Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

16.1 Introduction and Literature Review

The primary goal of reading is to construct a mental representation of meaning from a text (Grabe and Stoller 2002). This meaning constructing activity involves coordination of multiple levels of sub-component processes, including lower-level processes and higher-level processes (Grabe 2009). Theoretical models of reading place different emphases on the roles of lower-level processing and higher-level processing in reading comprehension. On one side of the coin, some researchers stress the importance of efficiency of lower-level processing, suggesting that inefficiency in word processing inhibits higher-level reading comprehension processes, such as uses of reading strategies and text comprehension (e.g. Perfetti 1988; Perfetti and Hart 2001). In the present study, this hypothesis is referred to as the inhibition hypothesis. On the other side of the coin, other researchers emphasize the role of strategic processing in reading comprehension, maintaining that, as long as readers have sufficient time to carry out the reading task, inefficiency in lower-level processing does not normally hinder reading comprehension, as readers are able to use strategies to compensate for processing and/or language problems (e.g. Walczyk 2000; Walczyk et al. 2007). In the present study, this hypothesis is referred to as the compensation hypothesis.

16.1.1 Theoretical Framework

The inhibition and compensation hypotheses will be explained by using two models: the Verbal Efficiency Model (VEM) (Perfetti 1988; Perfetti and Hart 2001); and the Compensatory-Encoding Model (C-EM) (e.g. Walczyk 2000; Walczyk et al. 2007).

16.1.1.1 Verbal Efficiency Model and the Inhibition Hypothesis

According to Perfetti’s (1988, 1999) VEM model, subcomponents in reading are arranged in a hierarchical manner and different reading processes have ranges of processing efficiency, with lower-level processes having more potential to become automatic through extensive practice than other processes. VEM assumes that the cognitive resources (attention and WM) necessary for good reading comprehension are limited in capacity. Therefore, efficient lower-level processing skills allow cognitive resources to be used for other higher-level comprehension processes. Efficient word processing skills are central to the model and VEM maintains that the inefficient word processing skills often inhibit readers’ problems with higher-level comprehension skills (e.g. building a coherent interpretation of text content, and use of reading strategies). VEM suggests that even in the adult population, for whom reading is presumably a well-practiced skill, there exists considerable variation in terms of efficiency of word processing.

16.1.1.2 Compensatory-Encoding Model and the Compensation Hypothesis

In proposing the C-EM model of reading, Walczky and his associates (Walczyk 2000; Walczyk et al. 2007) adopted basic assumptions of the VEM model but added compensatory mechanisms. Compensatory mechanisms are metacognitive in nature (Walczyk 1995), and are controlled processes, which have characteristics of slowness, serial in nature, and attention-demanding (Walczyk 2000; Wickens 1984). The establishment of the C-EM attempts to explain “the interplay between automatic and control processes” in reading (Walczyk 2000, p. 35). According to the C-EM, in fluent reading, lower-level processing tends to be carried out automatically and they make few demands on attention and WM. As a result, attention and WM can be used for higher-level comprehension processes. In situations where word processing is inefficiently or readers with a small WM, the C-EM assumes that readers are more likely turn to compensatory strategies.

One of the important predictions proposed in the C-EM model is that when there is no time pressure in reading, inefficient word processing and small WM “does not normally affect performance during reading because compensatory mechanisms operate routinely during performance” (Walczyk 1993, p. 127).

16.1.2 Empirical Evidence for Inhibition and Compensation in L1 Reading

In L1 reading, the common finding is that word processing is a good predictor of reading comprehension for beginning readers (Perfetti 1985), and word recognition among children is a major factor contributing to their later reading abilities (Adams 1999; Perfetti et al. 2005).

Whether word processing inhibits the reading comprehension of older L1 readers who are beyond the period of acquiring reading skills is more ambiguous. On the one hand, some studies showed that word processing is a good predictor for reading comprehension even among adult readers. On the other hand, other studies have found that there was only a weak link or no link between word processing and reading comprehension for older L1 readers (e.g. Walczyk 1995; Walczyk and Raska 1992).

A handful of studies with adult native English readers exist that provide some empirical evidence for the compensation hypothesis. Walczyk (1995) found that in a no time pressure condition, none of the measures of lexical access and WM (speed and accuracy) was correlated with comprehension. The results provided some evidence for the C-EM model, that under no time pressure, word processing and WM did not predict reading comprehension.

In two studies (Walczyk et al. 2001; Walczyk and Taylor 1996), Walczyk and his colleagues provided partial support for the predictions made in the C-EM. The results of the two study showed that compensatory mechanisms (i.e. behaviours and strategies) were negatively correlated with the speed of lexical access and the speed measure of WM, suggesting that readers of inefficient lower-level processing used more frequently compensatory mechanisms. However, Walczyk and Taylor (1996) found that speed of lexical access did negatively correlate with text comprehension, meaning the faster to retrieve the meaning of words leads to better reading comprehension.

