Unraveling Predictors of Reading Difficulties in EFL 6th Graders: Insights from latent profile analysis and multinomial logistic regression

doi:10.21203/rs.3.rs-5323631/v1

Download PDF

Research Article

Unraveling Predictors of Reading Difficulties in EFL 6th Graders: Insights from latent profile analysis and multinomial logistic regression

https://doi.org/10.21203/rs.3.rs-5323631/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

This study examines predictors of reading difficulties in 6th-grade EFL students via latent profile analysis (LPA) and multinomial logistic regression. Six profiles were identified, including the Disinterested in English Group, which faces severe reading challenges due to a poor home English environment (HEE), low parental support (PS), minimal English exposure, and low interest; the English Necessity Recognition Group, which struggles despite understanding English’s importance; the Excellent Parental Support Group, which experiences nonword reading issues despite strong PS and HEE; and the Excellent English Environment Group, which has the best reading outcomes. The key predictors of reduced reading difficulty are a robust HEE, high English exposure, and genuine interest. The study highlights the need for integrating supportive factors, such as improving HEE and PS, and emphasizing phonics education to enhance reading skills and motivation. Future research should explore additional variables and diverse populations.

Linguistics

Educational Psychology

reading difficulties

home English environment

parental support

English exposure

latent profile analysis

multinomial logistic regression

Reading proficiency in English is a critical predictor of academic success and a fundamental building block. In early reading development, word decoding ability is particularly crucial. This process involves recognizing letters and their corresponding sounds to derive meaning from words. Studies indicate that children who excel in word decoding during the early stages of reading are more likely to develop strong reading skills later on (Adams, 1990; Chall, 1983; Stahl & Murray, 1994). Consequently, providing appropriate reading materials to beginner learners is essential. Approaches focusing on high-frequency vocabulary and phonics-based methods are commonly employed, each offering distinct advantages. Utilizing texts with phonetic characteristics has proven effective in early reading education (Grossen, 1997).

A recently emerged sociocultural perspective argues that variations in students' reading abilities are deeply rooted in diverse English education backgrounds (Lee, 2018). This viewpoint suggests that socioeconomic and family backgrounds influence disparities in learners' reading ability (Cheung & Anderson, 2003; Coleman, 1988; D'Angiulli et al., 2004) and the presence and duration of private tutoring in English education (Choi & Lee, 2019). Additionally, the home literacy environment and the affective domains of learners play significant roles. Various sociocultural factors, such as the home environment, educational setting, social influences, and cultural elements, significantly impact the development of children's reading skills. Parental education levels, a reading culture within the home, and economic resources notably influence reading ability. Furthermore, the resources available in schools, teacher quality, and student interactions within the educational environment are crucial. Social factors such as the extent of exposure to English and social support, as well as cultural factors such as cultural capital and reading-related cultural practices, also play important roles.

The dual approach of using nonword reading and word reading tasks provides a robust framework for assessing reading difficulties in EFL learners. It allows educators and researchers to diagnose specific deficits accurately and develop effective, tailored interventions to support reading development in this unique learner population. By understanding and addressing these deficits, educators can better support EFL learners in developing strong reading skills and achieving academic success.

Given the seriousness of this issue, there is a pressing need for research exploring the development of reading proficiency and the various sociocultural factors affecting it. This study is significant because it has the potential to identify and address reading proficiency deficiencies within distinct sociocultural groups. Accordingly, this study aims to analyze the sociocultural characteristics of prominent schools in lower-income districts within the EFL classroom environment in Korea to predict the likelihood of reading proficiency deficiencies within these distinct groups. Additionally, this study determined reading deficiencies by measuring the most fundamental aspect of reading ability: decoding skills. Since decoding skills are not naturally acquired in an L2 context, they are essential skills that must be mastered to achieve proficiency in reading.

Therefore, this study has established the following research question:

1. How can sixth-grade students in Korea be categorized into distinct profiles on the basis of their English learning environments and experiences, and what are the characteristics of these profiles?

2. To what extent do the identified profiles predict reading difficulties in English reading abilities, specifically in decoding skills?

The Critical Role of Decoding Ability in Early Reading Success

In the EFL context, decoding ability, which involves the recognition of letters, sound generation, and word comprehension, is fundamental for children's early reading success. This skill is crucial for building a strong reading foundation and significantly predicts future reading proficiency (Adams, 1990; Chall, 1983; Stahl & Murray, 1994). As children progress, decoding becomes automatic, leading to quicker word recognition and reading fluency (LaBerge & Samuels, 1974). Early reading instruction typically uses two approaches: high-frequency vocabulary-based methods and phonics-based methods. While vocabulary-based methods improve sight word recognition, they may not fully develop decoding skills. Phonics-based methods, however, focus on the relationship between letters and sounds, helping students decode unfamiliar words systematically. Research supports the effectiveness of phonics in enhancing reading skills. Identifying and addressing decoding deficits is essential, as difficulties in this area can lead to challenges in reading fluency and comprehension, impacting overall academic success. Nonword reading depends on phonemic awareness and basic decoding strategies, whereas word reading relies on vocabulary recognition and semantic understanding.

