Language Learning ISSN 0023-8333 EMPIRICAL STUDY The Lexical Basis of Second Language Reading Comprehension: From (Sub)Lexical Knowledge to Processing Efficiency Mona G. Alshehri a a,b and Dongbo Zhangb Taif University b University of Exeter Abstract: This study compared how distinct lexical competences, including lexical knowledge as well as processing skills at both word/lexical and sublexical/ morphological levels, collectively and relatively predict reading comprehension in adult learners of English as a foreign language (EFL). The participants were 220 Arabicspeaking EFL learners in a Saudi university. A battery of paper- and computer-based tests was administered to measure the participants’ lexical competences, reading comprehension ability, and working memory. Hierarchical regression analyses revealed that over and above working memory, both lexical and sublexical knowledge were significant and unique predictors of reading comprehension, and sublexical processing efficiency, as opposed to lexical processing efficiency, predicted reading comprehension significantly. In addition, among the measured lexical competences, lexical knowledge was the strongest predictor, and the two knowledge variables collectively had a far greater influence on reading comprehension than did the two processing efficiency variables. These findings are discussed in light of the lexical basis of text comprehension. Keywords reading comprehension; English as a foreign language; lexical quality; lexical knowledge; lexical processing This study was part of the project for the first author’s PhD dissertation completed at the University of Exeter. The first author wishes to thank Taif University for sponsoring her study in the United Kingdom. We thank Laura Ciaccio, Theres Grüter (Language Learning Associate Editor), Emma Marsden (Language Learning Journal Editor), and anonymous reviewers for their comments and suggestions that helped improve this paper. Any errors are our own. Correspondence concerning this article should be addressed to Mona G. Alshehri, Department of Education, Taif University, Al Hawiyah, Taif 26571, Kingdom of Saudi Arabia. Email: mona.sh@tu.edu.sa The handling editor for this article was Theres Grüter. Language Learning 00:0, xxxx 2021, pp. 1–40 © 2021 Language Learning Research Club, University of Michigan DOI: 10.1111/lang.12478 1 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension Introduction The lexical quality hypothesis (Perfetti, 2007) contends that high-quality representations of lexical and sublexical features are fundamental for efficient word recognition and word-to-text integration and, consequently, for text comprehension. This hypothesis underscores the important role of diverse lexical processes and, accordingly, readers’ lexical competences in reading comprehension. Essentially, efficient text comprehension necessitates not only rich knowledge of word meanings but also an ability to process printed words and access their meanings rapidly (i.e., lexical processing or word recognition efficiency). Previous studies involving diverse groups of second language (L2) readers have confirmed the importance of lexical knowledge, notably vocabulary size/breadth, in reading comprehension (Choi & Zhang, 2021; Grabe, 2009; Zhang, 2012). Yet, less is known about the role of knowledge of sublexical features, notably morphological knowledge that, in light of the lexical quality hypothesis, should also play an important role in L2 reading comprehension. Limited research has concurrently considered both lexical and sublexical knowledge in adult L2 readers of English (see Zhang & Koda, 2012, for an exception; and see Ke, Miller, Zhang, & Koda, 2021, for a review and metaanalysis on the role of morphological awareness in biliteracy development in children). More important, the literature on L2 reading comprehension has paid little attention to lexical and sublexical processing efficiency that theoretically should also be fundamentally important given that efficient comprehension is a goal of reading (Grabe, 2009; Koda, 2005). Further research is thus warranted on how diverse lexical competences, which have been defined in this study to include not only lexical and sublexical knowledge but also lexical and sublexical processing efficiency, contribute to L2 reading comprehension. To this end, this study measured distinct lexical competences using a battery of paperand computer-based tests in a large group of Arabic-speaking learners of English as a foreign language (EFL) in a Saudi university and compared how the measured competences collectively and relatively predicted these learners’ L2 reading comprehension. Background Literature (Sub)Lexical Knowledge in Reading Comprehension Reading comprehension can be understood as “the process of simultaneously extracting and constructing meaning through interaction and involvement with written language” (RAND Reading Study Group, 2002, p. 11). The construction-integration model (Kintsch, 1988) contends that the process Language Learning 00:0, xxxx 2021, pp. 1–40 2 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension of text comprehension starts with a reader’s accessing and integrating word meanings for establishing a text model, and then the reader builds a situation model through activation of background knowledge and various inferencing processes. Reading comprehension thus arguably necessitates various linguistic processes including, notably, lexical processes that also underpin Perfetti’s (1985) verbal efficiency theory and the lexical quality hypothesis that Perfetti (2007) subsequently developed. The lexical quality hypothesis places lexical representations and processes at the center of a reading systems framework (Perfetti & Stafura, 2014) and posits that high-quality representations of lexical and sublexical features are fundamental to text comprehension (Perfetti, 2007). These representations involve the features of four constituents of word identity: orthography, phonology, semantics, and morphosyntax (Perfetti, 2007). Together, the quality of these four features and the coherence among them facilitate the rapid, lowresource retrieval of lexical word identities and their integration into a mental model of a text (Perfetti, 2007; Perfetti & Stafura, 2014). The above theoretical outline of the lexical underpinnings of reading comprehension is largely situated in the first language (L1) context but should pertain to L2 reading comprehension as well (Grabe, 2009). Words are the building blocks of texts. To comprehend a text, L2 readers need to know the meanings of the words that make up the text. The knowledge of word meanings, defined in this study as lexical knowledge, should thus play a critically important role in text comprehension. This instrumentalist view (Anderson & Freebody, 1981) of the importance of lexical knowledge in text comprehension can be well understood from a strand of L2 research that has focused on lexical coverage and adequate comprehension of texts (e.g., Hu & Nation, 2000; Schmitt, Jiang, & Grabe, 2011). Other studies of L2 readers of English have revealed strong positive correlations between vocabulary size/breadth of knowledge (i.e., the number of words whose meanings are known) and reading comprehension ability (e.g., Farran, Bingham, & Matthews, 2012; Qian, 1999; Zhang, 2012). Grabe (2009) highlighted that correlation coefficients could be greater than .90. Jeon and Yamashita’s (2014) meta-analysis showed that vocabulary knowledge is one of the strongest correlates of L2 reading comprehension (only next to grammatical knowledge; on average, r = .79). Compared to the wide recognition of and strong empirical evidence for the importance of lexical knowledge (i.e., knowledge of word meanings in the context of this study) in L2 reading comprehension, attention in the L2 literature has been limited to the important role of knowledge of sublexical features encapsulated in the lexical quality hypothesis (Perfetti, 2007). As we mentioned 3 Language Learning 00:0, xxxx 2021, pp. 1–40 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension earlier, high-quality representations of lexical and sublexical features are fundamental to efficient word recognition and word-to-text integration (Perfetti, 2007). Additionally, the binding of constituent features also plays an essential role. Morphology (both inflectional and derivational), in particular, has been underscored by some scholars as an important constituent binding mechanism, and morphological representations have a strong implication for reading acquisition (e.g., Ke et al., 2021; Kirby & Bowers, 2017). For example, in addition to modifying the meaning and (sometimes) the part of speech of the base word to which a suffix is added, English derivation is often characterized by phonological and/or orthographic change to the base word as well (e.g., apply → applicable). Theoretically, morphological knowledge (and processing, which is discussed in the next section) should also play an important role in the comprehension of English texts where multimorphemic words are prevalent (Nagy & Anderson, 1984). In fact, Jeon and Yamashita’s (2014) meta-analysis revealed that, on average, morphological knowledge had a correlation of .61 with L2 reading comprehension. It should be noted, however, that the number of effect sizes/correlation coefficients meta-analyzed (k = 6) for morphological knowledge was notably smaller than the number for vocabulary knowledge (k = 31). Thus, even though morphological knowledge, like orthographic knowledge (r = .51) and phonological awareness (r = .48), was categorized in the metaanalysis as a low-evidence predictor of reading comprehension because of the small number of correlations retrieved from the literature, the moderate average correlation did seem to lend clear support to the importance of morphology in L2 reading comprehension. In a more recent meta-analysis, Ke et al. (2021) reported a mean correlation of .52 between L2 morphological awareness and reading comprehension (k = 17). The issues that wait to be further explored in the L2 literature, however, are how morphological knowledge is important for reading comprehension and whether it predicts L2 reading comprehension over and above lexical knowledge. Theoretically, morphological knowledge such as knowledge of roots and affixes can contribute to text comprehension, independently of lexical knowledge, through at least two major mechanisms. On the one hand, the reader can apply morphological knowledge for more accurate and rapid recognition of morphologically complex words in a text by, for example, dividing those words into their morphemic constituents; on the other hand, morphological knowledge serves as a reliable strategy for the reader to unlock meanings of unknown words in textual reading, that is, instantaneous resolution of Language Learning 00:0, xxxx 2021, pp. 1–40 4 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension vocabulary gaps during reading or “on the spot vocabulary learning” (Nagy, 2007, p. 64). The empirical literature, however, has produced inconsistent findings. Zhang and Koda (2013), for example, found young Chinese-speaking EFL learners’ English morphological awareness that covered both derivation and compounding predicted their reading comprehension, over and above vocabulary knowledge (or lexical knowledge as defined for the purpose of the present study). Similar findings were also reported in some studies on young bilingual readers (e.g., Kieffer & Lesaux, 2008; Zhang, 2017). Yet, a significant, unique effect did not surface in Farran et al. (2012) study of Grades 3 and 5 Arabicspeaking bilingual readers of English in Canada. English morphological awareness barely explained any additional amount of variance in English reading comprheneison after vocablary knoweldge was also included in the regression model (vocabulary knowledge was actually the strongest predictor of reading comprehension; see Table 7, p. 2175). Likewise, in a study of adult Chineseand Korean-speaking learners of English in Canada, Qian (1999) found that morphological knowledge—knowledge of English affixes and stems that was intended to be one of the measures for vocabulary depth—did not uniquely and significantly predict reading comprehension. In Zhang and Koda’s (2012) study of adult Chinese-speaking EFL learners, derivational knowledge did not surface as a unique and significant predictor of reading comprehension when they controlled for vocabulary knowledge. (Sub)Lexical Processing Efficiency and Reading Comprehension Although it is essential that readers possess diverse linguistic knowledge for text comprehension, comprehension would be hampered if readers have not automatized lower-level linguistic processes. Comprehension requires simultaneous orchestration or execution of a number of processes (Perfetti, 1999); yet working memory