We recruited adults (18 years or more) and children (17 years of age or younger) without a history of neurological diagnosis through community advertisements and word of mouth. In this report, we present data from right-handed participants only. Thirty-six adults (15 male, age range 18–44, mean = 22.9 years, SD 5.9 years) and 19 children (10 male, age range 8–16 years, mean = 11.8 years, SD 2.6) successfully completed the scanning procedure. Three adults recruited to the study were removed from the current analyses; based on their age at assessment they were statistical outliers (± 2SD greater than the group mean) and so data are presented in that group on 33 adults (13 male, age range 18–30, mean = 21.3, SD 2.7).
Adults and parents/guardians of minors provided written informed consent for their child and all children assented to the study. The ethics committees of Aston University (#888) and University of Birmingham (ERN_11-0429), UK, approved the work.
Verbal Memory Paradigm
The fMRI paradigm comprised an incidental verbal encoding task presented binaurally. The overall study design involved three phases: (1) out of scanner familiarization, (2) in-scanner encoding, (3) out of scanner recognition. Words were selected on the basis that they contained 1 to 2 syllables and had an age of acquisition (AoA) rating of less than 600. The latter was derived from an open resource (http://websites.psychology.uwa.edu.au/school/MRCDatabase/mrc2.html) that multiplied the rating scale of Gilhooly and Logie (1980) by 100. Half of the words selected were ‘living’ and half were ‘non-living’ and these were balanced for AoA, familiarity, imageability, and length. During familiarization, participants listened to two words presented binaurally alternately, five times via headphones. Participants pressed a button to indicate whether words were living or non-living to ensure that rehearsal strategies were not employed differentially in adults and children and encourage deep encoding via semantic access. This established them as the ‘familiar’ words for the subsequent ‘novel’ versus ‘familiar’ contrast, because there is strong evidence of enhanced hippocampal activation to novelty (Gabrieli et al. 1997; Stern et al. 1996).
During scanning, words were presented in alternating blocks of ‘novel’ and ‘familiar’ (i.e. repeated) items (5 items per block; 4 blocks novel, 5 blocks familiar), with participants responding ‘living’ or ‘non-living’ via button-press. To overcome scanner noise, words were presented via headphones during the TR delay. To minimize head movement, participants were instructed to focus on a black cross hair presented in the center of a back-projected screen which was located at the rear of the scanner. To reduce the time of scanning, two separate runs were performed using different novel words and these were counterbalanced across participants. The total scanning time per memory run was approximately three minutes. Two runs were completed.
Following scanning and in a room within the imaging suite but outside the scanner, a surprise post-recognition memory test (‘old’ fMRI words versus ‘new’ foils) determined which items were successfully encoded, for later analysis. This was achieved by asking participants to listen to words on a laptop and indicate via button press whether the word was ‘old’ (ie they heard it during when they were inside the scanner) or ‘new’ (no, they did not hear it before now). Only correctly recognized words were included in further analyses by creating custom timing files for each participant to indicate which were correctly encoded.
Data were acquired at two centers, Birmingham University Imaging Centre (BUIC; Philips Achieva 3 T MRI) and Aston Brain Centre (ABC; Siemens Trio Tim, 3 T system), Birmingham, United Kingdom. Whole brain, high-resolution T1-weighted images were acquired for image co-registration in the sagittal plane at both sites (BUIC parameters: TR 8.4 s; TE 3.8 s; voxel size 1 mm isotropic; flip angle 8°; 175 slices. ABC parameters: TR 1.9 s; TE 3.4 s; voxel size 1 mm isotropic; flip angle 15°; 176 slices).
Functional MRI data were acquired with an echo-planar acquisition protocol to measure the blood-oxygen level dependent (BOLD) response (ABC: TR 4110 ms, delay 2300 ms; TE 30 ms; 2 mm isotropic voxels; flip angle 90°; 26 slices. BUIC: TR 4000 ms, delay 2300 ms; TE 35 ms, 2.3 × 2.3 × 2.5 mm voxels; flip angle 80°; 26 slices). To maximize sensitivity to mesial temporal structures data were acquired in a restricted region with the field of view covering the hippocampus and neighboring temporal lobe. On sagittal views the anterior–posterior axis was aligned with the long axis of the hippocampus and the body of the hippocampus was in the center.
All data were analyzed using statistic parametric mapping (SPM12) running in Matlab R2012a (http://www.fil.ion.ucl.ac.uk/spm/). Dummy scans were included prior to the first trial acquisition thus no volumes were removed from the series prior to analysis. Slice time correction was applied, followed by realignment where the mean image was used as reference. 6 motion parameters were extracted during realignment and used as regressors in the whole-brain analysis. A pre-established exclusion criterion was movement exceeding 2 mm (~ 1 voxel size) total displacement. The functional data were initially co-registered to individual anatomical images and then transformed to the standard SPM12 MNI EPI template to facilitate group comparisons. Finally, a 6 mm3 full width by half maximum (FWHM) isotropic gaussian kernel was used to spatially smooth the normalized images.
Statistical analysis at the individual level was estimated using the general linear model (GLM) in SPM12. The first model regressor was input as the novel condition and second regressor as the familiar condition. Movement regressors obtained during preprocessing were included in the model to control for variance associated with in-scanner movement. A high pass filter of 128 s cut-off was implemented to remove low frequency noise components. The key contrast was between encoded novel versus familiar words for whole-brain analyses. Group analysis was performed using a random-effects model and scanner site was included as covariate.
Region of Interest Analyses
Regions of interest (ROIs) were defined for each participant using the automatic anatomic labels (Tzourio-Mazoyer et al. 2002) in the Wake Forest Pick Atlas (Maldjian et al. 2003, 2004). Two ROIs were created for each participant; bilateral mesial temporal masks (MTL; including hippocampi and parahippocampal gyri) and bilateral superior temporal masks (STG; including middle and superior temporal gyri).
Memory asymmetry indices were estimated for the MTL and STG ROIs at individual and group level using the LI Toolbox, where threshold-independent lateralization indices were computed using a bootstrap method (Wilke and Lidzba 2007). Lateralization was categorized as left, right, or bilateral based on a commonly used value of 0.20. ROIs were categorized individually as left lateralized if LI ≥ 0.20, bilateral if LI < │0.20│, or right if LI ≤ − 0.20.
Analysis of variance (ANOVA) was used to assess group differences in MTL and STG laterality between adults and children to provide information on the average LI scores. Chi-square analysis was used to assess whether the proportion of left/right/bilateral LI scores differed between the groups. Given that the resultant sample of children and adults included cases at the age boundary (i.e. 16 for children and 18 for adults), we also exampled the relationship between age and LI scores using Pearson’s correlation across the whole sample. Only individuals with activation clusters of at least 5 voxels were included in the LI analyses (Wilke and Lidzba 2007).