Sociocultural Factors Affecting Reading Difficulties

Differences in learners' reading ability are influenced by various factors, including their English education background, socioeconomic status, and family environment. Higher parental education levels and a reading-rich home environment positively affect children's reading skills, whereas economic resources impact access to educational tools and support. The educational environment, including school resources and teacher quality, also plays a crucial role in reading development. In South Korea, the prevalence of private tutoring creates educational inequalities, with higher spending on tutoring linked to better English proficiency but exacerbating disparities for lower-income students. Affective factors such as motivation and self-efficacy further influence reading success. Addressing these factors through comprehensive interventions can help improve decoding abilities and support struggling readers. Table 1 summarizes the references to sociocultural background factors in learning.

Table 1

*References on Socio-Cultural Background Factors in Learning*
Factor	Researchers (Year)
Home Literacy Environment	Brown (2007); Duursma et al. (2007); Inoue et al. (2018); Jeoung & Kim (2008); Liu et al. (2018); Puglisi et al. (2017); Sénéchal (2006); Sénéchal & LeFevre (2014); Van Bergen et al. (2017); Weigel et al. (2006)
Socioeconomic and Family Differences	Cheung & Anderson (2003); Coleman (1988); D'Angiulli et al. (2004); Duncan & Brooks-Gunn (2000); Jung et al. (2011); Kim (2005); Kim (2017); Kwak (2006); Park (2011); Pong (1997); Rothstein (2010); Sirin (2005)
Private Tutoring Presence and Duration	Choi & Lee (2019); Kim & Lee (2015); Kim & Lee (2018); Lee (2003); Lee (2006); Lee (2018); Oh & Kang (2012); Park (2012)
Affective Domain	Mega et al. (2014); Pekrun et al. (2002); Sainsbury & Sc Hagen (2004); Schunk et al. (2012); Wigfield et al. (2008); Oh & Cha, 2017

Participants

The study recruited 1,104 sixth-grade students from nine elementary schools in middle- to low-income areas across five provinces in Korea. These students had been studying English as a regular subject for more than three years. However, 196 students were excluded because of absences, missing test values, or failure to submit questionnaires, resulting in a final sample of 908 participants. The data were collected over three years, with measurements taken at similar times each year to ensure consistency. All the participants attended public schools following the Korean National English Curriculum, with a shared achievement goal. The data were collected at the start of the second semester in 2021 (N = 598), 2022 (N = 64), and 2023 (N = 246), reflecting the typical English proficiency and learning experiences of Korean sixth-grade students. The schools were selected to represent the characteristics of sixth graders across different regions, with each school having approximately 20 students per class and more than 80 sixth-grade students, indicating that they were large-scale schools.

Procedure

Tests were prepared and administered by Korean and native-speaker teachers (in 3 schools by both Korean and native-speaker teachers and in 6 schools by Korean teachers only). Extensive pilot work was conducted with researchers in each region to ensure the appropriateness of the tests, particularly for the phonological awareness tests formulated for this study.

To administer the TOWRE Nonword and word Reading Test, the examiner ensured a quiet environment free from distractions and prepared a stopwatch and scoring sheet. The examiner handed the list of nonwords and words to the students, instructing them to read as many nonwords and words as possible, as quickly and accurately as possible, within 45 seconds. The examiner started the stopwatch when the student began reading, marking each nonword and word read correctly on the scoring sheet. If the student skipped a word or remained silent for more than 3 seconds, the examiner marked the word incorrectly and prompted them to move to the next one, continuing until 45 seconds elapsed.

Standardized instructions were provided to ensure consistency across testers. Before beginning each task, the students received verbal and visual instructions and three examples of the task. The first reading test was administered two weeks after the students began the first grade, and the second test was administered two weeks later. The average of the two test scores was used. Surveys were conducted at each school to investigate the students' learning backgrounds. Written consent was obtained from the children's parents and teachers at the start of the study.

Measurement Variables

The study evaluated students' second language (L2) reading-related abilities, specifically decoding skills, via nonword and word reading tests. These tests aimed to determine reliance on phoneme‒grapheme correspondence (phonological route) and word recognition (lexical route).

Nonword Reading Test: Administered using two forms (A and D) of the Test of Word Reading Efficiency (TOWRE-2). The participants had 45 seconds to read as many pronounceable nonwords as possible. The number of correctly read nonwords was used to measure phonological decoding efficiency. The average score from both forms provided insights into participants' phonological decoding abilities.

Sight Word Efficiency (SWE) Test: In addition, from TOWRE-2, this test involved reading a list of high-frequency words within 45 seconds. The participants' scores were based on the number of correctly read words, indicating their sight word vocabulary size and ability to recognize words quickly. Two forms (A and D) were used, and the test was reliable, with Cronbach’s alpha values of 0.925 for nonword reading and 0.938 for word reading. The students were categorized into three groups on the basis of the Z scores derived from the tests. Table 2 provides details on the classification of reading ability groups by standard scores.

Table 2

*Classification of Reading Ability Groups by Standard Scores*
Group Classification	General Reading (GR)	Reading Difficulty (RD)	Severe Reading Difficulty (SRD)
Z score range	Z score > -1.0	-1.0 ≥ Z score ≥ -1.5	Z score <-1.5

Analysis methods

The analysis was conducted in two main stages to address the research questions. First, profile analysis techniques, ANOVA, and chi-square tests were used to examine the common characteristics of the learners' educational backgrounds. This initial step involved developing several latent profile models. To ensure accurate results, skewed variables were normalized via transformation methods in SPSS (version 24.0), and all continuous variables were standardized. The students were then grouped into different latent profiles on the basis of their reading learning backgrounds. These classifications were saved and analyzed concerning various covariates. Multinomial logistic regression analysis was employed for the second research question, which explored the relationship between the identified profiles and English reading proficiency. This analysis assessed how the different profiles related to students' English reading ability.

