Beginning with Tallal's 1980 study of the temporal order judgment of children with specific language impairments (SLI), research has explored the idea that the primary deficit of dyslexia could lay in deviant auditory processing skills.

Early research has related the interpretation of “temporal processing” strictly to rapid succession or short durational cues, as measured by gap detection tasks (Tallal, 1980). However, recent studies have demonstrated that the deficits observed in dyslexic readers are not mainly linked to the processing of short, rapidly presented stimuli, but especially to the processing of dynamic acoustic features such as frequency modulation (FM) and sound rise time (RT) (Goswami et al., 2002; Witton et al., 2002; Hämäläinen et al., 2005; Fraser et al., 2010; Boets et al., 2011; Poelmans et al., 2011; Law et al., 2014).

The processing of speech requires the interpretation and recognition of high-level perceptual units, such as words, sentences, and utterances. These perceptual units are an amalgam of various acoustic-phonetic cues that correspond to a time scale, specific to various phonological grain size units. For example, time windows of 0.14–0.33 s correspond to segmental information relating to syllable recognition, whereas phoneme identification relies upon the perception of shortened time scales of 0.02–0.08 s (Obrig et al., 2010). It is thought that during the pre-literate phase of development, a deficit in the perception, and processing of these acoustic-phonetic cues could ultimately limit a person's ability to isolate and reflect upon basal phonological information, thus resulting in inaccurate phonological representations (Nittrouer, 2006; Boets et al., 2007). This cascade of effects from a disruption in auditory processing, through speech, to the development of phonological representations has come to be known as the temporal auditory processing deficit theory (Ghesquière et al., 2014).

Supporting this theory, a growing body of research has provided evidence of a relationship between measures of dynamic auditory processing, phonology, and literacy achievement in pre-schoolers (e.g., Boets et al., 2011), in school-aged children (Talcott et al., 1999; Witton et al., 2002; Poelmans et al., 2011) as well as in adults (Hämäläinen et al., 2005; Law et al., 2014).

Speech perception research in the dyslexic population has primarily relied upon two experimental paradigms: (i) the perception of speech presented in back ground noise and (ii) categorical perception of stop consonant, often utilize optimal listening conditions, and do not involve whole word perception (for a review see Vandermosten et al., 2010). Past studies utilizing the more ecological speech-in-noise measure have demonstrated that children (Snowling et al., 1986; Wible et al., 2002; Bradlow et al., 2003; Ziegler and Goswami, 2005; Boets et al., 2011) and adults with dyslexia (Dole et al., 2012; yet see Hazan et al., 2009; Law et al., 2014) exhibit pronounced difficulty on speech perception tasks under noisy background conditions, while often not demonstrating any impairment of speech perception in silent conditions (Bradley and Bryant, 1983; Bradlow et al., 2003). Additional support for speech-in-noise deficits of individuals with dyslexia has been provided through neurophysiologic studies. Anomalies have been observed in the neural encoding of speech-in-noise stimuli of individuals with dyslexia when compared to normal reading controls (Wible et al., 2002), yet no differences were found between groups in quite listening conditions (Cunningham et al., 2001).

In accordance with the temporal auditory processing theory, we would expect measures of dynamic auditory processing to relate with performance on speech-in-noise perception measures. In an investigation of pre-reading children by Boets et al. (2011) a clear relationship between a measure of auditory processing and speech-in-noise perception was found. Results of this study demonstrated that children who went on to develop dyslexia were already impaired in dynamic FM sensitivity and speech-in-noise perception prior to reading instruction. These measures were also found to uniquely predict later growth in reading. Yet in two more recent studies no clear evidence was found supporting a relationship between dynamic auditory processing and speech-in-noise perception in kindergarten and 6th grade (Poelmans et al., 2011; Vanvooren et al., submitted).