16.1.3 Empirical Evidence for Inhibition and Compensation in FL Reading

In FL reading, whether inefficient word processing inhibits reading comprehension has produced inconsistent results. On the one hand, word processing was found to positively correlate with reading comprehension in FL reading, suggesting the more proficient a FL reader process at word or sub-word level, the better he/she can achieve in comprehension (e.g. Koda 1992; Nassaji 2003; Nassaji and Geva 1999). These positive and significant correlations obtained in these studies suggest that inefficient word processing inhibits reading comprehension. For instance, Nassaji and Geva (1999) and Nassaji (2003) found that both processing beneath word level (orthographic processing) and word level processing significantly correlated with reading comprehension for adult FL readers of English speaking Farsi as their L1.

On the other hand, in other studies, word processing has been found not to influence comprehension significantly (e.g. Haynes and Carr 1990; Stevenson 2005; van Gelderen et al. 2004). For instance, Haynes and Carr (1990) found that among Chinese EFL learners, although word processing variables (i.e. word decoding variable and lexical access variable) in English reading positively correlated with reading speed, they did not correlate with levels of comprehension. Similarly, in a think-aloud study by Stevenson (2005) with Dutch adolescent EFL readers, the results suggested that word processing efficiency did not significantly correlate with levels of global reading comprehension.

16.1.4 The Present Study

L1 reading studies have showed that time pressure influences the contribution of lower-level processing to L1 reading comprehension (e.g. Walczyk 1995). Whether time pressure influences the role of lower-level processing to FL reading comprehension needs to be investigated empirically, because a handful of FL studies to date have demonstrated inconsistent results on how lower-level processing contributes to FL reading. The current study aims to investigate this issue with Chinese learners of English as a foreign language (EFL) at university level.

The study asks two research questions:

  1. 1.

    To what extent does lower-level processing (i.e. LA and WM) relate to reading comprehension in (a) untimed and (b) timed FL reading for Chinese EFL learners at university level?

  2. 2.

    To what extent does lower-level processing (i.e. LA and WM) contribute to reading comprehension in (a) untimed and (b) timed FL reading for Chinese EFL learners at university level?

16.2 Method

16.2.1 Research Design

A repeated measures design is used in which the same participant was required to read two English texts in two different reading conditions (no time pressure and time pressure conditions).

16.2.2 Setting and Participants

The study was conducted in a national university in China with 404 Chinese undergraduates (138 males and 266 females). The ages of the participants ranged from 18 to 23, with a Mean of 20.22 years old. The participants had received on average 7.5 years of English instruction: 6 years in secondary school and 1.5 year in university.

16.2.3 Instruments

16.2.3.1 Lexical Access Test

To measure word processing, the present study employed a computerized lexical access (LA) test adapted from Haynes and Carr’s (1990) paper test. The LA test required learners to decide as quickly as possible whether a pair of words were synonyms or antonyms. This test was delivered using DMDX software (version 3.3.1.1), that recorded both accuracy and reaction time (RT) in milliseconds (Forster and Forster 2003).

There were 60 word pairs, half of which were synonyms and half of which were antonyms. The lexical relationship between the words was checked in an online Thesaurus (www. thesaurus.com). The reliability – Cronbach’s alpha was .94, indicating a very high reliability for the LA test.

16.2.3.2 Working Memory Test

A modified computerized Operation Span Task (OSpan) (Unsworth et al. 2005), was used to measure WM. The OSpan task was delivered using DMDX software (version 3.3.1.1) (Forster and Forster 2003). There were 40 items organised into 10 sets ranging from 2 items to 6 items in one set. The test asks participants to judge simple mathematical equations at the same time to memorize isolated English words. The WM test collected three aspects of information: students’ accuracy of judgment on the correctness of arithmetic equations, students’ RTs on the judgment, and the number of correctly recalled English words. The three aspects were formed composite Z-scores as indicators of WM. The reliability – Cronbach’s alpha was .91, suggesting the WM test was quite reliable.

16.2.3.3 Reading Comprehension Test

Reading comprehension was measured through four expository texts. In each condition, two texts were used. The four texts were adapted from College Reading Workshop (Malarcher 2005). Efforts were also made to maintain a similar level of text readability across the texts. Reading comprehension was measured using a multiple choice format, which is the most commonly used task for assessing reading comprehension (Brantmeier 2005; Phakiti 2008). For each text, ten multiple choice questions were constructed with four possible choices. The reading test obtained a reliability of .83.

16.2.4 Data Collection Procedure

The gathering of data was carried out in two stages. The first stage was the collection of the reading comprehension in the two reading conditions in English classes. The second stage was the collection of LA and WM data in a quiet computer laboratory.

16.2.5 Data Analysis

For research question 1, a series of bivariate Pearson product moment correlation were carried out separately for the two reading conditions. In order to answer research question 2, separate regression analyses were carried out in the two reading conditions with reading comprehension as dependent variables, and LA and WM as independent variables.