Latent profile analysis (LPA) was used to classify students into latent subgroups on the basis of their learning backgrounds. This technique, which is suitable for heterogeneous populations, estimates the probability of individuals belonging to latent subgroups and provides precise classification indices. Model fit was evaluated via several criteria:

The information criteria used were the Akaike information criterion (AIC), Bayesian information criterion (BIC), and sample size-adjusted BIC (aBIC), with lower values indicating a better fit.

Entropy: Values close to one indicate good classification accuracy.

Statistical Testing: Parametric bootstrap likelihood ratio difference (BLR) and Lo‒Mendell‒Rubin adjusted (LMR) tests, with significance (p < .05) favoring models with more profiles.

LPA was conducted via Mplus 8.7, which handles missing data with full information maximum likelihood (FIML). Models with different numbers of profiles were compared on the basis of their absolute and relative fits. Chi-square tests and ANOVA were used to examine differences in individual, home, regional, and school characteristics among the derived profiles.

To address the second research question, multinomial logistic regression was used to analyze the impact of six independent variables on students' reading profiles:

Home English environment: Measured through items on English and Korean books owned or borrowed.

Parental Support: Evaluated on the basis of parental involvement in English education.

Preschool English Experience: Assessed through items about English learning in kindergarten.

Postschool English Experience**: Includes items on English learning outside of school.

Interest in English: Rated through one item on interest in English.

Necessity of English: Rated through one item on the perceived necessity of English.

All the variables were rated on a 4-point Likert scale, where higher scores indicated better conditions or higher levels. The regression model aimed to predict students' reading profiles on the basis of these learning background factors.

Descriptive statistics of the students’ L2 reading ability measurements

The descriptive analysis of all L2 reading-related measures, including the mean, standard deviation, skewness, kurtosis, and Cronbach’s alpha, has been reported. To satisfy the assumptions for preliminary analysis and maximum likelihood estimation, the skewness and kurtosis values were checked to confirm the normality of the data. Table 3 shows descriptive statistics of each test.

Table 3

*Descriptive Statistics on Reading Abilities (N = 908)*
Measures	Total	Mean	SD	Min	Max	Skewness	Kurtosis	Reliability
NWR	66	29.24	15.195	0	64	.032	.610	.925
WR	108	49.80	20.507	0	90	− .514	.287	.938
Note. NWR = Nonword Reading; WR = Word Reading; Total = Number of final words in the list

Research Question 1: Categorizing into Distinct Profiles and Identifying the Characteristics of Each Profile

The evaluation of fit indices indicates that the three-class model is more suitable than the two-class model due to significantly lower AIC, BIC, and saBIC values, suggesting a better characterization of the study population by heterogeneous groups rather than a single homogeneous group. While the AIC, BIC, and saBIC values decrease as the number of classes increases from two to seven, the seven-class model was not supported by the LMR index, which was a limitation for this model.

The six-class model was selected as the final model, supported by both LMR and BLRT tests, and demonstrated the lowest AIC, BIC, and saBIC values among the models with two to six classes. It also had an entropy value of .90, indicating low classification error and high accuracy in distinguishing between profiles. The six-class model had clear categorization, with classification rates ranging from 6.1–40.1%, and each profile represented at least 5% of the sample.

The elbow plot of profiles, shown in Fig. 1, validated the fit of the six-class model by indicating a sharp decrease in fit indices up to the sixth profile, followed by a gentler reduction. This visual inspection supports the choice of the six-class model, reflecting an optimal balance between model complexity and fit. Table 4 provides the model fit indices of the latent profile analysis.

Table 4 Model Fit Indices of Latent Profile Analysis

Number of Profile	AIC	BIC	SABIC	LMR	BLRT	Entropy	Latent Classification Rate(%)
Number of Profile	AIC	BIC	SABIC				1	2	3	4	5	6	7
2	14427.8	14519.1	14458.8	.0	.0	.90	72.4..4	27.6
3	14237.3	14362.4	142709.8	.3614	.0	.75	22.5	50.7	26.7
4	13981.4	14140.1	14035.3	.0439	.0	.82	19.6	8.5	47	24.9
5	13790.6	13983.0	13856.0	.901	.0	.89	24.8	12.3	7.7	34.1	21.1
6.	13719.4	13945.4	13796.2	.0008	.0	.90	7.1	19.7	6.1	40.1	6.8	20.2
7	13655.9	13915.61	13744.11	.167	.0	.91	7.5	19.4	6.0	40.4	6.3	18.5	2.0
Note. AIC = Akaike information criterion; BIC = Bayesian information criterion; SABIC = Sample adjusted BIC; LMR = adjusted Lo‒Mendell‒Rubin likelihood ratio; BLRT = Bootstrap Likelihood Ratio Test (p ≺ .001), and the correlation coefficients are all significant at the .001 level. NW = L2 nonword reading; WR = L2 word reading; PR = L2 passage reading; WK = L2 word knowledge; SK = L2 syntactic knowledge; RC = L2 reading comprehension.

The latent profiles identified in the study were named on the basis of the analysis of patterns across six indicators and comparison with previous research. Figure 2 illustrates the patterns of these six profiles on the basis of standard scores. Table 5 lists the names assigned to each profile along with their characteristics, and Fig. 3 provides a diagrammatic representation of these profile features.

