Cognitive tasks engage a set of distributed cortical areas that work together toward accomplishing a desired goal. Each neural center has its own specialization; activation is synchronized across participating areas. In regard to reading ability, the neural network can be generally associated with neuroanatomical areas within the left perisylvian region. Areas within this region have their own cognitive specialization and vary in their importance to reading ability (Shaywitz et al., 2004).

In terms of functional neuroimaging, an abundant literature links poor word recognition skill to differences in the neuronal functioning of three neuroanatomical areas within the perisylvian network: occipito-temporal cortex (OTC), temporo-parietal cortex (TPC), and inferior frontal gyrus (IFG). A variety of studies have shown that the neural signature of brain function in individuals with reading disability (RD) is characterized by decreased activity in the left OTC (Brunswick et al., 1999; Paulesu et al., 2001) and left TPC (Rumsey et al., 1997; Horwitz et al., 1998; Brunswick et al., 1999) and increased activity in right TPC (Shaywitz and Shaywitz, 2003) region. However, this functional pattern of RD is malleable: Children with RD who respond to intervention exhibit significantly increased activation of left hemisphere TPC and decreased activation of right hemisphere TPC (Simos et al., 2002, 2006; Aylward et al., 2003; Shaywitz et al., 2004). Such differences in functional activation serve to elucidate the neurobiological phenotype for responsiveness to intervention (RTI). This is an important point because under the new Federal educational policy enacted in the reauthorization of the Individuals with Disabilities Education Act (IDEA; P.L. 108-446; IDEA, 2004), RTI may be used to identify students with disabilities. An important question is the degree to which this classification can be connected to, and in a sense validated by, anomalous neurobiological characteristics.

While functional MRI (fMRI) studies have provided information about neuronal response to word stimuli, coordinated transfer of information between brain regions within a network is also an important component of information processing. White matter (WM) tracts connecting different brain regions support communication between brain regions, though these have been less frequently investigated in imaging studies. Until recently, the in vivo investigation of WM integrity was not possible. Now, with the advent of diffusion tensor imaging (DTI) WM integrity can be evaluated in adults and children. DTI is a technique that examines the structure of WM in the brain by measuring the molecular diffusion of water. Cell membranes and myelin hinder the diffusive movement of water perpendicular to those membranes. Hence diffusion rates are higher parallel to axons than perpendicular to them. Information on the directional preference of diffusion is used to perform fiber tracking and estimate connectivity (Conturo et al., 1999; Jones et al., 1999; Mori et al., 1999; Basser et al., 2000).

How highly connected gray matter regions are to one another appears to be associated with reading ability. In fact, several studies comparing children with reading difficulty to normal readers have reported significant differences in WM integrity in and between the temporal-parietal regions (Klingberg et al., 2000; Beaulieu et al., 2005; Deutsch et al., 2005; Niogi and McCandliss, 2006). While differences between the WM of good and poor readers have been shown, the relationship between WM connectivity and intervention has only recently been investigated (Keller and Just, 2009). Using a voxel-based analysis of diffusion tensor parameters, Keller and Just (2009) showed changes in left medial superior frontal WM in 8–10 year olds after 100 h of tutoring. But, they did not examine WM architecture and connectivity as related to an individual's classification as a responder or non-responder to intervention. It should be noted that a few neuroimaging studies actually have addressed this issue, but only using functional neuroimaging (Simos et al., 2006; Odegard et al., 2008; Davis et al., in press). Whether a similar correlation exists between responsiveness to instruction and WM brain structures is not clear. Use of a complementary neuroimaging modality that captures WM connectivity thus can provide important and unique information.

The purpose of the current study was to investigate the potential relationship between WM connectivity and responsiveness to classroom based instruction in a group of children in the first grade, using voxel based analysis of diffusion tensor magnetic resonance imaging data. To investigate the relationship between WM connectivity and responsiveness to instruction, we identified nine cortical regions of interest that have strong theoretical justification for playing a part in reading skill (Table 1). Each region is involved with several different processes; therefore, we list in Table 1 only the behaviors of interest to the current study. We calculated the connectivity strength (defined below) between these gray matter regions and correlated the strength estimate with participants’ scores on reading measures. We hypothesized that we would replicate previous findings of differences in WM related to reading skill. In addition, we anticipated that children's responsiveness to reading intervention would provide additional information. To our knowledge, this is the first cortical connectivity study with children who differ in their response to a year-long reading intervention. However, this is an exploratory study with a small sample size and further studies with a larger sample size are needed.