Two of the more sensitive dynamic auditory measures in differentiating between individuals with dyslexia and controls have been shown to be the FM and RT tasks. FM detection assesses an individual's ability to detect fluctuations in a carrier frequency at a certain modulation rate. Individuals with dyslexia have been shown to have a reduced sensitivity to FM detection when compared to control groups, demonstrating the ability of FM tasks to differentiate between adult, school aged and pre-reading dyslexics from normal readers (Witton et al., 1998, 2002; Ramus et al., 2003; Boets et al., 2007; yet see Halliday and Bishop, 2006; Stoodley et al., 2006; for a review see Hämäläinen et al., 2013). In addition to the findings of group differences, a study by Witton et al. (1998) found that the phonological decoding skills of both dyslexics and controls significantly correlated with FM sensitivity of 2 and 40 Hz.

More recently RT detection tasks, another measure of dynamic auditory processing, have been shown to be a more sensitive measure in discriminating between populations of dyslexic and normal readers. RT discrimination tasks measure an individual's ability to detect subtle differences in the rate of change of an amplitude envelope. RT tasks allow for an indirect assessment of how well an individual can detect the onset of syllables which are necessary for speech perception (Goswami et al., 2002; Goswami, 2011; Poelmans et al., 2011) and are utilized in the segmentation of the speech signal into its base parts, such as syllables, and onset/rime (Goswami et al., 2010). Detection of such cues has been shown to be associated with the reading, writing and the phonological skills of adult and child populations (Goswami et al., 2002; Hämäläinen et al., 2005; Thomson et al., 2006; Pasquini et al., 2007; Thomson and Goswami, 2008; Fraser et al., 2010; Goswami, 2011; Law et al., 2014, 2016). Goswami et al. (2002) demonstrated that 25% of the unique variance in the reading and spelling of children could be predicted by individual differences in RT sensitivity, when IQ and age are controlled for. Additionally, findings demonstrating the relationship between RT and reading have also remained consistent across different orthographies (Goswami, 2011).

Yet not all studies have been able to replicate support for the temporal auditory processing theory (Halliday and Bishop, 2006; Stoodley et al., 2006; White et al., 2006). Though theoretically appealing, the temporal auditory processing deficit theory of dyslexia has faced criticism regarding the use of adequate controls for the psychophysical tasks in addition to questions concerning directionality and the lack of a clear association between speech-in-noise perception tasks and auditory processing deficits.

Questions have been raised regarding the observed poor performance of individuals with dyslexia in psychophysical studies, in that such observations may be a function of a general difficulty with task completion, thus resulting in the misinterpretation of non-sensory difficulties, such as those with attention or general task difficulty, as sensory ones (Stuart et al., 2001; Roach et al., 2004). To address such concerns Poelmans et al. (2011) utilized an intensity discrimination (ID) task, matched in design and methodology to the other experimental dynamic auditory processing tasks to act as a control measure. The inclusion of an ID measure permitted Poelmans and colleagues to rule out related task demands, attention, and cognitive aspects as driving factors of the observed auditory problems. This is in line with the observation that group differences between typical (normal) and dyslexic readers are often not found in measures of ID (see Hämäläinen et al., 2013).

Additional criticism has been drawn regarding the directionality and causality of the proposed theory. Arguments have been put forth stating that the processing of basic auditory stimuli may be affected in a top-down manner through poorly specified phonemic representations and are a consequence of the poor reading experiences (Bishop et al., 2012). Evidence to support such a top-down relationship has been provided in two studies, suggesting that the auditory system gets tuned into listening for particular frequency and /or amplitude changes, during speech perception, thus creating a situation where the individual favors the processing of speech-specific auditory cues (Nittrouer and Miller, 1997; Mayo et al., 2003). For instance, both studies have demonstrated that mature cue weighting strategies for speech perception develop in childhood as a result of increasing phonological awareness. Yet, a study by Johnson et al. (2009) noted evidence of a bidirectional relationship of phonological awareness and auditory processing. As most studies have centered on a single time point and populations of adults and school aged children after the onset of literacy instruction (for a review see Hämäläinen et al., 2013), questions of directionality and causality are difficult to address. In one of the few studies which investigated pre-reading auditory processing deficits in children longitudinally and could provide evidence of directionality, Boets et al. (2011) retrospectively explored the temporal auditory processing deficit theory in a population of pre-reading Dutch speaking children who later developed dyslexia. Through a series of partial cross-lagged correlations Boets and colleagues could not support a reliable interpretation of directionality, leading to the conclusion of a probable bidirectional relationship between auditory processing, speech-in-noise perception, and phonological awareness. Supporting these findings a longitudinal study of pre-reading children by (Vanvooren et al., submitted) found no evidence for a unidirectional causal link between auditory processing, speech-in-noise perception and phonological awareness during the first stages of reading acquisition.