Table 5

*Profile Characteristics and Naming*
Profile	Characteristics	Naming
Profile 1	- Poor home English environment and low parental support. - Very low pre- and postschool English experience. - Low interest in and necessity of English.	Disinterested in English Group
Profile 2	- Poor home English environment but slightly better parental support. - Low pre- and postschool English experience. - Recognizes the necessity of English despite low overall scores.	English Necessity Recognition Group
Profile 3	- Poor home English environment and low parental support. - Low preschool English experience but increased postschool experience. - Low interest in and necessity of English.	Post-School English Experience Increase Group
Profile 4	- Average home English environment and parental support. - Slightly positive pre- and postschool English experience. - Moderate interest in and necessity of English.	Above Average English Environment Group
Profile 5	- Excellent home English environment and good parental support. - Low postschool English experience. - Moderate interest in and necessity of English.	Excellent Parental Support Group
Profile 6	- Excellent home English environment and above-average parental support. - High pre- and postschool English experience. - High interest in and necessity of English.	Excellent English Environment Group

Additional data were collected and analyzed to better understand the differences between the various reading difficulty profiles, and Table 6 shows the differences across profile groups.

Table 6

*Differences across Profile Groups*
Categories		Profiles							χ² (Chi-squared)
		1(66)	2(175)	3(52)	4(354)	5(64)	6(197)	Total	χ² (Chi-squared)
Personal Characteristics
Gender	Female	33 (50.0)	77 (44.0)	7 (32.7)	153 (43.2)	31 (48.4)	107 (54.0)	18 (46.0)	χ²=10.82*
Gender	Male	33 (50.0)	98 (56.0)	5 (67.3)	201 (56.8)	33 (51.6)	91 (46.0)	91 (54.0)	χ²=10.82*
Region	Capital	1 (0.5)	28 (13.1)	5 (2.3)	97 (45.3)	23 (10.7)	60 (28.0)	214 (100)	χ²=64.59 ***
	Capital Region	61 (11.6)	115 (21.9)	37 (7.1)	190 (36.3)	29 (5.5)	92 (17.6)	524 (100)
	Provincial Small City	4 (2.3)	32 (18.7)	10 (5.8)	67 (39.2)	12 (7.0)	46 (26.9)	171 (100)
School Characteristics
After- School Programs	No	4 (2.5)	20 (12.7)	3 (1.9)	60 (38.2)	24 (15.3)	46 (29.3)	157 (100)	χ²=38.08 ***
After- School Programs	Yes	62 (8.2)	155 (20.6)	49 (6.5)	294 (39.1)	40 (5.3)	152 (20.2)	752 (100)	χ²=38.08 ***
Co teaching	No	62 (10.2)	139 (22.9)	45 (7.4)	224 (36.8)	30 (4.9)	108 (17.8)	608 (100)	χ²=70.57 ***
Co teaching	Yes	4 (1.3)	36 (12.0)	7 (2.3)	130 (43.2)	34 (11.3)	90 (29.9)	301 (100)	χ²=70.57 ***
Innovative School	No	63 (7.7)	167 (20.3)	50 (6.1)	321 (39.1)	53 (6.4)	168 (20.4)	822 (100)	χ²=20.39 ***
Innovative School	Yes	3 (3.4)	8 (9.2)	2 (2.3)	33 (37.9)	11 (12.6)	30 (34.5)	87 (100.0)	χ²=20.39 ***
Home Environment
Home English Environment		1.53 (.533)	1.61 (.500)	1.46 (.503)	1.76 (.461)	3.56 (.500)	3.59 (.493)	2.22 (.996)	F = 580.051***
Parental Support		2.20 (.898)	2.57 (.847)	2.31 (.875)	2.56 (.826)	3.00 (.735)	2.87 (.892)	2.62 (.870)	F = 11.108***
Start Time for English Education
Preschool		1.77 (.856)	2.06 (.923)	.65 (.861)	2.40 (.963)	2.50 (1.113)	2.76 (.912)	2.33 (.996)	F = 21.955***
Postschool		1.47 (.503)	1.35 (.478)	.40 (.495)	3.65 (.478)	1.53 (.503)	3.83 (.378)	2.92 (1.174)	F = 988.558***
Affective Domain
Interest in English		1.64 (.671)	2.22 (.767)	.71 (.800)	2.46 (.817)	2.52 (.776)	2.78 (.856)	.38 (.866)	F = 30.604***
Perception of English Necessity		1.65 (.480)	3.32 (.536)	.71 (.536)	.40 (.523)	3.08 (.741)	3.41 (.705)	.14 (.816)	F = 173.128***
Note. Capital City: Seoul; Nearby Metropolitan Areas: Incheon and Suwon; Regional Small and Medium-Sized Cities: Jeonju, Changwon; whether the school offers after-school programs such as remedial support for struggling students and after-school English classes; whether there is coteaching with native English-speaking teachers; and whether the school has a flexible curriculum.

Research Question 2: Predicting English Reading Difficulties and Related Educational Background Factors

In this study, both profile-based and factor-based multinomial logistic regression analyses were employed simultaneously due to the unique insights each approach provides. The profile-based analysis evaluates the risk of reading difficulties on the basis of the characteristics of specific groups, whereas the factor-based analysis helps understand the detailed impact of individual factors. By combining these approaches, this study contributes to a comprehensive understanding of reading difficulties and severe reading challenges and aids in the development of effective intervention strategies.

