The Role of Executive Functions in Reading
Fluency in the First and Second Language
Researchmasterthesis Psychology
L.G. Naaktgeboren, 10154175
Developmental Psychology
Supervisors:
P. Snellings (Developmental Psychology)
M.J.H. van Koert (Developmental Psychology)
Second assessor:
1
AbstractThe ability to read is important in our society, because we are confronted with words every day. However, some people have more difficulties with reading than others. One important aspect of reading is reading fluency. Word recognition speed and vocabulary knowledge are well-known contributors to reading fluency. The current study investigates whether executive functioning is an additional component to sentence reading fluency, in the first language (L1) and the second language (L2), and whether reading fluency in L1 influences reading fluency in L2. This was studied in 94 children in grade 4 (age 8-10), who were proficient in their L1, Dutch, and just started to learn their L2, English. It was found that in both L1 and L2, executive functioning had no additional influence on reading fluency. In L1, L1 word recognition speed and L1 vocabulary knowledge influenced reading fluency, although only L1 word recognition speed influenced reading fluency when executive functioning was added. In L2, only L2 word recognition speed contributed to reading fluency. This suggests that the factors influencing reading fluency are different for L1 and L2, at least when the experience with L2 is limited. When the influence of L1 reading fluency on L2 reading fluency was studied, L1 reading fluency was a significant contributor to L2 reading fluency, as were L2 word recognition speed and L2 vocabulary knowledge. Executive functioning had no additional influence on L2 reading fluency. The findings suggest that in children in grade 4, executive functioning is no significant contributor of reading fluency in both L1 and L2, while word recognition speed is a significant contributor in both L1 and L2, and vocabulary knowledge on contributes to reading fluency in L1. Furthermore, in line with the interdependence hypothesis (Cummins, 1979), the reading fluency of L2 is dependent on the reading fluency in L1.
Introduction
The ability to read is important in our society. The western society is adapted to the fact that people can read; everywhere around us we are confronted with words. This requires an expert ability level of reading. All of us learn to read in our first language (L1), but most people also have to learn a second language (L2), to be able to communicate with foreigners. In 2002, the European Commission even decided that in the future, everyone should learn at least two languages apart from their first language (European Council, 2002).
Some people have more difficulties with reading than others. Since such a high level of reading ability is needed in our society, it is important to study which factors can influence reading acquisition in L1 and L2, and which factors put children at risk for poor reading acquisition. The current study investigates one of the most important aspects of reading: reading fluency, in particular sentence reading fluency. Word recognition speed and vocabulary are well-known contributors to reading fluency, and this study explores whether executive functioning is an additional component.
Reading fluency concerns the speed with which a sentence or a word can be recognized and produced (Beck, Perfetti, & McKeown, 1982; Fuchs, Fuchs, Hosp, & Jenkins, 2001). The importance of reading fluency is reflected in the finding that even when the accuracy of word recognition is high, slow reading still accompanies reading comprehension problems (Perfetti, 1999). Hence, in order to focus our attention to understanding the things we read, we first need to speed up word
identification (Pikulski & Chard, 2005; Beck et al., 1982). Models on reading comprehension show this order of processes as well: reading comprehension can be divided in lower and higher order skills, where lower order skills consist of word and letter recognition skills and higher order skills consist of skills needed for text comprehension (Van Gelderen et al., 2003).
At least two factors, word recognition and vocabulary, influence reading fluency. First, Stanovich (1991) found that the lower order skill word recognition has a strong correlation with the higher order skill reading comprehension and he argues this is a causal link. This suggests that when word recognition skills improve, speed of sentence comprehension also increases. Second,
participants who were trained on target words had faster reaction times in a sentence
2
whether word recognition and vocabulary are the only factors that influence reading fluency. Executive functioning could be an additional factor.
According to Miyake et al.’s (2000) model, executive functioning consists of three components: updating (working memory), inhibition, and cognitive flexibility (task switching). Working memory involves the storage and processing of information (Baddeley, 1992; 2003; Baddeley & Hitch, 1974), inhibition concerns the inhibition of automatic or dominant responses (Miyake et al., 2000) and cognitive flexibility involves the switching of attention between multiple tasks or mental sets (Garon, Bryson, & Smith, 2008). Although Miyake et al. (2000) argued that executive functioning is a factor model with working memory, inhibition and cognitive flexibility as its components, the individual subtests measuring working memory, inhibition and task switching do not always correlate (Bull & Scerif, 2001). It is possible that executive functioning influences reading fluency directly and indirectly via word recognition and vocabulary.
Baddeley (2003) has shown that a reciprocal relationship exists between working memory and vocabulary, where a good phonological memory results in a greater memory capacity, which makes it easier to learn new vocabulary. In this way vocabulary size increases, which subsequently makes it easier to remember words. As of yet, few findings link executive functioning to reading fluency in sentences, but non-fluent word recognition has already been linked to a limited working memory capacity (Stanovich, 1991). Furthermore, executive functions have been found to influence the level of text comprehension. A lower working memory has been shown to predict lower
comprehension rates (Carretti, Borella, Cornoldi, & De Beni, 2009) and word recognition rates (Swanson & Ashbaker, 2000; Christopher et al., 2012). The same effect has been found for inhibition: poor inhibition correlated with poorer comprehension (Cain, 2006). Cain (2006) suggests that
children with weaker inhibitory skills cannot inhibit irrelevant information during text reading, which impairs text comprehension, and possibly also sentence comprehension. Finally, cognitive flexibility was lower in children with comprehension problems than in children with normal levels of
comprehension and decoding (Cartwright, 2015).
The findings above all illustrate the connections between reading fluency, vocabulary, word recognition skills and executive functioning in L1. To the author’s knowledge, few studies have been carried out into these connections for L2 learners. It has been shown that bilingual verbal proficiency in sequentially bilingual Turkish-Dutch children predicted the working memory capacity at age 6 (Blom, Küntay, Messer, Verhagen, & Leseman, 2014). Concerning inhibitory control, 5- to 8-year old children who learn a second language at school are in between monolinguals on the one hand, and bilinguals and trilinguals on the other hand (Poarch & van Hell, 2012), because when a second (or third) language is learned, more inhibitory control is needed to inhibit the other language(s).