capacity is limited (Baddeley, 2007). A lack of automatized lower-level processes would constrain the participation of higher-order processes such as textual inferencing for effective construction of a mental model. From a lexical perspective, because words are intended for use in the real world, including for reading texts, knowing a word should not be simply about an ability to “recognize it in connected speech or in print” and “to access its meaning” but should entail the competence “to do these things within a fraction of a second” (Nagy & Scott, 2000, p. 273). The lexical quality hypothesis (Perfetti, 2007), and its predecessor the verbal efficiency theory, embodies “a capacity theory of comprehension” (Just & Carpenter, 1992, p. 122). It underscores high-quality representations of (sub)lexical features because they are 5 Language Learning 00:0, xxxx 2021, pp. 1–40 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension fundamental to rapid recognition of printed words and word-to-text integration processes. (Sub)lexical processing efficiency is an essential element of the reading comprehension process (Perfetti & Stafura, 2014). In the L1 English reading literature, particularly studies of school children or developing readers, sight word recognition efficiency and word decoding fluency have been found to be critical determinants of reading comprehension (Garcia & Cain, 2014). Theoretically, the above emphasis on efficient lexical and sublexical processing should not pertain to L1 or monolingual readers only. In fact, word recognition efficiency, that is, accurate and rapid recognition of printed words, has been recognized as essential to L2 reading comprehension (Grabe, 2009; Koda, 2005). Empirically, however, compared to the L1 reading literature, research that has considered fluency-related lexical competences has appeared much less often in the literature on L2 English reading, and the existing body of research has often approached the issue from diverse perspectives and generated mixed findings. On the one hand, some studies of young ESL learners or bilingual children, like those of monolingual children, considered the contribution of word decoding fluency to reading comprehension. Proctor, Carlo, August, and Snow (2005), for example, found that after they had controlled for oral vocabulary, English decoding fluency was not a unique and significant predictor of fourthgrade Spanish-speaking ESL learners’ reading comprehension in the United States. Yet, in Pasquarella, Gottardo, and Grant’s (2012) study of adolescent L2 readers of English in Canada, real and pseudoword decoding fluency significantly predicted reading comprehension after the researchers had controlled for vocabulary knowledge. On the other hand, there has been a small number of studies, mostly of foreign language learners of English, that approached the issue of lexical processing efficiency in light of readers’ rapid lexical/semantic decision. As part of the NELSON project, van Gelderen et al. (2004), for example, measured adolescent Dutch-speaking EFL readers’ speed of word recognition with a lexical decision task, that is, a task that asked learners to decide as fast as they could whether a letter string presented on a computer screen was an existing word (see also Harrington, 2018, where lexical decision tasks were intended to measure L2 lexical facility). Reaction times (RTs) and accuracy of responses were both recorded. Among the five concurrent predictors of English reading comprehension, only vocabulary knowledge, in addition to metacognitive knowledge, uniquely and significantly predicted reading comprehension. A significant, unique effect did not surface for the RTs or for word recognition speed. Yamashita’s (2013) study of Japanese-speaking university EFL learners, Language Learning 00:0, xxxx 2021, pp. 1–40 6 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension however, found that reading comprehension was significantly predicted by learners’ efficiency of decoding (judgement on whether a nonce word could be read as an English word) and lexical meaning access (judgment on whether words in a pair were antonyms) measured with a paper-based, timed yes/no decision task. It should be noted, however, that Yamashita, unlike van Gelderen et al. (2004), did not concurrently consider the students’ lexical knowledge. It thus remains unclear whether the significant effect identified for the processing efficiency measures would remain had a lexical knowledge measure been included. To date, very little research has aimed to test whether sublexical processing efficiency, particularly morphological processing efficiency, would be a dimension of lexical competence that may uniquely predict L2 reading comprehension along with other dimensions (lexical vs. sublexical/morphological knowledge, on the one hand, and lexical processing efficiency, on the other hand). Overall, despite increasing interest in morphological knowledge and L2 reading comprehension (e.g., Ke et al., 2021; Kieffer & Lesaux, 2008; Zhang & Koda, 2012) and L2 morphological processing and lexical representation (Clahsen, Felser, Neubauer, Sato, & Silva, 2010; see also Ciaccio & Clahsen, 2020), little effort has been expended to combine the two lines of research and to examine how morphological processing efficiency may have a unique role to play during text reading. Logic suggests that if morphological knowledge is important for lexical inferencing and/or word decoding fluency during text comprehension, as some L2 studies have suggested (e.g., Zhang & Koda, 2012), the use of or access to this knowledge must occur in a rapid manner for comprehension to be smooth and efficient. Zhang and Ke (2020) underscored the importance of morphological decoding fluency in L2 reading comprehension. If efficient morphological processing, which entails quick access to morphological features such as morphological structure and meanings of morphemic constituents, is not in place, fluent morphological decoding would not be possible. In other words, morphological knowledge is necessary but insufficient for efficient processing or recognition of multimorphemic words in print. Empirically, as in the case of lexical knowledge versus lexical processing efficiency, studying morphological processing efficiency in conjunction with morphological knowledge would be warranted in order to explore their hypothetically unique contribution to L2 reading comprehension. The Present Study This study set out to address the aforementioned gaps and explore the lexical basis of L2 reading comprehension in light of the lexical quality hypothesis. 7 Language Learning 00:0, xxxx 2021, pp. 1–40 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension The overarching question to be answered was: How do distinct lexical competences collectively and relatively predict L2 reading comprehension? Three sets of research questions were further posed to guide this study. The first set examined the contribution of lexical versus sublexical predictors; the second set examined the contribution of knowledge versus processing efficiency predictors; and the last set examined the collective and relative contributions of the four lexical competences. 1. How does lexical versus sublexical knowledge, on the one hand, and lexical versus sublexical processing efficiency, on the other hand, relatively predict L2 reading comprehension? How does lexical-level competence (knowledge and processing efficiency) versus sublexical competence (knowledge and processing efficiency) relatively predict L2 reading comprehension? 2. How does lexical knowledge versus processing efficiency, on the one hand, and sublexical knowledge versus processing efficiency, on the other hand, relatively predict L2 reading comprehension? How does knowledge (lexical and sublexical) versus processing efficiency (lexical and sublexical) relatively predict L2 reading comprehension? 3. How do the four lexical competences—lexical and sublexical, on the one hand, and knowledge and processing efficiency, on the other hand—collectively and relatively predict L2 reading comprehension? Method Participants This study was conducted at a women’s university in Saudi Arabia. The participants were 268 Arabic-speaking first-year students in the university. For various random reasons such as absence from class or schedule conflict, 48 students missed one or more of the testing sessions. The analyses for this study were, therefore, based only on the data from those who attended all the testing sessions described below (N = 220). The students’ age ranged between 17 and 22 years (Mage = 20 years). The background questionnaire showed that a large majority of the participants started learning English when they were about 12 years old. The participants represented a range of undergraduate majors offered by the Saudi university, including, for example, media, English, chemistry, nutrition, and computer science. English is generally the medium of instruction in Saudi universities, particularly for science and engineering majors. Before proceeding to their discipline studies in English, which typically starts with Language Learning 00:0, xxxx 2021, pp. 1–40 8 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension the second year, Saudi university students need to go through a whole year of intensive English learning to enhance their English proficiency, particularly English for academic purposes. This was the case for the participants of this study, who were first-year students. They all participated on a voluntary basis. Measures We administered a battery of paper- and computer-based tests that are described in detail below, in groups or individually, to measure the participants’ distinct lexical competences, reading comprehension, as well as their working memory. We piloted all instruments with 30 other first-year students who were studying at the same university but who did not participate in the study. We subsequently modified some of the instruments, and we also collected feedback from some students from the pilot testing to help with the modification process. All tasks (Alshehri & Zhang, 2021a, 2021b, 2021c, 2021d, 2021e) except the reading comprehension task are available on https://www.iris-database.org. The reading comprehension task is not publicly available because it was from the Gates-MacGinitie Reading Tests (MacGinitie, MacGinitie, Maria, & Dreyer, 2000), which are copyright-protected and can be purchased from https://riversideinsights.com. All measures showed a fair to high level of reliability (see Table 1; Brown, 2014). In light of recent discussions on and recommendations for instrument reliability based on internal consistency in the literature on psychoeducational assessment and applied linguistics (e.g., McNeish, 2018; Plonsky & Derrick, 2016), we have reported McDonald’s omega, which does not assume Tauequivalence and which we calculated using the structural equation modeling method following Hayes and Coutts’s (2020) recommendations. We have also reported Cronbach’s alpha because of its wide, albeit increasingly contested, use in the literature. Reading Comprehension We measured reading comprehension with a standardized reading test, Form S from the Gates-MacGinitie Reading Tests (MacGinitie et al., 2000). We selected this test because it considers different types of texts and assesses literal as well as inferential comprehension. Another consideration was that, as opposed to any retired standardized tests that target nonnative speakers of English (e.g., IELTS), the participants were unlikely to have taken this test. Although this test is more commonly used for L1 populations, it has been widely used to measure L2 readers’ comprehension as well (e.g., Akamatsu, 2003; Li & Kirby, 2015). We selected four short reading passages (mean length about 120 9 Language Learning 00:0, xxxx 2021, pp. 1–40 Measure M SD 95% CI αb ωc 21 8.15 3.47 [7.69, 8.61] .630 .641 12 10 10 30 7.78 6.29 4.41 18.44 2.92 2.57 2.47 5.62 [7.39, 8.17] [5.94, 6.63] [4.09, 4.74] [18.03, 19.47] .755 .747 .661 .812 – 1,580.70 343.20 – 2,806.00 1,286.90 24 14.01 3.99 [1,538.60, 1,622.80] [2,635.10, 2,977.00] [13.55, 14.59] – 1,966.50 423.70 – 3,631.80 1,298.60 72 40 38.10 30.27 14.58 7.73 [1,910.20, 2,023.40] [3,459.20, 3,804.30] [36.16, 40.04] [29.67, 31.65] Skewness SE Kurtosis SE 0.479 0.154 −0.161 0.306 .757 .750 .670 .795 −0.370 −0.392 −0.696 −0.525 0.152 0.152 0.153 0.151 −0.603 −0.638 −0.102 −0.203 0.303 0.303 0.306 0.302 –d –d −0.593 0.152 1.126 0.302 –d –d 1.994 0.164 5.357 0.327 .684 .689 0.087 0.153 −0.424 0.304 –d –d −0.829 0.151 1.038 0.302 –d –d 1.413 0.164 3.720 0.327 .949 .904 .951 .911 0.320 −0.882 0.151 0.151 −0.490 0.182 0.300 0.301 (Continued) 10 Lexical Basis of L2 Reading Comprehension Reading comprehension Affix form Affix meaning Affix function Separability (accuracy) Separability (raw RT) Separability (IESe ) Combinability (accuracy) Combinability (raw RT) Combinability (IESe ) Vocabulary levels Lexical decision (accuracy) Itemsa Alshehri and Zhang Language Learning 00:0, xxxx 2021, pp. 1–40 Table 1 Measures and descriptive statistics for all competency measures Measure M SD 95% CI αb ωc Skewness SE – 973.95 186.50 –d –d −0.298 0.151 0.050 0.301 – 1,358.30 537.10 –d –d 1.720 0.165 4.071 0.328 20 14.02 3.73 [951.20, 996.70] [1,302.10, 1,535.00] [13.84, 14.76] .754 .736 −1.110 0.151 1.740 0.300 – 1,988.30 320.70 –d –d −0.079 0.152 −0.046 0.302 – 3,039.80 1278.50 –d –d 2.397 0.164 5.671 0.327 [1,949.00, 2,027.60] [2,870.00, 3,209.70] Kurtosis SE Note. N = 220. Affix form, affix meaning, and affix function = sublexical knowledge; separability and combinability = sublexical processing; vocabulary levels = lexical knowledge; lexical decision = lexical processing; RT = reaction time; IES = inverse efficiency score. a Number of items in the measure. b Cronbach’s α estimate of reliability. c McDonald’s ω estimate of reliability, which does not assume Tau-equivalence, was calculated using the structural equation modeling method in Mplus (Version 8.0; Muthén & Muthén, 1998–2017; see Hayes & Coutts, 2020). d Reliability could not be calculated for raw RTs for computer-based measures because participants showed diverse patterns of correct “yes” reactions across real-word stimuli. Reliability also could not be calculated for IES RTs because there was only one holistic score for each participant. e IES was calculated by dividing participants’ raw mean RT by the percentage of their correct responses. Lexical Basis of L2 Reading Comprehension Language Learning 00:0, xxxx 2021, pp. 1–40 Lexical decision (raw RT) Lexical decision (IESe ) Working memory (accuracy) Working memory (raw RT) Working memory (IESe ) Itemsa Alshehri and Zhang 11 Table 1 (Continued) Alshehri and Zhang Lexical Basis of L2 Reading Comprehension words) from Level 5, based on our expert knowledge about Saudi university students’ reading proficiency and on our pilot study. Of the four passages that we selected, we deliberately chose two to be informational and the other two to be narrative. Each passage was accompanied by five or six multiple-choice questions (each with four response options), with a total of 21 questions across the four passages. This test was paper-based and administered in two class sessions, two passages in each session and each session about 25 minutes. The participants were instructed to read the passages silently and to circle an answer for each question. Lexical Knowledge We narrowly defined lexical knowledge as learners’ knowledge of meanings of individual words. We intended lexical knowledge to represent the participants’ vocabulary breadth and measured it with a modified version of the Vocabulary Levels Test (Schmitt, Schmitt, & Clapham, 2001; Webb, Sasao, & Ballance, 2017). The test for this study covered only four levels of word frequency: 1,000, 2,000, 3,000, and 5,000 words. From each frequency level, we randomly sampled six clusters of words; each cluster consisted of a list of six words and three meaning choices. Different from the original Vocabulary Levels Test, we translated the three meaning choices and presented them in Arabic, the native language of the participants. The participants were asked to select a word to match each meaning choice. This test was administered in one class session; the participants were given 20 minutes to complete it. Sublexical Knowledge We were particularly interested in learners’ sublexical knowledge that pertains to morphology or more specifically derivation. We measured the participants’ morphological knowledge with a task that we modeled on the Word Part Levels Test (Sasao & Webb, 2017). Although we retained the format, the instructions, and the scoring method of the original test, we redesigned some items, giving consideration to Saudi university students’ English learning experience and knowledge of English prefixes and suffixes. The test consisted of three sections that assessed knowledge of form, meaning, and use/function of English affixes (e.g., -less and super-). The first section consisted of 12 items that measured knowledge of the correct written forms of common English prefixes and suffixes. The participants were presented with four orthographically similar options, only one of which was a correct affix and should thus have been circled (e.g., multi-, mul-, mlt-, tui-). The second section consisted of 10 items that measured knowledge of Language Learning 00:0, xxxx 2021, pp. 1–40 12 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension meanings of affixes. The participants were asked to select from four choices a simple English word that conveyed the meaning of a target prefix or suffix. For each affix (e.g., un-), two words (e.g., unhappy and unfair) were given as examples to contextualize the use of the affix; the Arabic translation of the four English word choices (e.g., for un-: again, no, back, and new) was also provided. The last section included 10 items that measured knowledge of how an affix indicates the part of speech of a derivational word (i.e., the syntactic properties of affixes). For each item, a prefix or suffix (e.g., -ish) was presented together with a derivational word (e.g., selfish) to show its use. The participants were asked to select “noun,” “verb,” “adjective,” or “adverb” to demonstrate their understanding about how the target affix indicated the part of speech of a word to which it was attached. Arabic translations of the terms noun, verb, adjective, and adverb were also provided. The test was administered to the participants in a separate class session of about 20 minutes. Lexical Processing Efficiency Although the lexical knowledge measure described earlier aimed to assess how many words that the participants knew (specifically, their written receptive vocabulary size), lexical processing efficiency in the context of this study concerned visual word recognition efficiency, that is, how rapidly the participants could recognize a printed word that they knew. To measure lexical processing efficiency, we adopted a computer-based lexical decision task. The lexical decision task consisted of 40 real words as well as 20 decodable pseudowords (e.g., toag) as fillers. The real words were randomly selected from the 1,000 word level of the most frequent words based on the Corpus of Contemporary American English (Davies, 2008; and see www.wordfrequency.info for the relevant information) and should thus have been known to the participants. The order of these real words and pseudowords was randomized. The participants were asked to indicate whether they knew a word (i.e., knew its [partial] meaning) displayed on the screen by pressing as quickly as possible the “yes” or “no” key marked on the keyboard (cf. the literature on measuring efficiency of vocabulary recognition using computer-based, yes/no tests such as Harrington, 2018, and Pellicer-Sánchez & Schmitt, 2012). Both RTs and yes/no responses were recorded. Details on the testing procedure are provided later in the Data Collection Procedure section. Sublexical Processing Efficiency The sublexical processing efficiency measure focused on morphological processing. Two computer-based tasks were included. In the separability task, 13 Language Learning 00:0, xxxx 2021, pp. 1–40 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension following Koda’s (2000) study, the participants were asked to decide, as quickly as possible, whether a word presented in the center of the computer screen could be separated into two or more meaningful components (i.e., stem and affix). There were 30 stimulus words that we assumed that the participants would know. Those words were selected from an initial list created by the first author based on her many years of teaching in the university and familiarity with the participants’ curriculum. They were also checked by the participants’ English teachers and later piloted with a separate group of students, as we mentioned earlier. Fifteen of the stimuli were actual derivational words, such as government and disappear, that can be segmented into govern and -ment and dis- and appear, respectively. The other 15 words were monomorphemic words that included a letter or a string of letters resembling an English affix, for example, power and kitchen. Conversely, the combinability task asked the participants to decide, as quickly as possible, whether the two word-parts presented on the computer screen could be combined to make a meaningful longer English word. There were 24 items in this task, including 12 items that were combinable, such as fear and less, and 12 items that were not, such as un and home. Working Memory Text comprehension necessitates the execution of a number of processes, the efficiency of which depends heavily on readers’ mental capacity (Just & Carpenter, 1992). Working memory capacity, in particular, is a significant correlate of L2 reading comprehension (Harrington & Sawyer, 1992). (Sub)lexical processing itself also depends on working memory (Tokowicz, 2014). To obtain a more accurate understanding of the effect of lexical competences, particularly that of processing efficiency, we also measured the participants’ working memory and later included working memory as a covariate when, in regression analysis, we used different lexical competences to predict reading comprehension. We measured working memory with a computerized digit span task, which is one of the most widely used types of working memory tests (Richardson, 2007). The test for this study consisted of 20 numerical sequences— 10 for forward span and 10 backward span—assessing short-term storage of the stimulus sequences. For the forward span items, the participants were asked to decide, as quickly as possible, whether a digit sequence presented on the computer screen was the one that they had just seen immediately beforehand and in the same given order. Likewise, for the backward span items, they were to decide whether a digit sequence was the one that they Language Learning 00:0, xxxx 2021, pp. 1–40 14 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension had just seen immediately beforehand but that had the order reversed. For both types of span, there were five sets of random numerical digits increasing in length of sequence (i.e., number of digits in a sequence; it started with two-digit sequences and ended with six-digit sequences). Each set consisted of two items: one with the order matched and the other with the order not matched. Data Collection Procedure For the paper-based lexical and sublexical knowledge measures and the reading comprehension test, the first author negotiated with the participants’ English language teachers to administer them in groups in four to six class sessions (each session about 20–30 minutes) as it was convenient for the classes. The working memory and the lexical and sublexical processing efficiency measures were administered individually on a laptop computer and run on PsychoPy (Version 3.0; Peirce et al., 2019). The computer-based testing was conducted in a quiet space on the university campus in one session that lasted about 15 minutes. Data collection was completed over a period of four weeks. Task instructions were given in Arabic (or Arabic and English) to ensure the participants’ full understanding of them. The paper-based tests were carried out first, followed by the computerbased tests. For the computer-based testing, the working memory test was administered first, followed by the (sub)lexical processing efficiency measures. For all measures, testing began with an on-screen instruction and some practice items. The participants were asked to give a response for an item presented in the center of the computer screen by pressing as quickly as possible “yes” (the left arrow key) or “no” (the right arrow key) marked with stickers on the keyboard. Both RTs and yes/no responses were recorded. RT was calculated as the interval between the onset of an item and the time of “yes” or “no” being pressed. For the working memory test, the participants began the test by seeing a digit sequence for a fixed length of 1,000 ms. Upon the offset of the stimulus sequence, the question “Is this (a digit sequence) the number you saw in the given order?” or “Is this (a digit sequence) the number you saw in the reverse order?” appeared on the screen and was presented in Arabic. For all the computer-based tests, the pressing of a key automatically activated the next item. If no key was pressed for an item or no response was detected after a certain period of time, the item would automatically disappear and the next item would appear. The time assigned for an item to be answered before disappearing ranged from 1,000 to 2,000 milliseconds, with the baseline time for each item estimated on the basis of the pilot study. 