A dynamic approach towards dyslexia and reading: A
study on pink noise in L1 and L2 reading fluency
Myrte Anne Kristiane Maathuis
S1871358
Abstract
This small-scaled study has investigated the influence of dyslexia on reading fluency and reading comprehension in the first and second language. A self-paced reading task was conducted in the L1 and the L2 to study the reading process from a dynamic perspective. This entails that variability, which in conventional research is often considered to be random noise and therefore discarded or minimized, can provide valuable information about the underlying complex systems of language learning. The presence or absence of pink noise in the data series can tell us something about the degree of coordination and self-organization of the cognitive system. The computer software E-prime (2002) was used to measure response times and the variability patterns in these response times were analyzed for each individual. This was done using spectral analysis. Spectral slopes were an-alyzed and it was expected that compared to the nondyslexic learner, the dyslexic learner would show shallower slopes indicative of white noise, especially in the L2. The results showed that, con-trary to our expectations, in data series of poor readers with long response times pink noise can be present. Because this study was a case study, the results cannot be generalized for a larger popu-lation, but they do bring up the question whether the presence of pink noise is representative for reading fluency, and whether the self-paced reading task is suitable for distinguishing between good and poor readers.
Introduction
research, Vellutino, Fletcher, Snowling & Scanlon (2004) pointed out that “English presents a signif-icantly greater challenge to the beginning reader than other more regular alphabetic systems that contain consistent mappings between letters and sounds and are described as transparent (or shal-low) orthographies” (p. 17). For native speakers of Dutch, an orthographically transparent language, learning English as a second language can be a great struggle, especially spelling and pronuncia-tion. It can be assumed that English spelling and pronunciation is even more difficult for dyslexic learners. Helland and Kaasa (2005) and Miller-Guron and Lundberg (2000) examined the influence of dyslexia on L2 reading skills and found major differences within the dyslexic learner groups. Miller-Guron and Lundberg (2000) found that factors such as motivation and anxiety had a major influence on the preferences of individual learners. Lundberg (2002) also pointed out the importance of indi-vidual differences with regard to foreign language learning. Lundberg showed that several studies showed that reading in the L2 can be mastered fully by less experienced learners, and that these learners may even outperform native speakers of the L2 (Elley, 1994; Wagner, Spratt, and Ezzaki, 1989; cited in Lundberg, 2002). Lundberg (2002) presented a model which summarizes the factors that underlie the individual differences between learners. These findings by Helland and Kaasa (2005), Miller-Guron and Lundberg (2000) and Lundberg (2002) all emphasized the importance of the individual differences between (dyslexic) learners. Additionally, it needs to be taken into account that differences in the native and foreign languages can be of great influence on the language learn-ing development.
its components as well as external influences from the environment. This entails that individual fac-tors such as language aptitude, L1, age, motivation and impairments as well as education, resources and language teaching are essential to a learner’s development. It is very difficult to predict the course of individual language development. Language development is not a linear process; there is no end-state in the process which can show growth as well as decline. Well-coordinated complex systems tend to self-organize into attractor states in which the system components can function optimally.
and analyzed in the original order; one study investigated the influence of dyslexia on reading (Wijnants et al., 2012) and another study examined scaling relations in second language develop-ment (Lowie, Plat, & De Bot, 2014). These studies have provided evidence for the relevance and usefulness of non-linear analyses in language studies. They have also created awareness of the need for more non-linear studies on dyslexia and second language development.
This study aims to investigate what non-linear analysis of self-paced reading task response times can tell us about the individual differences between dyslexic and non-dyslexic learners with regard to reading fluency. It is predicted that this study will duplicate Wijnants et al.’s (2012) findings that showed convincing differences in variability patterns in the word naming task response time data of dyslexic readers and non-dyslexic readers. Where non-dyslexic readers’ variability patterns were indicative of well-coordinated underlying sub-systems, it was found that for dyslexic readers, the patterns were more random and less coordinated. In this study, a self-paced reading task was conducted to measure reading fluency, and it was carried out in two languages because we also wanted to be able to investigate how second language reading fluency is influenced by dyslexia and how this differs from the effects that dyslexia has in the native language. The tasks were carried out with a variety of learners: dyslexic, non-dyslexic, less experienced, beginning learners, and more experienced, advanced learners, to see for which type of learner which degree of coordination could be established. In addition, these results were compared to participants’ scores on more conven-tional reading fluency tests to see whether non-linear analysis yields similar results as the currently used tests for dyslexia diagnosis.
finally, the conclusion section which will present the main conclusions and recommendations for further research.
Theoretical Background
Reading development in the L1 and the L2
Learning how to read is essential to language learning. Several components are involved in reading ability. Vellutino (2004) tested a convergent skills model of reading development which included three major components of reading development: reading comprehension, context-free word identi-fication and language comprehension, which were in their turn connected to the following compo-nents: phonological coding ability, phonological awareness, phonological decoding and spelling, and visual coding ability, visual analysis, semantic knowledge and syntactic knowledge (p. 6). The find-ings were consistent with the model, which implicated that for less experienceder learners, context-free word identification and the phonological skills involved in this process are predictive of reading comprehension. For older learners, the relationship between reading comprehension and language comprehension skills was stronger. They also found that for both learner groups, semantic knowledge was important for reading comprehension, whereas syntactic knowledge was not. Read-ing comprehension has also been associated with the ability to construct and update mental models of situations; working memory capacity is considered to play a crucial role in this process (Barnes, Raghubar, Faulkner and Denton, 2014; Van der Schoot, Reijntjes, and Van Lieshout, 2014). Other research by Perez, Majerus and Poncelet (2011) has also linked verbal short term memory capacity with early reading skills. In their study, it was found that order short term memory capacity is predic-ative of long-term reading decoding abilities.
Yan, Siegel, Wade-Woolley, 2001). Miller-Guron and Lundberg (2001) pointed out that it is generally assumed that second language aptitude is dependent on L1 language aptitude and that because of this assumption problems in second language learning are often attributed to language L1 deficits (p. 42). Mastering these skills is more difficult in the L2 due to factors such as orthography and typology, which can differ greatly between the L1 and the L2 (Miller-Guron & Lundberg, 2000). Es-pecially for learners whose native language is orthographically transparent it can be very problematic to learn an opaque language such as English. Lundberg (2002) addressed the issue of whether there is a critical period for second language learning and argued that although evidence shows that older learners benefit from the advantage of having „greater cognitive maturity”, it is also the case that „[i]t seems, however, as if the phonological and prosodic aspects of L2 can be acquired with ease and efficiency at an earlier age” (p. 169). Older learners can benefit from acquired linguistic awareness and a good L1 proficiency, but less experienced learners are more likely to become native-like (Lundberg, 2002, p. 169). A case study by Dimroth (2008) provided evidence for the claim that language development of less experienced learners is different from that of more experienced learners with regard to order and rate of the acquisition process. Less experienced learners learn faster and are less dependent on conservative learning strategies compared to the more experi-enced learners, resulting in a different end product (p. 146-47).