Furthermore, a study involving task switching revealed that the speed of attention control in L2 was a significant contributor to L2 proficiency (operationalized as lexical access), in addition to speed of attention in L1 (Segalowitz & Frenkiel-Fishman, 2005). The authors argue that on the one hand, it could mean that as a second language learner improves his attention control skills, he learns his language more efficiently, and thus becomes more proficient. On the other hand, it could mean that as a second language learner becomes more proficient in a language, he learns to control the attention to the relevant information. Thus, of these studies, some studies state that language proficiency influences executive functioning (Blom et al., 2014; Poarch & van Hell, 2012), and one study states that executive functioning influences language proficiency (Cain, 2006), while another study states the direction could be in both ways.
The connection between L1 and L2 reading skills is the focus ofs the interdependence hypothesis (Cummins, 1979), which states that the reading competence in L2 is dependent on the reading competence in L1. The linguistic coding differences hypothesis (Sparks & Ganschow, 1991; 1993; 1995) also stresses that L1 reading skills are important for L2 learning, but has a bigger focus on cognitive skills. This hypothesis assumes that there is one core component that underlies the reading skills across all languages, which means that the reading skills that are important in L1 are also important in L2.
3
All of the findings above indicate that there is a connection between word recognition skills, vocabulary knowledge and reading fluency, but it is unclear whether executive functioning
additionally contributes to reading fluency. As noted before, different studies have been done studying connection between some of these factors, but to the author’s knowledge, none of them combined all factors in one study. Furthermore, the connections between these factors are unclear for a second language that is just starting to emerge. The current study aims to clarify if and to what extent executive functioning influences reading fluency, on top of vocabulary and word recognition speed, in both L1 and L2. Furthermore, it aims to clarify the influence of L1 reading fluency on L2 reading fluency. The connections between word recognition, vocabulary, executive functioning and reading fluency in L1 and L2 were investigated in children in grade 4 of primary school, in which the participants just started to learn a second language. With this sample the effect of the different factors on reading fluency could be studied at the start of learning a new language. It is hypothesized that executive functioning is a significant contributor to reading fluency, in addition to word
recognition skills and vocabulary in the first and second language, as all executive functions have been found to have a negative influence on reading fluency, where better executive functions result in faster reaction times. The expectation is that this relationship is the same in L1 as in L2. In
addition, it is expected that executive functioning will contribute to word recognition skills and vocabulary. Finally, it is hypothesized that L1 factors will contribute to their L2 counterparts (Cummins, 1979).
Method Participants
The data were gathered as part of the ORWELL-project, which is an acronym for ORal and Written English Language Learning. This is a longitudinal study which examines Dutch (L1) and English (L2) reading and writing skills, intelligence and executive functions in children in grade 4, 5 and 6 of primary school. The current data were gathered from February until May 2017, and involved the children in grade 4 (N = 298) of seven different schools across the Netherlands. Since all children spoke Dutch fluently, we use L1 to refer to Dutch as the dominant language (but not necessarily the first language), and L2 to refer to English as the foreign language (but not necessarily the second language).
Instruments
A total of nine instruments were used, of which six were developed by the research team. These consisted of two vocabulary knowledge tests (one for L1 and one for L2), four speed tests (two for L1 and two for L2) and three executive functioning tests. The speed tests measured speed of word recognition and sentence verification, and contained well-known words and simple sentences, in order to ensure that these tests measured individual differences in speed processing instead of linguistic knowledge. The vocabulary tests measured individual differences in linguistic knowledge with more difficult items. The executive functioning tests were developed to measure inhibition and task switching, and an existing task was used to measure working memory.
The independent variables were the executive functioning tests, word recognition speed tests and the vocabulary tests (both measured in L1 and L2). The dependent variable was reading fluency, measured with a sentence verification speed test in both L1 and L2.
All self-developed instruments (the four speed tests and two of the executive functioning tests) were piloted and revised to reach the intended reliability and difficulty. Furthermore, the difficulty of the instructions was tested, so that upon testing it was clear that all children understood the task.
Reading fluency. Reading fluency was measured with a sentence verification task. This task
consisted of 100 sentences, of which 50 sentences made sense and 50 did not. Participants had to decide as fast as they could whether a sentence made sense or not. An example of a sensible item is:
Seventeen is a number. An example of a nonsense item is: Parents are younger than their children.
4
knowledge. In the task, a sentence was presented on a laptop screen and participants had to respond whether the sentence made sense with the left index finger or whether it did not make sense with the right index finger (counterbalanced across participants). Reaction time and accuracy were recorded. For this analysis, only the reaction time of correctly answered sensible sentences were included. This task was administered in Dutch and English.
Word recognition. To test whether the participant could discriminate between words and
non-words, a word-recognition task was administered. This task consisted of 100 items, of which 50 items were words and 50 were pseudowords. Each word had a pseudoword counterpart, which differed one to three letters from the existing word. The English pseudowords were checked against an English-Dutch dictionary. Participants had to decide as fast as they could whether a letter string was a word or not. An example of an existing word is: follow. An example of a pseudoword is: monnow. In this task, a letter string was presented on a laptop screen and the participants had to respond whether it was a word with their left index finger or a pseudoword with their right index finger (counterbalanced across participants). Reaction time and accuracy were recorded. For this analysis, only the reaction time of correctly answered existing words were included. This task was
administered in Dutch and English.
Vocabulary. To test the vocabulary of the participants in both Dutch and English, several sets
of the Peabody Picture Vocabulary Test (PPVT) were administered (PPVT-IIINL for Dutch, PPVT-IVEN for English). In this test, a prerecorded word was played and the participants had to choose the right meaning of the word by tapping one of the four pictures presented on a tablet. The Dutch words were recorded by a native speaker of Dutch, and the English words by a native British English speaker. In the Dutch version, set 7 to 11 were administered, with a total of 60 items. In the English version, set 1 to 7 were administered, with a total of 84 items.
Executive functions. Verbal tasks were chosen to measure executive functioning, because this
study aimed to examine whether executive functioning could predict language proficiency. It was expected that the verbal components of executive functioning were a better predictor than the non-verbal ones (Rogers & Monsell, 1995; Meuter & Allport, 2009; Linck, Osthus, Koeth, & Bunting, 2014; Kidd & Hulmes, 2015; Verhagen & Leseman, 2016).
