GLM resultsTouch vs. baseline

To understand touch processing in the dog brain, we first ran t-contrasts in SPM using an anatomical mask of the SI and SII as an inclusive mask.

Contrasting Touch (bilateral) vs. Baseline, we found bilateral clusters located in the somatosensory cortex of the dogs, in a left hemispheric cluster (peak location: -20 -4 12) with k = 65, pFWE < 0.001, T = 4.98, a right hemispheric cluster (peak location: 20 − 6 12) with k = 23, pFWE < 0.05, T = 5.65, and a medial cluster (only at trend level, peak location: 2 0 19, k = 21, pFWE = 0.053, T = 4.53), see Fig. 3A. There is a stronger activation in the touch condition, as opposed to the baseline condition. In the reverse contrasts (Baseline > Touch), investigating relative deactivation during the Touch vs. BL, we found two clusters, one right hemispheric (peak location: 12 2 19) with 105 voxels, pFWE < 0.001, T = 5.26, and one left hemispheric (peak location: -16 2 14) with 63 voxels, pFWE < 0.005, T = 5.05, see Fig. 3B (i.e. higher activation in the baseline condition than during Touch).

Fig. 3

(A) Activations in response to Touch stimulation (whole-brain), N = 22. Bilateral SI and SII activation, as well as medial cluster (whole-brain), presumably in the cruciate sulcus and precruciate, as well as cingulate, gyrus. (B) Stronger responses in baseline than touch predominantly in postcruciate gyri. Image cluster thresholded at 0.005, 40 voxels

On the whole-brain level (without any anatomical or any other mask), all three cortical clusters reached highly significant levels (all pFWE < 0.005), for both activations and deactivations. However, the clusters with stronger activation during Touch vs. BL showed strongly increased extent in the whole-brain analysis, e.g. more than twice the size for the first cluster, and even larger differences for the other clusters (see Table 1), and were located more posteriorly than the clusters in the reverse contrast (see Fig. 3B). In addition to the cortical clusters, we found a cerebellar cluster (peak location: 5–39 4, vermis) with k = 42, pFWE = 0.038, T = 4.81, with stronger activation during Touch vs. BL.

Table 1 Significant clusters in the wholebrain analysis, contrasting Touch (both sides) with Baseline. Clusters are similar to those found in the SVC (see peak locations), however much larger and have generally higher T-values and better significanceLeft-sided touch vs. baseline

Contrasting touch on the left flank vs. Baseline, we found a large cluster in the right hemisphere (peak location: 20 − 6 12) with k = 74, pFWE < 0.001, T = 6.9, as well as a smaller left hemispheric cluster, which only reached trend level in the SVC analysis (peak location: -16 -8 12) with k = 18, pFWE = 0.073 (cluster level), T = 4.75, and a central cluster (peak location: 5 0 19) with k = 30, pFWE = 0.016, T = 4.68, which showed stronger activation during Left touch than BL. In the opposite direction, we found one right hemispheric cluster (peak location: 14 4 20) with k = 70, pFWE < 0.001, T = 6.18, showing stronger activation in Baseline than Left touch, see Fig. 4).

Fig. 4

Blood oxygenetation level-dependent (BOLD) response to touch on left flank (A + B) and right flank separately (C + D). Strong activation in posterior parts of SI and SII and rostro-medial regions. De-activation in postcruciate gyrus (B + D). Image cluster thresholded at 0.005, 50 voxels. L, R, A, P: left, right, anterior, posterior, respectively

On the whole-brain level, a similar picture as in the Touch vs. BL analysis arose: descriptively all clusters were larger with smaller p-values (see Table 2).

Table 2 Significant clusters in the whole-brain analysis of left flank touch vs. baselineRight vs. baseline

Contrasting touch on the Right flank with Baseline, we found a large cluster in the left hemisphere (peak location: -19 -4 12) with k = 83, pFWE < 0.001, T = 5.73, with stronger activation for touch on the Right flank vs. Baseline (see Fig. 4C). In the opposite direction, we found one right hemispheric cluster (peak location: 12 2 18) with k = 79, pFWE < 0.001, T = 5.97, with stronger deactivation during Right flank touch, see Fig. 4D.