Reading fluency is another important aspect of reading ability. In receptive reading, speed, accuracy and automaticity are important components. Reading speed is generally considered to reflect reading fluency as it is often used to operationalize the construct reading fluency. Words-per-minute is often the measurement used for reading fluency (Geva, 2000). In the Netherlands, the conventional reading fluency tests such as the One-Minute-Test measure accuracy and speed and use the words-per-minute rate to establish reading fluency. Taguchi, Gorsuch, and Sasamoto (2006) provided the following definition: „fluent readers are able to identify words in text quickly and accu-rately with a minimal amount of attention” (p. 1). In the reading process, lexical knowledge plays an important role. Lexical knowledge is needed to comprehend a (second) language. Conventionally, lexical knowledge is considered to be stored in the mental lexicon. Very influential was Levelt’s (1989) perception of the lexicon. In Levelt’s models of the lexicon, it included separate modules and was considered to be lexically driven. Levelt's (1999) model of speech production only accounts for the mental lexicon of the L1. The question whether there are separate L1 and L2 lexicons, or whether there is one big multilingual lexicon has been the subject of discussion for the past decades. Kroll and Stewart (1994) provided empirical evidence in favor of the separate L1 and L2 lexicon theory, but this study was criticized by Kroll and Dijkstra (2002) who proposed the idea of one bilingual lexicon instead. De Bot, Lowie, and Verspoor (2005) stated that although the idea of separate lexi-cons for L1 and L2 used to be the common assumption, more recently the paradigm has shifted from a static representation of the mental lexicon towards the activation metaphor, which implies that the lexicon is basically a network consisting of interconnected entries which „vary in their degree of activation” (p. 42). With regard to theories about the organization of the multilingual mental lexicon, De Bot et al. (2005) claimed that:
There is a clear need for a more dynamic model that can take into account all these different factors, some of which are continuously changing, In view of this observation, the activation metaphor seems to be the most at-tractive alternative, as the level of activation can continuously change for each individual lexical item. (p. 44-45)
is much more dynamic than is currently assumed by most bilingual models of speech production, and that the bilingual system can operate in different language activation states" (p. 1074).
Recently, the static, linear representation of language acquisition has been challenged by an emergentist, more dynamic approach towards language acquisition. Research by linguists who took a usage-based approach towards language development emphasized the importance of context in language learning. The lexicon, which had often been explained with the dictionary metaphor, a static collection of lexical entries in the mind, was now approached from a more dynamic perspective. It is presented as a network of episodic memory. Episodic or implicit memory entails that memories (and therefore lexical knowledge) are based on personal experiences rather than based on facts or acquired skills, which is the case for semantic or declarative knowledge (Tulving, 1983). In his chap-ter on episodic memory, Tulving (2002) pointed out the following dynamic characchap-teristics of memory systems:
Intervention with the operation of a even if it occurs through a single component unique to that system--affects all those learning and memory performances that depend on that system .... Different systems have emerged at different stages in the evolution of the species, and they emerge at different stages in the development of individual organisms. (Tulving, 1985, as cited in Tulving, 2002, p. 6)
Elman's findings. In one of his studies, Elman (2011) demonstrated the possibility of lexical knowledge without a lexicon. His dynamic approach to lexical knowledge emphasizes the im-portance of interaction between words and their environment and related to this, their context sen-sitivity.
Dyslexia
When a child’s failure to read without difficulty cannot be explained by their intelligence, learning environment or other known causes, it is often diagnosed with specific development dyslexia or, shortened, dyslexia (Snowling, 1987). The complexity of dyslexia makes it difficult to put forward a clear definition. According to the definition by the Dyslexia Institutes of America (DIA), is "a specific learning disability that is neurological in origin. It is characterized by difficulties with accurate and/or fluent word recognition, and by poor spelling and decoding abilities". In their studies on prevalence of specific reading retardation, Rutter et al. examined the intelligence and reading skills of a large group of children and were able to distinguish two groups of backward readers: those whose im-paired reading skills could be explained by their IQ scores, the generally backward readers, and those readers who underperformed on the reading test in comparison to their IQ scores, the specif-ically retarded readers (Rutter et al., 1970; Yule et al., 1974; as cited in Snowling, 1987). Snowling (1987) pointed out that the Isle of Wight studies have only investigated reading skills and that spelling is also an essential factor in language development which needs to be accounted for in research (Snowling, 1987). However, it has been found, as research by Boder (2008) has shown, that the correlations between reading and spelling are so consistent that reading and spelling skills can be considered to be predictive of one another. A common approach towards dyslexia is the deficit ap-proach which aims to identify the underlying cognitive deficits in dyslexia. As Morrison and Manis (1982) pointed out, the main problem with this approach is the fact that it can never be said whether any of these cognitive deficits is the cause or consequence of dyslexia (as cited in Snowling, 1987). Cossu (1999) argued that sparsity of data can lead to flawed conclusions due to failure to interpret correctly what he defined as „the highly specialized architecture of cognition and in particular the specificity of reading processes themselves” (p. 214).
the long-term memory. In a study on early reading development in children who have a familial risk for dyslexia by Pennington and Lefly (2001), it was found that letter-name knowledge or phoneme awareness as children become older is an important predictor of reading ability in less experienced children. Similar results were reported on in a study by Poualakanaho, Poikkeus, Ahonen, Tolvanen and Lyytinen (2004) in which it was emphasized that the assessment of phonological awareness can be carried out independently from general language skills. Catts, Adlof and Weismer (2006) argued that poor readers differ from poor comprehension in the fact that poor readers suffer from phonological processing and memory deficits whereas poor comprehends do not (p. 279). Where Catts et al. (2006) distinguished between poor reading comprehenders and learners with dyslexia, other research has shown that both groups show deficits in the working memory capacity. Van der Schoot et al. (2014) argued that poor reading comprehension skills are caused by readers’ inability to update the representation or situation model with current information from the key situational di-mensions involved in reading. They aimed to examine whether this inability appeared in the con-struction process or in the updating process. They did this by having participants read a story to which inconsistencies were added. Eye fixation and self-paced reading tasks were used to track readers’ reading process. They found that both types of readers read the texts with inconsistencies in the story line more slowly. In addition, they noticed that readers with good reading comprehension tended to revisit the parts of the story that contained the inconsistencies, whereas poor readers did not do this as frequently. The results from the comprehensions questions that followed the reading task showed that the poor comprehension readers made more mistakes. From their findings, they concluded that children with poor reading comprehension lack the working memory capacity to de-tect or keep track of inconsistencies in text. Taking into account these findings, it becomes apparent that, in contrary to Catts et al.’s (2006) findings, Van der Schoot et al. (2014) did find that there is an overlap in deficits between the poor comprehenders and the dyslexic learners.
out, the main problem with the deficit approach is the fact that it can never be said whether any of these cognitive deficits is the cause or consequence of dyslexia (as cited in Snowling, 1987). Cossu (1999) argued that sparsity of data can lead to flawed conclusions due to failure to interpret correctly what he defined as „the highly specialized architecture of cognition and in particular the specificity of reading processes themselves” (p. 214). Research by Matisse, French and Rapin (2008) provided evidence for the claim that dyslexia is not caused by a single cognitive deficit but that it involves multiple cognitive factors. This theory of multiple cognitive factors is in line with the DST approach, and it is an argument in favor of a more holistic approach towards language learning. It is also an argument in favor of a more individual-focused approach towards language learning; if multiple cog-nitive factors are involved in language learning and language impairments, the composition of these factors might differ from individual to individual.
