An audiovisual integration deficit underlies reading failure in nontransparent writing systems: An fMRI study of Chinese children with dyslexia

Highlights

• Audiovisual integration is impaired in Chinese dyslexic children.

• Dyslexics showed reduced activation in the left AG, MFG and SFG.

• Dyslexics showed altered functional connectivity of the left AG.

Abstract

Establishing the associations between speech sounds and visual scripts is an essential step for reading acquisition. Audiovisual integration deficits have been proposed to be a cause of developmental dyslexia, and this has been evidenced by behavioral and neural findings from alphabetic languages with relatively transparent orthographies. Here, using functional magnetic resonance imaging (fMRI) in a lexical decision task, this study examined the neural correlates of audiovisual integration deficits in Chinese children with dyslexia and age-matched controls. The results indicated that the children with dyslexia showed a reduced audiovisual congruency effect in the left angular gyrus, left superior temporal gyrus and left middle and superior frontal gyrus. Moreover, dyslexics showed reduced functional connectivity for the congruency effect between the left angular gyrus and both the left lingual gyrus and left cerebellum. The regional activation and functional connectivity abnormalities might underlie the ill-developed correspondence between orthographic and phonological information in Chinese dyslexia. This study, for the first time, illustrates the neural mechanisms of the audiovisual integration deficit in dyslexia in a nontransparent logographic writing system, extending our understanding of the neural basis of dyslexia.

Introduction

Developmental dyslexia is characterized by a significantly low reading achievement based on chronological age, despite adequate intelligence and socioeconomic opportunity, and affects approximately 5%–17.5% of school children (S. E. Shaywitz, 1998). To date, the etiology of dyslexia remains unclear. Successful acquisition of a reliable correspondence between auditory sound and visual forms of written scripts is a critical step in developing fluent reading skills in alphabetic languages (Ehri, 2005; Ziegler & Goswami, 2005). Such audiovisual integration processing would progressively become automatic with the development of reading and in turn help to tune the phonological representation of the isolated speech sounds (Vera Blau et al., 2010). Given the vital role of audiovisual integration in reading acquisition, one hypothesis posits that the difficulty in audiovisual integration of speech sound and visual script is a causal factor of dyslexia (Birch & Belmont, 1964), which is supported by subsequent behavioral (Francisco, Jesse, Groen, & McQueen, 2017; González et al., 2015; Harrar et al., 2014) and neural findings (Vera Blau et al., 2010; Blau, van Atteveldt, Ekkebus, Goebel, & Blomert, 2009; Froyen, Willems, & Blomert, 2011; Richlan, 2019).

Using fMRI, previous neuroimaging studies have identified several brain regions necessary for audiovisual integration processing in typical readers. Typically, two strategies have been adopted to isolate the brain substrates necessary for audiovisual integration. The first strategy is to compare the differences in neural responses between audiovisual bimodal and visual or auditory modal stimuli (Beauchamp, 2005). The second strategy is to compare the differences in neural responses evoked by consistent and inconsistent bimodal audiovisual stimuli. The bilateral posterior superior and middle temporal gyrus/sulcus (STG and MTG/STS) (Beauchamp, Argall, Bodurka, Duyn, & Martin, 2004; Calvert, 2001; Calvert, Campbell, & Brammer, 2000; Raij, Uutela, & Hari, 2000; Saito et al., 2005; Taylor, Moss, Stamatakis, & Tyler, 2006; Van Atteveldt, Formisano, Goebel, & Blomert, 2004) and left inferior frontal gyrus (IFG) (Kronschnabel, Brem, Maurer, & Brandeis, 2014; Skipper, Nusbaum, & Small, 2005; Taylor et al., 2006) have consistently been found to be involved in audiovisual integration, despite that there may be anatomical variations due to differences in task paradigms.

