Fourteen healthy right-handed male subjects (mean age 24.6 ± 0.71 years) participated in the fMRI experiment. None of the subjects reported a loss of tactile sensation; a history of major medical or neurological illness, such as epilepsy; significant head trauma; or a lifetime history of alcohol dependance. This study was carried out in accordance with the recommendations of Ethics Committee of Human and Animal Experiments, Kyoto University and Okayama University, Japan with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Ethics Committee of Human and Animal Experiments, Kyoto University and Okayama University, Japan.

Five custom-built plastic raised angle stimuli (from 30° to 150° proceeded by 30°) were used in this experiment (Figure 1A). These angle stimuli were raised by 0.5 mm from a 40.0-mm square base as described in our previous study (Wu et al., 2010). An MRI-compatible apparatus was used in this experiment, which can hold 16 raised angle stimuli at once, and all angle stimuli can be presented to the subject automatically by using a set of ultrasonic motors.

We used an event-related fMRI paradigm to assess brain activity during the tactile angle matching processing. Prior to the initiation of the fMRI experiment, all subjects were trained outside of the MR scanner until they felt comfortable performing the task. Then, the subject laid supine in the MRI tunnel with earplugs and was instructed to relax. The subject’s right arm was extended to the device and was comfortably supported by cushions. The subjects placed their right index finger lightly on the surface, with the other fingers resting on a plastic frame. The left index and middle fingers were placed on each of the two buttons of the response box. As shown in Figure 1B, three angle stimuli were moved under subject’s right index finger during a 6-s encoding phase. Then, after a 4-s inter-stimulus interval (ISI), one angle stimulus was presented for 2 s. The subjects were asked to identify whether the last angle stimulus had been presented during the encoding phase by using the response key during a 4-s response phase. The total duration of one trial was 16 s, and one of three intervals (8 s or 10 s or 18 s) followed each trial. Each subject performed a total of 72 trials, which were randomly separated into 10 runs.

Functional MRIs were acquired on a 3T Siemens Trio whole-body MRI system. Standard sequence parameters were used to obtain the functional images as follows: gradient-echo EPI; repetition time (TR) = 2000 ms; echo time (TE) = 30 ms; flip angle = 85; 32 axial slices of 3 mm thickness with 20% slice gap; matrix = 64 mm × 64 mm; and in-plane resolution = 3.0 mm × 3.0 mm. A T1-weighted high-resolution anatomical image volume was obtained from each participant (voxel size = 1 × 1 × 1 mm³) before the acquisition of the functional data.

We used the Statistical Parametric Mapping (SPM12) package (Friston et al., 2007) implemented in Matlab R2007b (MathWorks, US) to process and analyze the fMRI data. Functional images from each run were realigned to the first image and then realigned to the mean image following the initial realignment. Slice-timing correction was then performed to adjust for differences in slice-acquisition times. All realigned images were co-registered to the T1-weighted anatomical image. The T1-weighted anatomical image was normalized to Montreal Neurological Institute (MNI) space using the DARTEL procedure (Ashburner, 2007). The parameters from the DARTEL procedure were then applied to each functional image as well as to the T1-weighted anatomical image. The normalized functional images were filtered using a Gaussian kernel of 8 mm full width at half-maximum (FWHM) in the x-, y- and z-axes.

After the preprocessing, a general linear model (GLM) was fit to the fMRI data for each subject (Friston et al., 1994; Worsley and Friston, 1995). The blood-oxygen-level-dependent (BOLD) signal for all tasks was modeled using box-car functions convolved with the canonical hemodynamic response function (HRF). The design matrix of each subject included 10 runs. The time series for each voxel was high-pass filtered at 1/128 Hz. Assuming a first-order autoregressive model, the serial autocorrelation was estimated from the pooled active voxels using a restricted maximum likelihood procedure and was used to whiten the data. Motion-related artifacts were minimized via the incorporation of six parameters (three displacements and three rotations) from the rigid-body realignment stage into each model. For each subject, we evaluated the linear contrasts of each task phase (encoding and recognition) relative to the baseline. We then obtained the contrast images that were used for the random-effects group analysis. To confirm bilateral SII activation during the tactile angle encoding and recognition phase, we performed a one-sample t-test for each contrast. The height threshold was set at p = 0.005 (uncorrected), and the extent threshold of activation was p < 0.05, which was familywise error (FWE) corrected for multiple comparisons over the whole brain. Coordinates in MNI space were labeled according to probabilistic maps (Eickhoff et al., 2005) in MNI space or the Talairach atlas after coordinate transformation into Talairach space (Lancaster et al., 2000, 2007). To confirm the activation of bilateral SII for different phases, we subsequently conducted an ROI analysis and used the SPM12 to extract the BOLD signal from the whole bilateral anatomical SII areas (Eickhoff et al., 2006a,b) of subjects.