To test the hypothesis that reading performance was correlated with connectivity of key gray matter regions, we correlated each element of the connectivity matrix with the participants’ scores on standardized tests of reading proficiency. Figure 4 is a visual representation of the correlations between participants’ connectivity estimates and their performance on each of the standardized measures. Within these matrices, left hemisphere connections are in the bottom triangle and right hemisphere connections are in the top triangle. The gray scale denotes significance level, with lighter colors indicating more significant correlations. Because of the small sample size, significant correlations were reported at p < 0.05 uncorrected. Connections with significant correlation at the p < 0.05 level are starred. We investigated age and gender influences on the behavioral and imaging results using two separate one way ANOVAs. Results showed a single significant effect: The left TRI to IFS pathway had a significant gender effect, in which girls had a significantly greater number of WM fibers connecting these regions. In addition, we correlated participant's performance on the behavioral tasks with the volume of each region. We found a positive correlation between the PDE subtest of the TOWRE and the left planum temporal (p = 0.02) and the left superior temporal cortex (p = 0.03).

To facilitate interpretation of the significant connections, scatterplots for each finding are shown in Figures 5 and 6. These scatterplots also represent the relationship with response to instruction. In these plots, the NR participants are red, R participants are blue, and C participants are green.

As shown in Figure 5, increased reading score on each of the reading measures is related to greater connectivity between the left ANG and INS; the NR participants generally have lower connectivity than R or C participants. Similarly, performance on the WAT subtest correlated positively with connectivity values between TRI to STC (Figure 6). Scatterplots of this correlation showed that NR participants had lower connectivity than R or C participants (although only one R dataset survived outlier rejection). Alternatively, three negative correlations were found in which poorer readers, and more particularly, NR participants, had a greater magnitude of connectivity between brain regions than the other two groups. Two correlations involve connections to the THAL, yet in different hemispheres of the brain: the first was right THAL to right TRI and the second was left THAL to left STC. The third remaining region was also in the right hemisphere – TRI to STC.

Reading relies on a diffuse network of brain regions spread throughout the entire brain (Pugh et al., 2001). Reading skill is likely a result of the level of integration (and potentially segregation) of different gray matter regions within a network. Previous studies converge on the finding that good and poor readers have significantly different FA scores in a left hemisphere temporo-parietal region within this network (Klingberg et al., 2000; Beaulieu et al., 2005; Deutsch et al., 2005; Niogi and McCandliss, 2006; Rimrodt et al., 2010). From these results, the field has gained critical information on the structural relationship that exists between the brain and reading skill. However, the previous studies provide information on the contribution of a single tract to reading, effectively ignoring the connections between gray matter regions which are ultimately responsible for the functional processes that underlie reading skill. It may be that these findings are the critical pieces of the connectivity network that differs in good and poor reading. The present study attempted to clarify this issue by investigating connectivity strength between gray matter regions critical to reading, thus examining more directly than the previous studies the putative relationship between gray matter (function) and WM (structure).

Specifically, in the current study we used probabilistic tractography to calculate the strength of WM connections and investigate the potential relationship between WM connectivity and reading skill in the first grade. Results demonstrated that eight connectivity probability estimates correlated with participants’ reading skill, four of which were connectivity estimates between the angular gyrus and insula. We are confident that these results are not attributable to differences in participants’ gender or age. Furthermore, although we found evidence of a correlation between variability in the volume of the brain regions and participants’ behavioral scores, these findings involved different regions than the connectivity findings. Our results extend the previous findings of WM differences related to reading skill to identify specific connections between gray matter regions. In particular, we propose that our findings of connectivity differences between insular cortex to angular gyrus in the left hemisphere corresponds to the temporo-parietal WM region reported in previous papers (Klingberg et al., 2000; Beaulieu et al., 2005; Deutsch et al., 2005; Niogi and McCandliss, 2006; Rimrodt et al., 2010).