The multinomial logistic regression analysis revealed significant improvements in model fit for NWR difficulties. The final model's -2 Log Likelihood improved from 152.782 (baseline) to 47.786 (final), indicating substantial fit enhancement (Chi-square = 104.996, df = 10, p < .001). The model explained 15.7% of the variance (Nagelkerke R²). Profiles 1 (Disinterested in English Group) and 2 (English Necessity Recognition Group) had significant positive effects on NWRD, with Profile 1 having an Exp(B) of 8.095, making participants approximately eight times more likely to face difficulties. For severe nonword reading difficulty (SNWRD), Profile 1 was the strongest predictor (Exp(B) = 36.667), whereas Profiles 2 and 5 (Excellent Parental Support Group) also had significant effects, indicating greater odds of severe difficulty.

For WRD, the model fit improved significantly, with the − 2 Log Likelihood decreasing from 162.533 (baseline) to 48.532 (final) (Chi-square = 114.001, df = 10, p < .001), explaining 16.9% of the variance (Nagelkerke R²). Profiles 1 (Disinterested in English Group) and 2 (English Necessity Recognition Group) were significant predictors for WRD, with Profile 1 having an Exp(B) of 10.531, indicating an approximately tenfold greater likelihood of difficulties. For severe WRD, Profile 1 was again a strong predictor (Exp(B) = 30.167), with Profile 2 also significantly impacting severe difficulties (Exp(B) = 4.585). Table 7 shows the profiles and their impacts.

Table 7

*Profiles and Their Impact*
Profile	Variables	NWR		Variables	WR
Profile	Variables	Exp(B)	95% CI	Variables	Exp(B)	95% CI
Profile 1	NWRD	8.095	3.58–18.28	WRD	10.531	4.394–25.242
Profile 1	SNWRD	36.667	11.73-114.59	SWRD	30.167	11.274–80.722
Profile 2	NWRD	3.746	1.94–7.21	WRD	4.081	1.977–8.423
Profile 2	SNWRD	3.898	1.21–12.51	SWRD	4.585	1.781–11.806
Profile 3	NWRD	2.143	.81-5.63	WRD	1.913	.632-5.795
Profile 3	SNWRD	2.143	.38-12.09	SNWRD	2.105	.506-8.753
Profile 4	NWRD	1.686	.89-3.18	NWRD	1.635	.802-3.331
Profile 4	SNWRD	1.033	.29-3.57	SNWRD	.870	.305-2.484
Profile 5	NWRD	.952	.30-3.01	NWRD	1.863	.658-5.274
Profile 5	SNWRD	4.167	1.08–16.06	SNWRD	2.277	.619-8.368
Note. 1. Reference = Profile 6, General NWR, General WD group; 2. NWRD = Nonword Reading Difficulty; WRD = Word Reading Difficulty; SNWRD = Severe NonWord Reading Difficulty; SWRD = Severe Word Difficulty; 3. Coefficients in bold are statistically significant at p < .05 or lower.

The multinomial logistic regression analysis was used to assess predictors of nonword reading difficulty (NWRD) and severe nonword reading difficulty (SNWRD). The model significantly improved from the baseline, with the − 2 Log Likelihood decreasing from 917.450 to 827.268, and was statistically significant (Chi-square = 90.182, df = 12, p < .001). The model explained 13.6% of the variance (Nagelkerke R² = .136). Significant predictors for NWRD included higher scores on Environmental Variables (EV) (Exp(B) = 0.673, p = .001), Academic Experience (AEX) (Exp(B) = 0.791, p = .006), and Interest in English (IN) (Exp(B) = 0.753, p = .032), indicating that higher scores on these factors are associated with lower odds of NWRD. For SNWRD, the significant predictors were AEX (Exp(B) = 0.511, p < .001) and the necessity of English (NE) (Exp(B) = 0.631, p = .012), where higher scores corresponded to reduced odds of severe difficulties.

The analysis for word reading difficulty (WRD) and severe word reading difficulty (SWRD) revealed a significant model improvement, with the − 2 Log Likelihood decreasing from 929.976 to 824.150 (Chi-square = 105.826, df = 12, p < .001). The model explained 15.8% of the variance (Nagelkerke R² = .158). Significant predictors for WRD included higher scores on EV (Exp(B) = 0.679, p = .003) and AEX (Exp(B) = 0.705, p < .001). For SWRD, significant predictors included AEX (Exp(B) = 0.515, p < .001) and NE (Exp(B) = 0.673, p = .015), indicating that higher scores on these factors reduce the odds of severe word reading difficulties. Overall, higher scores on EV and AEX are associated with reduced odds of WRD, whereas higher AEX and NE scores are linked to lower odds of SWRD. Table 8 shows the variables and their impacts.

Table 8

*Variables and Their Impact*
	Variable	NWR			WR
	Variable	Exp(B)	95% CI		Exp(B)	95% CI
RD	HEE	.673	.530 .855		.679	.525	.879
	PS	1.073	.845	-1.362	1.223	.943	1.586
	PRE	112	.902	-1.371	.948	.755	1.192
	PEE	.791	.668	− .936	.705	.588	.846
	IE	.753	.581	− .975	.846	.638	1.122
	NE	1.123	.862	-1.462	1.079	.811	1.434
SRD	HEE	1.004	.715	-1.410	.887	.649	1.214
	PS	.914	.637	1.312	.850	.616	1.174
	PRE	.717	.507	1.016	.816	.605	1.102
	PEE	.511	.389	.672	.515	.406	.654
	IE	.757	.499	1.149	.698	.483	1.009
	NE	.631	.441	.904	.673	.489	.926
Note. 1. NWR = Non-Word Reading; WR = Word Reading; RD = Reading Difficulty; SRD = Severe Reading Difficulty 2. HEE = Home English environment; PS = Parental support; PRE = Preschool English experience; PEE = Postschool English experience; IE = Interest in English; NE = Necessity of English. 3. Reference = General nonword reading group, general word reading group. 4. Coefficients in bold are statistically significant at p < .05.