Lastly, the temporal auditory processing deficit theory of dyslexia has received criticism relating to the lack of a clear association between speech-in-noise perception tasks and auditory processing deficits in the literature, thereby calling into question the viability of the theory (Rosen, 2003). Although studies have demonstrated deficits independently in the dynamic processing and speech-in-noise perception in individuals with dyslexia, only a handful of studies have assessed measures of both in the same population (Boets et al., 2011; Poelmans et al., 2011; Law et al., 2014; Vanvooren et al., submitted). Using RT and FM discrimination measures Poelmans et al. (2011) examined the same population of children of the longitudinal study of Boets et al. (2011) in at the age of 11, and although a relationship among dynamic auditory processing and speech-in-noise perception was present at an earlier age (Boets et al., 2011), Poelmans et al. found no clear evidence supporting a relationship at a later age. Additionally, using similar measures, Law et al. (2014) was unable to support such a relationship in an adult population. Such results suggest that the observed auditory processing problems and their association with speech perception skills in individuals with dyslexia are present at birth through early childhood, thus contributing to early phonological deficits (Corriveau et al., 2010). However, auditory processing problems may diminish through development and eventually become resolved. The diminishing of the severity of the auditory impairment and its association with speech perception through time may obscure potential effects of this deficit in later reading achievement and related skills (Galaburda et al., 2006). Therefore, to fully understand the relationship of auditory processing and speech-in-noise perception early in development a replication of the findings pertaining to FM discrimination of Boets et al. (2011) and the inclusion of the more sensitive RT measure pre-reading is required.

The aim of the current longitudinal study is to address the above criticism of the temporal auditory processing deficit theory and to offer evidence in support of the theorized cascade of effects from auditory processing through speech in noise perception to phonological awareness and ultimately disrupting reading. Although previously examined in other languages this study will attempt to replicate earlier findings in an English speaking, pre-reading population. In addition to the FM detection task used in Boets et al. (2011), the more sensitive measure of RT was added as an assessment of speech envelope cues and to reflect the growing body of evidence of the importance of such cues in the early development of phonological awareness (Corriveau et al., 2010). Similar to Poelmans et al. (2011), an ID task was included in the testing battery to act as a means of control for attention difficulties and task related demands.

The objectives of this paper are 3-fold. First, to determine the relation between the kindergarten measures of auditory processing and speech-in-noise perception tasks, and the cognitive and literacy outcome measures at grades 1 and 2. Secondly, this study will attempt to address the question of directionality through an examination of pre-reading RT discrimination, FM sensitivity and PA ability to reliably predict later literacy achievement. Lastly, we investigate the presence of performance differences between groups based on behaviourally observed literacy problems across three time points: pre-reading kindergarten, grade one, and grade two.