Diagnosis of dyslexia
static measures differentiated between normal readers and dyslexic readers with high sensitivity and specificity.
From a component-dominant approach to a interaction-dominant approach
Conventionally, cognitive behavior has been studied from a component-dominant perspective. Influ-enced by the computer metaphor, cognition has been investigated as if it consisted of separate modules in which fixed procedures took place. As stated by Carello and Moreno (in Riley and Holden, 2005):
Linear interactions mean that the effect of an unobservable component can be recovered in an overall measure like response time because each component effect spans a sub-interval of response time. Interaction existed only between the output produced by separate modules. (p. 6)
Interaction was acknowledged to occur within the modules and between the output produced by these modules but interaction between different modules was not considered to be part of the pro-cess. Cognitive behavior was seen as the result of straightforward cause-and-effect relations. How-ever, this view is not shared by all researchers. Van Orden, Holden, & Turvey (2003) argued that „[c]omponent-dominant dynamics protect the integrity of component effects” (335), but that this ap-proach towards purposive behavior is too limited to account for the complexity of behavior. The cause-and-effect metaphor fails to include the self-organizing properties of complex systems and the ability of the system to select its own patterns.
among processes. It is the interactions rather than the interacting components which need to be studied. Several studies have reported on the studies of language development from this interaction-dominant approach (e.g. Van Orden, et al., 2003; Wijnants, Bosman, Hasselman, Cox, & Van Orden, 2009; Lowie et al., 2014). In a quantitative study investigating interaction-dominant behavior in hu-man cognition, Ihlen and Vereijken (2010) claimed that whereas before, there was no sufficient evi-dence for the ubiquity of pink noise in human cognition being proof of interaction-dominant dynamics, their study provided „quantitative support for a paradigm shift toward interaction-dominant dynamics in human cognition” (p. 436).
A Dynamic Systems Theory approach towards language development
The interaction-dominant approach is essential to Dynamics Systems Theory (DST), a theory which regards language acquisition as an ongoing non-linear process without a final stage or end goal in which both growth and decline occur. It is for that reason that Larsen-Freeman (1997), who was one of first to apply DST to language acquisition, replaced the term „acquisition” with „development”. Language development is regarded to change over time due to complex interactions between both internal and external resources (Verspoor, De Bot, and Lowie, 2011). The focus of study is therefore not on the end product but on the developmental processes which change over time.
environment without sacrificing too much flexibility” (p. 88). The ability to self organize allows the complex system the flexibility to adapt to its environment. According to Van Orden et al. (2003), „ [i]n living systems, criticality itself emerges spontaneously, self-organized criticality (Bak, 1996). Living systems self-organize to stay near critical states” (p. 333). Self organized criticality (SOC) is the optimal condition for a complex system in which flexibility and stability are perfectly balanced. There are different types of attractor states. Shallow, less stable and thusly more flexible attractor states allow for the system to adapt more easily to occurring changes in the environment whereas deep attractor states that show great stability lose flexibility and might subsequently lose the ability to adapt to new changes, causing the ongoing processes to stagnate.
Dynamical patterns in cognitive behavior Van Orden et al. (2005) predicted that:
(a) cognition—whatever its nature— does not divide into statistically independent processes, and (b) the same processes govern cognitive performance in very short and very long time frames (p. 122)
Van Orden, Kloos and Wallot (2009) showed that dynamic patterns in variability across repeated measures can be found in human behavior. According to Verspoor et al. (2011) „the degree of vari-ability in any particular sub-system is seen as an intrinsic property of the developmental process” (p. 65). By analyzing patterns of variability in the order in which they occur, insight into the dynamics of development can be gained. Van Orden, Holden, and Turvey (2003) claimed that:
Fractals are organized in such a way that they are self-similar; structurally, each smaller part resem-bles the whole. Fractals frequently occur in nature; earthquakes, coastlines, rivers, and mountains have fractal properties. Daw (2011) claims that "trial-by-trial analyses are particularly suitable to de-veloping a more detailed and dynamic picture of learning than was previously available” (p. 2). By keeping the time-order of of repeated measurements intact, scaling relations can be discovered that represent optimal self-organization, which is manifested in Pink Noise or 1/f scaling1. 1/f scaling
entails that frequently occurring fluctuations have a low amplitude, and high amplitude fluctuations have a low frequency. As mentioned above, research has shown that this optimal condition, or pink noise, can be detected in a variety of human behavioral tasks (Van Orden et al., 2003; Wijnants et al., 2009; Lowie et al. 2014). Wijnants et al. (2009) give the following definition of 1/f scaling:
1/f scaling indicate[s] that the magnitude of variation in response latencies is proportional to the timescale on which it is measured, thus composing a complex sequence effect spanning over the entire time course of an experiment. (…) It follows that a 1/f scaling relation can be expressed as a relation between the size of changes (power) and how often changes of that size occur (frequency), which is inversely proportional on logarithmic scales. (3)
As phrased by Liebovitch and Shehadeh (in Riley and Van Orden, 2005) „instead of measuring a property (the kinetic rate constant) at one time scale, we measure how a property (the effective kinetic rate constant) changes when we measure it at different time scales” (213). They claimed that:
For a fractal object that extends over many scales, in space or time, a property depends on the scale at which it is measured. There is no one measurement that best describes the object. The object is best described by how the property measured depends upon the resolution at which it is measured. This relationship is characterized by a parameter called the fractal dimension. The fractal dimension can be calculated from the slope of this loga-rithmic-logarithmic graph. (219)
Spectral analyses can be carried out to examine how patterns change over time. Plotted on a log-log scale, pink noise will result in a negative slope between 0 and -1. A slope close to zero would represent Gaussian or white noise. Randomization of the arrangement of trial-to-trial series would
Dynamic measures of dyslexia
dyslexia in the L2 because it managed to reduce the influence of L2 vocabulary and the amount of schooling. The origin of dyslexia, however, is still subject to discussion. Wijnants et al. (2012), pointed out that poor phonological skills are commonly thought to underlie dyslexia but that other theories have emphasized the importance of morphological awareness to reading acquisition and others regarded visuo-attentional, auditory deficits or motor problems to contribute to dyslexia. Wijnants et al. (2012) claimed that because multiple studies have shown that there is such a great variety of cognitive components and processes that may be of contribution to dyslexia the „intrinsic dynamics of the components themselves may matter less than the mutual interdependence among those components” (p. 2). They therefore took a holistic, interaction-dominant approach towards dyslexic and non-dyslexic word-naming performance in beginning readers and studied the back-ground noise of the response times. They found that the dynamic measures they used, 1/f scaling and recurrence quantification analysis, are useful in differentiating between dyslexic and non-dys-lexic readers. They also found strong correlations between the complexity measures and the severity of the participants' reading deficiency (Wijnants et al., 2012). These dynamic measures showed strong correlation with the static measures which are conventionally used to diagnose dyslexia. Statement of purpose
useful and reliable information about the influence of dyslexia on the reading process. Research by Lowie et al. (2014) has shown that spectral analysis can provide valuable information about L1 and L2 processing. This research on dyslexia will investigate and compare the patterns of variability in the response times of in self-paced reading tasks performed by a dyslexic less experienced language learner, non-dyslexic less experienced language learners and non-dyslexic more experienced lan-guage learners. The participants will perform the self-paced reading tasks in their L1, Dutch, and the L2, English, a language which the less experienced learners have only just started to learn and the more experienced learners are expected to have acquired up to a decent level as they are currently university students. Conventional reading fluency tests are carried out with two of the less experi-enced learners, the dyslexic learner and a non-dyslexic learner, to set a baseline measurement with the non-linear results can be compared.