This study classified the characteristics of the learning backgrounds of EFL sixth-grade students into profiles via LPA and analyzed the predictive variables of the classification via multinomial logistic regression analysis. By examining profiles based on factors such as the home English environment (HEE), parental support (PS), preschool English experience (PRE), postschool English experience (PEE), interest in English (IE), and necessity of English (NE), the study revealed significant variations in reading challenges among students.

The study identified six distinct profiles with varying levels of reading difficulties: Disinterest in English Group (Profile 1) presented the highest likelihood of severe reading difficulties, highlighting the need for targeted, intensive interventions due to a poor home English environment; low parental support; minimal exposure to English before and after school; and low levels of interest in English and perceptions of necessity. The English necessity recognition group (Profile 2) faced significant reading challenges despite recognizing the importance of English and having slightly better parental support than did Profile 1, suggesting that acknowledgment of English's importance alone is insufficient without supportive home conditions and increased English exposure. The excellent parental support group (Profile 5) demonstrated a high likelihood of severe nonword reading difficulties despite an excellent home English environment and strong parental support, likely due to low preschool English experience and moderate interest in English. This finding indicates that critical, favorable home conditions and parental involvement must be complemented by additional English exposure and motivation to address reading difficulties effectively. The excellent English environment group (Profile 6) presented the most favorable conditions for reducing reading difficulties, with an excellent home English environment, above-average parental support, high levels of preschool and postschool English exposure, and a strong interest in English and the perception of necessity. This underscores the effectiveness of a holistic approach that integrates multiple supportive factors to foster reading development.

The key predictors of reading difficulties identified in this study are a substantial home English environment, increased exposure to English before and after school, and genuine interest in English. These predictors highlight the importance of creating a supportive English-speaking atmosphere at home and extending learning opportunities beyond formal schooling. The findings contrast with those of Kim's (2023) research, which identified postschool English education as the most significant predictor. Kim's study focused on general students, whereas this study concentrated on EFL learners with reading difficulties, which may explain the differing results. Moreover, interest in English was the only predictor in the nonword reading difficulties group, which indicates that phonics education needs to be strengthened to improve decoding skills. This finding also shows that phonics education should be further studied to increase learners' interest, as the learning difficulty group has a low interest in English. Studies have shown that systematic phonics education is helpful for enabling students to recognize letters, make sounds, understand the meaning of words, and improve reading fluency and reading comprehension support (Grossen, 1997; Bond & Dykstra, 1967). Some studies suggest that phonics education can improve students' confidence and lead to learning motivation and more frequent reading activities (National Reading Panel [NRP], 2000; Torgesen, 2000). Furthermore, the need for English emerged as a significant predictor of severe reading difficulties, indicating its role in driving the need for language proficiency.

This study contributes to understanding EFL learners with reading difficulties in a human-centered way. This study also has practical educational implications, especially for students who are disinterested in English and those in the English necessity recognition group, who face severe reading challenges. First, the research results revealed that the causes of reading difficulties are extensive and complex; thus, treatment for this specific group should consider not only school education but also the development of materials to improve the English environment at home and linked education, which considers both regular and nonregular education. It is necessary to provide opportunities such as workshops to parents and caregivers and distribute materials that can be used at home to create an environment where continuous exposure is possible. Second, to expand educational opportunities inside and outside of school, opportunities for after-school activities or primary academic supplementary education should be provided, and programs that can be provided in the local community should be studied to prevent students' academic deficits.

However, despite these contributions, some limitations must be supplemented for follow-up research. First, since this study was cross-sectional, only a cross-sectional factor analysis of learners was possible. Therefore, it is necessary to supplement the identified predictive factors and causal relationships from a long-term perspective. Second, since the study of students' learning backgrounds was conducted on the basis of their questionnaire responses, somewhat subjective results were obtained. Therefore, follow-up research should also include questionnaires from parents to obtain more objective data. Third, this study did not consider other potential factors, such as learners' cognitive ability, socioeconomic status, or quality of education. Therefore, if follow-up studies attempt to analyze these aspects, it is expected that a broader understanding of reading difficulties will be possible.

Ethics Approval: This study was conducted in accordance with the ethical standards of the relevant institutional and national committees and with the 1964 Helsinki Declaration and its later amendments. Approval was obtained from the Institutional Review Board (IRB) of Seoul National University.

