined for changes in performance over time before being used to predict performance on a broad measure of reading comprehension in two subsequent years. This study demonstrates the heterogeneity of word level and reading comprehension skill development in adolescents with and without SLD. Performance by profile remained consistent for adolescents with SLD but varied for those without. The present findings add to a growing body of evidence demonstrating the majority of adolescents with reading comprehension difficulties, regardless of disability status, also experience difficulty with word level skills. This underscores the importance of employing reading achievement assessments for adolescents that are sufficiently comprehensive to identify weaknesses in foundational reading skills that may be contributing to reading comprehension difficulties. This data can then be used to help align intervention to meet the needs of adolescent populations of readers.