Again, on the whole-brain level, we replicated all clusters we found in the SVC analysis, with larger cluster sizes. Additionally, we found a central cluster as well as a left hemispheric cluster (see Table 3).

Table 3 Significant clusters in the whole-brain analysis, contrasting Right touch and Baseline. Note the additional central and left hemispheric clusters, absent in the SVC analysisInvestigating activation outside of anatomical masks

Through the GLM analysis and the notably larger clusters in the whole-brain analysis, it is apparent that the overlap between the observed activation and the regions we expected to be activated (as included in our anatomical masks) was not optimal. Therefore, we ran an exploratory SVC analysis using our anatomical mask as an exclusive mask, purely to better understand and illustrate the location of the activation outside of the anatomical mask. We found 4 significant clusters, one left hemispheric cluster, located close under SII, pFWE = 0.001, k = 84, T = 6.27, peak location at -22 -8 10, one right hemispheric cluster, pFWE = 0.038, k = 42, T = 5.16, peak location at 20 − 6 10. Additionally, we found one rostro-central cluster, pFWE < 0.005, k = 73, T = 4.91, peak location at 0 4 22, and a cerebellar cluster, pFWE = 0.038, k = 42, T = 4.87, at 5–39 4 (all p from cluster level inference). There were no significant clusters in the reverse contrast (BL vs. Touch). The left and right hemispheric clusters were located in the rostral and caudal sylvian gyri (left and right respectively), the central cluster comprised parts of the right cingulate gyrus and right precruciate gyrus. The cerebellar cluster was located in the right hemisphere and parts of the (right) vermis. Figure 5 shows all activations and deactivations in the main contrast (Touch vs. BL) with the contours of the anatomical mask superimposed. Please note that this analysis was not preregistered. We used the same inference levels as for all other analyses.

Fig. 5

Activation in Touch vs. Baseline contrast, whole-brain. Purple contours: somatosensory cortex mask bounds (SI + SII). Clusters outside of SI + SII are ventral and caudal, as well as central. Image cluster thresholded at 0.005, 40 voxels

FIR analysisFirst vs. last time bins

As above, we used an anatomical mask of SI and SII for SVC in these analyses, when not specified differently. Contrasting the first second of Touch stimulation on the right flank with the last (4th ) second of Right touch stimulation, we found only one cluster at trend level, pFWE = 0.071, k = 35, T = 4.02, with peak location at -16 -6 18. This activation is significant when exploring the data using a mask that only included the contralateral left hemispheric SI and SII, with pFWE < 0.005, see Fig. 6. There were no significant clusters in the opposite direction (4th time bin vs. first time bin). For the touch stimulation of the left flank, there was again a trend level cluster, pFWE = 0.078, k = 17, T = 3.89, with the peak location at the exact contralateral side as found in the Right flank simulation, at 16 − 6 18. Again, using small-volume correction for the right hemispheric SI + SII, this cluster reached significance, pFWE < 0.005. There were no significant clusters in the opposite direction (last time bin vs. first time bin). None of the clusters survived on the whole-brain level.

Fig. 6

BOLD response during first second of touch stimulation contrasted to the last second of stimulation. Separately for right and left flank touch. Image of right touch cluster thresholded at 0.1 (to capture the trend level activations), 15 voxels. For left touch, 0.05 and 10 voxels

Time bins vs. Baseline

In addition to contrasting the two time bins where we expected the strongest difference (first and last), we looked at activation differences against Baseline for each time bin, on a whole-brain level. In particular, we were interested in seeing whether the activation visibly shifts across the somatosensory cortex as time progresses. For the first time bin (first second of stimulation), we found four clusters for the Left touch (two in the right hemisphere, one medial, one in the left hemisphere, see Table 4. All clusters survived the whole-brain level. For the first second of Right touch stimulation we found a large left-hemispheric cluster (k = 131), which not only survived on the whole-brain level, but was notably larger (k = 356), see Table 4. For the second and third time bins, there was at minimum one cluster in the hemisphere contralateral to the touch stimulation, which survived whole-brain level, see Tables 5 and 6. There were no significant clusters in the fourth (last) time bin.

