The results of this study evidenced a difference in the perception of a paRol/paRl contrast between children with dyslexia and typically reading children. The paRol/paRl boundary was located at shorter /o/ durations for the children with dyslexia compared to the typically reading ones.
Such perceptual differences did not seem to be related to the difference in chronological age between the DYS and CTL children in this study. The comparison between CTL children and adults suggests that the effect of age is to shift the boundary towards a longer duration. If age was responsible for the difference in boundary location between the DYS and the CTL children in this study, the paRol/paRl boundary should have been located at a longer /o/ duration for the DYS children. Although a nonlinear effect of age remains possible, this would mean that the boundary would decrease between the ages of seven and nine years old (the mean ages of the CTL and DYS children in the present study) before rising between nine years and adulthood. While a nonlinear effect of age on allophonic perception has indeed been reported in the literature [30
], the inflection point was located around seven years old and is therefore not compatible with the inflection at nine years old that would underlie a non-linear effect in the present study. Finally, the difference in boundary location between the seven year-old children and the adults might be due to non-perceptual factors that covary with age. However, non-perceptual factors might also be at the origin of the boundary difference between the seven and nine year-old children. Whatever its origin, an effect of age on the location of the boundary should go in the same direction for both comparisons (i.e., those between children and between children and adults).
These results are impressive for two reasons. Firstly, the relationship between dyslexia and the perception of a paRol/paRl contrast was entirely based on theoretical grounds. The results followed a prediction of the allophonic theory without any prior empirical evidence. Remember that following allophonic theory, the paRol/paRl boundary should be located at a shorter /o/ duration for people with dyslexia because they attribute a separate identity to the transitional segments between /R/ and /l/, and that they should rely on these allophonic segments to perceive a vowel inside the /Rl/ cluster at relatively short /Rol/ durations, a prediction that was verified in the present results.
Secondly, the high degree of reliability of the difference in boundary location along the parol/parl continuum suggests that it is closely related to the core difference in phonemic perception between children with dyslexia and typical-reading children.
Allophonic theory readily accounts for the enhanced sensitivity of DYS children to vocalic segments inside a consonant cluster. However, another possible explanation is that people with dyslexia perceive the acoustic signal with a higher temporal acuity. According to the Temporal Sampling Framework (TSF), brain generators oscillating at higher rates would be at the origin of a large set of deficits evidenced in dyslexia, including allophonic perception [31
]. The atypical neural responses of the DYS to relative high frequency modulations (in the beta-gamma range) would have as direct consequence in that they perceive “subphonemes” (i.e., allophonic segments) [33
]. If higher temporal acuity was at the origin of allophonic perception, the enhanced perception of short speech segments by the DYS should not depend on their spectral content. DYS should also be more sensitive to a spectrally uniform vowel inside a consonant cluster, but this remains to be proved (Preliminary data suggest that there is no difference in boundary location between children with dyslexia and chronological age control children along a paRol/paRl continuum that was generated by temporal reduction of a /o/ segment with constant formant frequencies (i.e., the /R
/ segment was replaced by a segment generated by reduplication of a central period at the middle of the segment).
The enhanced sensitivity of DYS children to allophonic segments fits pretty well with the results of previous studies that evidenced an enhanced sensitivity to allophonic features in children or adults with dyslexia, in at least three different languages (French, Dutch, Spanish; see Introduction). Overall, taking account of the present results on segmental allophony with the previous results on contrastive allophony, a majority of the children affected with dyslexia display allophonic perception. These are probably those who also display a deficit in phoneme awareness and are classically considered as affected by “phonological” dyslexia. One pending question however is whether allophonic perception is directly responsible for their specific failure in reading acquisition.
A recurring question in the recent literature concerns the implications of allophonic perception for learning to read. On purely deductive grounds, perceiving more sounds than there are letters to represent them (featural allophony) and perceiving extra sounds between two successive letters (segmental allophony) should raise serious difficulties in capturing the regularities that govern a writing system. In a study with English-speaking school-age children, brain event-related potentials to a Finnish phoneme contrast were better correlated to reading skills than those with an English phoneme contrast; the higher the correlation with the foreign contrast the poorer the reading skills [34
]. However, there are hints that some children might overcome these difficulties while others might not. The effect of a Categorical Perception deficit (CP deficit, a proxy for allophonic perception, e.g., [35
]) on reading was mediated by a deficit in phonemic awareness in a study with French children of about 10 years old with dyslexia [36
]. However, there was no evidence for a cascading effect from a CP deficit to reading fluency in a study with Dutch children of about 9 years old at familial risk for dyslexia [37
]. For these children, the relationship between the CP deficit and reading was mediated by RAN (Rapid Automatized Naming). A possible interpretation of both the French and Dutch results is that the consequences of allophonic perception for reading depend on other factors, and that the latter vary with the writing system. Remember that the Dutch writing system is more transparent that the French one [38
], and that specific reading difficulties are mainly a matter of reading fluency in languages with a fairly transparent orthography, whereas they are also a matter of reading accuracy in languages with a more opaque orthography [39
]. Consequently, the effect of allophonic perception on reading might be modulated either by timing constraints, explaining the mediation by RAN in Dutch, and also by the access to phoneme representations, explaining the mediation by phonemic awareness in French.
Timing constraints might impact the synchronization of grapheme and phoneme decoding in the temporal cortex, which is vital for learning to read [24
]. Phoneme decoding is slower than allophonic decoding in children with dyslexia [21
], a difference that might be larger in children with a RAN deficit. The temporal delay between phonemic and allophonic decoding might also vary with individual differences in the access to phonological representations in the prefrontal cortex, such differences being correlated with phonemic awareness [9
]. A higher degree of phonemic awareness might correspond to a quicker access to the prefrontal cortex and, in turn, a faster transmission of the phonemic codes to the temporal cortex. Finally, it should be stressed that it is not only the access to the prefrontal cortex but also the nature of the prefrontal representations (phonemic vs. allophonic: [40
]) that might affect the transmission of the phonemic codes to the temporal cortex.