15 Language Learning 00:0, xxxx 2021, pp. 1–40 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension Scoring and Missing Values For the paper-based measures, one point was awarded for each correct answer; an incorrect answer or a missing response did not receive any points. The maximum score possible was thus 21 for the reading comprehension test and 72 for the lexical knowledge test. For the three sections of the sublexical (or morphological) knowledge test, the maximum scores possible were 12, 10, and 10, respectively. The scoring and handling of missing data for the computer-based tasks were less straightforward. There are no consistent methods for handling the data of decision tasks like those in the present study (Jiang, 2012). In the literature on L2 reading comprehension, although some studies have considered both accuracy and RTs of responses (e.g., Cremer & Schoonen, 2013), others have incorporated only RTs (e.g., van Gelderen et al., 2004). In the present study, for all computerized measures, we adopted RTs for our analysis. This choice was also in line with our purpose of comparing (sub)lexical knowledge against processing efficiency for which it was of greater interest to examine the participants’ response or decision latency. We, however, accommodated accuracy rate in the final RT calculation for all computerized measures (see the discussion on the inverse efficiency score or IES in the next paragraph; see also the Limitations and Future Research section). For the lexical processing efficiency task, we focused only on the 40 real words; for those words, we relied on the RTs of correctly answered items. To calculate the right RTs for analysis, for each participant, we recoded the RT of a “no” decision for a real word as missing; and the RT of a missing decision was also coded as missing. Then we calculated the mean RT for each item. A RT that was above or below the item mean by two or more standard deviations was subsequently considered to be an outlier and further recoded as missing. This was followed by computing the mean RT of correctly answered items for each participant. Finally, to accommodate the rate of correct responses, a raw RT was replaced by an inversed value (Ratcliff, 1993). For each participant, the IES was calculated by dividing the raw mean RT by the percentage of correct responses (Townsend & Ashby, 1983). In this respect, the participants with a low RT but a low accuracy rate as well would be penalized for the low accuracy. The same procedure was largely followed for calculating the adjusted RTs for the two sublexical/morphological processing efficiency tasks as well. The only exception was that, unlike the filler items or pseudo words in the lexical decision task, the monomorphemic items for the separability task and the items not combinable for the combinability task were not excluded for RT calculation. This was because a “no” decision on those items was considered Language Learning 00:0, xxxx 2021, pp. 1–40 16 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension to also show the participants’ attention to morphological features and thus to reflect their morphological processing efficiency. Results Descriptive Statistics, Reliabilities, and Normality All statistical analyses, unless stated otherwise, were performed with the SPSS (Version 26) software. Table 1 presents the means, standard deviations, reliabilities (McDonald’s omega and Cronbach’s alpha), and skewness and kurtosis values of all measured competences (Appendix S1 in the online Supporting Information presents elaborated descriptive statistics that include the minimum and maximum scores and the range for each variable). The accuracy rates, raw RTs, as well as IESs/adjusted RTs are shown for the (sub)lexical processing efficiency measures although, for the reason that we mentioned earlier, only IES RTs were used for the subsequent bivariate correlation and regression analyses. For most measures, the skewness and kurtosis estimates were generally below the rule-of-thumb values for univariate normality (e.g., ±2 for both skewness and kurtosis) as well as the critical values that may result in significant deviation from multivariate normality (e.g., ±2 and ±7 for skewness and kurtosis, respectively; Curran, West, & Finch, 1996). The kurtosis of the IES RTs was higher than that of the raw RTs and than that of the paper-based measures. Nonetheless, the kurtosis was within the acceptable range for multivariate normality. Normality of residuals was checked and confirmed through the examination of histograms and probability-probability plots (cf. Gelman & Hill, 2007, who argued that normality is one of the least important assumptions to check when running generalized linear models). Alpha was set at .05 for all correlations and regression analyses. Bivariate Correlations Table 2 shows the bivariate correlations between all the variables. Reading comprehension correlated positively and significantly with all knowledge variables, notably lexical knowledge (i.e., vocabulary knowledge; r = .643), which also produced the highest correlation with reading comprehension. Reading comprehension also correlated negatively and significantly with working memory (r = −.169) and the two sublexical processing efficiency tasks of the separability and combinability (r = −.183 and −.193, respectively). The correlation between reading comprehension and lexical processing efficiency (i.e., lexical decision) was negative as well (r = −.084); however, it did not achieve statistical significance. 17 Language Learning 00:0, xxxx 2021, pp. 1–40 Competency 1. Reading comprehension 2. Affix form 3. Affix meaning 4. Affix function 5. Separability 7. Vocabulary levels 8. Lexical decision 9. Working memory 2 r (p)95% CI .391 (< .001) [.276, .497] .369 (< .001) [.261, .462] .519 (< .001) [.401, .613] −.183 (.007) [−.290, −.062] −.193 (.004) [−.317, −.061] .643 (< .001) [.549, .713] −.084 (.213) [−.330, .054] −.169 (.012) [−.260, −.075] _ .463 (< .001) [.365, .557] .506 (< .001) [.415, .585] −.254 (< .001) [.058, −.374] −.220 (.001) [−.333, −.101] .405 (< .001) [.286, .514] −.137 (.043) [−.328, −.044] −.182 (.007) [−.324, −.046] 3 r (p)95% CI 4 r (p)95% CI 5 r (p)95% CI 6 r (p)95% CI 7 r (p)95% CI 8 r (p)95% CI _ .518 (< .001) [.428, .592] −.305 (< .001) [−.417, −.195] −.294 (< .001) [−.402, −.190] .447 (< .001) [.333, .548] −.228 (.001) [−.398, −.152] −.266 (< .001) [−.393, −.130] _ −.291 (< .001) [−.385, −.197] −.331 (< .001) [−.424, −.234] .547 (< .001) [.422, .643] −.218 (.001) [−.367, −.149] −.176 (.009) [−.273, −.071] _ .369 (< .001) [.252, .490] −193 (.004) [−.289, −.084] .247 (< .001) [.144, .479] .202 (.003) [.053, .353] _ −.206 (.002) [−.322, −.084] .273 (< .001) [.144, .512] .144 (.033) [.021, .274] _ −.227 (.001) [−.365, −.146] −.207 (.002) [−.311, −.066] _ .087 (.197) [.013, .228] Note. N = 220. Affix form, affix meaning, and affix function = sublexical knowledge; separability and combinability = sublexical processing; vocabulary levels = lexical knowledge; lexical decision = lexical processing. The 95% confidence intervals are based on 1,000 bootstrap samples. Bolding indicates correlation coefficients that are significant at p < .05. 18 Lexical Basis of L2 Reading Comprehension 6. Combinability 1 r (p)95% CI Alshehri and Zhang Language Learning 00:0, xxxx 2021, pp. 1–40 Table 2 Bivariate correlations for all measured competences Alshehri and Zhang Lexical Basis of L2 Reading Comprehension It is also important to note that the three measures of sublexical knowledge were all significantly correlated with each other. Knowledge of affix forms significantly positively correlated with knowledge of affix meanings (r = .463) and knowledge of affix function (r = .506); and knowledge of affix meaning and knowledge of affix function also showed a significant positive correlation (r = .518). All three sublexical knowledge measures also significantly positively correlated with lexical knowledge, r = .405, .447, and .557, respectively, for the affix form, meaning, and function tasks. Finally, all the (sub)lexical knowledge measures negatively and significantly correlated with all the (sub)lexical processing efficiency measures, which made sense because the processing efficiency measures had a focus on speed (i.e., the lower the value, the greater the speed). The two sublexical processing efficiency measures and working memory were also positively and significantly correlated. The correlation between lexical processing efficiency and working memory was also positive but not statistically significant (r = .087). Contribution of Lexical Competences to Reading Comprehension A series of hierarchical regression analyses (Jeon, 2015) was performed to examine how different dimensions of lexical competences—lexical versus sublexical and knowledge versus processing efficiency—collectively and relatively contributed to L2 reading comprehension over and above working memory. For all analyses, working memory was entered first into the regression equation as a covariate (it explained about 2.9% of the variance in reading comprehension), followed by different lexical competences entered individually or as a block. The three sublexical knowledge measures were always entered as a block; likewise, the RTs for the morphological separability and combinability tasks were also entered as a block to represent sublexical processing efficiency. The order of entry was also switched for different predictors to test and compare their unique contribution to reading comprehension. Multicollinearity was diagnosed for multiple regression analysis through variable inflation factors, which ranged from 1.108 to 1.864 and were smaller than the lowest bound of rule-of-thumb values (i.e., 2.5) for indicating the presence of multicollinearity (Allison, 1999). Comparing Lexical and Sublexical Predictors The first set of research questions sought to compare lexical and sublexical predictors of reading comprehension. We conducted three sets of regression analyses for this purpose. We first examined how lexical knowledge and sublexical knowledge predictors relatively contributed to reading comprehension, 19 Language Learning 00:0, xxxx 2021, pp. 1–40 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension Table 3 Comparing lexical and sublexical predictors of reading comprehension Steps Predictors R2 95% CI 1 Working memory .029 [.001, .084] Lexical vs. Sublexical knowledge 2 Lexical knowledge .414 [.315, .493] 3 Sublexical knowledge .459 [.353, .527] 2 Sublexical knowledge .300 [.193, .380] 3 Lexical knowledge .459 [.353, .527] Lexical vs. Sublexical processing 2 Lexical processing .034 [.000, .087] 3 Sublexical processing .068 [.008, .125] 2 Sublexical processing .067 [.011, .130] 3 Lexical processing .068 [.008, .125] Lexical vs. Sublexical (knowledge & processing) 2 Lexical knowledge & .419 [.316, .494] processing 3 Sublexical knowledge & .467 [.353, .529] processing 2 Sublexical knowledge & .300 [.185, .374] processing 3 Lexical knowledge .467 [.353, .529] &processing R2 adjusted R2 p .024 .029 .012 .409 .446 .287 .446 .386 .044 .271 .159 < .001 .001 < .001 < .001 .025 .050 .055 .050 .005 .034 .039 .000 .297 .021 .012 .876 .411 .390 < .001 .447 .049 .002 .280 .271 < .001 .447 .167 < .001 and then we analyzed how lexical processing efficiency and sublexical processing efficiency relatively contributed to reading comprehension. Finally, we compared how the two lexical-level competences (i.e., knowledge and processing efficiency together) and the two sublexical competences (also knowledge and processing efficiency together) relatively predicted reading comprehension. Estimates of the regression coefficients in the final regression model can be seen in Appendix S2 in the online Supporting Information for each set of analyses. The upper panel of Table 3 shows that, when we controlled for working memory, lexical knowledge additionally significantly explained 38.6% of the variance in reading comprehension. Over and above working memory and lexical knowledge, sublexical knowledge also significantly predicted reading comprehension; it, however, only additionally explained 4.4% of the variance. When sublexical knowledge was entered into the regression equation as the second step, it significantly added 27.1% to the variance explained. The unique Language Learning 00:0, xxxx 2021, pp. 1–40 20 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension effect of lexical knowledge remained significant; over and above working memory and sublexical knowledge, it explained 15.9% of the variance in reading comprehension. Thus lexical knowledge explained a far greater amount of unique variance than did sublexical knowledge although the unique effect of both predictors was significant. The middle panel of Table 3 shows the unique contribution of lexical versus sublexical processing efficiency. After controlling for working memory, lexical