Author:r4 3ds
Abstract—To remedy the paucity of studies on the relationship between vocabulary knowledge and speaking proficiency, we examine the degree to which second language (L2) speaking proficiency can be predicted by the size, depth, and speed of L2 vocabulary among novice to intermediate Japanese learners of English. Studies 1 and 2 administered vocabulary tests and a speaking test to 224 and 87 L2 learners, respectively. Analyses using structural equation modeling demonstrated that a substantial proportion of variance in speaking proficiency can be explained by vocabulary knowledge, size, depth, and speed. These results suggest the centrality of vocabulary knowledge to speaking proficiency.(by r4 3ds)
I. INTRODUCTION
Vocabulary has long been recognized as a vital component and a good indicator of second language (L2) performance and proficiency (e.g., Schmitt, 2010; Stæhr, 2009). However, compared to the numerous studies on associations between L2 vocabulary and reading (e.g., Qian, 2002; Van Gelderen et al., 2004), little research has been conducted into the relationships between L2 vocabulary and other L2 skills (Stæhr, 2009). Examples include Stæhr (2009) for listening, Schoonen et al. (2003) for writing, and De Jong, Steinel, Florijn, Schoonen, and Hulstijn (2012) for speaking. The current article focuses on the relationship between L2 vocabulary knowledge and L2 speaking proficiency among novice- to intermediate-level Japanese learners of English, by conducting two studies that use structural equation modeling (SEM).
A. Vocabulary Knowledge and Its Predictive Power
While researchers generally agree with regard to the multicomponential nature of vocabulary knowledge, various proposals have been put forward regarding what exactly constitutes vocabulary knowledge (e.g., Meara, 2005; Schmitt, 2010). One classification frequently employed involves the size and depth of vocabulary (e.g., Qian, 2002). Size, or breadth, expresses a quantitative dimension involving knowledge of a word form and a primary meaning, also described as the form-meaning link. Depth represents a qualitative dimension, defined as ―how well a learner knows individual words or how well words are organized in the learner’s mental lexicon (Stæhr, 2009, p. 579), and includes knowledge of partial to precise meaning, word frequency, affix knowledge, syntactic characteristics, and lexical network.
In addition to size and depth, another lexical aspect that has recently attracted attention and been incorporated into vocabulary frameworks is speed of processing, or how fast learners can recognize and retrieve knowledge stored in the mental lexicon (e.g., Meara, 2005). Processing speed (often referred to as automaticity, efficiency, or fluency) of lexical access and retrieval is considered to play a crucial role in the use of vocabulary in real-life situations, as well as in L2 proficiency (e.g., Van Moere, 2012). This may be true especially of listening and speaking, which require on-line processing (Schmitt, 2010).
Of these multidimensional lexical aspects, size has been considered primary, because of the importance of the form-meaning link for vocabulary use (e.g., Laufer, Elder, Hill, & Congdon, 2004; Schmitt, 2010). A number of empirical studies have been conducted to examine the relative importance of size versus depth and speed in terms of predictive powers of L2 skills. Qian and Schedl (2004) investigated vocabulary knowledge and reading comprehension among 207 L2 learners of English at intermediate and advanced levels, and reported that 57% of variance of L2 reading scores was explained by size, with an additional 4% of variance explained by depth. A similarly large variance (54%) predicted solely by size was indicated by Qian (2002), with an additional 13% explained by depth (n = 217). Finally, Stæhr (2009) provided further support for these results, showing that 49% of L2 listening variance was accounted for by size, but just 2% by depth (n = 115). In sum, previous studies suggest that size can predict much of reading and listening variance, while depth contributes relatively little.
It should be noted that the proportion of variance explained by variables changes, depending on the order in which independent variables are entered into the regression equation. The results from the studies described above were derived when size was entered first, followed by depth. The effect of this is that depth is able to predict only the remaining variance, that is, whatever was not predicted by size. Since size was highly correlated with depth, sharing a large variance with it (49% in Qian, 2002, r = .70; 71% in Qian & Schedl, 2004, r = .84; 64% in Stæhr, 2009, r = .80), the variance that could have been predicted by depth was already predicted by size. As a result, the predictive power of depth appears much lower than size. Therefore, the small proportion of variance explained by depth does not indicate that depth is less important for predicting reading and writing skills. To the contrary, when depth was the first variable entered into the regression equation, it could predict a much higher proportion of reading and listening variance, while size added only a small percentage (59% depth and 8% size in Qian, 2002; 55% and 6% in Qian & Schedl, 2004; 42% and 9% in Stæhr, 2009). These results suggest that size and depth in fact predict reading and listening proficiency in similar ways. Unlike reading and listening, however, the relative contributions of size and depth to speaking and writing skills remain unclear. In the present article, this constitutes the basis for conducting Study 1.