Working memory. Working memory was measured with the digit span task from the Wechsler
Intelligence Scale for Children (WISC-IIINL). In this task a prerecorded sequence of numbers was played and participants had to repeat this sequence in the same order (digit span forward) or in the reverse order (digit span backward). The sequence started with two numbers and ended with nine numbers. For each sequence amount participants had two attempts. When both attempts were wrong, the subtest was ended. Digit span forward consisted of a maximum of 16 items, and digit span backward consisted of a maximum of 14 items. The total score was the score on digit span forward plus the score on digit span backwards.
Inhibition. Inhibition was measured with an auditory Simon task (Xiong & Proctor, 2016),
which uses the speech sounds /a/ and /o/. The speech sounds were presented monaurally to the left or right ear. Participants had to respond to /a/ with their left index finger and to /o/ with their right index finger (counterbalanced across participants). This created congruent trials (in which the side of the stimulus and response were the same) and incongruent trials (in which the stimulus and
response sides differed). Inhibition was measured as the Simon effect (Hedge & Marsh, 1975). This effect is the finding that participants tend to be faster on congruent trials (in which the location of the stimulus and the response are the same) than on incongruent trials (in which the location of the stimulus and the response differ). It is suggested that this effect represents the tendency to respond to the location of the stimulus (Simon, Acosta, & Mewaldt, 1975). In this task, the Simon effect was obtained by subtracting the median reaction time on congruent trials from the median reaction time on incongruent trials. The task contained 3 blocks of 100 trials, which made a total number of 300
5
trials. Reaction time and accuracy were measured. In this analysis, only reaction times on correct trials were included.
Task switching. The ability to switch between different tasks was measured with an
adaptation of the alternating run paradigm (Nicolay & Poncelet, 2015; Gladwin, Lindsen, & de Jong, 2006). In this paradigm, tasks A and B were performed in an AABBAABB… sequence. Participants saw the letter S or H and heard the binaural speech sound /a/ or /o/ simultaneously. A 2x2 matrix was continuously present and the letters appeared in a clockwise sequence in the four cells. If the letters appeared in the upper part of the matrix, participants had to respond to the letters (task A), and when the letters appeared in the lower part they had to respond to the speech sounds (task B). Participants had to respond to the S and the /a/ with their left index finger and to the H and the /o/ with their right index finger (counterbalanced across participants). This mixing block contained 96 items. Two pure blocks of both 96 items, in which participants did one of the two tasks, were administered as well, but were not included in the analysis. Task switching ability was measured by the switch cost or switch effect. This effect was based on the finding that participants tend to be slower on a trial where they have to do another task than the trial before (a switch trial) than on a trial where they have to do the same task as the trial before (a repeat trial; Monsell, 2003). In this task, the switch effect was obtained by subtracting the median reaction time on repeat trials from the median reaction time on switch trials.
Feedback was provided during the whole task. Reaction time and accuracy were measured. In this analysis, only reaction times on correct trials were included.
Procedure
All tests were administered by well-trained test assistants. For the whole ORWELL-project, the children took part in five classroom sessions of 45 minutes each, and six individual sessions of 30 minutes each. The classroom sessions contained the PPVT-NL and PPVT-EN tests, among other tests, while the individual sessions contained the sentence verification tasks, lexical decision tasks, the WISC Digit span, the Simon task and the task switching task, among other tasks. The order of sessions was quasi-random; at some schools the sessions with English tasks were administered first, while on the other schools the sessions with Dutch tasks were administered first. The tasks within each session were administered in a random order.
Scoring
For the speed tests (word recognition and sentence verification) the median reaction times of the correct responses on the existing words and sensible sentences were calculated. The median was chosen since reaction time distributions are usually right skewed, which can have a large influence on the mean. Reaction times were screened for outliers. When the reaction time was faster than the quickest reaction time of an expert (native speaker of L1 and advanced speaker of L2), the reaction time was seen as an outlier (960 ms for Dutch sentence verification, 706 ms for English sentence verification, 250 ms for Dutch and English word recognition). The expert reaction time for English sentence verification was faster than that of Dutch sentence verification, since the English sentences were easier and shorter than the Dutch sentences. Long reaction times were not seen as outliers. Outliers and incorrect responses were scored as missing values. Since we wanted to measure individual differences in speed processing and not linguistic knowledge, only the reaction time of items with a correct rate of at least 80% was taken into account. This resulted in the deletion of 13 Dutch sentences, 26 English sentences, 2 Dutch words and 20 English words. It seemed like the English word recognition task, and both the English and Dutch sentence verification tasks were difficult for the children. Only very simple words or sentences seemed suitable to test individual differences in speed processing. Furthermore, as in Van Gelderen et al. (2004), who used the same kind of speed tests, if a participant scored less than 62.5% correct on a speed test, the whole test was scored as missing, to prevent the mean reaction time to consist off too few measurements. This resulted in the deletion of 9 participants for Dutch sentence verification, 18 participants for English sentence verification, 1 participant for Dutch word recognition and 11 participants for English word
6
recognition. Two of the executive functioning tasks used reaction time as well. For these tasks, the Simon task and the task switching task, reaction times below 250 ms were scored as outliers. Outliers and reaction times of incorrect responses were scored as missing. Apart from the procedure
described above, some participants had not completed all tests at the moment of writing this paper. For these participants whole test scores were missing as well.
Data analysis
Since executive functioning was measured with different tests for working memory, inhibition and task switching, the original idea was to perform a factor analysis to load these scores on the factor executive functioning. However, the different executive functions did not correlate with each other (see Results), which is why they were entered into the regression separately. Subsequently, two hierarchical regressions were performed to test the first two main questions. Because the influence of word recognition skills and vocabulary on reading fluency is already established (Stanovich, 1991; Beck et al., 1982), these predictors were added first, and subsequently, the executive functions were added to see if executive functioning could additionally explain unique variance between
participants. To determine whether executive functioning had an indirect effect on reading fluency via word recognition and vocabulary, four additional multiple regressions were conducted with speed of word recognition and vocabulary of each language as dependent variable, respectively, and working memory, inhibition and task switching as independent variables. To determine whether L1 reading fluency contributes to L2 reading fluency, a multiple regression was performed with L1 reading fluency, L2 word recognition speed, L2 vocabulary knowledge, working memory, inhibition and task switching as predictors.