processing efficiency did not predict reading comprehension significantly, whether it was entered before or after sublexical processing efficiency. It barely explained any additional variance in reading comprehension when sublexical processing efficiency was already in the model. However, sublexical processing efficiency uniquely explained a small yet significant proportion of variance in reading comprehension. Specifically, when working memory and lexical processing efficiency were in the regression model, sublexical lexical processing efficiency additionally significantly explained 3.4% of the variance. Last, we compared the effects of the two lexical predictors with those of the two sublexical predictors. As shown in the bottom panel of Table 3, the lexical predictors (knowledge and processing entered as a block) had a far greater unique effect on reading comprehension than did the sublexical predictors, although the unique effect of both was statistically significant. Specifically, over and above working memory and the lexical predictors, the sublexical predictors additionally and significantly explained about 4.9% of the variance in reading comprehension. However, the lexical predictors, when entered into the regression model at the last step, significantly explained about 16.7% of the variance in reading comprehension. Taken together, the findings suggested that lexical-level competences overall had a stronger effect on reading comprehension than did sublexical competences, and this advantage seemed to be attributed to the large effect of lexical knowledge on reading comprehension. With respect to processing efficiency, the effect at the sublexical level, though small, was greater. Comparing Knowledge and Processing Efficiency Predictors The second set of research questions aimed to compare the effects of knowledge and processing efficiency predictors. Three sets of regression analyses again were conducted. We first compared these two types of competence at the lexical level and then at the sublexical level. Last, we compared the effects of lexical and sublexical knowledge (i.e., the two levels together) and those of lexical and sublexical processing efficiency (together). Estimates of the regression 21 Language Learning 00:0, xxxx 2021, pp. 1–40 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension Table 4 Comparing knowledge and processing efficiency predictors of reading comprehension Steps Predictors R2 95% CI 1 working memory .029 [.001, .084] Lexical knowledge vs. Lexical processing 2 Lexical knowledge .414 [.315, .493] 3 Lexical processing .419 [.316, .494] 2 Lexical processing .034] [.000, .087 3 Lexical knowledge .419 [.316, .494] Sublexical knowledge vs. Sublexical processing 2 Sublexical knowledge .300 [.193, .380] 3 Sublexical processing .300 [.185, .374] 2 Sublexical processing .067 [.011, .130] 3 Sublexical knowledge .300 [.185, .374] Knowledge vs. Processing (lexical & sublexical) 2 Lexical & sublexical .459 [.353, .527] knowledge 3 Lexical & sublexical .467 [.353, .529] processing 2 Lexical & sublexical .068 [.008, .125] processing 3 Lexical & sublexical .467 [.353, .529] knowledge R2 adjusted R2 p .024 .029 .012 .409 .411 .025 .411 .386 .004 .005 .385 < .001 .214 .297 < .001 .287 .280 .055 .280 .271 .000 .039 .232 < .001 .997 .012 < .001 .446 .430 < .001 .447 .009 .341 .050 .039 .032 .447 .400 < .001 coefficients can be seen in Appendix S3 in the online Supporting Information for each set of analyses. The upper panel of Table 4 shows the results of the first comparison. Controlling for working memory and lexical processing efficiency, lexical knowledge significantly explained a unique proportion of variance in reading comprehension (about 38.5%). Conversely, however, a unique effect did not surface for lexical processing efficiency when it was entered last into the model, and minimal additional variance of reading comprehension was explained (0.4%). The middle panel of Table 4 presents the results of the second comparison. Sublexical knowledge, whether entered in the model before or after sublexical processing efficiency, significantly predicted reading comprehension. As the last predictor entered in the model, sublexical knowledge uniquely explained about 23.2% of the variance of reading comprehension. Conversely, Language Learning 00:0, xxxx 2021, pp. 1–40 22 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension although working memory capacity was controlled for, sublexical processing efficiency significantly predicted reading comprehension (3.9% of the variance explained), but it failed to significantly predict reading comprehension when sublexical knowledge was also in the model (0.0% of the variance explained). Finally, as shown in the bottom panel of Table 4, the two knowledge measures (lexical and sublexical combined) collectively, uniquely, and significantly explained about 40% of the variance in reading comprehension when working memory and the two processing efficiency measures (lexical and sublexical) were also in the model. Conversely, entered after working memory and the two knowledge predictors, the two processing efficiency measures barely explained any additional variance in reading comprehension (0.9%). Taken together, the above findings seemed to suggest that knowledge was a far stronger predictor of reading comprehension than processing efficiency, which was true for both the lexical and the sublexical level or disregarding the level of competence. Unique Contribution of Each Predictor Distinct from the first two questions, the last set of research questions focused on the unique and relative contribution of each predictor. Regression coefficient estimates can be seen in the bottom panel of either Appendix S2 or Appendix S3 in the online Supporting Information. Table 5 shows that the four lexical competences collectively explained over 40% of the variance in reading comprehension. The top section of the table shows the results of the unique contribution of sublexical knowledge, and sublexical processing efficiency when all the other predictors (working memory included) were in the model. The unique contribution was significant for sublexical knowledge (4.0% of the variance explained), but not for sublexical processing efficiency (0.1% of the variance explained). Likewise, the bottom section of Table 5 shows the unique contribution of lexical knowledge, and lexical processing efficiency, when all the other predictors were in the model. The unique contribution of lexical knowledge was significant (16.5% of the variance explained); yet a significant, unique effect did not surface for lexical processing efficiency (0.8% of the variance explained). Based on the unique proportion of variance explained for reading comprehension (i.e., R2 ), lexical knowledge appeared to be the strongest unique predictor, followed by sublexical knowledge. With the presence of the knowledge predictors and working memory in the model, lexical and sublexical processing efficiency barely contributed to reading comprehension. 23 Language Learning 00:0, xxxx 2021, pp. 1–40 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension Table 5 The unique contribution of each predictor of reading comprehension Steps Predictors R2 95% CI 1 Working memory .029 [.001, .084] Unique contribution of sublexical knowledge vs. processing 2 Lexical knowledge .414 [.315, .493] 3 Lexical processing .419 [.316, .494] 4 Sublexical processing .427 [.318, .498] 5 Sublexical knowledge .467 [.353, .529] 4 Sublexical knowledge .467 [.358, .532] 5 Sublexical processing .467 [.353, .529] Unique contribution of lexical knowledge vs. processing 2 Sublexical knowledge .300 [.193, .380] 3 Sublexical processing .300 [.185, .374] 4 Lexical processing .302 [.184, .373] 5 Lexical knowledge .467 [.353, .529] 4 Lexical knowledge .459 [.346, .523] 5 Lexical processing .467 [.353, .529] R2 adjusted R2 p .024 .029 .012 .409 .411 .413 .447 .452 .447 .386 .004 .008 .040 .048 .001 < .001 .214 .219 .001 < .001 .866 .287 .280 .279 .447 .441 .447 .271 .000 .002 .165 .159 .008 < .001 .997 .413 < .001 < .001 .068 Discussion The present study set out to investigate how four distinct dimensions of lexical competence—lexical versus sublexical, on the one hand, and knowledge versus processing efficiency, on the other hand—collectively and relatively contribute to reading comprehension in adult learners of English so as to shed light on the lexical basis of L2 reading comprehension. The four lexical competences collectively explained over 40% of the variance in the participants’ reading comprehension. Compared to the processing efficiency predictors, the knowledge predictors had a predominant association with reading comprehension. In fact, when the effects of the knowledge predictors were taken into consideration, those of the processing efficiency predictors were no longer significant. Additionally, the lexical predictors, when examined collectively, had a greater effect on reading comprehension than did the sublexical predictors; yet this overall effect did not seem to hold specifically for processing efficiency in that sublexical processing efficiency seemed to have a larger effect on reading comprehension than lexical processing efficiency (nonetheless, the effect of both processing efficiency predictors was very small). Finally, among the four lexical competences, lexical knowledge was the strongest predictor, followed by sublexical knowledge and then the processing efficiency predictors. Language Learning 00:0, xxxx 2021, pp. 1–40 24 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension Lexical Versus Sublexical Knowledge in Reading Comprehension The lexical quality hypothesis (Perfetti, 2007) contends that high-quality representations of lexical and sublexical features are fundamentally important for text comprehension. The lexical basis of reading comprehension that the hypothesis underscores (Perfetti & Hart, 2001) has been largely supported in the L2 (as well as L1) reading comprehension literature. Notably, a strong association has consistently been found between vocabulary knowledge and reading comprehension (Choi & Zhang, 2021; Grabe, 2009; Jeon & Yamashita, 2014). This relationship was confirmed in the present study. The lexical knowledge measure that targeted vocabulary size explained nearly 40% of the variance in reading comprehension (when the effects of working memory and sublexical knowledge were concurrently considered; see Table 4). This finding corroborates previous findings that were derived largely from speakers of first languages other than Arabic (e.g., Japanese, Chinese, Spanish) and that supported a strong association between vocabulary knowledge and reading comprehension. Taken together, these findings suggest that regardless of learners’ L1 background, lexical knowledge or knowledge of word meanings should not be neglected when trying to understanding the components that contribute to L2 reading comprehension. An issue understudied in the literature pertains to the (unique) importance of knowledge of sublexical features encapsulated in the lexical quality hypothesis. In the present study, we focused on (derivational) morphological features because morphology could serve to bind other sublexical features, including orthography, phonology, semantics, and grammar (see Kirby & Bowers, 2017). In fact, this study attended to several aspects of morphological knowledge that touched on orthography (the affix form measure), semantics (the affix meaning measure), and grammar (the affix function measure). In the L2 literature, despite an increasing interest in the role of morphology in reading comprehension, attention to morphology has overall been limited and most existing studies have focused on young EFL learners or bilingual children (e.g., Ke et al., 2021; Kieffer & Lesaux, 2008; Zhang & Koda, 2013). Few studies have attended to this issue in adult learners of English (see Zhang & Koda, 2012, for an exception). In the present study that focused on adult Arabic-speaking EFL learners, all three measures of morphological knowledge significantly and positively correlated with reading comprehension, and, collectively, they significantly predicted reading comprehension over and above lexical knowledge (i.e., vocabulary size), even though the unique effect was much smaller than that of lexical knowledge (see Table 3). This finding thus lends support to the emphasis that the lexical quality 25 Language Learning 00:0, xxxx 2021, pp. 1–40 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension hypothesis places on the importance of sublexical representations for text comprehension. The finding also suggests that, for adult learners of English, morphological knowledge is uniquely important for reading comprehension independent of lexical knowledge. Yet, it seems to differ from the findings of two previous studies that also focused on adult learners. Zhang