An additional concern is that the three abovementioned studies—Qian (2002), Qian and Schedl (2004), and Stæhr (2009)—all used the Word Associates Test (WAT) format (Read, 1993), which is designed to assess synonyms and collocations. According to Schmitt (2012), relationships between size and depth can vary depending on what specific areas of depth researchers target. The similar relationships of size and depth to reading and listening skills that these studies suggest may be due to their use of the same test format, and perhaps also because the synonyms that the WAT assessed overlapped with the size aspect. Therefore, studies employing formats different from the WAT are desirable.
Regarding the predictive power of lexical processing speed in relation to size, Van Gelderen et al. (2004) showed that size and speed were both moderately correlated with L2 reading comprehension (r = .63 and –.47, respectively), and that size was more effective (40%) than speed (22%) in predicting L2 reading, when each was separately entered into the regression equation. This pattern has also been observed in studies of L2 writing (Schoonen et al., 2003) and L2 speaking (De Jong et al., 2012, in press). Previous studies suggest that, unlike the case of size and depth, where their degree of predictive power is roughly the same, the predictive power of speed, albeit still substantial, is smaller. One reason for this may be the existence of a threshold level of speed: Speed may be strongly related to reading, writing, and speaking proficiency until learners reach a certain threshold level of sufficient speed, after which point further increase in speed does not entail greater speaking proficiency.
To conclude, size seems to hold considerable power in predicting L2 proficiency, when it is the first variable entered into the regression equation, while depth and speed contribute limited predictive powers for the remainder of the proficiency. However, when depth or speed is entered into the regression first, depth tends to exhibit a predictive power similar to size, whereas speed may have a predictive power less than size. This indicates the complicated nature of the contribution that these three lexical aspects make to language proficiency; thus far, however, only a limited number of studies have investigated this issue. To our knowledge, only Uenishi (2006) has tested the three aspects separately in relation to speaking, and even Uenishi’s study is limited to novice speakers, and furthermore does not report test reliability or details about the tests and analysis. Thus, the report of Study 2 presented in this paper, inspecting the relationships between the factors of size, depth, speed, and L2 speaking proficiency, fills a significant gap in current research.
B. Relationships between L2 Vocabulary Knowledge and L2 Speaking
The well-known models of the speaking process proposed by Levelt (1989) and Kormos (2006) describe three main stages of speech production: conceptualization, formulation, and articulation. During the first stage, speakers form preverbal messages in the conceptualizer. In the formulator, they search for and retrieve necessary vocabulary from the mental lexicon, which contains information related to vocabulary and syntactic structures, in order to produce utterances with syntactic and phonological information. In the final stage, they utter the speech that they have formulated. Levelt stated that L1 speakers conduct these processes in parallel and automatically, without using substantial cognitive resources. However, L2 speakers experience much greater difficulty in executing such processes, a fact that prompted Kormos (2006) to propose an L2 speaking model.
According to both models, vocabulary holds a central position in formulating an utterance with the appropriate meanings, although other types of knowledge, including syntactic, morphological, and phonological knowledge, as well as nonlinguistic world knowledge and communication strategies, are also indispensible. The models indicate further the necessity of size, depth, and processing speed of vocabulary knowledge in speaking, because speakers use both form-meaning links (i.e., size) and the syntactic and morphological information associated with each word in the mental lexicon (depth), and because automatic, or at least relatively fast, lexical retrieval (speed) is required for smooth and effective communication.
In addition to the theoretical importance of vocabulary for speaking, empirical studies have been conducted into the relationship between vocabulary knowledge and speaking. Table 1 summarizes nine previous studies that have quantitatively investigated the relationships between L2 vocabulary knowledge and speaking. We did not include studies into the relationships between vocabulary knowledge and lexical complexity, because of the difficulty in identifying measures that can be interpreted with high validity when analyzing short texts (see Koizumi & In’nami, 2012).
For instance, De Jong et al. (2012) investigated to what degree ―L2 knowledge skills‖ and ―L2 processing skills‖ explain L2 speaking proficiency (specifically functional adequacy), and whether the contributions of linguistics skills are different between more and less successful L2 learners. They administered eight speaking tasks and nine tests of linguistic skills to 181 adult learners of Dutch at intermediate and advanced levels, including a test of vocabulary knowledge (combining size and depth) and another of speed of lexical retrieval. They assessed size by requiring participants to supply L2 single-word forms appropriate to the sentence context, with one letter provided as a hint (90 items), and depth (specifically collocation) through a format that elicited L2 ―prepositional phrases and verb-noun collocations‖ appropriate to the sentence (p. 17; 26 items). However, although they combined two formats to produce the total vocabulary knowledge scores, the depth items accounted for only 22% (26/116). Thus, we consider that their vocabulary knowledge test assessed mostly the aspect of size. In addition, a test of lexical retrieval speed measured the time it took participants to produce L2 forms corresponding to pictures provided. SEM analysis showed that vocabulary knowledge (size and depth combined) and intonation rating predicted 75% of speaking proficiency, and that speed contributed little to the prediction.