Table 4 Significant clusters contrasting the first time bin against Baseline on a whole-brain levelTable 5 Significant clusters contrasting the second time bin against Baseline on a whole-brain levelTable 6 Significant clusters contrasting the third time bin against Baseline on a whole-brain levelVisual inspection of activation shift

While statistical differences between time bins were observed, we also wanted to get an idea of the activation changes as a function of time. To this end we inspected data visually, by plotting the significant clusters found in each time bin onto the Touch vs. BL activation, see Fig. 7 below. From the visual inspections, the following qualitative observations seem to be possible: (1) activation is particularly strong in the first bin, and virtually nonexistent in the final bin. (2) Activation shifts to more central areas from the first to the third bin. (3) The peak of cerebellar activity happens during the second time bin. (4) Activations in the second and third time-bin seem more concentrated on central and dorsal areas, compared to activation during the first bin.

Fig. 7

Clusters of activation by time bin, plotted on top of the Touch vs. BL contrast. Red outlines mark the activation in response to touch in the first time bin (0–1 s), green outlines activation in the second time bin (1–2 s), and blue the activation of the third time bin. No significant clusters were found in the fourth time bin. The figure illustrates both a relative overlap of activations, but also a shift towards more dorsal parts of SI and SII over the course of time

Comparison of first and third time-bin

Since there were no significantly activated clusters in the 4th time bin, we decided to run a comparison between the first and the third bin. Please note that this analysis was not preregistered or planned, but arose from the analysis of the data, i.e. the result that no significant clusters were found in the 4th time bin.

Comparing the first and third time bin on the left and right hemisphere separately, again using SVC with an anatomical mask of SI and SII, we found a left hemispheric cluster, pFWE < 0.001, k = 94, T = 6.12 with peak at -19 -4 12 in response to Right flank touch, and a right hemispheric cluster, pFWE < 0.001, k = 67, T = 7.63 with peak location at 20 − 4 12, and a rostro-central cluster, pFWE = 0.02, k = 23, T = 4.75, peak location at 2 0 19, in response to Left flank touch, see Fig. 8.

Fig. 8

BOLD response during first second of touch stimulation contrasted to the third stimulation second. Separately for Right and Left flank touch. Image cluster thresholded at 0.05, 10 voxels

Laterality quotients

Out of the sample of 22 dogs, 19 showed a lateralization, 9 towards the right hemisphere and 10 towards the left hemisphere (values smaller than − 0.1 or bigger than 0.1 respectively). The mean absolute lateralization quotient was 0.49. For the dogs whose touch processing was lateralized to the left hemisphere, the mean was 0.54 (± 0.36 std), and for the right-hemisphere-processing dogs 0.58 (± 0.4 std).

The sample showed a sample wide asymmetry, pFWE < 0.001, T21 = 5.9 (undirected). This was also true for the left (p < 0.001, T9 = 4.71) and right (pFWE < 0.005, T8 = -4.32) lateralized dogs separately.

Since there is some debate as to how to interpret negative mean beta weights, we plotted the left and right hemispheric cluster means per dog, see Fig. 9 including areas where LQ quotients above 0.1 and below − 0.1 can be found.

Fig. 9

Relationship between mean left and right activation and laterality indices (with absolute values in the denominator), see 2.5.2. Circles represent female, triangles male dogs. Blue markers indicate no lateralization (values between − 0.1 and 0.1), green a rightward processing preference (LQ higher than 0.1), and red a leftward processing preference (lower than − 0.1)