Connectivity between left insula and angular gyrus correlated with all four reading tasks, indicating that better readers in general had greater connectivity estimates between these two brain regions on both timed and untimed reading tasks. Related to this, relative to the other two groups, NR had the fewest fiber pathways connecting these two regions. Despite the small sample size in the current study, the redundancy of findings across different measures of word reading skill is promising. However, interpretation of the findings is challenging, as the insula and the angular gyrus have very different roles. Within the cognitive domain, functional activation occurs in the left insula during language, speech, working memory, and attention tasks. Evidence exists that the left insula plays an integrative role between purely cognitive and other systems, including sensorimotor, social-emotional, and olfacto-gustatory systems (Dolan, 2002; Johansen-Berg and Matthews, 2002; Chen et al., 2007; Frith and Singer, 2008). By contrast, several studies have concluded that the left angular gyrus is important for the lexical processes that occur when reading words (e.g., Binder et al., 2009); though, it is possible that the angular gyrus is involved in phonological processing, especially in early years, as inferior parietal lobe is often associated with mapping phonological representations to their orthographic representations (Pugh et al., 2000). These results may indicate that connectivity between these two brain regions is a crucial aspect of the neural network that supports word recognition skill.

Alternatively, it may be that the insula and/or angular gyrus are simply acting as way stations (or relay stations), connecting the gray matter regions that are actually performing the domain specific tasks such as phonological processing and sight word recognition. For example, WM tracts traveling from planum temporale (Wernicke's area) in the superior temporal cortex to Broca's area in the frontal cortex are critical for speech development and normal reading (Hickok et al., 2003). These WM tracts may interact with the insula as they travel past on their way to frontal cortex. Related to this, it is possible that our results would change if a more liberal fiber tracking method were used. In the current study, WM tracts were counted if they connected a pair of gray matter regions and did not touch additional gray matter regions. This method was used because connectivity between two regions cannot be established with a high degree of reliability once the WM tract passes through additional gray matter region(s). Therefore, it would be fruitful for future imaging studies on reading skill to investigate both the direct connection between the insula and angular gyrus, while also considering the indirect role these regions play in the larger perisylvian language network. Additionally, in the current study, the entire insula was used as the seed region. However, insular cortex is often divided into separate regions with distinct functions (Dolan, 2002; Dupont et al., 2003; Kurth et al., 2010). Future connectivity studies with the insula would benefit from dividing it based on its functional differentiation.

In addition, a finding of interest that was unexpected was two negative correlations with thalamic connectivity estimates and the WAT subtest, a measure of pseudoword reading skill, in which NR participants had increased connectivity values compared to both R and C participants. Because the thalamus acts as an information processing way station for the brain, relaying sensation and motor signals and contributing to the regulation of arousal (Portas et al., 1998), it is logical that there would be connections between the thalamus and reading related areas (STG and IFG). In fact, previous studies have reported individual variability in functional (Price et al., 1994; Fiebach et al., 2002; Turkeltaub et al., 2002) and structural (e.g., Galaburda and Eidelberg, 1982) aspects of the thalamus that relate to differences in reading skills, suggesting linkages between the thalamus and reading ability. For people who are poor at reading, thalamocortical connections may be more strongly related to reading ability than good readers because poor readers rely on these connections to a greater extent than those who are skilled. This would explain a negative correlation between strength of connection and reading skill (i.e., poorer reading = more connections). Although the sample size in the current study is small, these negative correlations, coupled with previous negative correlation findings and thalamic findings, are intriguing and suggest a need for further exploration. This finding requires follow up studies with larger numbers of participants to fully interpret its meaning. Other studies have also shown negative correlations between reading skill and neuroimaging findings (Brunswick et al., 1999; Sarkari et al., 2002; Shaywitz et al., 2002, 2003; Turkeltaub et al., 2003; Brem et al., 2009), which has been interpreted as utilization of alternative pathways. However, the exact meaning of these negative correlations is not known. It may be that they demonstrate that poor readers implement a compensatory strategy via the use of an alternative pathway (e.g., rely on memory for sight word reading; articulation based word decoding, etc.), or further investigations may reveal that they are in fact a result of error in measurement (e.g., multiple comparisons; performance differences between ability groups, etc.). For this reason, further investigation of this topic is needed to fully understand it.

An additional finding worth noting is that the scatterplots for each of the significant correlations showed a potential trend between participants’ brain connectivity and their level of responsiveness to instruction. Although each correlation showed group differences in WM connectivity, the strength of the connections varied with the pair of regions under investigation. In other words, compared to children who responded to intervention (both R and C groups), the NR group had increased connectivity in some regions and decreased connectivity in others. Although caution is warranted when interpreting these results as the study was based on a small sample size, this is a promising finding, which demonstrates the importance of considering responsiveness to instruction in future studies of RD.

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.