Results
In total, 94 participants (N = 94) completed all tests and had no missing test scores. The vocabulary knowledge scores of the included participants did not significantly differ of the scores of participants with one or more missing test scores in L1 (t(288) = -1.61, p = .108) and L2 (t(292) = -.18, p = .859). The descriptive statistics for all variables in L1 (Dutch) and L2 (English) are presented in Table 1. As can been seen, for the sentence verification speed tests, it seems that the participants were faster on average in L2 than in L1. For the word recognition speed tests, the participants seem to be equally fast in L1 as in L2. However, this comparison should be taken with caution since the test statistics are based on a different number of items and the tests were created with a different level of difficulty in mind for Dutch and English. The difference in speed between sentence verification and word
recognition can be explained by extra time needed to read a whole sentence. The reliability of the tasks was poor (sentence verification speed L1: Cronbach’s α = .59, L2: Cronbach’s α = .46, word recognition speed L1: Cronbach’s α = .55, L2: Cronbach’s α = .67). The vocabulary test in both L1 and L2 and the working memory test were of appropriate difficulty and have enough variance in their scores. Furthermore, the reliability of both vocabulary tests is good (L1: Cronbach’s α = .77, L2: Cronbach’s α = .86). Table 1 shows that there were visible Simon and Switch effects, however, the variance of both effects was large.
7
Table 1. Medians/means, standard deviations, reliability, number of items and number of
participants per test
Variable L1 (Dutch) L2 (English) Mdn/M SD α (N) No. of items Mdn/M SD α (N) No. of items Sentence verification speed a 4196.5 1129.8 .588 (61) 37 1885.8 610.3 .458 (84) 24 Word recognition speed a 852.5 184.9 .552 (92) 48 851.0 188.1 .663 (93) 30 Vocabulary knowledge 37.9 6.4 .774 (94) 60 56.9 9.0 .864 (94) 84 Working memory b 11.8 2.2 - c 30 Variable
Congruent/repeat trials Incongruent/switch trials
Accuracy RT (ms) Accuracy RT (ms) Simon/ Switch effect (SD) M SD Mdn SD No. of items M SD Mdn SD No. of items Inhibition 135.9 16.5 588.5 86.3 150 130.7 16.9 630.0 85.2 150 41.5 (57.7) Task switching 34.3 6.8 797.8 113.9 48 28.8 5.6 881.0 137.2 48 83.2 (92.1)
Note. N = 94. The maximum score is equal to the number of items, except for the speed tests, were there is no maximum
score. For the inhibition and the task switching task, the number of congruent/repeat trials and incongruent/switch trials is half of the total number of items. L1 = main language, L2 = foreign language.
a Median reaction time in milliseconds (ms). b Was administered in Dutch.
c There was too little variance to calculate Cronbach’s α.
Main language (L1)
Table 2 contains the correlation matrix of all L1 (Dutch) variables and the executive functions. As can be seen, the only significant correlations were the correlation between sentence verification speed and word recognition speed (r = .56, p < .001); and the correlation between sentence verification speed and vocabulary knowledge (r = -.33, p = .019). Unexpectedly, vocabulary knowledge did not significantly correlate with word recognition speed. Furthermore, none of the executive functions correlated significantly with one another. Therefore, it was decided not to construct a latent variable executive functioning based on these three tests. In addition, none of the executive functions significantly correlated with sentence verification speed, word recognition speed or vocabulary knowledge, which makes it unlikely that executive functions will directly or indirectly influence sentence verification speed. To show this, a multiple regression analysis was performed. In the subsequent regression, the executive function tasks were entered separately.
8
Table 2. Correlation matrix of L1 variables and executive functions
Variable Sentence verification speed L1 Word recognition speed L1 Vocabulary knowledge L1 Working
memory Inhibition Task switching Sentence verification speed L1 1 Word recognition speed L1 .511** 1 Vocabulary knowledge L1 -.241* -.123 1 Working memory -.137 -.115 .199 1 Inhibition .123 -.055 -.172 -.124 1 Task switching -.065 .011 .089 .069 .034 1
Note. N = 94, * p < .05, ** p < .01. L1 = main language, L2 = foreign language.
A multiple linear regression was conducted to predict sentence verification speed based on word recognition speed, vocabulary knowledge, working memory, inhibition and task switching. As the influence of word recognition speed and vocabulary knowledge on sentence verification speed is already established from the literature (Stanovich, 1991; Beck et al., 1982), these variables were entered first. To explore the difference in unique explained variance of those variables, in step 1a only vocabulary knowledge was added to the model, and in step 1b only word recognition speed was added to the model. In the second model, vocabulary knowledge and word recognition speed were added together. In the third model, working memory, inhibition and task switching were added to the model in addition to word recognition speed and vocabulary knowledge. The results of model 1a show that vocabulary knowledge explained a significant amount of variance in sentence verification speed (F(1, 92) = 5.688, p = .019, R2 = .058, R2Adjusted = .048). The analysis shows that vocabulary knowledge significantly predicted the sentence verification speed (β = -.24, t(92) = 8.48, p < .001). When only word recognition speed was added to the model (step 1b), it was found that word recognition speed explains a significant amount of variance in sentence verification speed
(F(1, 92) = 32.569, p < .001, R2 = .261, R2Adjusted = .253). The analysis shows that word recognition speed significantly predicted sentence verification speed (β = .51, t(92) = 5.71, p < .001). For the second model, where vocabulary knowledge and word recognition speed were added together, the explained variance increased with .235 from step 1a to step 2, which was significant (F(1, 91) = 30.34,
p < .001); and .032 from step 1b to step 2, which was significant (F(1, 91) = 4.16, p = .044, R2 = .294,
R2Adjusted = .278). The analysis shows that word recognition speed significantly predicted the sentence verification speed (β = .49, t(91) = 5.51, p < .001), as did vocabulary knowledge (β = -.18,
t(91) = -2.04, p = .044). When working memory, inhibition and task switching were added to the
model, the explained variance increased with .019, which was not significant (F(3, 88) = .82, p = .485,
R2 = .313, R2Adjusted = .274). The analysis showed that word recognition speed significantly predicted the sentence verification speed (β = .50, t(88) = 5.53, p < .001). However, in this analysis vocabulary knowledge could not predict sentence verification speed (β = -.15, t(88) = -1.60, p = .114). Furthermore, working memory (β = -.03, t(88) = -.34, p = .734), inhibition (β = .12, t(88) = 1.36,
p = .177) and task switching (β = -.06, t(88) = -.67, p = .508) could not predict sentence verification
speed either. The unstandardized and standardized regression coefficients of the fitted hierarchical regression models are presented in Table 3.