Fifty-eight preschool children ranging in age from 4 to 5 years old and attending Senior Kindergarten (SK) in the Ontario Canada public school system were originally selected for the study. At the completion of the third year of the study, 44 children remained. Three children were absent due to relocation to a school district not included in the study and one child's parents chose not to participate in the second phase of data collection. Additionally, to reduce the influence of second language learning on the sample, 10 children were removed from the study after enrolling in a French immersion education program. Children were initially recruited to meet one of two classifications, either being at high-risk (HR) for developing dyslexia, or being at low risk (LR). The high-risk group was selected based on the child having at least one first-degree family relative with an official diagnosis of dyslexia. The low-risk group consisted of children with no family history of reading difficulties. Groups were matched on measures of intelligence, socioeconomic status, gender, age, hyperactivity and educational environment (see Table 1). All participants were reported through parental surveys to possess no signs of brain damage or long term auditory or visual impairments and were native English speakers. Only children considered “pre-readers” were initially included in the study. For the purposes of this study a “pre-reader” was defined as a child who had not received any formal reading instruction prior to the first testing period. Additionally children who demonstrated ceiling performance on our measure of letter knowledge proficiency along with the ability to decode three consecutive words from the target word list of the Wide Range Achievement Test (WRAT3) (Snelbaker et al., 2001; e.g., red, milk, was) were excluded from the study.

Participation was voluntary. Upon registering parents completed an online questionnaire which informed the study of the child's medical history, behavior and family history of reading and spelling (dis)abilities. The parental questionnaire also included screening for potential hyperactivity or behavioral problems, using questions taken from the Strengths and Difficulties Questionnaire (SDQ) (Goodman, 2001). Additionally, parental educational levels were measured using the seven point ISCED-scale (UNESCO, 1997). Groups were found not to differ on measures of age, IQ, socioeconomic status (SES) and parental educational level, as can be seen in Table 1.

Statistical analyses were performed with SPSS 20.0 software (IBM Corp, 2011). Data from all variables were checked with Shapiro-Wilk's test for normality. All data were found to be normally distributed (p > 0.05) with the exception of some auditory processing data: FM and RT in kindergarten in addition to FM at both first and second grades as well as ID at first grade. In order to approach a normal distribution, variables were transformed by a logarithmic transformation. The assumption of homogeneity of variance was assessed by Levene's Test for Equality of Variances. Group comparisons were investigated based on an independent-samples t-test. Correction for multiple testing was applied across all group comparisons to avoid the likelihood of false positive conclusions through the application of the False Discovery Rate (FDR) procedure (Benjamini and Hochberg, 1995). The FDR procedure is a simple sequential Bonferroni-type procedure that has been demonstrated to control for the FDR for independent test statistics. Pearson correlations between kindergarten measures of auditory processing and speech perception tasks and cognitive measures at grades 1 and 2 were calculated. In addition Partial Pearson correlations with age, IQ and group as covariates were calculated.

To address the questions of directionality between the dynamic auditory processing measure of RT and PA a series of cross-lagged partial correlations while controlling for autoregressive effects, in addition to the effects of age, IQ and group, were performed. This method provides a way of drawing tentative causal conclusions regarding directional effects of auditory processing (Kenny, 1975).

In order to assess the predictive factors relating to first and second grade literacy measures (reading and spelling), four sets of simultaneous linear regression analyses were calculated across both groups. For each model later literacy performance in grade one and two was predicted by kindergarten measures of phonological awareness (PA), letter knowledge (LK), and dynamic auditory processing (RT) after controlling for Age and IQ.

Table 2 shows concurrent and predictive relationships between all measures of dynamic auditory processing, speech-in-noise perception, phonological awareness, and measures of literacy.

Of the two kindergarten dynamic auditory processing measures only RT correlated significantly with PA and the reading composite scores at all grade levels. Additionally, RT in first grade was found to be significantly correlated with PA, while it was found to be approaching significance with reading at grade one and two. However, speech-in-noise was not found to relate to any of the assessed measures across all time points. As would be expected from the auditory processing deficit theory both measures of auditory processing (RT and FM) were found to be significantly correlated within and between each grade level. However, auditory processing measures were not found to be related at any time point with measures of speech-in-noise.

When group, IQ, and age were introduced across all subjects to control for any spurious effects (see lower left half of Table 2) the majority of the relations were maintained with the exception of the relationship of kindergarten RT with Kindergarten PA, r = 0.197, p = 0.224; as well as the relationship of letter knowledge with grade 1 and 2 reading (r = 0.293, p = 0.067, and r = 0.270, p = 0.092).