A few neuroimaging studies of dyslexia have investigated the neural correlates of audiovisual integration dysfunction in dyslexia in alphabetic languages (Vera Blau et al., 2010; Blau et al., 2009; Francisco et al., 2018; Kast, Bezzola, Jancke, & Meyer, 2011; Kronschnabel et al., 2014; Pekkola et al., 2006). Using fMRI, Pekkola et al. (2006) examined audiovisual integration in Finnish adults with dyslexia and found that dyslexics, relative to fluent readers, exhibited hyperactivity in the bilateral IFG, the supplementary motor area (SMA) and the left inferior parietal lobule (IPL). The increased brain activation in motor and visual-related regions was interpreted as a compensatory mechanism for perceptual difficulties in dyslexic readers who might rely more on motor-articulatory and visual strategies (Pekkola et al., 2006). Blau et al. (2009) used the congruency effect of letters and sounds as a probe to examine audiovisual integration in Dutch adults with dyslexia and found that the dyslexics showed a reduced congruency effect in the bilateral STG. Moreover, the audiovisual integration impairment was found to underlie the phonological perception deficit in dyslexic readers (Blau et al., 2009). The opposite pattern of dysfunction in people with dyslexia reported by Pekkola et al. (2006) and Blau et al. (2009) is likely associated with attentional control, which is more involved in the active task used by Pekkola et al. (2006) than the passive task used by Blau et al. (2009). In addition, Kronschnabel et al., (2014) noted that the single letters and speech sounds used previously (Blau et al., 2009; Pekkola et al., 2006) were different from the speech sounds and orthographies in actual reading. Thus, some studies have utilized speech-like long units (i.e., letter strings, words) to investigate the audiovisual integration deficit in dyslexia under a more realistic reading context (Kast et al., 2011; Kronschnabel et al., 2014). For instance, using disyllabic words and pseudowords in a lexical decision task, Kast et al. (2011) found that German adults with dyslexia, compared to nondyslexic adults, showed reduced brain activation in the left supramarginal gyrus (SMG) and the right STS in both audiovisual and unimodal visual and auditory conditions, while they showed increased brain activation in the right anterior insula during audiovisual integration. Moreover, using both single and three-letter stimuli, Kronschnabel et al., (2014) reported a group difference in the audiovisual congruency effect in the left IFG, angular gyrus (AG) and the inferotemporal cortex between Swiss dyslexic adolescents and controls. Importantly, such differences were found to be more pronounced in three-letter strings than in single letters, suggesting that naturalistic or word-like stimuli were more sensitive to the audiovisual integration deficit in dyslexia (Kronschnabel et al., 2014).

In addition to brain localization, two recent studies examined the audiovisual integration deficit in dyslexia from the perspective of neural networks (Rüsseler, Ye, Gerth, Szycik, & Münte, 2018; Ye, Rüsseler, Gerth, & Muente, 2017). Using independent component analysis (ICA), an fMRI study demonstrated group differences between German adults with and without dyslexia in a neural component including the STG, Rolandic operculum, MTG and anterior cingulate cortex, suggesting that these regions might contribute to the audiovisual integration deficit in dyslexia (Ye et al., 2017). Similarly, another study using ICA found that German dyslexic readers differed from fluent readers in two neural components associated with the integration of disyllabic words. The first component includes the fusiform gyrus (FG) and occipital gyrus, and the other component includes the left STS and frontal cortex (Rüsseler et al., 2018). The aforementioned two studies suggested that the examination of neural network abnormalities could provide more complete maps of brain dysfunction related to the audiovisual integration deficits in dyslexia, although these types of studies have been scarcely conducted.

Another critical issue is that the empirical evidence from adults (Blau et al., 2009; Kast et al., 2011; Pekkola et al., 2006) or adolescents (Kronschnabel et al., 2014) with dyslexia can hardly discriminate whether the neural correlates of audiovisual integration deficits are a fundamentally causal factor or a consequence of a lifetime of reading difficulties. Until now, only one fMRI study has examined the audiovisual integration deficit in children with dyslexia (Vera Blau et al., 2010). Blau et al. (2010) examined the audiovisual integration of letters and speech sounds in Dutch children with dyslexia and found that children with dyslexia showed an impaired congruency effect in the left planum temporale/Heschl's sulcus (PT/HS) in the audiovisual bimodal condition, as well as reduced brain activity in the FG with visual letters and the anterior STG, PT/HS and STS with the auditory stimuli (Vera Blau et al., 2010). This study replicated the findings from adults with dyslexia (Blau et al., 2009), suggesting that the audiovisual integration deficit might be a causal factor of dyslexia.