9
Table 3. Linear model of the predictors of speed of sentence verification in L1
B (CI) SE B β p Step 1a Constant 5810.00 (4448.95, 7171.06) 685.29 .000 Vocabulary knowledge L1 (-77.91, -7.11) -42.51 17.82 -.24 .019 Step 1b Constant 1456.31 (481.36, 2431.27) 490.89 .004 Word recognition speed L1 (2.04, 4.21) 3.12 .55 .51 .000 Step 2 Constant 2784.92 (1174.01, 4395.83) 810.98 .001 Word recognition speed L1 (1.91, 4.07) 2.99 .54 .49 .000 Vocabulary knowledge L1 (-62.95, -0.82) -31.88 15.64 -.181 .044 Step 3 Constant 2680.35 (719.10, 4641.60) 986.90 .008 Word recognition speed L1 (1.95, 4.13) 3.04 .55 .50 .000 Vocabulary knowledge L1 (-58.23, 6.32) -25.96 16.24 -.15 .114 Working memory -16.33 (-111.41, 78.76) 47.85 -.03 .734 Inhibition 2.42 (-1.11, 5.94) 1.78 .12 .177 Task switching -.73 (-2.90, 1.45) 1.09 -.06 .508
Note.N = 94. R2 = .058 for Step 1a (p = .019), R2 = .261 for Step 1b (p < .001), R2 = .294 for Step 2 (p < .001), (ΔR2 = .235 (p < .001) from Step 1a to Step 2, ΔR2 = .032 (p = .044) from Step 1b to Step 2)), R2 = .313 for Step 3 (ΔR2 = .019 (p = .485). L1 = main language, L2 = foreign language.
To test whether the executive functions had an indirect effect on sentence verification speed via word recognition speed or vocabulary knowledge, two multiple regression were conducted with the latter two variables as dependent variables and the executive functions as independent variables. It was found that working memory, inhibition and task switching could not explain a significant amount of the variance in word recognition speed (F(3, 90) = .57, p = .638, R2 = .019, R2Adjusted = -.014). The analysis shows that working memory (β = -.12, t(90) = -1.19, p = .238), inhibition (β = -.07,
t(90) = -.68, p = .502) and task switching (β = .02, t(90) = .21, p = .833) could not predict word
recognition speed. Furthermore, it was found that working memory, inhibition and task switching could not explain a significant amount of the variance in vocabulary knowledge (F(3, 90) = 2.20,
p = .094, R2 = .068, R2Adjusted = .037). The analysis shows that working memory (β = .17,
t(90) = 1.70, p = .094), inhibition (β = -.15, t(90) = -1.49, p = .139) and task switching (β = .08, t(90) = .80, p = .423) could not predict vocabulary knowledge.
Foreign language (L2)
Table 4 contains the correlation matrix of all L2 (English) variables and the executive functions. As can be seen, there were significant correlations between sentence verification speed and word recognition speed (r = .63, p <.001); between sentence verification speed and vocabulary knowledge (r = -.36, p <.001); and between vocabulary knowledge and word recognition speed
10
with one another. Therefore, it was decided not to construct a latent variable executive functioning based on these three tests. In addition, none of the executive functions significantly correlated with sentence verification speed, word recognition speed or vocabulary knowledge, which makes it unlikely that executive functions will directly or indirectly influence sentence verification speed. To show this, a multiple regression analysis was performed. In the subsequent regression, the executive function tasks were entered separately.
Table 4. Correlation matrix of L2 variables and executive functions
Variable Sentence verification speed L2 Word recognition speed L2 Vocabulary knowledge L2 Working
memory Inhibition Task switching Sentence verification speed L2 1 Word recognition speed L2 .625** 1 Vocabulary knowledge L2 -.360** -.410** 1 Working memory -.129 -.136 .171 1 Inhibition .051 .080 .088 -.124 1 Task switching -.084 -.070 .173 .069 .034 1
Note. N = 94, ** p < .01. L1 = main language, L2 = foreign language.
A multiple linear regression was conducted to predict sentence verification speed based on word recognition speed, vocabulary knowledge, working memory, inhibition and task switching. As the influence of word recognition speed and vocabulary knowledge on sentence verification speed is already established from the literature (Stanovich, 1991; Beck et al., 1982), these variables were entered first. To explore the difference in unique explained variance of those variables, in step 1a only vocabulary knowledge was added to the model, and in step 1b only word recognition speed was added to the model. In the second model, vocabulary knowledge and word recognition speed were added together. In the third model, working memory, inhibition and task switching were added to the model in addition to word recognition speed and vocabulary knowledge. The results of model 1a show that vocabulary knowledge explained a significant amount of variance in sentence verification speed (F(1, 92) = 13.706, p < .001, R2 = .130, R2Adjusted = .120). The analysis shows that vocabulary knowledge significantly predicted the sentence verification speed (β = -.36, t(92) = -3.70, p < .001). When only word recognition speed was added to the model (step 1b), it was found that word recognition speed explains a significant amount of variance in sentence verification speed
(F(1, 92) = 58.963, p < .001, R2 = .391, R2Adjusted = .384). The analysis shows that word recognition speed significantly predicted sentence verification speed (β = .63, t(92) = 7.68, p < .001). For the second model, where vocabulary knowledge and word recognition speed were added together, the explained variance increased with .274 from step 1a to step 2, which was significant (F(1, 91) = 41.79,
p < .001); and .013 from step 1b to step 2, which was not significant (F(1, 91) = 1.98, p = .163, R2 = .404, R2Adjusted = .390). The analysis shows that word recognition speed significantly predicted the sentence verification speed (β = .57, t(91) = 6.46, p < .001), while vocabulary knowledge did not (β = -.13, t(91) = -1.41, p = .163). When working memory, inhibition and task switching were added to the model, the explained variance increased with .002, which was not significant (F(3, 88) = .08,
p = .973, R2 = .405, R2Adjusted = .371). The analysis shows that word recognition speed significantly predicted the sentence verification speed (β = .57, t(88) = 6.26, p < .001), while vocabulary
knowledge (β = -.12, t(88) = -1.28, p = .204), working memory (β = -.03, t(88) = -.34, p = .733), inhibition (β = .01, t(88) = .15, p = .882) and task switching (β = -.02, t(88) = -.26, p = .797) could not predict sentence verification speed. The unstandardized and standardized regression coefficients of the fitted hierarchical regression models are presented in Table 5.