Research Question
The main research question in this study is: what can non-linear measures tell us about reading processes in general, and the influence of dyslexia on the reading process in the L1 and the L2? Do different types of learners display different patterns of variability? Non-linear spectral analyses were used to analyze the self-paced reading task response times in order to examine whether a) differ-ences in variability patterns are found between the different types of learners b) there are differdiffer-ences to be found between variability patterns in L1 Dutch, an orthographically transparent language, and the L2 English, an orthographically opaque language for different types of learners and c) the results on the L2 self-paced reading task are different for the less experienced learners and the more expe-rienced learners.
pseudo word test and the L2 word tests the prediction is that the non-dyslexic learner will outperform he dyslexic learner. Correlations between the reading fluency tests and the reading comprehension tests are predicted based on the findings in earlier research.
Method
Participants
This study has taken into account Vellutino’s (1979) remarks that „dyslexia is best studied in children who have average or above average intelligence, intact (or corrected) sensory acuity, no severe neurological damage or other debilitating physical disabilities, and who have not been hampered by serious emotional or social problems, socioeconomic disadvantage or inadequate opportunity for learning (as cited in Snowling, 1987, p. 13). These conditions have therefore been met in all partici-pants. This study contained seven participants in total (n = 7). Four of the participants in this study are in their first years of regular Dutch secondary schools. They are aged between 11 and 14 year. In addition, two more experienced university students were tested for comparison. Precise ages are not known for these participants.
Two of the participants, Participant dyslexic learner (DL) and Participant non-dyslexic learner (NDL), have been selected on the basis of their scores on the Cito test conducted in the last year of primary school. Both participants scored between 525 and 530 (out of 550). One of the participants has been diagnosed with dyslexia2. The score percentage on the language component of the Cito
test is evidence of the learning impairment; he scored below 10 percent on this part of the test. He has no record of other learning impairments. The other participant scored above 49 percent on the language component of the Cito test and has no record of dyslexia or any other learning impairment. The test results of these two participants will be most suitable for analysis because conditions have been controlled for these two learners. The four other participants will also be used for comparison. Measures
In this research, the independent variables are the diagnosis dyslexic or non-dyslexic, proficiency level: beginners and advanced learners, and age: less experienced and more experienced learners.
2 Either in primary school or during the first year of secondary school tests were conducted. Either way, dyslexia had been officially
These last two independent variables correlate with each other because the beginning learners in this study are less experienced and the advanced learners are more experienceds. These variables will therefore be treated as one. To be able to examine the complex process involved with reading, two constructs were selected. What we want to investigate in the reading process has been divided into two constructs: reading fluency and reading comprehension. These constructs are operational-ized in two different sets of tasks. The fluency tests that were conducted measure reading fluency as a score of words per minute. The self-paced reading tasks are used to measure reading fluency operationalized as response times in milliseconds. Additionally, as a control variable reading com-prehension is tested in the self-paced reading task by a set of multiple choice questions. The con-struct of reading comprehension is therefore operationalized as a score of correct answers in per-centage.
Receptive reading fluency test: the self-paced reading tasks
readers (Hales, 1994; Huntingdon & Bender, 1993; Paget & Reynolds, 1984). In order to assure that learners did read the text and not just randomly pressed the space bar without reading, a selection of reading comprehension questions were asked after each text. Readers could only answer these questions if they had read the entire story.
Productive reading fluency tests
In addition, several reading fluency tests were conducted to investigate participants’ performances with regard to productive reading speed and accuracy on the following language aspects: (second) language reading, text reading, word reading fluency and phonemic awareness. A set of tests was used that is conventionally used diagnose dyslexia. A One-Minute-Test (EMT) was conducted in the L1 (Dutch) and the L2 (English) to examine word reading fluency. Participants were asked to read out as many words as possible within the timespan of one minute. Because the use of English loan words is common in Dutch, an English loan words test was carried out. In addition, the Klepel A test was conducted to measure participants’ reading fluency and accuracy of pseudo-words.
Procedure
Participants were orally instructed during the experiment. Written instructions were also provided by the self-paced reading task, but only after everything had been explained to the participants. Partic-ipants were instructed to take their time and it was emphasized that they were not under any time pressure. Participants were assured to ask questions before, during or after the task if there was anything that was unclear. The tests were conducted inside the school environment. Since the tests were conducted by the students’ own English teacher, the participants were explicitly told that they would not be graded on their performance.
Analysis
each story were then analyzed using spectral analysis. The frequencies and amplitudes were pre-sented on a log-log scale. The slope was calculated to be able to "capture the relation between amplitudes and frequencies of variation in the data signal" (Wijnants et al., 2009, 85). A slope close to 0 signifies white, random noise, a slope close to -1 is indicative of pink noise, and a slope close to -2 indicates brown, rigid noise.
Descriptive analysis were carried out on the data gathered from the productive reading flu-ency tests (the EMTs, the Klepel test and the English loan words test) because in this case study, complete data sets were only available for two of the participants. With regard to the EMTs in English and Dutch, the raw scores of the participants were compared. These scores are based on the num-ber of words read correctly within the time limit of one minute. The Klepel test results have been standardized according to the Stanine scale, and the standardized scores on the Klepel A pseudo-words test as established by Van den Bos, Lutje Spelberg, Scheepstra, and De Vries (1994).
Results
Receptive reading fluency test: the self-paced reading tasks
An overview of the means and standard deviations of the response times of all participants and the less experienced learners and more experienced learners respectively in the L1 and the L2 can be found in Table 1. An overview of the group averages and standard deviations can be found in Table 2. L1 MEAN SD L2 MEAN SD Partici-pant NDL 456.78 219.87 629.87 342.30 Partici-pant NDL 608.72 771.83 531.29 306.09 302 601.42 270.77 528.42 215.27 316 424.86 86.41 424.35 85.28 401 458.17 320.83 468.59 304.85 402 241.50 39.14 307.35 57.34 403 333.06 94.92 321.68 109.31 MEAN 446.36 257.68 458.79 202.92
Group MEAN SD
L1 task 446.36 257.69
L2 task 458.79 202.92
Less experienced learners
L1 task 522.95 377.22
L2 task 528.48 237.24
More experienced learners
L1 task 344.24 151.63
L2 task 365.87 157.17
Table 2. Overview of group means and standard deviations of participants’ response times on the self-paced reading task for the L1 and the L2 for the group total and the separate less experienced learners and more experienced learners group.
