The Effects of CALL Versus Traditional L1 Glosses on L2 Reading Comprehension

ALAN TAYLOR
Brigham Young University-Idaho

Abstract:
Meta-analytic research of experiments conducted on the effects of native-language (L1) glosses on second-language (L2) reading comprehension have revealed a significant difference between groups of studies with traditional and computer-assisted L1 glosses. This means that learners provided with L1 glosses comprehend significantly more text�through the medium of a computer�than learners with traditional, paper-based L1 glossing aids. The average weighted effect size of studies with computer-assisted L1 glosses is large (g = 1.09) and differs significantly (p < .001) when compared to studies with traditional L1 glosses (g = .39). This article explores factors explaining how and when computer-assisted L1 glosses can be effective in L2 reading comprehension.

KEYWORDS

L2 Reading Comprehension, L1 Glossing, CALL Glossing, CALL Reading, Meta-analysis

INTRODUCTION

I just did the French, I mean not the French, the English definition, `cause that'd be the easiest for me to understand. I mean I don't speak French all that well, and I'd hadda look-up French, I'd a had more French to read, which I might not understand either, so I just used the English `cause it usually summarized it in just a coupla words. - Dave. (Hayden, 1997, pp. 146-147)

Computer-assisted language learning (CALL) studies have found that native-language (L1) glosses are used by readers because they can be an efficient tool for understanding a second-language (L2) text and they are easily tailored to fit learner lexical needs (Bell & Leblanc, 2000; Hayden, 1997). CALL L1 glossing is defined in the present study as an instant look-up capability—dictionary or linked—to which readers have immediate access to meaning in their native language. This meta-analytic study focuses on such glosses and their effectiveness with or without a computer.

309

It is important to consider studies conducted on glossing in light of research concerning the extent to which the L1 can be used in L2 reading and the fairly recent debate concerning general L1 use in L2 learning (e.g., Cook, 2001; Turnbull, 2001; Wells, 1998). This debate continues for several reasons. First, it is unclear how much the L1 should be used in the CALL classroom, given the dynamic social nature of CALL classrooms and that different social contexts may require differing amounts of L1 use (see Taylor, 2002b). Second, much experimental research, especially in the field of reading comprehension, suggests that the L1 can have a significant effect on L2 learning. Yet, this seems contrary to the goal of using the L2 in the classroom as much as possible.

This meta-analysis has found that CALL L1 glosses have a significantly stronger effect on L2 reading comprehension than traditional glosses. Its rationale is described in a series of questions: “Why glosses?”, “Why CALL glosses?”, and, finally, “Why L1 glosses?”

Why glosses?

Bernhardt and Kamil (1995) suggested that linguistic knowledge (linguistic threshold) influences L2 reading comprehension to a greater degree than does L1 reading skill competence (linguistic interdependence). Thus, general background knowledge may not have as essential a role in L2 reading comprehension as linguistic knowledge. Some researchers claim the lexicon plays a central role in second language acquisition (see Gettys, Imhof, & Kautz, 2001). From a bottom-up perspective of reading, such a point of view seems valid. From an interactive reading standpoint, top-down strategies such as activating background knowledge and noticing text structure both are important but are often dependent on lexical knowledge (Eskey, 1988). Glossing is essentially bottom-up lexical help providing direct support for L2 lexical acquisition and perhaps more generally for reading comprehension.

Why CALL glosses?

CALL glosses are easily accessible. Further, using software that provides immediate access to L1 glosses is not difficult. Software, such as Hachette dictionnaire Oxford iFinger (1997), can be easily used while consulting online texts. Instead of looking up a word, all learners need to do is either type in a word or simply click on it. There are several studies that have compared traditional, paper-based L1 glosses with CALL L1 glosses (e.g., Goyette, 1995; Stoehr, 1999). With regard to CALL effectiveness, Chapelle and Jamieson (1986) wrote, “Clearly, CALL effectiveness cannot be looked at as though CALL represented one form of instruction and all students were in need of that kind of instruction” (p. 42). The present study does not claim that CALL L1 glosses are the one factor that will make all the difference in L2 learning. However, we do claim that CALL L1 glosses can make an important difference in L2 reading comprehension.

Given the proliferation of computer technology, use, and skills, there is increasing access to authentic L2 texts. Thus, more research is needed on how computer-mediated reading comprehension of L2 texts can be enhanced by certain kinds of help such as L1 glosses.

Why L1 glosses?