PPIs provide estimates of context-specific changes in effective connectivity between a seed region and other brain regions (Friston et al., 1997; O’Reilly et al., 2012; for review see Friston, 2011). To assess task-dependent contributions of bilateral SII to activity in other brain regions, we performed a generalized form of context-dependent PPI (gPPI) analysis (McLaren et al., 2012). We first determined the coordinates of the bilateral SII in the group analysis by evaluating mean maximum of the encoding phase and recognition phase. We then searched for the participant-specific maxima that were located within the same anatomical region (OP) and within 8 mm from the local group maximum (individual maximum). All voxels depicted by the same contrast (encoding + recognition > baseline, at a threshold of p < 0.05, uncorrected) within 8-mm diameter around the individual maximum served as the seed region for each subject. Time-series data were then extracted. We then calculated the PPI terms between the seed region and psychological factors in the following three steps. First, the extracted MR signal from each seed region was deconvolved with the canonical HRF. The resulting time series represented an approximation of neural activity. Second, the neural time series data were centered and multiplied by the psychological factors of (encoding phase > baseline) and (recognition phase > baseline). Finally, the interaction time series was convolved with the HRF, representing an interaction variable at the hemodynamic level (PPI term). The design matrix at the individual level included not only the PPI regressor but also the time series of the seed region, the phase effect, and regressors of no interest. We evaluated the linear contrast of the PPI regressor for each subject, and the obtained contrast image was used for subsequent group analysis (paired t-test). The height threshold was set at p = 0.005 (uncorrected), and the extent threshold of activation was p < 0.05, which was FWE corrected for multiple comparisons over the whole brain.

All statistical analyses related to the fMRI data were performed using the SPM12 package (Friston et al., 2007). The R package (R Core Team, 2014) was used for all additional statistical analyses. A one sample t-test was performed using the function ‘t.test()’ to evaluate whether the mean task accuracy exceeded the chance level. A repeated measures within-subject analysis of variance (ANOVA) was performed using the function ‘aov car()’ to evaluate the bilateral SII activation at different time points. In addition, a follow up post hoc test was used ‘lsmeans’() function. The significance level for all statistical tests was 0.05. The reported p values for the interaction, simple effect and main effect were Bonferroni-corrected.

To confirm the task performance, we calculated the mean response time (2.27 ± 0.09 s) and accuracy (64.48 ± 1.57%). We performed one sample t-test to compare the mean accuracy with the chance level (50%). We found that the mean accuracy significantly exceeded the chance level (t₍₁₃₎ = 9.208, p < 0.001).

Initially, we confirmed that the encoding and recognition phase of the tactile angle matching task (relative to the baseline) activated a widespread set of brain regions, including the bilateral PoCG, precentral gyrus (preCG), middle frontal gyrus (MFG), medial frontal gyrus (mFG), OP/Insula, inferior parietal lobule (IPL), intra-parietal sulcus (IPS), superior parietal lobule (SPL) and precuneus. In addition to these regions, the contrast of the recognition > baseline additionally indicated activation in the bilateral inferior frontal gyrus (IFG).

As shown in Figure 2A, both the encoding and recognition phase significantly activated bilateral SII. Then, two-way repeated measures ANOVA (2 regions × 14 levels of time points) of activity in bilateral SII revealed a significant main effect of time point (F_(13,169) = 26.667, p < 0.001), but we found no main effect on region (F_(1,13) = 0.657, p = 0.432). Moreover, we also found a significant interaction between the region and time point (F_(13,169) = 12.405, P < 0.001). A post hoc comparison revealed that the BOLD signal changes of the left SII (Figure 2B) for the encoding phase were significantly higher than those for the right SII (i.e., time points 8 [p = 0.0005], 10 [p < 0.0001] and 12 [p < 0.0001]); however, bilateral SII activated at the same level for the recognition phase (time point 18 [p = 0.101]). We also performed a repeated measures ANOVA (2 regions × 2 phases) to examine the difference between encoding and recognition phases for each region. We found a significant main effect of region (F_(1,13) = 10.286, p = 0.007), but we found no main effect on phase (F_(1,13) = 1.820, p = 0.200). In addition, we also found a significant interaction between the region and phase (F_(1,13) = 24.684, p < 0.001). A post hoc comparison revealed that the BOLD signal changes of the left SII for the encoding phase was significantly higher than that for the recognition phase (p = 0.039); however, right SII activated at the same level for the encoding and recognition phases (p = 1).

In this analysis, bilateral SII regions were regarded as a seed region, while the task phase (recognition vs. encoding) was regarded as a psychological factor. We excluded the data of one participant from this analysis because no areas of activation met the criteria of seed selection. As shown in Figure 3A and Table 1, the PPI analysis with a left SII seed region revealed regions of significant activation in the left PoCG extended to the PreCG and right PreCG extended to the MFG. In contrast, the right SII seed PPI analysis revealed regions of significant activation in the left posterior parietal cortex (PPC), including SPL, IPL, IPS and precuneus (Figure 3B).