16.3 Results

16.3.1 Descriptive Statistics

Table 16.1 presents the descriptive statistics.

Table 16.1 Descriptive statistics for reading comprehension, LA and WM
Full size table

16.3.2 Results for Research Question 1

Table 16.2 presents the results of the correlation analyses.

Table 16.2 Results of Pearson product moment correlation analysis (reading comprehension)
Full size table

Table 16.2 showed that firstly, there was no significant correlation between students’ LA and reading comprehension in untimed reading condition (r = −.09, p = .07), whereas a small and negative relationship was found between LA and reading comprehension in timed reading (r = −.22, p < .01). Since RTs was used to measure LA, a negative relationship between LA and reading comprehension means that readers who were slower to access meanings of English words (who had longer RTs) tended to achieve poorly in timed FL reading; whereas readers who were faster to get access to meanings of English words (who had slower RTs) had a tendency to obtain better comprehension in timed FL reading. Secondly, WM was shown to be correlated positively with students’ reading comprehension in both untimed (r = .11, p < .05) and timed reading conditions (r = .20, p < .01), and both the values of correlation were small. The magnitude of correlation between WM and comprehension in untimed reading was smaller than that between WM and comprehension in timed reading. This means that students who had larger WM were more likely to be associated with better reading comprehension in both reading conditions. But WM had stronger association with comprehension in timed reading than in untimed reading. Additionally, LA was found to be significantly and negatively correlated with WM (r = −.28, p < .01). Lastly, reading comprehension in untimed reading was positively related to that in timed reading, and the magnitude of correlation was moderate (r = .43, p < .01). The above results between LA, WM and reading comprehension in the two reading conditions seemed to support the prediction made by the C-EM model, which maintains that when there is no time pressure in reading, inefficient word processing and small WM does not normally affect reading performance (Walczyk 1993).

16.3.3 Results for Research Question 2

Since the results of correlation analysis indicated no significant relationship between LA and reading comprehension in untimed reading, therefore, for untimed reading, only a simple regression was performed with WM as an independent variable. For timed reading, both LA and WM were used as independent variables in a multiple regression analysis. The effect size of regression analyses f 2 was also calculated and reported. The results of the regression analyses are displayed in Tables 16.3 and 16.4 separately.

Table 16.3 Results of simple regression analysis (untimed reading comprehension)
Full size table
Table 16.4 Results of multiple regression analysis (timed reading comprehension)
Full size table

The results of the simple regression analysis in Table 16.3 showed that WM was a significant predictor of English reading comprehension in untimed reading condition (β = .11, R 2 = .01, p < .05, f 2 = .01), accounting for only about 1 % variance, and the effect size was small. The results of the multiple regression in Table 16.4 revealed that the variable of LA alone could explain about 5 % variance of reading comprehension in timed reading condition (β = −.18, R 2 = .05, p < .01, f 2 = .05). The variable of WM was also a significant factor for explaining the reading performance in the timed reading condition, contributing to about 2 % of variance (β = .15, R 2 = .02, p < .01, f 2 = .02). Both the values of the effect size attributable to the LA and WM were small, with .05 and .02 respectively. The above results suggested that the two variables, LA and WM, together could account for about 7 % variance of reading comprehension in timed-reading condition.

16.4 Discussion and Conclusion

The results of the present study seem to suggest that in untimed reading condition, LA does not inhibit reading comprehension, but in timed reading condition, the efficiency of LA inhibited reading comprehension to some extent. While there have been no studies in FL reading comparing the relationship between the efficiency of LA and reading comprehension in different reading conditions, two of the previous studies have found that the relationship between LA and reading comprehension was not significant when readers read without time imposed on them.

In a think-aloud study, Stevenson (2005) found that there was no significant relationship between the speed of English word processing and reading comprehension among 22 adolescent Dutch EFL learners. Using think-aloud protocols gave readers sufficient time for them to complete their reading tasks at hand, this situation simulated the untimed reading in the present study. However, the direct comparison of the results between the two studies needs to be made with caution. Stevenson (2005) tested the efficiency of learners’ decision on whether letter strings are real or pseudo English words. This measure measurement is different to the one used in the current study, because the former measurement does not require learners to retrieve the meaning of the letter strings, while the latter measurement ask learners not only to recognize the word but also to obtain the meaning of the word from their mental lexicon.

Using the same type of LA test, Haynes and Carr (1990) found that the efficiency of LA did not significantly correlate with reading comprehension among a group of Taiwanese EFL learners. The results of the current study seem to corroborate the results of Haynes and Carr (1990)’s study among the learners with the same language background. While Haynes and Carr’s study only used paper and pencil test for LA, which might not be that accurate in terms of recording RTs, the present study used computer software to capture RTs to milliseconds.

In conclusion, the results of the present study seem to support the compensation hypothesis that when reading in no time pressure condition, the lower-level processing do not seem to affect FL readers’ reading comprehension.