It can be seen that on average, the participants did not read faster in the L1 than they did in the L2. When divided into two groups: beginning learners (Participant NDL, Participant NDL, participant 302, participant 316) and advanced learners (participant 401, participant 402, participant 403), it was found that the advanced learners did read faster than the beginning learners. The difference between the average response times of the beginning learners and the advanced learners turned out to be significant (t(4) = 2.26; p < .05).
Spectral Analysis
Group differences
An overview of the power spectrum on log-log scales of all participants can be found in Appendix C. The spectral slopes of all individuals and the overall averages in the L1 and the L2 can be found in Table 3. The group average slopes can be found in Table 4.
Table 3. The outcomes of the spectral analysis: slopes for the response time data series of the self-paced reading task in the L1 and the
L2.
Group L1 L2
Beginning learners -0.13 -0.30
More experienced learners -0.16 -0.28
Table 4. The outcomes of the spectral analysis: group average slopes for the response time data series of the self-paced reading task in the L1 and the L2.
When looking at the spectral slopes in L1 self-paced reading task response times data series, it was observed that on average, the advanced learners showed steeper spectral slopes in comparison with the beginning learners (M = -0.16, SD = 0.09 versus M = -0.13, SD = 0.09 respectively), but a t-test revealed that this difference was not statistically significant. With regard to the L2 self-paced reading task response times, the average spectral slope found in the data series of the beginning learners was slightly steeper than the average spectral slope of the advanced learners (M = -0.30,
Participant Spectral Slope
SD = 0.15 versus M = -0.28, SD = 0.23 respectively), but a t-test revealed that this difference was not statistically significant. On average, the group of beginning learners showed steeper slopes in the L2 (M = -0.30) than in the L1 (M = -0.13), but a t-test revealed that this difference was not significant either. For the advanced learners, the same observation was made; learners showed steeper spectral slopes in the L2 in comparison to the L1 (M = -0.28 versus M = -0.16), but a t-test proved that this difference was not statistically significant either.
Individual differences
As can be read in Table 1, the steepest slope found in the L1 self-paced reading task data series belonged to participant 316, a beginning learner. The steepest spectral slope in the L2 data series belonged to participant 402, an advanced learner. A series of t-tests were conducted to see whether there were significant differences within-subjects or between-subjects with regard to the response times on the L1 and the L2 task. Two of the participants read significantly faster on the L2 self-paced reading task than on the L1 self-paced reading task; on average, this difference turned out to be statistically significant for Participant NDL (t(1022) = 2.11; p < 0.05 and participant 403: (t(1022) = 1.78; p < .05). For one of the participants that read faster than the dyslexic participant (Participant NDL) on the L1 self-paced reading task and a t-test revealed that on average, this difference turned out to be statistically significant (t(1022) = -3.06; p < .005). T-tests showed that the other participants did not read significantly faster on either of the two self-paced reading tasks and except from partic-ipant 316, none of the particpartic-ipants outperformed the dyslexic particpartic-ipant.
Productive reading fluency tests
English loan words
Klepel A test EMT L1 raw scores
EMT L2 raw scores
Participant DL average below average (-2) 76 43
Participant NDL very good above average
(+4)
88 85
Table 5. Standardized scores on the productive reading fluency tests: the English loan words test and the Klepel A test and raw scores
on the L1 EMT and the L2 EMT.
Discussion
better on the self-paced reading tasks and display more pink noise than the less experienced learn-ers. This expectation was based on the fact that these learners have had more years of (foreign) language learning and therefore a higher proficiency level in the L1 as well as the L2 can be as-sumed. In this study it was found that dyslexia, age and selected language did not have as much effect on the appearance of pink noise in the data series as expected. In this study, in which self-paced reading task response times were used for spectral analysis, there was no clear relationship between productive reading fluency tested by the one-minute-tests and receptive reading fluency tested by the self-paced reading task or age or proficiency level and receptive reading fluency. This raises the following questions: a) What does pink noise say about dyslexia b) what does pink noise tell us about reading fluency c) is the self-paced reading task used in this experiment suitable for testing reading fluency for dyslexic learners? and if not, how can it be improved?
Pink noise and dyslexia
Wijnants et al. (2012) found strong correlations between pink noise patterns and the severity of dyslexia. In our case, the findings differed from the expectation that similar correlations would be found. In this case study, based on the spectral analysis, it appeared that for Participant NDL, dys-lexia did not indicate less optimal reading processing. In fact, his cognitive behavior tended towards pink noise which would suggest that the cognitive system is fully able to self-organize and coordinate the processes involved this reading task. This was in complete contrast with the linear analysis of the self-paced reading task in which Participant DL was clearly the slowest reader. To be able to explain this, we need to examine the reading fluency results further.
Productive reading fluency tests
that the dyslexic Participant DL gained an average score on the English loanwords test, and per-formed below average (-2) on reading pseudo or nonsense words (the Klepel test) whereas Partici-pant NDL scored very well on the English loan words test and performed above average (+4) on the Klepel test. So far, these results fit in with our prediction that Participant NDL’s reading fluency is lower than that of a non-dyslexic learner. The L1 EMT raw score of Participant DL did not differ greatly from Participant NDL’s scores (76 versus 88 respectively) but the L2 EMT raw score did (43 versus 85 respectively). This is also in line with our expectations; Participant DL was expected to score lower on the L2 EMT both compared to Participant NDL’s scores on the L2 EMT as well as in comparison with his own scores on the L1 EMT. For Participant NDL, there is hardly any difference between his score on the L1 EMT and his score on the L2 EMT (88 versus 85 respectively). Based on these results, it could be expected L2 reading might not be very problematic for him. For Partici-pant NDL, the results indicate that his reading skills in L1 might be average but that L2 reading is probably more problematic. On the basis of these results, one would expect similar results for the self-paced reading task response times; Participant NDL was expected to display little difference between the tasks in the two different languages whereas for Participant DLit was expected that the English reading task will be more challenging. It was expected that Participant NDL’s response times would be lower than Participant NDL’s performance on the self-paced reading task. It was predicted that this would show in the spectral analysis and that Participant NDL’s data point series would show steeper spectral slopes than Participant NDL’s, and in addition, it was hypothesized that Participant DL would display less pink patterns and shallower slopes in the L2 compared to the L1. This was, however, not the case.