11
Table 5. Linear model of the predictors of speed of sentence verification in L2
B SE B β p Step 1a Constant 3449.49 (2698.31, 4200.67) 378.22 .000 Vocabulary knowledge L2 (-37.33, -11.26) -24.30 6.56 -.36 .000 Step 1b Constant 292.84 (-176.27, 761.95) 236.20 .218 Word recognition speed L2 (1.50, 2.55) 2.03 .26 .63 .000 Step 2 Constant 917.58 (-80.62, 1915.78) 502.52 .071 Word recognition speed L2 (1.29, 2.43) 1.86 .29 .57 .000 Vocabulary knowledge L2 (-20.32, 3.47) -8.42 5.99 -.13 .163 Step 3 Constant 1009.50 (-125.02, 2144.01) 570.89 .080 Word recognition speed L2 (1.26, 2.44) 1.85 .30 .57 .000 Vocabulary knowledge L2 (-20.48, 4.44) -8.02 6.27 -.12 .204 Working memory -8.18 (-55.78, 39.42) 23.95 -.03 .733 Inhibition .13 (-1.64, 1.90) .89 .01 .882 Task switching -.14 (-1.24, .96) .55 -.02 .797
Note.N = 94. R2 = .130 for Step 1a (p < .001), R2 = .391 for Step 1b (p < .001), R2 = .404 for Step 2 (p < .001), (ΔR2 = .274 (p < .001) from Step 1a to Step 2, ΔR2 = .013 (p = .163 from Step 1b to Step 2)), R2 = .405 for Step 3 (ΔR2 = .002 (p = .973). L1 = main language, L2 = foreign language.
To test whether the executive functions had an indirect effect on sentence verification speed via word recognition speed or vocabulary knowledge, two multiple regression were conducted with the latter two variables as dependent variables and the executive functions as independent variables. It was found that working memory, inhibition and task switching could not explain a significant amount of the variance in word recognition speed (F(3, 90) = .82, p = .489, R2 = .026, R2Adjusted = -.006). The analysis shows that working memory (β = -.12, t(90) = -1.17, p = .246), inhibition (β = .18, t(90) = .64,
p = .524) and task switching (β = -.06, t(90) = -.61, p = .544) could not predict word recognition speed.
Furthermore, it was found that working memory, inhibition and task switching could not explain a significant amount of the variance in vocabulary knowledge (F(3, 90) = 2.12, p = .104, R2 = .066,
R2Adjusted = .035). The analysis shows that working memory (β = .17, t(90) = 1.68, p = .096),
inhibition (β = .104, t(90) = 1.01, p = .316) and task switching (β = .16, t(90) = 1.54, p = .127) could not predict vocabulary knowledge.
Influence of L1 on L2
Table 6 contains the correlation matrix of all L1 (Dutch) variables, L2 (English) variables and the executive functions. As can be seen, there were significant correlations between sentence
verification speed of both languages (r = .66, p < .001) and between word recognition speed of both languages (r = .59, p <.001). However, the correlation between vocabulary knowledge of both languages was not significant (r = .19, p = .068). Word recognition speed in L1 significantly correlated
12
with sentence verification speed in L2 (r = .55, p < .001), as did word recognition speed in L2 with sentence verification speed of L1 (r = .36, p < .001). Furthermore, vocabulary knowledge in L1 significantly correlated with sentence verification speed in L2 (r = -.23, p = .025), but not with word recognition speed in L2 (r = -.18, p = .086). In addition, vocabulary knowledge in L2 did not
significantly correlate with sentence verification in L1 (r = -.11, p = .301), but it significantly correlated with word recognition speed in L1 (r = -.22, p = .033).
Table 6. Correlation matrix of L1 variables, L2 variables and executive functions
Variable Sentence verification speed L1 Sentence verification speed L2 Word recognition speed L1 Word recognition speed L2 Vocabulary knowledge L1 Vocabulary knowledge L2 Working
memory Inhibition Task switching Sentence verification speed L1 1 Sentence verification speed L2 .660** 1 Word recognition speed L1 .511** .553** 1 Word recognition speed L2 .361** .625** .591** 1 Vocabulary knowledge L1 -.241* -.231* -.123 -.178 1 Vocabulary knowledge L2 -.108 -.360** -.220* -.410** .189 1 Working memory -.137 -.129 -.115 -.136 .199 .171 1 Inhibition .123 .051 -.055 .080 -.172 .088 -.124 1 Task switching -.065 -.084 .011 -.070 .089 .173 .069 .034 1
Note. N = 94, * p < .05, ** p < .01. L1 = main language, L2 = foreign language.
To test whether L1 sentence verification speed directly contributed to L2 sentence verification speed, a multiple regression was conducted with L1 sentence verification speed, L2 word recognition speed, L2 vocabulary knowledge, working memory, inhibition and task switching as predictors. To explore the unique explained variance of sentence verification speed in L1, in step 1 only sentence
verification speed in L1 was added to the model. In step 2, the additional variables were added to the model. The results of model 1 show that sentence verification speed in L1 explained a significant amount of variance in sentence verification speed in L2 (F(1, 92) = 70.96, p < .001, R2 = .435,
R2Adjusted = .429). The analysis shows that sentence verification speed in L1 significantly predicted the sentence verification speed in L2 (β = .66, t(92) = 8.42, p < .001). When the additional variables were added (model 2), the explained variance increased with .191, which was significant
(F(5, 87) = 8.92, p < .001, R2 = .627, R2Adjusted = .601). The analysis shows that L1 sentence
verification speed (β = .51, t(87) = 7.19, p < .001), L2 word recognition speed (β = .38, t(87) = 4.99, p < .001) and L2 vocabulary knowledge (β = -.15, t(87) = -1.99, p = .049) could significantly predict L2 sentence verification speed, while working memory (β = .01, t(87) = .21, p = .834), inhibition (β = -.03,
t(87) = -.42, p = .673) and task switching (β = .00, t(87) = .03, p = .978) could not. The unstandardized
and standardized regression coefficients of the fitted hierarchical regression model are presented in Table 7.