CALL studies have shown that the L2 is seldom used when put next to the L1. Researchers of CALL studies (e.g., David & Lyman-Hager, 1997; Goyette, 1995; Hayden, 1997; Bell & Leblanc, 2000) all report quantitative results suggesting negligible usage of other types of glosses on L2 reading comprehension by L2 readers. First and second-year learners who have the option of choosing between L1 and L2 glosses generally choose the L1.

THE LITERATURE

Aweiss (1994) investigated whether there was a causal relationship between L1 CALL reading supports and L2 reading comprehension in English-speaking learners of Arabic. Aweiss' results revealed that those with CALL L1 glosses recalled significantly more pausal units that those without glosses. Stoehr (1999) found that participants with L1 glosses recalled a significantly higher amount of L2 text than those without glosses. Hayden (1997) observed that CALL L1 glossing, when presented to readers along with several other options of glossing had a significant effect on L2 reading comprehension. Hayden also found that the other glossing options were hardly consulted when placed next to L1 glosses. Moreover, Hayden's results suggested that CALL L1 glosses encouraged more look-ups of words.

The ease and rapidity of access of helps made possible by computer-mediated text may actually promote a greater amount of lookup behaviors, as is associated with bottom-up processing. It is far easier to simply “click” on a word or expression and instantly see a definition than to locate each expression in a printed dictionary. (p. 215)

Goyette (1995) observed that learners with L1 glosses did better than those without glosses, although the difference was not significant. Goyette also found that there was no significant difference in amount of reading time in comparing the online with the hard-copy dictionary treatments. Moreover, even though the reading times were comparable, the online dictionary was used significantly more than the hard-copy dictionary. Goyette (1995) concluded

The large differences in the number of words accessed by dictionary condition and by language indicate that, as expected, the speed and ease of access of the on-line dictionary do increase the frequency of dictionary consultation. On-line dictionary look-ups were found to be particularly prevalent in second-language reading. (p. 96)

Goyette suggested that online dictionaries encouraged readers to look up significantly more words, stating that “speed and ease of access of electronic dictionaries do encourage readers to consult dictionaries more often” (p. 110). It is interesting to note that, with the exception of Goyette's study, glossing had an enormous effect on L2 reading comprehension. Nonsignificant results can easily be explained by confounding variables and Type II error.

THE META-ANALYSIS

Method

Studies relevant and included in the present meta-analysis had to meet four criteria: (a) all studies up to the year 2002 were included, (b) the study had to be either an experimental or a quasi-experimental, (c) at least one of the dependent variables of the study had to be reading comprehension, and (d) the effect of access to glosses—whether traditional or CALL—versus no access to glosses had to be tested. Studies failing to meet any of these criteria were not included in the meta-analysis. Lomicka's (1998) classic study, for example, mixed L1 with L2 glosses and thus did not fit the criteria.

A variety of electronic sources were searched to find relevant studies, including theses and dissertations, for the meta-analysis. The most important of these were Dissertation Abstracts International (DAI), Languages and Literatures Behavior Abstracts (LLBA), The Educational Resources Information Center (ERIC), Psychology Information (Psych INFO), and the World Wide Web (WWW). A total of 18 study reports (outcomes) met the four criteria for inclusion in the present meta-analysis (see Table 1 below).

In order to provide a more comprehensive explanation of what the studies suggest, quantitative meta-analysis is an effective means by which statistical results can be combined and the overall effectiveness of treatments ascertained. Quantitative meta-analysis makes use of a statistic called the “effect size,” which is characterized as large (g = .80 or above), medium (g = .50-.79), small (g = .20- 49), or of no practical importance (less than .20). An effect size of .20 means that on average, the learners provided with the experimental treatment (independent variable, in this case, the CALL L1 glosses) perform two-tenths of a standard deviation (on a dependent variable, e.g., a reading comprehension test) above those participants who did not receive the experimental treatment. The effects size, g, is calculated using the formula where  is the mean of the experimental group,  is the mean of the control group, and Sp is the pooled standard deviation of the control and experimental groups. The effect size is a more accurate measure of the effect of the degree of effect of a treatment on participants as measured by the testing instrument than a statistical test of significance because such tests are strongly dependent on sample size and do not provide a standardized indication of the impact of the independent (treatment) variable on the dependent variable (test instrument). Analyzing the literature according to whether or not statistical significance was obtained can often lead to the wrong overall conclusion, as demonstrated by measurement experts Hedges and Olkin (1980).