It should be pointed out that all the aforementioned studies were conducted in transparent/semitransparent writing systems, such as Dutch (Vera Blau et al., 2010; Blau et al., 2009), Finnish (Pekkola et al., 2006) and German (Rüsseler et al., 2018; Ye et al., 2017), in which the high consistency of correspondence between visual letter and speech sounds determines the importance of letter-sound associations for reading development. A recent fMRI study of English letters showed that brain activation of audiovisual integration depends on the extent of orthographic transparency. Specifically, brain activity in the superior temporal cortex, a well-known core region for audiovisual integration of letters/words, was detected with transparent pairs (letters and their names and numerals and number names) but not with less transparent pairs (letter and speech sound) (Holloway, van Atteveldt, Blomert, & Ansari, 2013). Correspondingly, a recent behavioral study demonstrated that audiovisual integration skills did not differ between those with English dyslexia and age- and reading-matched controls (Nash et al., 2017), casting on doubt on the audiovisual integration hypothesis of dyslexia. Consequently, it is necessary to test the audiovisual association deficit in dyslexia and its neural substrates in conditions with different degrees of orthographic transparency.

Chinese provides an excellent opportunity to address this question as Chinese is an opaque logographic writing system that differs dramatically from alphabetic languages. The basic unit of written Chinese is the character, a combination of radicals that are formed by strokes. Most modern-day use Chinese characters (about 90%) are compound characters with a phonetic and a semantic radical (Tan, Hoosain, & Peng, 1995), and the phonetic implies the character's pronunciation and the semantic implies the meaning. For example, the character 沐 (mu4) consists of the phonetic radical '木' and the semantic redical '氵'. The syllable is the basic phonological unit in Chinese, and each character corresponds to a syllable. Critically, there is no consistent correspondence between orthography and phonology in Chinese. Furthermore, Chinese includes a large number of homophones with a single syllable shared by many characters, forming extremely complex correspondences between orthographic and phonological forms. Behavioral (Ho, Chan, Lee, Tsang, & Luan, 2004; Shu, McBride-Chang, Wu, & Liu, 2006) and neuroanatomical studies (Cao et al., 2017; Siok, Niu, Jin, Perfetti, & Tan, 2008; Siok, Perfetti, Jin, & Tan, 2004) have reported deficits in phonological processing of Chinese characters that were visually presented in those with Chinese dyslexia, providing indirect evidence for an impairment in orthographic-phonological binding of characters in Chinese dyslexia.

Using fMRI, this study sought to explore the neural basis of audiovisual integration deficits in Chinese children with dyslexia and age-matched controls. We used a lexical decision task with real word-level stimuli, as previous studies have suggested that speech-like stimuli would be more sensitive to the audiovisual integration deficit in dyslexia (Kronschnabel et al., 2014). In line with previous studies (Vera Blau et al., 2010; Blau et al., 2009; Pekkola et al., 2006), we adopted the congruency effect (comparison between congruent and incongruent bimodal stimuli) as a probe, which has the advantage of controlling for attentional factors in group comparisons. We expected to observe atypical brain activation and functional connectivity in the posterior superior temporal cortex in those with Chinese dyslexia, findings which have been consistently reported in previous fMRI studies with alphabetic languages (Vera Blau et al., 2010; Blau et al., 2009; Kast et al., 2011; Kronschnabel et al., 2014; Ye et al., 2017). We also expected to detect the culture-specific neural basis of the audiovisual integration deficit in Chinese dyslexia. For example, abnormal brain activation and structure in the left middle frontal gyrus (MFG) have been reported in Chinese children with dyslexia (Siok et al., 2008; Siok et al., 2004). This region has been suggested to be associated with orthography-to-phonology transformation in the processing of Chinese characters (Tan, Laird, Li, & Fox, 2005).