Reviewing these studies, we found varied results that may be explained by three main factors. First, the nine studies in Table 1 differed in the tasks/tests they administered and the aspects of vocabulary knowledge and speaking that they targeted. In terms of vocabulary aspects, four studies assessed size only (Funato & Ito, 2008; Hilton, 2008; Koizumi, 2005; Milton et al., 2010), with two studies measuring depth or speed (Ishizuka, 2000; Segalowitz & Freed, 2004) and three studies integrating size, depth, and processing speed (De Jong et al., 2012, in press; Uenishi, 2006). Regarding speaking aspects, six studies assessed overall speaking proficiency (De Jong et al., 2012; Funato & Ito, 2008; Ishizuka, 2000; Koizumi, 2005; Milton et al., 2010; Uenishi, 2006), while three assessed fluency (De Jong et al., in press; Hilton, 2008; Segalowitz & Freed, 2004). Strong correlations were found between size and overall speaking proficiency in three studies (e.g., r = .79 in De Jong, 2012) but not in two other studies (e.g., r = .27 in Funato & Ito, 2008). Additionally, weak or moderate correlations were found in some combinations: for example, between size and oral fluency in most studies (e.g., r = .67 in Hilton, 2008).
Second, some studies (e.g., Funato & Ito, 2008) failed to report their test and/or rater reliability, and a large measurement error may have led to underestimation of the strengths of relationships. SEM is a more appropriate tool than correlation or regression analysis for modeling relationships between variables with measurement error controlled for, in order to obtain rigorous and trustworthy results. Among the nine previous studies, only De Jong et al. (2012) used SEM, and they demonstrated strong relationships between size and overall speaking proficiency (r = .79).
Third, participants in the previous studies had different ranges of proficiency: novice only (e.g., Uenishi, 2006), novice to intermediate (Funato & Ito, 2008), novice to advanced (e.g., Hilton, 2008), and intermediate to advanced (e.g., De Jong et al., 2012). These differences in proficiency levels may have affected the results. For example, among five studies into associations between size and overall speaking proficiency, the two that used only intermediate and advanced learners showed high correlations (r = .79 in De Jong et al., 2012), whereas three that included learners at the novice level reported weak, moderate, or strong correlations (e.g., r = .53 in Uenishi, 2006). According to De Jong et al. (2012), a relatively wide range of proficiency levels should be incorporated when modeling proficiency, and studies dealing with only novice learners may not have sufficient variation, perhaps leading to weaker correlations. Further, the relative contribution model (e.g., Adams, 1980) posited that vocabulary plays a more important role in speaking proficiency among lower-level learners and that the impact of vocabulary becomes weaker as proficiency levels rise. This suggests that the contribution of vocabulary knowledge to speaking would be stronger among novice and intermediate than intermediate and advanced learners.
The mixed results generated by previous studies warrant further research into the relationships between L2 vocabulary knowledge and speaking proficiency, and particularly into the relative contribution of size, depth, and speed to L2 speaking proficiency. This article attempts to cover wider aspects of vocabulary knowledge (size, depth, and speed) and speaking (overall speaking, fluency, accuracy, and syntactic complexity [SC]), using SEM to account for measurement error, and including learners of a relatively wide range of proficiency levels. We employ novice- and intermediate-level learners, and compare our results with those of De Jong et al. (2012), who employed intermediate- and advanced-level learners, in order to examine the relative contribution model (e.g., Adams, 1980).
C. Present Study
Although vocabulary knowledge is only one among the many variables that affect oral production (De Jong et al., 2012, in press), the literature review above shows that it is theoretically indispensable. However, although empirical investigations have generally supported this, the limited number of studies conducted justifies further research. This study conceptualizes vocabulary knowledge according to three aspects: size, depth, and speed. We also followed Housen and Kuiken (2009) in regarding speaking proficiency as consisting primarily of fluency, accuracy, and SC.
We conducted two studies: Study 1 examines the relationships between size, depth, and speaking proficiency, while Study 2 adds speed to the design of Study 1. Our research question is to what extent L2 speaking proficiency is predicted by L2 vocabulary knowledge, in terms of overall knowledge, size, depth, and speed. Drawing on previous studies (e.g., De Jong et al., 2012; Milton et al., 2010; Qian & Schedl, 2004), we hypothesize that vocabulary knowledge contributes substantially to the prediction of speaking proficiency, and that size and depth predict speaking similarly to each other, and to a greater degree than speed.
Author:r4 3ds