13
Table 7. Linear model of the predictors of speed of sentence verification in L2
B SE B β p Step 1 Constant 570.02 (204.78, 935.25) 183.90 .003 Sentence verification speed L1 (.27, .44) .36 .04 .66 .000 Step 2 Constant 348.19 (-573.96, 1270.33) 463.95 .455 Sentence verification speed L1 (.20, .35) .28 .04 .51 .000 Word recognition speed L2 (.75, 1.74) 1.25 .25 .38 .000 Vocabulary knowledge L2 (-19.91, -.03) -9.97 5.00 -.15 .049 Working memory 4.01 (-34.06, 42.09) 19.15 .01 .834 Inhibition -.30 (-1.71, 1.11) .71 -.03 .673 Task switching .01 (-.87, .89) .44 .00 .978
Note.N = 94. R2 = .435 for step 1 (p < .001), R2 = . 627 for Step 2 (p < .001), ΔR2 = .191 (p < .001) from Step 1 to Step 2.
L1 = main language, L2 = foreign language.
Discussion
The aim of the current study was to determine whether executive functioning contributes to reading fluency in both L1 and L2, in addition to the well-known contributors vocabulary knowledge and word recognition speed. To this end, the sample of the present study contained children of grade 4 of primary school, who were proficient readers of Dutch and who learned to read in a new foreign language, English. Correlations showed that in L1, reading fluency only correlated with word recognition speed and vocabulary knowledge, respectively, and that word recognition speed and vocabulary knowledge were uncorrelated, which was unexpected. However, a possible explanation for this result is that the frequency of items of the word recognition task significantly differed from the frequency of the words in the PPVT-NL. The words in the word recognition task were used more frequent then the words in the PPVT-NL. In L2, only reading fluency, word recognition and vocabulary knowledge were correlated. Furthermore, the executive functions did not correlate with each other. Although Miyake et al. (2000) stated that executive function is a factor model with working memory, inhibition and task switching as its components, other studies suggest that the individual tests measuring those components do not always correlate with each other (Bull & Scerif, 2001). This could explain why there was no correlation between the different executive functions. Furthermore, the executive functions did not correlate with any of the language skills in both languages.
Three multiple regression models were fitted, one on the L1 language skills and the executive functions, one on the L2 language skills and the executive functions, and one on the influence of L1 reading fluency on L2 reading fluency. The first model showed that while word recognition speed was a predictor of L1 reading fluency, vocabulary knowledge and the executive functions working
memory, inhibition and task switching were not. However, as an individual predictor or together with word recognition speed, vocabulary knowledge did influence reading fluency. A possible explanation could be that a part of the unique variance that could be explained by vocabulary knowledge in earlier steps of the model was shared with the executive functions. Furthermore, this result means that executive functioning had no direct effect on L1 reading fluency, which was unexpected. The executive functions did not have an indirect effect on L1 reading fluency as well.
The second model showed that only word recognition speed was a predictor of L2 reading fluency, while vocabulary knowledge and the executive functions working memory, inhibition and
14
task switching were not. Vocabulary knowledge only influenced reading fluency when it was the only predictor; combined with word recognition speed, vocabulary knowledge did not influence reading fluency anymore. A possible explanation for this effect is that the English sentences were very simple, and mostly consisted of cognates that resembled Dutch words. This is also shown by the significant correlation between L1 vocabulary knowledge and L2 reading fluency. This could make it hard to distinguish children with lower vocabulary knowledge from high vocabulary knowledge on speed measures. Although the correlation between L2 vocabulary knowledge and L2 reading fluency was significant, it was no significant predictor anymore when L2 word recognition speed was added to the model. This indicates that the variance that vocabulary knowledge could explain, was shared with word recognition speed, and that possibly, the L2 sentence verification task and the L2 word recognition task were to similar, which is also indicated by the strong correlation between those tasks. In addition, executive functioning had no direct or indirect effect on L2 reading fluency.
These results indicated that executive functioning had no direct or indirect effect on reading fluency in both L1 and L2. Since this result was found in both languages, it cannot be explained by the level of English of the children. It could be that the executive functions task that were used were too difficult for the children. The Simon and Switch effects were small, with large standard deviations. It could be that these verbal tasks were too difficult for the children and therefore the size of the effects differed between children. Furthermore, previous reliability research in the WISC manual showed that the working memory task we used, was less reliable in our age-range. These findings are inconsistent with Stanovich (1991), Carretti et al. (2009) and Cain (2006) who showed the link
between executive functions, word recognition and reading comprehension.
The third model showed that L1 reading fluency, L2 word recognition speed and L2 vocabulary knowledge were significant predictors of L2 reading fluency, but working memory, inhibition and task switching were not. The influence of L1 reading fluency on L2 reading fluency is in line with the interdependence hypothesis of Cummins (1979), which states that L2 reading competence is a function of L1 reading competence. Furthermore, these results are in accordance with linguistic coding differences hypothesis (Sparks & Ganschow, 1991; 1993; 1995), since L1 reading fluency explained the most variance in L2 reading fluency. This indicates that there might be a cognitive component that underlies L1 and L2 reading fluency.
The current study had a few limitations. First, almost two third of the participants had to be excluded because they had one or more missing scores on a subtest. Although the vocabulary knowledge did not differ between the included and excluded participants, it is unclear in which way the exclusion influenced the other variables. Future studies could use a multilevel design, in which data of a participant that has one or more missing test scores can still be used. Second, the reliability of the designed tests (word recognition speed and sentence recognition speed) was low. This could have influenced the results and that is why caution is needed in the interpretation of the regressions.
In conclusion, this study suggests that in children in grade 4, in both L1 and L2 reading fluency is influenced by word recognition speed, but not by vocabulary knowledge, working memory, inhibition and task switching. Furthermore, L1 reading fluency influences L2 reading fluency.
However, caution is needed and future research with more reliable tests should confirm the current results.
References
Baddeley, A. (1992). Working memory. Science, 255(5044), 556-559.
Baddeley, A. (2003). Working memory and language: An overview. Journal of communication
disorders, 36(3), 189-208.
Baddeley, A.D. & Hitch, G.J. (1974). Working Memory. In: G.A. Bower (Ed.), Recent advances in
15
Beck, I. L., Perfetti, C. A., & McKeown, M. G. (1982). Effects of long-term vocabulary instruction on lexical access and reading comprehension. Journal of Educational Psychology, 74(4), 506-521. Blom, E., Küntay, A. C., Messer, M., Verhagen, J., & Leseman, P. (2014). The benefits of being
bilingual: Working memory in bilingual Turkish–Dutch children. Journal of experimental child
psychology, 128, 105-119.