Akaike, H. (1974). A new look at the statistical model identification. IEEE Transactions on Automatic Control, 19(6), 716-723. https://doi.org/10.1109/TAC.1974.1100705
Bishop, D. V., & Adams, C. (1990). A prospective study of the relationship between specific language impairment, phonological disorders and reading retardation. Journal of Child Psychology and Psychiatry, 31(7), 1027-1050.
Bond, G. L., & Dykstra, R. (1967). The cooperative research program in first-grade reading instruction. Reading Research Quarterly, 5, 5-142.
Brasseur-Hock, I. F., Hock, M. F., Kieffer, M. J., Biancarosa, G., & Deshler, D. D. (2011). Adolescent struggling readers in urban schools: Results of a latent class analysis. Learning and Individual Differences, 21(4), 438-452.
Brasseur-Hock, I. F., Hock, M. F., Kieffer, M. J., Biancarosa, G., & Deshler, D. D. (2011). Distinguishing features of adept and less adept adolescent readers: A latent class analysis of reading proficiency. Journal of Educational Psychology, 103(3), 632-647. https://doi.org/10.1037/a0023513
Brighton, C. M., Moon, T. R., & Huang, F. H. (2015). Advanced readers in reading first classrooms: Who was truly “left behind”? Considerations for the field of gifted education. Journal for the Education of the Gifted, 38(3), 257-293.
Chall, J. S., & Jacobs, V. A. (1983). Writing and reading in the elementary grades: Developmental trends among low SES children. Language Arts, 60(5), 617-626.
Chen, F., Curran, P. J., Bollen, K. A., Kirby, J. B., & Paxton, P. (2009). An empirical evaluation of the use of fixed cutoff points in RMSEA test statistic in structural equation models. Sociological Methods & Research, 37(3), 464-494. https://doi.org/10.1177/0049124108322236
Cheung, S. K., & Anderson, J. R. (2003). The effect of socioeconomic status, home environment, and parenting on the development of children’s early literacy. Early Childhood Research Quarterly, 18(4), 474-495.
Cheung, S. Y., & Andersen, R. (2003). Time to read: Family resources and educational outcomes in Britain. Journal of Comparative Family Studies, 34(3), 413-433.
Choi, J.-H. (2022). Teachers’ experiences using phonics instruction in Korean elementary school classrooms (Doctoral dissertation). Walden Dissertations and Doctoral Studies, 13266. Retrieved from https://scholarworks.waldenu.edu/dissertations/13266/
Choi, S. Y., & Lee, K. H. (2019). The effects of private tutoring on students' academic achievement in Korea. Asia Pacific Education Review, 20(2), 293-310.
Clemens, N. H., Hsiao, Y.-Y., Griffiths, A.-J., Capin, P., & Simmons, L. E. (2017). Using latent profile and latent transition analyses to assess early reading skills in Spanish-speaking ELs. Journal of Educational Psychology, 109(4), 575-589. https://doi.org/10.1037/edu0000162
Clemens, N. H., Simmons, D., Simmons, L. E., Wang, H., & Kwok, O. M. (2017). The prevalence of reading fluency and vocabulary difficulties among adolescents struggling with reading comprehension. Journal of Psychoeducational Assessment, 35(8), 785-798.
Collins, L. M., & Lanza, S. T. (2010). Latent class and latent transition analysis: With applications in the social, behavioral, and health sciences. Wiley. https://doi.org/10.1002/9780470567333
Collins, L. M., Fidler, P. L., Wugalter, S. E., & Long, J. D. (1993). Goodness-of-fit testing for latent class models. Multivariate Behavioral Research, 28(3), 375-389. https://doi.org/10.1207/s15327906mbr2803_4
Coleman, J. S. (1988). Social capital in the creation of human capital. American Journal of Sociology, 94, S95-S120.
D'Angiulli, A., Siegel, L. S., & Hertzman, C. (2004). Schooling, socioeconomic context, and literacy development. Educational Psychology, 24(6), 867-883.
D'angiulli, A., Siegel, L. S., & Maggi, S. (2004). Literacy instruction, SES, and word‐reading achievement in English‐language learners and children with English as a first language: A longitudinal study. Learning Disabilities Research & Practice, 19(4), 202-213.
Dolch, E. W. (1937). Side lights on a combined word list. The Elementary English Review, 14(1), 22-24.
Duursma, E., Augustyn, M., & Zuckerman, B. (2007). Reading aloud to children: The evidence. Archives of Disease in Childhood, 92(7), 554-557.
Grimm, R. P., Solari, E. J., McIntyre, N. S., & Denton, C. A. (2018). Developmental trajectories of reading fluency in at-risk and not-at-risk first-grade readers: A cross-classified random effects model. Journal of Educational Psychology, 110(7), 1013-1033. https://doi.org/10.1037/edu0000267
Grossen, B. (1997). 30 years of research: What we now know about how children learn to read. National Institute of Child Health and Human Development.
Hagenaars, J. A., & McCutcheon, A. L. (Eds.). (2002). Applied latent class analysis. Cambridge University Press. https://doi.org/10.1017/CBO9780511607500
Inoue, T., Georgiou, G. K., Muroya, N., Maekawa, H., & Parrila, R. (2018). Can earlier spelling predict later literacy skills? A 5-year longitudinal study. Reading and Writing, 31(2), 349-368.
Jeoung, Y., & Kim, H. (2008). The influence of family and school environmental factors on literacy achievements among Korean elementary school students. Journal of Educational Research, 22(2), 91-116.
Jung, S., & Lee, S. (2011). Socioeconomic status and academic achievement in Korean adolescents: The mediating role of self-efficacy and achievement motivation. Korean Journal of Child Studies, 32(5), 1-18.
Kim, E. J. (2023). Identifying distinct subgroups of English reading ability among Korean EFL 6th graders and predictive relationships with English learning backgrounds via latent profile analysis and multinomial logistic regression (Doctoral dissertation, Seoul National University).
