FMRI data was preprocessed and analyzed using the SPM12 software package (Wellcome Department of Imaging Neuroscience, University College London, UK, 2014). Images were spatially realigned to the first volume by means of rigid body transformation, and unwarped. A confound of testing very young children with an fMRI paradigm is that head motion is highly correlated with age and further developmental characteristics (Satterthwaite et al., 2012), leading to increased motion artifacts. Consequently, the Artrepair toolbox (Mazaika, Hoeft, Glover, & Reiss, 2009) was used to detect outlier volumes in which scan‐to‐scan motion was greater than 1.5 mm (Karipidis et al., 2017; Pleisch et al., 2019). Outlier volumes were replaced by interpolating with the preceding and following correct images. Children with more than 10% repaired images (Pleisch et al., 2019) were excluded from further analysis (N = 22). Overall, less than 2.8% of the data was repaired. Tissue probability maps for native space components of the structural images were created according to an age‐matched pediatric template using the Template‐o‐Matic toolbox (Wilke, Holland, Altaye, & Gaser, 2008). The nonlinear Fast Diffeomorphic Anatomical Image Registration Algorithm (DARTEL; Ashburner, 2007) was used to create a study‐specific template. Subsequently, transformation from this study‐group specific template to MNI space was estimated. Finally, functional images (voxel size 2 × 2 × 2 mm3) were spatially smoothed with an 8 mm (FWHM) Gaussian kernel.

In a next step, individual fixed‐effect models were computed using the default value of the high‐pass filter (128 s) which partitions out the confounding influence of physiological noise. Experimental conditions were entered into a general linear model (GLM) and motion parameters generated during realignment were included as regressors of no interest to control for overall motion effects. Basic contrast maps (target stimulus against null) were generated for each stimulus condition (baseline conditions, hereafter denoted as: checkerboards, houses, faces, written words, and spoken words). First, we examined the baseline contrast of each stimulus condition by entering the single‐subject maps into second level one‐sample t‐tests using the flexible factorial model of SPM. Similar to the results of Dehaene‐Lambertz et al. (2018), checkerboards elicited larger activation than the other categories (bilateral calcarine: k = 12,784, T = 23.8, p FWE < .001, [−8–90 4], right SPL: k = 351, p FWE < .03, T = 7.96, [24– 70 44]). This was most certainly due to perceived movement induced by the fast‐changing line orientations from trial to trial and checkerboards were discarded from further analysis. Consequently, the specific neural response to houses, faces and written words was extracted by subtracting the activation of the other two remaining visual conditions (houses > [faces, written words], faces > [written words, houses], written words > [faces, houses]; cf. Dehaene‐Lambertz et al., 2018; Monzalvo et al., 2012 for a similar approach). The differential contrast of written words > [faces, houses] did not yield any significant activation and was thus not included in the whole brain–behavior association analysis. In a second step, single‐subject contrast maps of the baseline contrasts (faces, written words, spoken words) and the differential contrast faces > [written words, houses] were entered into group level regression analyses. Here, results of the behavioral tests were entered as regressors to evaluate (A) the relationship between cognitive–linguistic prereading skills (PA, RAN) and basic visual and auditory processing and (B) the power of preliterate neural processing to predict reading fluency after two years of formal reading instruction. Significant neural activation was inspected on the whole brain level with an initial cluster‐defining threshold of p < .001 (uncorrected) and a second family wise error (FWE‐)corrected cluster‐level extent threshold, measured in units of contiguous voxels (k), of p FWE < .05 corrected for multiple comparisons across the set of analyzed voxels (Flandin & Friston, 2019; Mueller, Lepsien, Möller, & Lohmann, 2017; Woo, Krishnan, & Wager, 2014). To avoid false positives, regression analyses were additionally controlled for the number of tests performed resulting in p < .006 for the concurrent correlational analysis between neural activity, PA and RAN (4 fMRI contrasts × 2 cognitive–linguistic skills) and p < .0125 for the longitudinal prediction of reading (4 fMRI contrasts × 1 reading skill) denoted as p corr. For the sake of comparability with previous studies (e.g., Chyl et al., 2018; Dehaene‐Lambertz et al., 2018; Monzalvo et al., 2012; Pollack & Price, 2019), we report the results for both: the stricter P‐values additionally accounting for multiple testing and also the standard correction accounting for the number of voxels (FWE). Brain regions are reported according to the Montreal Neurological Institute (MNI) space brain atlas.

As stated above, the differential contrast written words > [faces, houses] did not yield any significant results on the whole brain level. Thus, we decided to run a more focal region of interest (ROI) analysis to capture more subtle effects of the brain–behavior relationship. Literature‐based ROI analysis was computed for the bilateral fusiform, MTG, and the STG; the left IOG, and the SPL. All ROIs were anatomically defined using the aal atlas of the wfupickatlas (Maldjian, Laurienti, & Burdette, 2004; Maldjian, Laurienti, Burdette, & Kraft, 2003; Tzourio‐Mazoyer et al., 2002). The significance threshold for the ROI analysis was set to p < .001 (uncorrected) and k > 10 voxels. To account for the number of regression models, we report the results at a stricter threshold of p corr < .0005 for the concurrent brain–behavior analysis (1 fMRI contrast × 2 cognitive–linguistic skills). For the longitudinal prediction of reading fluency, no further correction was required (1 fMRI contrast × 1 reading skill). A figure of the ROIs and the results of the ROI analyses for all other contrasts is provided in the supplementary materials (Figure S2; Tables S3 and S4).

Note: The content above has been extracted from a research article, so it may not display correctly.

Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.

We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.