Figure 1 displays concurrent, autoregressive and cross-lagged (partial) correlations. As FM was not found to significantly correlate with measures of PA, FM was excluded from this analysis. RT and PA in kindergarten and first grade were found to have a significant concurrent relationship. Significant predictive relationships of RT in kindergarten with first grade RT and PA measures were found and are depicted in Figure 1A. After controlling for autoregressive effects of kindergarten PA the predictive relationship of kindergarten RT and first grade PA was maintained, thus suggesting directionality.

Results of the regression analysis shown in Table 3 revealed that phonological awareness, and the dynamic auditory processing measure of RT uniquely contributed to reading at both first and second grades. RT was found to account for 6.8% of the variance of first grade reading after controlling for PA, age, IQ, and letter knowledge, while PA was found to explain 9.5% of the variance after controls were accounted for. In the case of second grade reading after accounting for the variance of controls, RT accounted for 7.1% additional variance, in addition to the 11.0% of the variance attributed to PA. PA in kindergarten was found to explain an additional 13.4% of the variance of first grade spelling after accounting for controls. In addition, kindergarten PA was found to account for 7.5% of the variance of second grade spelling. At both time points letter knowledge was found to offer no significant contribution reading, yet was found to account for 6.2 and 6.3% of the variance of first and second grade spelling.

To investigate the presence of performance differences between groups based on the behaviourally observed literacy problems, the sample was retrospectively divided. Two groups, children with dyslexia and unimpaired children, were created based on their performance on literacy tasks at the start of second grade. A classification of literacy impairment was based on a child performing below the 10th percentile on two of the three second grade literacy measures: word reading, spelling or non-word reading. The resulting dyslexic (Dys) group consisted of 17 high-risk children and 4 low-risk children. The literacy unimpaired (control) sample was constructed of 19 low risk children. Four children from the high-risk group did not meet the cut-off criteria of dyslexia. Past research has demonstrated that similar groups of high risk normal reading children differ across many measures from low risk controls, so it was decided to exclude these individuals from group analysis of control subjects (Pennington and Lefly, 2001; Snowling et al., 2003). Additionally, due to the small sample size of high risk normal reading children separate statistical analysis was not performed on these subjects.

Tables 4–6 show the performance of children on all auditory, speech-in-noise perception, phonological awareness and literacy tests according to their classification and age. Independent t-tests found no group differences across measures of age, gender, IQ, SES, and hyperactivity (p > 0.05).

Fitting with the auditory processing deficit theory of dyslexia, it was assumed that measures of speech-in-noise perception would be found to relate to both auditory perception and phonological measures. Yet this study was not able to demonstrate any evidence to support the existence of a speech-in-noise perception deficit in children with dyslexia. These results are contrary to past research (Snowling et al., 1986; Wible et al., 2002; Bradlow et al., 2003; Boets et al., 2007; Ziegler et al., 2009). In addition, the speech-in-noise measure was found to be unrelated to any of our measures of dynamic auditory processing, phonological awareness or literacy. Therefore, this study could not support the theorized directional pathway from auditory processing through speech-in-noise perception to phonological skills as proposed by the temporal auditory processing deficit theory. Three possible arguments can be made to explain these findings. Firstly, it could be argued that dynamic auditory processing either independently relates to reading measures or relates through phonological awareness and not through speech perception. However, this remains unlikely considering the prevalence of dynamic auditory cues in the speech signal. An alternative explanation offered by Poelmans et al. (2011) theorized that the developmental link between auditory processing and speech perception might diminish with age due to the effect of different developmental influences over time (also see the longitudinal study of Boets et al., 2011). Thus, the inability to discover a relationship between these measures may be a result of the age of assessemnt. At the age the children were assessed in our study, speech-in-noise perception not only relies on bottom-up auditory processing but also involves various top-down processes such as semantic and syntactic cues which may have masked the presence of a primary deficit. Past research has demonstrated the existence of a relationship between early auditory processing and later speech perception in infancy (Leppänen et al., 2010). In addition, it is known that a new-born's auditory processing is sensitive to all phonemic contrasts and quickly becomes constrained to acoustic features specific to their native language (Kuhl, 2004). Thus, auditory processing's influence on speech perception may be limited to the first year of life. As argued by Vanvooren et al. (submitted), impairment in the processing of speech specific auditory cues at a very early stage could potentially impede speech perception during early stages of language acquisition.