Receptive reading fluency tests
caused by outliers (data points more than three times the SD) but as these peaks might be part of the underlying reading process, these data points have not been removed from the data because they may tell us something about the reading process. The expectation that there is a correlation between reading fluency and response times in a reading comprehension task is not fulfilled in this study. The main difference between the one-minute tests and the self-paced reading tasks is the type of performance the learners have to deliver; in the one-minute tests the learners have to read out loud, a productive task whereas in the self-paced reading task only their receptive skills are tested. Research has shown that phonological awareness is often impaired in dyslexic learners, implicating poorer judgment of initial sounds which results in less accurate judgment (Desroches, Joanisse, & Robertson, 2006). This impairment is even more likely to influence dyslexic learners’ performances on a reading-out-loud test because this is a more complex task; learners do not only have to be able to recognize sounds but they also have to apply and produce these sounds, a pro-cess in which meta-phonological awareness is needed.
Differences in response times between the less experienced learners and the more experienced learners
The hypothesis that more experienced, advanced learners have a higher reading fluency than the less experienced, beginning learners was confirmed. It was found that on average, the university students did read faster than the secondary school students. These results were found for the self-paced reading task conducted in the L1 and for the self-self-paced reading task conducted in the L2. So far, his confirms our hypothesis.
L1. This is an unexpected outcome. This may be due to the fact that their current learning environ-ment, university, exposes the more experienced, advanced learners to the English language on a daily basis. Academic reading skills are required for most university courses in the L1 and in English. University students might even read more frequently in the L2 than they do in the L1. This might have been a factor in the fast response times of the more experienced, advanced learners. This leads us to the question whether this result can also be found in the spectral analysis. This was not the case; similar results were not found in the spectral slopes calculated from the variability patterns. In fact, on average, there was no significant difference between the spectral slopes found in the variability patterns of the response time data series of the different types of learners.
Fractal patterns
Taking these results into consideration, the most important question that arises is how we can ex-plain the difference between the reading fluency scores and the variability patterns. With regard to the group averages—although these groups were so small that ‘individuals with similar background characteristics’ would be more appropriate—it has already been mentioned that although for both age groups the average response times were faster in the L1, for both of the groups, there was no significant difference between the average spectral slopes found in the variability patterns of the response time data series in the L1 and in the L2. This entails that the performances on the tasks were equally well-coordinated in the L1 and the L2.
tasks (L1 and L2) of all participants. It seems to be the case that response time speed does not necessarily determine the spectral patterns found in the data either; in both the L1 and the L2 task, the slower reader displayed more pink noise patterns than the faster reader. With regard to the L2 tasks, Participant NDL outperformed Participant DL in terms of response time. In the L2, Participant DL is a slow reader, which can be expected of a dyslexic reader, and yet his L2 fractal pattern is clearly indicative of pink noise. Only one other participant showed a steeper spectral slope; partici-pant 402 showed a slope of -0.54 on the L2 reading task. This participartici-pant was an more experienced, advanced learner, participant 402, who also had the fastest response times on the L1 self-paced reading task and the L2 self-paced reading task.
process does not suffer from his dyslexia? Or was it the difference in the task set up that was crucial? The self-paced reading task was performed on a computer which showed the reader one word at a time. This reading experience is very different from reading lists of words that are to be read vertically rather than horizontally, which makes for awkward unnatural reading circumstances. Perhaps the fact that only one word at a time was displayed is actually a method of reading from which dyslexic readers can benefit. According to Just, Carpenter, and Woolley (1982), data results from self-paced reading tasks are comparable to data results from tasks in which the eye-tracking method is used because similar effects were observed on word-level. They found that readers process text not per sentence but word-by-word as soon as these words are encountered. Word function, frequency and length are of influence on readers’ response times. Research by Rinck, Gámez, Díaz, and De Vega (2003) brought forth similar results for sentence-by-sentence self-paced reading compared to the eye tracking method. In our self-paced reading task, the readers were only exposed to one word at a time and although for normal readers it might be a disadvantage that the task requires pressing a button for each word, which might obstruct the natural reading process (Rayner, 1998), this might just provide dyslexic readers with the extra time they need for processing the words. We have meas-ured reading fluency in multiple ways, but because our reading comprehension measurements were not very elaborate, it is difficult to confirm that all participants fully understood the storyline.
Reading fluency in the L1 and the L2
reading task data series for the L2, in which the spectral slope from Participant NDL’s data series was steeper than the spectral slope for two of the university students (spectral slopes of -0.44 for Participant DL and -0.10 and -0.20 for the university students respectively were measured. One could argue that less experienced learners might still possess the ability to learn a language implicitly and only be able to use activate that knowledge in receptively, but as Granena (2013) pointed out, this is only the case for learners at a very less experienced age. As learners become older, they start to rely on cognitive aptitude more and more. Granena (2013) did stress the important role that is played by individual differences in learning ability and cognitive aptitude, which can be predictive of later language acquisition and pointed out that „early learners’ L2 attainment is characterized by high interindividual variability” (p. 673-74). Our findings were in line with this phenomenon of high interin-dividual variability; both our dyslexic and or non-dyslexic participant did not performed as expected based on the results found in other, larger sized studies.
Limitations of the study and recommendations for future research
An important limitation of this study is the number of participants; a study this small scaled cannot be generalized because it is not representative of a larger population. Hardly any statistical analyses could be carried out, and therefore the conclusions from this research are tentative. However, the results are useful for follow up work in the field of research. Although the use of non-linear measure-ments has become more common in the field of cognitive linguistics, very few of these studies con-tained both the use of self-paced reading tasks for data collection and less experienced learners as participants. It is therefore difficult to compare the results of this small-scaled study with other em-pirical findings. It would have been very interesting to compare our findings for the dyslexic partici-pant with other dyslexic learners from different age categories. It would also be interesting to meas-ure how participants would perform on other languages next to Dutch and English. English is an orthographically opaque language and it might be the case that the orthographic nature of a lan-guage has a large influence on productive reading fluency tasks but not so much on receptive read-ing fluency tasks.
this way of reading has its limitations. Rayner (1998) argued that noncumulative self-paced reading is problematic because it does not allow readers to reread the text. In our study, we did not test the learners under „natural” reading conditions so it is hard to say whether the fact that our self-paced reading tasks were non-cumulative was problematic. Even if the reading tasks had been conducted under „natural” reading conditions, it would nonetheless be difficult to compare self-paced reading data to normal reading data because one of the main problems of self-paced reading is the fact the demands of the task, pressing a button to read, slows readers down. However, this does not matter for non-linear analyses; spectral analysis analyzes the data points in relation to the other data points within a subject. A recommendation for future research would be to conduct different types of reading tasks. An advantage of conducting two reading tasks under different reading circumstances would be that it would enable us to compare the variability patterns in the data series and see what effects different task set ups have on the reading processes involved. It would be interesting to compare word-by-word reading to sentence-by-sentence reading because, as pointed out by Wallot and Van Orden (2011), the characteristics of reading word-by-word may contribute to disorder in the data: „These aspects of self- paced reading may be sources of randomness” (p. 253). Nevertheless, it needs to be pointed out that for our dyslexic reader, the self-paced reading task yielded in unex-pected 1/f scaling. Lowie et al. (2014) also address the ‚whitening’ effect that self-paced reading tasks may have, but they argue that simple key pressing tasks are less likely to be affected by un-certainty because they are more predictable tasks than response tasks.