Bull, R., & Scerif, G. (2001). Executive Functioning as a Predictor of Children's Mathematics Ability: Inhibition, Switching, and Working Memory. Developmental Neuropsychology, 19:3, 273-293. Cain, K. (2006). Individual differences in children's memory and reading comprehension: An
investigation of semantic and inhibitory deficits. Memory, 14(5), 553-569.
Carretti, B., Borella, E., Cornoldi, C., & De Beni, R. (2009). Role of working memory in explaining the performance of individuals with specific reading comprehension difficulties: A meta-analysis.
Learning and individual differences, 19(2), 246-251.
Cartwright, K. B. (2015). Executive Function and Reading Comprehension. In Parris, S.R., & Headley, K. (Eds.), Comprehension instruction: Research-based best practices, (3rd Edition, pp. 56-71). New York/London: The Guilford Press.
Christopher, M. E., Miyake, A., Keenan, J. M., Pennington, B., DeFries, J. C., Wadsworth, S. J., Wilcutt, E., & Olson, R. K. (2012). Predicting word reading and comprehension with executive function and speed measures across development: a latent variable analysis. Journal of Experimental
Psychology: General, 141(3), 470-488.
Cummins, J. (1979). Linguistic interdependence and the educational development of bilingual children. Review of educational research, 49(2), 222-251.
European Council (2002). Presidency Conclusions. Barcelona European Council. 15 and 16 March. SN 100/1/02 REV 1. Accessed on 30 June 2017 on
http://www.consilium.europa.eu/uedocs/cms_data/docs/pressdata/en/ec/71025.pdf Fuchs, L. S., Fuchs, D., Hosp, M. K., & Jenkins, J. R. (2001). Oral reading fluency as an indicator of
reading competence: A theoretical, empirical, and historical analysis. Scientific studies of
reading, 5(3), 239-256.
Gladwin, T. E., Lindsen, J. P., & de Jong, R. (2006). Pre-stimulus EEG effects related to response speed, task switching and upcoming response hand. Biological Psychology, 72, 15-34.
Garon, N., Bryson, S. E., & Smith, I. M. (2008). Executive function in preschoolers: a review using an integrative framework. Psychological Bulletin, 134, 31-60.
Hedge, A. & Marsh, N.W.A (1975). The effect of irrelevant spatial correspondences on two-choice response-time. Acta Psychologica 39, 427-439.
Kidd, G.R., & Hulmes, L.E., (2015). Keeping track of who said what: Performance on a modified auditory n-back task with young and older adults. Frontiers in Psychology, 6, 978.
16
Linck, J. A., Osthus, P., Koeth, J. T., & Bunting, M. F. (2014). Working memory and second language comprehension and production: A meta-analysis. Psychonomic Bulletin & Review, 21(4), 861-883.
Meuter, R.F.I., & Allport, A. (2009.) Bilingual language switching in naming: asymmetrical costs of language selection. Journal of Memory and Language, 40, 25–40.
Miyake, A., Friedman, N., Emerson, M., Witzki, A., Howerter, A., & Wager, T.D. (2000). The unity and diversity of executive functions and their contributions to complex “frontal lobe” tasks: A latent variable analysis. Cognitive Psychology, 41, 49-100.
Monsell, S. (2003). Task switching. Trends in cognitive sciences, 7(3), 134-140
Nicolay, A. C., & Poncelet, M. (2015). Cognitive abilities underlying second-language vocabulary acquisition in an early second-language immersion education context: A longitudinal study.
Journal of Experimental Child Psychology, 115, 655-671.
Perfetti, C. A. (1999). Comprehending written language: A blueprint of the reader. In C. M. Brown & P. Hagoort (Eds.), The neurocognition of language (pp. 167– 208). Oxford: Oxford University Press.
Pikulski, J. J., & Chard, D. J. (2005). Fluency: Bridge between decoding and reading comprehension.
The Reading Teacher, 58(6), 510-519.
Poarch, G. J., & van Hell, J. G. (2012). Executive functions and inhibitory control in multilingual children: Evidence from second-language learners, bilinguals, and trilinguals. Journal of
experimental child psychology, 113(4), 535-551.
Rogers, R.D., & Monsell, S. (1995). Costs of a predictable switch between simple cognitive tasks.
Journal of Experimental Psychology: General, 124, 207– 231.
Segalowitz, N. & Frenkiel-Fishman, S. 2005. Attention control and ability level in a complex cognitive skill: attention-shifting and second language proficiency. Memory & Cognition, 33, 644–653. Simon, J.R., Acosta, Jr, E., & Mewaldt, S.P. (1975). Effect of locus of warning tone on auditory choice
reaction time. Memory and Cognition 3, 167-170
Sparks, R. L., & Ganschow, L. (1991). Foreign language learning differences: Affective or native language aptitude differences? The Modern Language Journal, 75(1), 3-16.
Sparks, R., & Ganschow, L. (1993). Searching for the cognitive locus of foreign language learning difficulties: Linking first and second language learning. The Modern Language Journal, 77(3), 289-302.
Sparks, R. L., & Ganschow, L. (1995). A strong inference approach to causal factors in foreign language learning: A response to MacIntyre. The Modern Language Journal, 79(2), 235-244.
17
Stanovich, K. E. (1991). Word recognition: Changing perspectives. In R. Barr, M. L. Kamil, P. B. Mosenthal, & P. D. Pearson (Eds.), Handbook of reading research (Vol. II, pp. 418–452). New York / London: Longman.
Swanson, H. L., & Ashbaker, M. H. (2000). Working memory, short-term memory, speech rate, word recognition and reading comprehension in learning disabled readers: Does the executive system have a role?. Intelligence, 28(1), 1-30.
Van Gelderen, A., Schoonen, R., De Glopper, K., Hulstijn, J., Snellings, P., Simis, A., & Stevenson, M. (2003). Roles of linguistic knowledge, metacognitive knowledge and processing speed in L3, L2 and L1 reading comprehension A structural equation modeling approach. International
Journal of bilingualism, 7(1), 7-25.
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(1), 19.
Verhagen, J., & Leseman, P. (2016). How do verbal short-term memory and working memory relate to the acquisition of vocabulary and grammar? A comparison between first and second language learners. Journal of Experimental Psychology, 141, 65-82.
Xiong, A, & Proctor, R.W. (2016). Decreasing auditory Simon effects across reaction time
distributions. Journal of Experimental Psychology: Human Perception and Performance, 42(1), 23-38.