Nevertheless, a more plausible explanation of the lack of findings could be due to specific task characteristics. The lack of group differences most likely were a function of the stationary speech weighted background noise used as a speech mask. Dole et al. (2012) noted that such masking noises are less effective in differentiating between dyslexic and normal readers than modulated noises and background speech masks. Therefore, offering an explanation for some of the heterogeneity of findings surrounding speech-in-noise perception deficits of individuals with dyslexia across development (Hazan et al., 2009; see Boets et al., 2011; Dole et al., 2012; Law et al., 2014) and including the results reported here.

Yet it is important to consider that these results reflect only one aspect of speech perception, that being speech-in-noise perception. Although this measure does represent a more natural measure of speech perception it must be noted that this task relies not only on basic acoustic perception but also elements of auditory attention or selective attention which may have influenced the results.

Although RT discrimination and FM detection measures were not found to relate to the speech-in-noise measure, a significant relationship was found between these two pre-reading measures of dynamic auditory processing. In addition, kindergarten RT was found to relate to concurrent and later phonological awareness and reading in grades one and two. The findings of a pre-literate relationship of measures of RT and phonological awareness are in line with other longitudinal studies that explored RT and early pre-reading phonological awareness (Corriveau et al., 2010). Yet, kindergarten and first grade measures of FM were not found to relate to later phonological awareness. The lack of kindergarten FM's relationship with phonological measures contradicted findings by Boets et al. (2011) who found FM in kindergarten to correlate with measures of phonology across all grade levels. As the FM detection measure of this study closely mirrored that used by Boets and colleagues, a potential explanation of the inconsistent results could rely on differences in the phonological awareness measures used. As the PA measure of this study consisted of a greater proportion of phonemic awareness tasks then syllable or rime awareness. The grain size level of the PA measure is of importance when considering its relations with speech specific auditory processing measures such as FM. As discussed earlier, time windows of 0.14–0.33 s correspond to segmental information relating to syllable recognition, while phoneme segmentation is reliant on the perception of shortened time scales of 0.02–0.08 s (Obrig et al., 2010). As the stimuli used within the FM task was based on a 2 Hz sinusoidal frequency modulation, it would be reasonable to expect measures of FM to more closely relate with a PA measure assessing grain size units at the rime and syllable level, as demonstrated in Boets et al. (2011).

Regression analyses of literacy measures accounting for letter knowledge and phonological awareness, and RT discrimination demonstrated kindergarten RT's ability to uniquely predict growth in reading achievement at grades one and two. Contrary to Boets et al. (2011), our dynamic auditory measure was found to uniquely predict variance in first and second grade reading suggesting that basic auditory processing skill's impact on reading development is not limited to the time point prior to reading instruction but extends through early stages of reading development. Results support the findings of Boets et al. (2011) in that individual differences in auditory processing are not simply a consequence of phonological awareness and early literacy achievement.

Although our results have demonstrated pre-reading RT measure's ability to predict later literacy skill, the variance explained within this model was limited supporting the findings of Plakas et al. (2013). Thus, highlighting auditory processing's role as one of many contributing risk factors in a multi deficit model of dyslexia, as theorized by Pennington (2006).

To address questions surrounding the directionality of the hypothesized causal pathway as predicted by the temporal auditory processing deficit theory, an investigation of the interrelations of auditory processing and phonological awareness across time points was conducted. Significant concurrent and predictive relationships were observed between the auditory processing measure of RT discrimination, and phonological awareness. Partial cross-lagged correlations, controlling for autoregressive effects, confirmed the directionality between dynamic auditory processing (specific to RT discrimination) and phonological awareness. Results demonstrated a larger impact of RT performance on future PA development than PA's influence on auditory processing development, thus supporting the bottom-up model proposed by Tallal (1980) within the first years of reading development.