and Koda (2012) found morphological knowledge only indirectly contributed to reading comprehension and only through vocabulary knowledge; when they controlled for vocabulary knowledge, the effect of morphological knowledge was not significant. Likewise, Qian (1999) did not report a unique and significant effect of the morphological knowledge predictor that was intended to measure an aspect of vocabulary depth knowledge. One reason for the discrepancy of findings might be that the lexical/vocabulary measures in both Zhang and Koda (2012) and Qian (1999) considered aspects of knowledge beyond that of individual word meanings. Specifically, both studies, in addition to vocabulary size (measured with a Vocabulary Levels Test), concurrently considered word association ability as a vocabulary depth measure, which was not the case in our study. Another reason might be that, in contrast to the two previous studies, our study had a more comprehensive consideration of aspects of morphological knowledge, including form, meaning, as well as function. Notably, the affix function task that targeted the participants’ knowledge of the syntactic properties (that is, partof-speech information of derivational affixes), had the highest, positive correlation with reading comprehension in this study (r = .519; see Table 2). This aspect of knowledge, which was not specifically considered in the two previous studies, is particularly underscored by Nagy (2007) as contributing to sentence parsing and reading comprehension. Whichever the reason might be, the above discussion suggests that morphological knowledge overall should be an important underpinning of reading comprehension (see also the size of correlations reported in Jeon & Yamashita, 2014). Yet, whether a unique effect can emerge, over and above lexical knowledge, may depend on what aspects of morphological knowledge are the focus, on the one hand, and what aspects of knowledge at the lexical level are the concurrent focus, on the other hand. This issue warrants further research. (Sub)Lexical Processing Efficiency in Reading Comprehension The processing efficiency measures generated a few very intriguing findings. Overall, when working memory and the two (lexical and sublexical) knowledge predictors were concurrently in the model, neither lexical nor sublexical Language Learning 00:0, xxxx 2021, pp. 1–40 26 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension processing efficiency predicted reading comprehension significantly. This was a surprising finding because, theoretically, for smooth text comprehension to happen, efficient word recognition and word-to-text integration are essential (Perfetti, 2007). In other words, text comprehension necessitates not only rich knowledge of word meanings and sublexical morphological features, which we discussed earlier, but also an ability to efficiently process printed words, including multimorphemic words, and access their meanings during text comprehension. The ability to quickly recognize a word and word parts (that is, the ease of accessing word knowledge), should, in theory, have an added value for reading comprehension (Nagy & Scott, 2000; Perfetti & Hart, 2001). Automatized lower-level processing skills are essential for enabling effective participation of higher-order processes for constructing mental models during text reading. This is in line with a capacity view of discourse comprehension (Just & Carpenter, 1992) and should pertain to any readers of English, whether English is their L1 or L2 (Grabe, 2009; Koda, 2005). We speculate that one interpretation for the lack of a unique and significant effect of the processing efficiency measures is that this finding may reflect what characterizes lexical involvement at the particular developmental stage of our participants. Although the participants had been learning English for at least six years (in a foreign language context), their English proficiency tended to be low. This can be partly seen from their low performance on the reading comprehension measure (the average score for the participants was about 8 out of 21 items; see Table 1). Level 5 of the Gates-MacGinitie Reading Tests, from which the passages and questions were sampled, actually targets fifth graders in an English-speaking context. In other words, for the participants to comprehend the passages, knowledge of word meaning (and knowledge of morphemic meanings for morphologically complex words) should reasonably be the dominant influence. But in the L1 reading literature at least, less skilled comprehenders compared to skilled comprehenders have tended to have problems with word processing or to show less immediate use of word meanings in the integration process (Nation & Snowling, 2004; Perfetti & Stafura, 2014). However, as we discuss in the Limitations section, this finding could also be affected by how RT scores may not adequately represent individual differences in (sub)lexical processing. Another aspect for attention might be that the comprehension test was not administered in a timed condition. Although the participants were asked to complete a test session within a specified period of time, that is, within 10 to 15 minutes per passage, this time restriction might be too relaxed (considering that each passage was only about 120 words long and followed by only 27 Language Learning 00:0, xxxx 2021, pp. 1–40 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension five questions) for processing efficiency to make a noticeable additional difference (above and beyond word knowledge), particularly when individual differences in working memory were also considered. The present finding, however, seems to corroborate findings from the NELSON project that studied adolescent learners of English in the Netherlands (e.g., Fukkink, Hulstijn, & Simis, 2005; van Gelderen et al., 2004). In these studies, word recognition speed was not found to uniquely and significantly predict reading comprehension. In addition, although word recognition training did improve word recognition speed, the effect did not transfer so as to benefit reading comprehension. Despite the weak, unique effects of the two processing efficiency predictors, their relative contribution shown in Table 3 deserves some attention. Specifically, when lexical processing efficiency was controlled for, sublexical processing efficiency had a significant, albeit small, effect on reading comprehension. Conversely, however, this significant effect did not surface for lexical processing. We speculate that this gap might be attributed to the psycholinguistic processes that could be differentially involved in the lexical decision task and the morphological processing tasks. Specifically, when learners make a decision about a highly frequent word such as sweet and visit (in a decontextualized task such as the lexical decision task in this study), they may rely only on orthographic processing with little meaning activation, which would be very different from processing those words in an actual text-reading situation where access to meanings is essential. In contrast, for the two morphological processing tasks, though also decontextualized, rapid semantic activation or attention to stem and affix meanings (e.g., inform and -ation for the stimulus word information) seemed unavoidable. Consequently, the required meaning activation process that seemed to favor the morphological processing tasks might have resulted in the relatively larger effect of sublexical processing efficiency than lexical processing in this study. Such an account also seems to be in line with that of Fukkink et al.’s (2005) result that was mentioned above, in that the improvement in the speed for recognizing decontextualized words as a result of the word recognition training might only represent enhanced orthographic (and phonological) processing and not capture the full lexical access that is required of reading comprehension. Limitations and Future Research We focused only on four major types of lexical competence to explore the lexical basis of reading comprehension. Although we considered both lexical and sublexical levels and both knowledge and processing efficiency dimensions and although these predictors explained over 40% of the variance in L2 Language Learning 00:0, xxxx 2021, pp. 1–40 28 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension reading comprehension, efficient reading comprehension does not depend solely on these dimensions. There are arguably other lexical knowledge and skills that underpin (L2) reading comprehension. In the L2 literature, there is, for example, an interest in the role of word or semantic association knowledge, which has often been studied as a type of vocabulary depth knowledge (Qian, 1999; Zhang, 2012; see Schmitt, 2014, and Zhang & Koda, 2017, for reviews). Cremer and Schoonen (2013) also distinguished between the availability and accessibility of semantic association knowledge. In both L1 and L2 reading literature, there has recently also been some attention given to knowledge of connectives (e.g., Crosson & Lesaux, 2013; Fraser, Pasquarella, Geva, Gottardo, & Biemiller, 2021) and knowledge of formulaic language or multiword lexical units (e.g., Kremmel, Brunfaut, & Alderson, 2017; Martinez & Murphy, 2011; Öksüz, Brezina, & Rebuschat, 2021). Collectively, these studies and ours contribute to a more comprehensive understanding about the lexical basis of (L2) reading comprehension. Nonetheless, it would seem too ambitious to accommodate all these dimensions in a single study. The relative contributions of different dimensions of lexical competence to reading comprehension may depend on learners’ L2 proficiency. Some researchers have split their sample of readers into proficient and less proficient subgroups and have aimed to examine whether the subgroups differed for any relational patterns (e.g., Cremer & Schoonen, 2013; Shiotsu & Weir, 2007). The present study did not perform this kind of ad hoc grouping because the participants were from the same learner population. Future research, however, might consider recruiting and comparing learners with distinct levels of language proficiency or at distinct developmental stages, or measuring and reporting proficiency rigorously (Park, Solon, Dehghan-Chaleshtori, & Ghanbar, 2021) so as to be able to use proficiency as a continuous variable in models, for example. Another limitation pertains to the relatively low reliability that we found for the portion of the Gates-MacGinitie Reading Tests (see Table 1) that we used with our participants (see the meta-analysis of reliability coefficients in L2 research by Plonsky & Derrick, 2016, where the median reliability for reading was .86). Although the Gates-MacGinitie Reading Tests are primarily intended for native English-speaking readers, they have been popular in research with L2 populations as well, such as Li and Kirby (2015) with adolescent Chinese-speaking EFL readers or Akamatsu (2003) with adult ESL readers. We thus have not speculated that the relatively low reliability could be attributed to the inappropriateness of the test for adult EFL learners. One reason for low reliability might be the relatively low number of passages and 29 Language Learning 00:0, xxxx 2021, pp. 1–40 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension questions included in this study. We sampled and administered only four passages because the tasks for the purpose of this study already required a commitment of over two hours, not to mention several other L2 reading tasks that we administered for other studies. Another reason might be the extreme difficulty of a couple of the questions. For example, only 13% of the participants correctly answered one question. To keep the test intact, we choose not to remove those items so as to augment the reliability for the present sample. Future research should consider adopting more passages with a larger number of questions. Finally, to accommodate the trade-off between speed and accuracy for computer-based tasks, we followed Townsend and Ashby’s (1983) method to calculate IESs and used those adjusted RTs as predictors of reading comprehension. Nevertheless, we recognize that IES is just one way to adjust RT; despite its wide use in the literature, including the language learning literature (e.g., Ke & Koda, 2017), efforts have been taken to explore other methods for better accommodation of the speed-accuracy interaction and debates on this topic are not uncommon (e.g., Davidson & Martin, 2013; Liesefeld & Janczyk, 2019; Vandierendonck, 2017). In addition, there have been explorations of and debates on the reliability of measuring and representing individual differences in nonnative lexical processing. Schmalz (2020), for example, explored a number of psycholinguistic marker effects. Word frequency, an important psycholinguistic marker, notably could modulate RT performance