A recommendation for future research would be to alter the way in which reading compre-hension was measured; in this study, only a set of multiple choice questions was used to determine whether or not readers had fully understood the texts. If the students were to have to summarize the stories in their own words in the L1, this would give a much more detailed and accurate indication of their comprehension of the task. Reading comprehension is perhaps even more important than read-ing fluency when learnread-ing outcomes are taken into consideration. Especially since our results indi-cated that poor reading fluency in terms of speed can still be the product of an underlying well-coordinated system.
average, less experienced learners and more experienced learners read faster in the L1 than they do in the L2. This difference in reading fluency turned out not to be statistically significant, but with larger group sizes a statistically significant difference between the two languages can be expected. Second of all, it was found that on average, more experienced learners read faster in the L21 as well as in the L2 than the less experienced learners do, but again, probably due to the small group sizes, this difference was not statistically significant but it can be expected that it would be the case with larger group sizes. However, our findings did not necessarily correspond with the presence of pink noise found in the variability patterns of these readers. On average, pink noise was more pronounced in the L2 data series than it was in the L1. This result was observed in both groups of learners and in both types of learners, dyslexic and non-dyslexic.
Conclusion
This study aimed to investigate whether spectral analysis of variability patterns are suitable for meas-uring reading fluency in the L1 and in the L2. It was argued that language development is a dynamic system and that the reading process can be regarded a complex system. Complex systems tend to self-organize, and in the variability patterns of well-coordinated systems there is a presence of fractal patterns, pink noise. A self-paced reading task was conducted to measure response times and these response times were analyzed and the findings of this study showed unexpected outcomes. Tradi-tional analyses showed that, as expected, on average, all participants read the L1 task faster than they read the L2 task. However, the expectation that the non-linear analysis would align with these findings was not met; on average, the variability patterns found in the L2 self-paced reading tasks showed more recognizable patterns of pink noise than the variability patterns found in the L1 self-paced reading task. In addition to that, when looking at the individual participants, it was found that the dyslexic reader showed pronounced patterns of pink noise in the L2 data series but not in the L1. The non-dyslexic learner who scored above average on the productive reading fluency tests and read significantly faster on the L1 self-paced reading task than on the L2 self-paced reading task did not show any signs of pink noise in either of the languages.
slow readers, and less experienced learners as well as more experienced learners can display fractal patterns in the L2, even if they perform below average on traditional reading fluency tests. What these patterns can say about language development needs to be investigated with a bigger sample size in order to be able to draw reliable conclusions. However, what these results have shown is that results gather from measurement of response times on a language task may not be representable of the underlying complex cognitive system. This case study has shown that as Wijnants et al. (2012) showed, the presence of pink noise in data can say something about reading abilities and in the case of dyslexic readers; it can reveal information about the severity of the impairment, but it can also contradict the findings of more traditional measurements and reveal differences between these scores and the underlying complex systems of language development. Additionally, it should be mentioned that our unexpected results might have been caused by the effects of the self-paced reading task setup. It could be that the setup of the reading task enhances the reading process for dyslexic readers; word-by-word reading omits a lot of distraction caused by other words in the texts. This is definitely worth looking into as dyslexic readers might benefit from this in everyday life.
References
Barbiero, C., Lonciari, I., Montico, M., Monasta, L., Penge, R., Vio, C., Tressoldi, PE., Ferluga, V., Bigoni, A., Tullio, A., Carrozzi, M., & Ronfani, L. (2012). The submerged dyslexia iceberg: How many school children are not diagnosed? Results from an Italian study. PLoS ONE, 7(10): e48082. DOI: 10.1371/journal.pone.0048082.
Barnes, M.A., Raghubar, K.P., Faulkner, H., & Denton, C.A. (2014). The construction of visual–spatial situation models in children’s reading and their relation to reading comprehension. Journal of Experimental Child Psychology, 119, 101–111.
Boder, E. (1973). Developmental dyslexia: A diagnostic approach based on three atypical reading-spelling patterns. Developmental Medicine & Child Neurology, 15(5), 663–687.
Breznitz, Z. (1997). Effects of accelerated reading rate on memory for text among dyslexic readers. Journal of Educational Psychology, 89(2), 289-297.
Brus, B. T. & Voeten, M. J. M. (1999). Eén-minuut-test: Vorm A en B: Verantwoording en handleiding: Schoolvorderingentest voor de technische leesvaardigheid, bestemd voor groep 4 tot en met 8 van het basisonderwijs [One-minute-test: Version A and B. Justifica tion and manual: Test for progress at school regarding decoding skill, from Grade 4 through Grade 8 of primary school]. Lisse, the Netherlands: Swets & Zeitlinger.
Carroll, J.M. & Iles, J.E. (2006). An assessment of anxiety levels in dyslexic students in higher education. British Journal of Educational Psychology, 76, 651–662.
Catts, H.W., Adlof, S.M., & Weismer, S.E. (2006). Language deficits in poor comprehenders: A case for the simple view of of reading. Journal of Speech, Language & Hearing Research, 49, 278-293.
Chomsky, N. (1965). Aspects of the theory of syntax. Cambridge, MA: MIT Press.
Daw, N. D. (2011). Trial-by-trial data analysis using computational models. Decision Making, Affect, and Learning, 1-26.
Desroches, A.S., Joanisse, M.F., & Robertson, E.K. (2006). Specific phonological impairments in dyslexia revealed by eyetracking. Cognition, 100, 32-42.
Dimroth, C. (2008). Age effects on the process of L2 acquisition? Evidence from the acquisition of negation and finiteness in L2 German. Language learning, 58(1), 117 -150.
Ditman, T., Holcomb, P.J., & Kuperberg, G.R. (2007). An investigation of concurrent ERP and self-paced reading methodologies. Psychophysiology, 44, 927–935.
Elbro, C., Daugaard, H.T., Gellert, A.S. (2012). Dyslexia in a second language? A dynamic test of reading acquisition may provide a fair answer. Annals of Dyslexia, 62(3), 172-85.
Elley, W.B. (Ed.). (1994). The IEA study of reading literacy: Achievement and instruction in thirty-two school systems. Oxford: Pergamon.
Ellis, N.C. (2008). The dynamics of second language emergence: Cycles of language use, language change, and language acquisition. Modern Language Journal, 92(2), 232-249. Elman, J.L. (2004). An alternative view of the mental lexicon. TRENDS in Cognitive Sciences 8(7),
301-306.
Elman, J.L. (2011). Lexical knowledge without a lexicon? The Mental Lexicon, 6(1), 1-33.
Geva, E. (2000). Issues in the assessment of reading disabilitird in L2 children—Beliefs and research evidence. Dyslexia, 6, 13-28.
Gottardo, A., Yan, B., Siegel, L.S., & Wade-Woolley, L. (2001). Factors related to English reading performance in children with Chinese as a first language: More evidence of
cross-language transfer of phonological processing. Journal of Educational Psychology, 93(3), 530-542.