Results contrasted with Boets et al. (2011) which demonstrated a lack of directionality between auditory processing (as measured through a FM discrimination task) and phonological awareness. It could be argued that RT sensitivity is less influenced by top-down processes during early stages of reading acquisition, and thus a more sensitive measure, when compared to FM, in establishing casual pathways as predicted by the theory.

In line with previous research (Pennington and Lefly, 2001; Snowling et al., 2003; Boets et al., 2011), children classified as dyslexic in grade two were found to differ significantly on all measures of phonological awareness, and literacy, across all three time points, when compared with typically developing readers.

As predicted by the temporal auditory processing deficit theory, group differences were expected across both measures of temporal auditory processing (RT and FM) but not for the non-temporal auditory ID control task. Group analyses demonstrated a statistically significant poorer performance of children later diagnosed with dyslexia on the measure of RT discrimination at the pre-reading phase, while a trend toward significance was observed for RT discrimination in first and second grade. Yet the same could not be said for measures of speech-in-noise perception, FM-detection or ID. The finding of poorer performance of children later found to be dyslexic on RT discrimination tasks prior to formal reading instruction indicates these problems are not consequential of the expressed literacy problems characteristic of dyslexia. These results were in line with the bulk of previous studies across age groups and languages (for a review see Hämäläinen et al., 2013).

The lack of significant group differences at each time point for the FM measure was unexpected as past research in both dyslexic children (Boets et al., 2011) and adults (Witton et al., 2002; Ramus et al., 2003) have demonstrated clear group differences. Similar to the results of this study, Law et al. (2014) unexpectedly reported a lack of group difference for FM in the presence of a RT-deficit. Law and colleagues suggested that such a difference in findings may imply the existence of a specific deficit in the perception of dynamic auditory cues related to the speech envelope, as measured through the RT discrimination task.

Several limitations regarding this work are worth noting. Although this study's sample size is comparable to the majority of the literature examining auditory processing and speech-in-noise perception in children with literacy impairments, the generalizability of the findings reported in this paper may be restricted due to the limited sample size of the study. Yet the validity of the conclusions of this study still remain valid because the group differences observed were confirmed by the correlational and regression analyses, i.e., pre-reading RT group difference is confirmed by unique predictive power on later reading achievement. Yet the restricted sample size did limit the statistical analysis we performed. A larger sample size would have permitted the use of structural equation modeling to allow for an analysis of the causal paths of the model we were investigating. It could be argued that sampling bias may have occurred during the recruitment. As enrolment for the study relied on parental responses to flyers sent home with children and did not involve a general sample, it could be argued that educationally motivated parents or parents concerned about their child's literacy success may had been more inclined to respond. The avoidance of this potential sampling bias was not possible due to restrictions placed on the solicitation of parent involvement by the school administration.

Results were not able to directly support the proposed cascade of effects as predicted by the temporal auditory processing deficit theory. Yet, dynamic temporal auditory processing was found to uniquely predict a proportion of later literacy achievement. Thus, extending the power of predicting future literacy outcomes to developmentally earlier precursors. Yet, this proposed deficit model was incapable of entirely explaining all of the expressed behavioral traits observed in a dyslexic population. Our findings taken together with past research that has demonstrated that not all individuals with auditory processing or phonological impairments develop dyslexia (see Snowling, 2008; Boets et al., 2011) support the proposed multiple deficit model of Pennington (2006) which stresses the need to explore a multifactorial etiology which accounts for multiple risk or protective factors. Thus, through an investigation of alternative cognitive factors, such as orthographic or morphological processing (Bekebrede et al., 2009; Law et al., 2015), alternative perceptual factors (Stein, 2001) and biological explanations (Nicolson et al., 2001), the variance and comorbid symptoms associated with the dyslexic population can be better understood.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.