and its representation for individual differences in lexical processing (see also Brysbaert, Mandera, & Keuleers, 2017). In the present study, the lexical decision task’s use of (correct responses to) words from the 1,000 most frequent words (unlike in the Vocabulary Levels Test or the lexical knowledge task, where a wider range of less frequent words was used) could have reduced the test’s ability to represent the participants’ individual differences in lexical processing efficiency. This is because the words may have been very efficiently recognized, thus creating a kind of ceiling effect and consequently limiting the test’s predictive power for reading comprehension. It is beyond the purpose of the present study to directly investigate these methodological issues, and as a result, we cannot rule out the possibility that the approach to handling the interaction between speed and accuracy and the reliability of the (adjusted) RT-based (sub)lexical processing measures might have influenced the research findings (e.g., the lack of a significant, unique effect for the lexical decision task). This can be a direction for future research. Language Learning 00:0, xxxx 2021, pp. 1–40 30 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension Conclusions In light of the lexical quality hypothesis, this study explored the lexical basis of L2 reading comprehension in a group of adult Arabic-speaking EFL readers by studying the collective and relative contributions of four distinct lexical competences: lexical versus sublexical and knowledge versus processing efficiency. Hierarchical regression analyses revealed that the four lexical predictors collectively explained over 40% of the variance in the participants’ reading comprehension. Compared to the processing efficiency predictors, the knowledge predictors had the predominant association with reading comprehension. When the knowledge predictors were not considered, sublexical (i.e., derivational morphological) processing efficiency, as opposed to lexical processing efficiency, significantly predicted reading comprehension, over and above working memory. Overall, among the four lexical competences, lexical knowledge was the strongest predictor, followed by sublexical knowledge and then the processing efficiency predictors. This study confirmed strong lexical involvement in L2 reading comprehension. It underscored the critical importance of knowledge of word meanings that had been found in many previous studies. Yet, it also showed that knowledge of sublexical morphological features can be important. Although the lexical processing efficiency measures did not significantly predict reading comprehension when lexical and sublexical knowledge were concurrently in the model, evidence emerged that the type of processing where meaning activation is mandated (e.g., judging whether word parts can combine) can also be important when accounting for reading comprehension. To our knowledge, the present study was the first of its kind that has concurrently considered both lexical and sublexical knowledge and processing efficiency to study reading comprehension in L2 learners. The findings enrich the current understanding about the fundamental role of lexical processes in L2 reading comprehension. They particularly shed light on how derivational morphological knowledge as well as processing skills may have a unique role to play for adult L2 learners of English. Final revised version accepted 9 September 2021 Open Research Badges This article has earned an Open Materials badge for making publicly available the components of the research methods needed to reproduce the reported procedure. All materials that the authors have used and have the right to share are 31 Language Learning 00:0, xxxx 2021, pp. 1–40 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension available at http://www.iris-database.org. All proprietary materials have been precisely identified in the manuscript. References Alshehri, M., & Zhang, D. (2021a). Digit span test. Materials from “The lexical basis of second language reading comprehension: From (sub)lexical knowledge to processing efficiency.” [Working memory test]. IRIS Database, University of York, UK. https://doi.org/10.48316/q7cd-mg79 Alshehri, M., & Zhang, D. (2021b). Lexical knowledge measure. Materials from “The lexical basis of second language reading comprehension: From (sub)lexical knowledge to processing efficiency” [Language test]. IRIS Database, University of York, UK. https://doi.org/10.48316/qk1w-7b42 Alshehri, M., & Zhang, D. (2021c). Lexical processing efficiency task. Materials from “The lexical basis of second language reading comprehension: From (sub)lexical knowledge to processing efficiency” [Language test]. IRIS Database, University of York, UK. https://doi.org/10.48316/xqtk-mh93 Alshehri, M., & Zhang, D. (2021d). Sub-lexical knowledge measure. Materials from “The lexical basis of second language reading comprehension: From (sub)lexical knowledge to processing efficiency” [Language test]. IRIS Database, University of York, UK. https://doi.org/10.48316/k4cp-jw68 Alshehri, M., & Zhang, D. (2021e). Sub-lexical processing efficiency task. Materials from “The lexical basis of second language reading comprehension: From (sub)lexical knowledge to processing efficiency” [Language test]. IRIS Database, University of York, UK. https://doi.org/10.48316/nnth-9h06 Allison, P. D. (1999). Multiple regression. Thousand Oaks, CA: Pine Forge Press. Anderson, R. C., & Freebody, P. (1981). Vocabulary knowledge. In J. T. Guthrie (Ed.), Comprehension and teaching: Research reviews (pp. 77–117). Newark, DE: International Reading Association. Akamatsu, N. (2003). The effects of first language orthographic features on second language in text. Language Learning, 53, 207–231. https://doi.org/10.1111/1467-9922.00216 Baddeley, A. (2007). Working memory, thought, and action. Oxford, UK: Oxford University Press. Brown, J. D. (2014). Classical theory reliability. In A. J. Kunnan (Ed.), The companion to language assessment (pp. 1165–1181). Oxford, UK: Wiley-Blackwell. https://doi.org/10.1002/9781118411360.wbcla054 Brysbaert, M., Mandera, P., & Keuleers, E. (2017). The word frequency effect in word processing: An updated review. Current Directions in Psychological Science, 27, 45–50. https://doi.org/10.1177/0963721417727521 Ciaccio, L. A., & Clahsen, H. (2020). Variability and consistency in first and second language processing: A masked morphological priming study on prefixation Language Learning 00:0, xxxx 2021, pp. 1–40 32 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension and suffixation. Language Learning, 70, 103–136. https://doi.org/10.1111/lang.12370 Choi, Y., & Zhang, D. (2021). The relative role of vocabulary and grammatical knowledge in L2 reading comprehension: A systematic review of literature. International Review of Applied Linguistics in Language Teaching, 59, 1–30. https://doi.org/10.1515/iral-2017-0033 Clahsen, H., Felser, C., Neubauer, K., Sato, M., & Silva, R. (2010). Morphological structure in native and nonnative language processing. Language Learning, 60, 21–43. https://doi.org/10.1111/j.1467-9922.2009.00550.x Cremer, M., & Schoonen, R. (2013). The role of accessibility of semantic word knowledge in monolingual and bilingual fifth-grade reading. Applied Psycholinguistics, 34, 1195–1217. https://doi.org/10.1017/S0142716412000203 Crosson, A. C., & Lesaux, N. K. (2013). Does knowledge of connectives play a unique role in the reading comprehension of English learners and English-only students? Journal of Research in Reading, 36, 241–260. https://doi.org/10.1111/j.1467-9817.2011.01501.x Curran, P. J., West, S. G., & Finch, J. F. (1996). The robustness of test statistics to nonnormality and specification error in confirmatory factor analysis. Psychological Methods, 1, 16–29. https://doi.org/10.1037/1082-989X.1.1.16 Davidson, D. J., & Martin, A. E. (2013). Modeling accuracy as a function of response time with the generalized linear mixed effects model. Acta Psychologia, 144, 83–96. https://doi.org/10.1016/j.actpsy.2013.04.016 Davies, M. (2008). The corpus of contemporary American English (COCA). Retrieved from https://www.english-corpora.org/coca/ Farran, L. K., Bingham, G. E., & Matthews, M. W. (2012). The relationship between language and reading in bilingual English-Arabic children. Reading and Writing, 25, 2153–2181. https://doi.org/10.1007/s11145-011-9352-5 Fraser, C., Pasquarella, A., Geva, E., Gottardo, A., & Biemiller, A. (2021). English language learners’ comprehension of logical relationships in expository texts: Evidence for the confluence of general vocabulary and text-connecting functions. Language Learning, 71, 872–906. https://doi.org/10.1111/lang.12453 Fukkink, R. G., Hulstijn, J., & Simis, A. (2005). Does training in second-language word recognition skills affect reading comprehension? An experimental study. Modern Language Journal, 89, 54–75. https://doi.org/10.1111/j.0026-7902.2005.00265.x Gelman, A., & Hill, J. (2007). Data analysis using regression and multilevel/hierarchical models. Cambridge, UK: Cambridge University Press. Grabe, W. (2009). Reading in a second language: Moving from theory to practice. New York, NY: Cambridge University Press. Harrington, M. (2018). Lexical facility: Size, recognition speed and consistency as dimensions of second language vocabulary knowledge. London, UK: Palgrave MacMillan. 33 Language Learning 00:0, xxxx 2021, pp. 1–40 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension Harrington, M., & Sawyer, M. (1992). L2 working memory capacity and L2 reading skill. Studies in Second Language Acquisition, 14, 25–38. https://doi.org/10.1017/S0272263100010457 Hayes, A. H., & Coutts, J. J. (2020). Use omega rather than Cronbach’s alpha for estimating reliability. But…. Communication Methods and Measures, 14, 1–24. https://doi.org/10.1080/19312458.2020.1718629 Hu, H. M., & Nation, P. (2000). What vocabulary size is needed to read unsimplified texts. Reading in a Foreign Language, 8, 689–696. Jeon, E. H. (2015). Multiple regression. In L. Plonsky (Ed.), Advancing quantitative methods in second language research (pp. 131–158). New York, NY: Routledge. Jeon, E. H., & Yamashita, J. (2014). L2 reading comprehension and its correlates: A meta-analysis. Language Learning, 64, 160–212. https://doi.org/10.1111/lang.12034 Jiang, N. (2012). Conducting reaction time research in second language studies. New York, NY: Routledge. Just, M. A., & Carpenter, P. A. (1992). A capacity theory of comprehension: Individual differences in working memory. Psychological Review, 99, 122–149. https://doi.org/10.1037/0033-295x.99.1.122 Ke, S., & Koda, K. (2017). Contribution of morphological awareness to adult L2 Chinese word meaning inferencing. The Modern Language Journal, 101, 742–755. https://doi.org/10.1111/modl.12428 Ke, S., Miller, R. T., Zhang, D., & Koda, K. (2021). Crosslinguistic sharing of morphological awareness in biliteracy development: A systematic review and meta-analysis of correlation coefficients. Language Learning, 71, 8–54. https://doi.org/10.1111/lang.12429 Kieffer, M. J., & Lesaux, N. K. (2008). The role of derivational morphology in the reading comprehension of Spanish-speaking English language learners. Reading and Writing, 21, 783–804. https://doi.org/10.1007/s11145-007-9092-8 Kintsch, W. (1988). The construction-integration model of text comprehension. Psychological Review, 95, 163–182. https://doi.org/10.1037/0033-295X.95.2.163 Kirby, J. R., & Bowers, P. N. (2017). Morphological instruction and literacy: Binding phonological, orthographic, and semantic features of words. In K. Cain, D. L. Compton, & R. K. Parrila (Eds.), Theories of reading development (pp. 437–462). Amsterdam, The Netherlands: John Benjamins. https://doi.org/10.1075/swll.15.24kir Koda, K. (2000). Cross-linguistic variations in L2 morphological awareness. Applied Psycholinguistics, 21, 297–320. https://doi.org/10.1017/S0142716400003015 Koda, K. (2005). Insights into second language reading: A cross-linguistic approach. New York, NY: Cambridge University Press. Kremmel, B., Brunfaut, T., & Alderson, J. C. (2017). Exploring the role of phraseological knowledge in foreign language reading. Applied Linguistics, 38, 848–870. https://doi.org/10.1093/applin/amv070 Language Learning 00:0, xxxx 2021, pp. 1–40 34 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension Liesefeld, H. R., & Janczyk, M. (2019). Combining speed and accuracy to control for speed-accuracy trade-offs(?). Behavior Research Methods, 51, 40–60. https://doi.org/10.3758/s13428-018-1076-x Li, M., & Kirby, J. R. (2015). The effects of vocabulary breadth and depth on English reading. Applied Linguistics, 36, 611–634. https://doi.org/10.1093/applin/amu007 MacGinitie, W. H., MacGinitie, R. K., Maria, K., & Dreyer, L. G. (2000). Gates-MacGinitie Reading Tests–Directions for Administration (levels 7/9 & 10/12) (4th Ed.). Rolling Meadows, IL: Riverside. Martinez, R., & Murphy, V. A. (2011). Effect of frequency and idiomaticity on second language reading comprehension. TESOL Quarterly, 45, 267–290. https://doi.org/10.5054/tq.2011.247708 McNeish, D. (2018). Thanks coefficient alpha, we’ll take it from here. Psychological Methods, 23, 412–433. https://doi.org/10.1037/met0000144 Muthén, L. K., & Muthén, B. O. (1998–2017). Mplus user’s guide (8th ed.). Los Angeles, CA: Author. Nagy, W. (2007). Metalinguistic awareness and the vocabulary-comprehension connection. In R. K. Wagner, A. E. Muse, & K. R. Tannenbaum (Eds.), Vocabulary acquisition: Implications for reading comprehension (pp. 52–77). New York, NY: Guilford Press. Nagy, W. E., & Anderson, R. C. (1984). How many words are there in printed school English? Reading Research Quarterly, 19, 304–330. https://doi.org/10.2307/747823 Nagy, W. E., & Scott, J. A. (2000). Vocabulary processes. In M. L. Kamil, P. B. Mosenthal, P. D. Pearson, & R. Barr (Eds.), Handbook of reading research (Vol. 3, pp. 269–284). Mahwah, NJ: Lawrence Erlbaum. Nation, K., & Snowling, M. J. (2004). Beyond phonological skills: Broader language skills contribute to the development of reading. Journal of Research in Reading, 27, 342–356. https://doi.org/10.1111/j.1467-9817.2004.00238.x Öksüz, D., Brezina, V., & Rebuschat, P. (2021). Collocational processing in L1 and L2: The effects of word frequency, collocational frequency, and association. Language Learning, 71, 55–98. https://doi.org/10.1111/lang.12427 Park, H. I., Solon, M., Dehghan-Chaleshtori, M., & Ghanbar, H. (2021). Proficiency reporting practices in research on second language acquisition: Have we made any progress? Language Learning, Retrieved from https://onlinelibrary.wiley.com/ action/showCitFormats?doi=10.1111%2Flang.12475 Pasquarella, A., Gottardo, A., & Grant, A. (2012). Comparing factors related to reading comprehension in adolescents who speak English as a first (L1) or second (L2) language. Scientific Studies of Reading, 16, 474–503. https://doi.org/10.1080/10888438.2011.593066 Peirce, J. W., Gray, J. R., Simpson, S., MacAskill, M. R., Höchenberger, R., Sogo, H., … Lindeløv, J. (2019). PsychoPy2: Experiments in behavior made easy. Behavior Research Methods, 51, 195–203. https://doi.org/10.3758/s13428-018-01193-y 35 Language Learning 00:0, xxxx 2021, pp. 1–40 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension Pellicer-Sánchez, A., & Schmitt, N. (2012). Scoring yes-no vocabulary tests: Reaction time vs. nonword approaches. Language Testing, 29, 489–509. https://doi.org/10.1177/0265532212438053 Perfetti, C. A. (1985). Reading ability. Oxford, UK: Oxford University Press. Perfetti, C. A. (1999). Comprehending written language: A blueprint for the reader. In C. Brown & P. Hagoort (Eds.), Neurocognition of language (pp. 167–208). Oxford, UK: Oxford University Press. Perfetti, C. A. (2007). Reading ability: Lexical quality to comprehension. Scientific Studies of Reading, 11, 357–383. https://doi.org/10.1080/10888430701530730 Perfetti, C. A., & Hart, L. (2001). The lexical basis of comprehension skill. In D. S. Gorfein (Ed.), On the consequences of meaning selection: Perspectives on resolving lexical ambiguity (pp. 67–86). Washington, DC: American Psychological Association. https://doi.org/10.1037/10459-004 Perfetti, C. A., & Stafura, J. (2014). Word knowledge in a theory of reading comprehension. Scientific Studies of Reading, 18, 22–37. https://doi.org/10.1080/10888438.2013.827687 Plonsky, L., & Derrick, D. (2016). A meta-analysis of reliability coefficients in second language research. The Modern Language Journal, 100, 538–553. https://doi.org/10.1111/modl.12335 Proctor, C. P., Carlo, M., August, D., & Snow, C. (2005). Native Spanish-speaking children reading in English: Toward a model of comprehension. Journal of Educational Psychology, 97, 246–256. https://doi.org/10.1037/0022-0663.97.2.246 Richardson, J. T. E. (2007). Measures of short-term memory: A historical review. Cortex, 43, 635–650. https://doi.org/10.1016/S0010-9452(08)70493-3 Qian, D. (1999). Assessing the roles of depth and breadth of vocabulary knowledge in reading comprehension. Canadian Modern Language Review, 56, 282–308. https://doi.org/10.3138/cmlr.56.2.282 RAND Reading Study Group. (2002). Reading for understanding: Toward an R&D program in reading comprehension. Santa Monica, CA: RAND. Ratcliff, R. (1993). Methods for dealing with reaction time outliers. Psychological Bulletin, 114, 510–532. https://doi.org/10.1037/0033-2909.114.3.510 Sasao, Y., & Webb, S. A. (2017). The Words Parts Levels Test. Language Teaching Research, 21, 12–30. https://doi.org/10.1177/1362168815586083 Schmalz, X. (2020). How (not) to measure individual differences: The case of statistical learning and lexical processing efficiency. Presentation given at the Clinic and Polyclinic for Child and Adolescent Psychiatry, Psychosomatics and Psychotherapy, University of Munich. Retrieved from https://osf.io/ctyqn/?show=revision Schmitt, N. (2014). Size and depth of vocabulary knowledge: What the research shows. Language Learning, 64, 913–951. https://doi.org/10.1111/lang.12077 Language Learning 00:0, xxxx 2021, pp. 1–40 36 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension Schmitt, N., Jiang, X., & Grabe, W. (2011). The percentage of words known in a text and reading comprehension. The Modern Language Journal, 95, 26–43. https://doi.org/10.1111/j.1540-4781.2011.01146.x Schmitt, N., Schmitt, D., & Clapham, C. (2001). Developing and exploring the behaviour of two new versions of the Vocabulary Levels Test. Language Testing, 18, 55–88. https://doi.org/10.1177/026553220101800103 Shiotsu, T., & Weir, C. J. (2007). The relative significance of syntactic knowledge and vocabulary breadth in the prediction of reading comprehension test performance. Language Testing, 24, 99–128. https://doi.org/10.1177/0265532207071513 Tokowicz, N. (2014). Lexical processing and second language acquisition. New York, NY: Routledge. Townsend, J. T., & Ashby, F. G. (1983). Stochastic modeling of elementary psychological processes. Cambridge, UK: Cambridge University Press. Vandierendonck, A. (2017). A comparison of methods to combine speed and accuracy measures of performance: A rejoinder on the binning procedure. Behavior Research Methods, 49, 653–673. https://doi.org/10.3758/s13428-016-0721-5 van Gelderen, A., Schoonen, R., De Glopper, K., Hulstijn, J., Simis, A., Snellings, P., & Stevenson, M. (2004). Linguistic knowledge, processing speed, and metacognitive knowledge in first-and second-language reading comprehension: A componential analysis. Journal of Educational Psychology, 96, 19–30. https://doi.org/10.1037/0022-0663.96.1.19 Webb, S., Sasao, Y., & Ballance, O. (2017). The updated Vocabulary Levels Test: Developing and validating two new forms of the VLT. ITL - International Journal of Applied Linguistics, 168, 34–70. https://doi.org/10.1075/itl.168.1.02web Yamashita, J. (2013). Word recognition subcomponents and passage level reading in a foreign language, Reading in a Foreign Language, 25, 52–71. Zhang, D. (2012). Vocabulary and grammar knowledge in second language reading comprehension: A structural equation modeling study. The Modern Language Journal, 96(4), 558–575. https://doi.org/10.1111/j.1540-4781.2012.01398.x Zhang, D. (2017). Derivational morphology in reading comprehension of Chinese-speaking learners of English. Applied Linguistics, 38, 871–895. https://doi.org/10.1093/applin/amv072 Zhang, D., & Ke, S. (2020). The simple view of reading made complex by morphological decoding fluency in bilingual fourth-grade readers of English. Reading Research Quarterly, 55, 311–329. https://doi.org/10.1002/rrq.287 Zhang, D., & Koda, K. (2012). Contribution of morphological awareness and lexical inferencing ability to L2 vocabulary knowledge and reading comprehension among advanced EFL learners: Testing direct and indirect effects. Reading and Writing, 25, 1195–1216. https://doi.org/10.1007/s11145-011-9313-z Zhang, D., & Koda, K. (2013). Morphological awareness and reading comprehension in a foreign language: A study of young Chinese EFL learners. System, 41, 901–913. http://doi.org/10.1016/j.system.2013.09.009 37 Language Learning 00:0, xxxx 2021, pp. 1–40 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension Zhang, D., & Koda, K. (2017). Assessing L2 learners’ vocabulary depth with word associates format tests: Issues, findings, and suggestions. Asian-Pacific Journal of Second and Foreign Language Education, 2, 1–30. https://doi.org/10.1186/s40862-017-0024-0 Supporting Information Additional Supporting Information may be found in the online version of this article at the publisher’s website: Appendix S1. Elaborated Descriptive Statistics and Reliability Estimates for Competency Measures. Appendix S2. Parameter Estimates of Regression Models for Lexical Versus Sublexical Predictors of Reading Comprehension. Appendix S3. Parameter Estimates of Regression Models for Knowledge Versus Processing Predictors of Reading Comprehension. Appendix: Accessible Summary (also publicly available at https://oasis-database.org) The importance of lexical competence for reading comprehension in a second language What this research was about and why it is important Words are the building blocks of texts. For good reading comprehension, second language (L2) learners need to know the meaning of a large number of words. Yet, this knowledge, often known as vocabulary breadth (or size), is insufficient. Written English is replete with affixed words where suffixes such as -ness and -ive are added to a word to modify its meaning (and often its grammatical status as well). This means that learners also need to have good affix or word part knowledge for recognizing those words during text reading. In addition, readers need to be able to access these types of knowledge efficiently (at speed). This study measured four distinct aspects of lexical competence in university Arabic-speaking learners of English as a foreign language in Saudi Arabia and tested the relative importance of these aspects of lexical competence for L2 reading comprehension. Vocabulary breadth had the strongest association with reading comprehension, followed by affix knowledge and then the speed-related lexical skills. Understanding the relative importance of different aspects of lexical competence is essential to appropriate instructional emphasis in L2 reading classrooms. Language Learning 00:0, xxxx 2021, pp. 1–40 38 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension What the researchers did r Participants were tested (in a group setting) on their vocabulary breadth, affix knowledge, and reading comprehension. r Participants were also tested (on an individual basis) on their speed of recognizing whole words and on their speed of segmenting words into meaningful parts or combining parts to make a word. r The reading comprehension test involved answering a total of 21 multiple choice questions about four short passages. r Statistical relations were calculated for the associations of the four distinct aspects of lexical competence with reading comprehension, and the strength of the associations was compared. What the researchers found r Vocabulary breadth was the most important of the four aspects of lexical competence in explaining the reading comprehension scores. r Affix knowledge was additionally important for accounting for reading comprehension scores. r Knowledge of words (vocabulary breath) and of affixes played a stronger role than did the two speed measures. r The speed of segmenting words and speed of combining word parts was also found to explain reading comprehension scores, though only to a relatively small extent. r The speed of word recognition was not found to play a notable role in comprehension scores. Things to consider r Diverse aspects of lexical competence can have distinct roles during text reading. r Knowing the meaning of a large number of words is important for reading comprehension, but knowledge of word parts such as affixes is also important. r It may help for L2 vocabulary instruction to go beyond simple meaning definitions, though the current study did not test this. r Further research should check whether the speed measures used in the current study reliably captured individual differences in learners’ access to words. Materials, data, open access article: Materials are publicly available at https: //www.iris-database.org. 39 Language Learning 00:0, xxxx 2021, pp. 1–40 Alshehri and Zhang Lexical Basis of L2 Reading Comprehension How to cite this summary: Alshehri, M. G., & Zhang, D. (2022). The importance of lexical competence for reading comprehension in a second language. OASIS Summary of Alshehri and Zhang (2022) in Language Learning. https://oasis-database.org This summary has a CC BY-NC-SA license. Language Learning 00:0, xxxx 2021, pp. 1–40 40