Granena, G. (2013). Individual differences in sequence learning ability and second language acquisition in early childhood and more experiencedhood. Language learning, 63(4), 665-703.
Grigorenko, E.L. (2001). Developmental dyslexia: an update on genes, brains, and environments. Journal of child psychology and psychiatry 42(1), 91 -125.
Hales, G. (1994). The human aspects of dyslexia. In G. Hales (Ed.), Dyslexia matters. (184–198). London: Whurr.
Helland, T. & Kaasa, R. (2005).Dyslexia in English as a second language. Dyslexia, 11(1), 41–60. Hermans, D., Ormel, E., Van den Besselaar, R., & Van Hell, J.G. (in press). Lexical activation in
bilinguals’speech production is dynamic: How language ambiguous word scan affect cross-language activation. Language and Cognitive Processes.
Holden, J.G. (2005). Gauging the fractal dimension of response times from cognitive tasks. In Riley, M.A., & Van Orden, G.C. (Eds.). Contemporary Nonlinear Methods for Behavioral Scientists: A webbook tutorial (pp 267-318).
Huntington, D. D., & Bender, W. N. (1993). Adolescents with learning disabilities at risk? Emotional well-being, depression, suicide. Journal of Learning Disabilities, 26(3), 159–166.
Ihlen, E.A.F., & Vereijken, B. (2010). Interaction-dominant dynamics in human cognition: Beyond 1/ƒα fluctuation. Journal of Experimental Psychology, 139(3), 436-463.
Jenkins, J.R., Fuchs, L.S., Van den Broek, P., Espin, C., & Deno, S.L. (2003). Sources of individual differences in reading comprehension and reading fluency. Journal of Educational Psychology, 95(4), 719-729.
Joanisse, M.F., Manis, F.R., Keating, P., & Seidenberg, M.S. (2000). Language deficits in dyslexic children: Speech perception, phonology and morphology. Journal of Experimental Child Psychology, 77(1), 30-60.
Jordan, J.S. (Ed.). (1998). System theories and a priori aspects of perception. Amsterdam: Elsevier Science & Technology.
Just, M.A., Carpenter P.A., & Woolley J.D. (1982). Paradigms and processes in reading comprehension. Journal of Experimental Psychology, 111(2), 228-38.
Kroll, J.F., & Dijkstra, A. (2002). The bilingual lexicon. In R. Kaplan (Ed.), Handbook of Applied Linguistics (pp. 301-322). Oxford : Ofxord UP.
Le Jan G., Le Bouquin-Jeannès R., Costet N., Trolès, N., Scalart, P., Pichancourt, D., Faucon, G., & Gombert, J.E. (2011). Multivariate predictive model for dyslexia diagnosis. Annals of Dyslexia, 61(1), 1-20.
Levelt, W. (1989). Speaking: From intention to articulation. Cambridge, MA: The MIT Press.
Levelt, W. (1999). A Blueprint of the speaker. In C. Brown & P. Hagoort (Eds.), The neurocognition of language (chapter 4). Oxford Press.
Liebovitch, L.S., & Shehadeh, L.A. (2005). Introduction to fractals. In Riley, M.A., & Van Orden, G.C. (Eds.). Contemporary Nonlinear Methods for Behavioral Scientists: A webbook tutorial (pp.178-266).
Lowie, W., De Bot, K., & Plat, R. (to appear). Pink noise in language production, a nonlinear approach to the multilingual lexicon. Ecological Psychology.
Lundberg, I. (2002). Second language learning and reading with the additional load of dyslexia. Annals of Dyslexia, 52(1), 165-187.
MacWhinney, B. (2006). Emergentism—Use often and with care. Applied Linguistics 27(4), 729– 740.
Mattis, S., French, J.H., and Rapin, I. (1975). Dyslexia in children and less experienced more experienced: Three independent neuropsychological syndromes. Developmental Medicine & Child Neurology, 17(2), 150–163.
Meara, P. (2004). Modelling vocabulary loss. Applied Linguistics, 25(2), 137-155.
Miller-Guron, L., & Lundberg, I. (2000). Dyslexia and second language reading: A second bite at the apple? Reading and Writing: An Interdisciplinary Journal, 12, 41-61.
Paget, K. D., & Reynolds, C. R. (1984). Dimensions, levels and reliabilities on the revised children’s manifest anxiety scale with learning disabled children. Journal of Learning
Disabilities, 17, 137–141.
Pennington, B.F. & Lefly, D.L. (2001). Early reading development in children at family risk for dyslexia. Child Development, 72(3), 816-833.
Perez, T.M., Majerus, S., & Poncelet, M. (2012). The contribution of short-term memory for serial order to early reading acquisition: Evidence from a longitudinal study. Journal of
Experimental Child Psychology, 111(4), 708–723.
Puolakanaho, A., Poikkeus, A., Ahonen, T., Tolvanen, A., & Lyytinen, H. (2004). Emerging phonological awareness differentiates children with and without familial risk for dyslexia after controlling for general language skills. Annals of Dyslexia, 54(2), 221-243.
Rayner, K. (1998). Eye movements in reading and information processing: 20 years of research. Psychological bulletin, 124(3), 372 -422.
Rutter, M, Tizard, J., & Whitmore, K. (eds). (1970). Education, Health and Behaviour. London: Longmans.
Rutter, M. & Yule, W. (1975). The concept of specific reading retardation. Journal of Child Psychology and Psychiatry, 16, 181-91.
Rinck, M., Gámez, E., Díaz, J.M., & De Vega, M. (2003). Processing of temporal information: evidence from eye movements. Memory & Cognition, 31(1), 77-86.
SchijfSchneider, W., Eschman, A., & Zuccolotto, A. (2002). E-Prime User’s Guide. Pittsburgh: Psy-chology Software Tools Inc.
Smith-Spark, J.H. & Fisk, J.E. (2007). Working memory functioning in developmental dyslexia. MEMORY, 15(1), 34-56.
Snowling, M. (1987). DyslexiaL A cognitive developmental perspective. New York: Basil Blackwell. Stolt, S., Haataja, L., Lapinleimu, H., & Lehotnen, L. (200*). Early lexical development of Finnish
children—A longitudinal study. First Language, 28(3), 259-279.
Taguchi, E., Gorsuch, G.J., & Sasamoto, E. (2006) Developing second and foreign language reading fluency and its effect on comprehension: A missing link. The Reading Matrix, 6, 1-18. Tulving, E. (2002). Episodic memory: from mind to brain. Annual Review of Psychology, 52, 1-25. Van den Bos, K.P., Lutje Spelberg, H.C., Scheepstra, A.J.M. & de Vries, J.R. (1994). De Klepel,
vorm A en B. Een test voor de leesvaardigheid van pseudowoorden. Verantwoording, handleiding, diagnostiek en behandeling. [The Klepel, version A and B. A test for pseudo words reading ability. Justification, manual, diagnosis and treatment. ] Nijmegen: Berkhout. Amsterdam: Harcourt Test Publishers.