Kim, J. H. (2005). Parental involvement and children's educational performance in Korea. Educational Review, 57(3), 369-382.
Kim, J. H. (2017). The effects of family background on students' academic achievement. Korean Journal of Sociology of Education, 27(1), 47-71.
Kim, J., & Lee, J. (2015). The effects of shadow education on students' academic performance in Korea. Economics of Education Review, 45, 1-12.
Kim, J., & Lee, J. (2018). Exploring the impact of private tutoring on academic outcomes: Evidence from Korea. KEDI Journal of Educational Policy, 15(1), 1-24.
Kornilov, S. A., & Grigorenko, E. L. (2018). Family-based genetic study of reading and language disabilities in Russian. Scientific Studies of Reading, 22(5), 390-403. https://doi.org/10.1080/10888438.2018.1483923
Kornilov, S. A., & Grigorenko, E. L. (2018). What reading disability? Evidence for multiple latent profiles of struggling readers in a large Russian sibpair sample with at least one sibling at risk for reading difficulties. Journal of Learning Disabilities, 51(5), 434-443.
Kwak, K. (2006). Self-concept, family systems, and family conflict in Korean-American adolescents. Journal of Adolescence, 29(5), 713-727.
LaBerge, D., & Samuels, S. J. (1974). Toward a theory of automatic information processing in reading. Cognitive Psychology, 6(2), 293-323.
Lee, B. M. (2002). An early English education: Its effects and significance in the Korean EFL context. Foreign Language Education Research, 5, 13-37.
Lee, J. Y. (2003). The influence of private tutoring on middle school students' academic achievement in Korea. Korean Journal of Educational Research, 41(3), 99-122.
Lee, J. Y. (2006). Private tutoring and its impact on students' academic achievement in Korea. Seoul: Korean Educational Development Institute.
Lee, J. Y. (2018). Analyzing the impact of private tutoring on students' academic performance in Korea. Seoul: Korean Educational Development Institute.
Lesaux, N. K. (2012). Reading and reading instruction for children from low-income and non-English-speaking households. The Future of Children, 73-88.
Lesaux, N. K., & Kieffer, M. J. (2010). Exploring sources of reading comprehension difficulties among language minority learners and their classmates in early adolescence. American Educational Research Journal, 47(3), 596-632. https://doi.org/10.3102/0002831209355469
Lo, Y., Mendell, N. R., & Rubin, D. B. (2001). Testing the number of components in a normal mixture. Biometrika, 88(3), 767-778. https://doi.org/10.1093/biomet/88.3.767
Lubke, G. H., & Muthén, B. O. (2007). Performance of factor mixture models as a function of model size, distribution, and class-specific parameters. Structural Equation Modeling, 14(1), 26-47. https://doi.org/10.1080/10705510709336601
Magidson, J., & Vermunt, J. K. (2004). Latent class models for classification. In Handbook of statistical modeling for the social and behavioral sciences (pp. 575-603). Springer. https://doi.org/10.1007/978-1-4419-9686-7_22
McLachlan, G. J., & Peel, D. (2000). Finite mixture models. Wiley. https://doi.org/10.1002/9780470316871
Muthén, B. (2001). Second-generation structural equation modeling with a combination of categorical and continuous outcomes. Political Analysis, 9(3), 225-249.https://doi.org/10.1093/oxfordjournals.pola.a004356
Muthén, B., & Muthén, L. K. (1998-2015). Mplus user’s guide (8.7th ed.). Muthén & Muthén.
National Reading Panel (NRP). (2000). Teaching children to read: An evidence-based assessment of the scientific research literature on reading and its implications for reading instruction. National Institute of Child Health and Human Development.
Nylund, K. L., Asparouhov, T., & Muthén, B. O. (2007). Deciding on the number of classes in latent class analysis and growth mixture modeling: A Monte Carlo simulation study. Structural Equation Modeling, 14(4), 535-569. https://doi.org/10.1080/10705510701575396
Schwartz, G. (1978). Estimating the dimension of a model. The Annals of Statistics, 6(2), 461-464. https://doi.org/10.1214/aos/1176344136
Sclove, S. L. (1987). Model-selection criteria are applied to some problems in multivariate analysis. Journal of the American Statistical Association, 82(399), 310-320. https://doi.org/10.2307/2289476
Tofighi, D., & Enders, C. K. (2007). Identifying the correct number of classes in latent class analysis: A comparison of methods. Proceedings of the American Statistical Association, 2007, 1254-1261.
Torgesen, J. K. (2000). Individual differences in response to early interventions: The role of phonemic awareness. Journal of Learning Disabilities, 33(2), 155-161.

The authors declare no competing interests.

Download PDF

Version 1

posted

You are reading this latest preprint version

Unraveling Predictors of Reading Difficulties in EFL 6th Graders: Insights from latent profile analysis and multinomial logistic regression

Status:

Version 1

Abstract

Figures

Introduction

Literature Review

The Critical Role of Decoding Ability in Early Reading Success

Sociocultural Factors Affecting Reading Difficulties

Methodology

Participants

Procedure

Measurement Variables

Analysis methods

Results

Descriptive statistics of the students’ L2 reading ability measurements

Research Question 1: Categorizing into Distinct Profiles and Identifying the Characteristics of Each Profile

Research Question 2: Predicting English Reading Difficulties and Related Educational Background Factors

Discussion and Conclusion

Declarations

References

Additional Declarations

Status:

Version 1