Correcting for multiple comparisons using cluster-size thresholding (Study 1)
In this method, an initial voxel-level (uncorrected) threshold is set. After that, thresholded maps are submitted to a whole-slab correction criterion based on the estimate of the map's spatial smoothness and on an iterative procedure (Monte Carlo simulation) for estimating cluster-level false positive rates. After 1000 iterations, the minimum cluster size that yielded a cluster-level false positive rate (α) of .05 (5%) was used to threshold the statistical maps of the interaction analysis. This method calculates the size that a cluster would need to be (the cluster threshold) to survive a correction for multiple comparisons at a given statistical level. Only activations whose sizes meet or exceed the cluster threshold are allowed to remain in the statistical maps. The above method has advantages as it is more sensitive than voxel-based analyses and it takes into consideration the fact that voxels are not independent units. This method does not test for false positive results of each voxel but of an entire region. However, it is now recommended to avoid using a liberal threshold of p<0.01, as it exposes the analysis to false positive results. Additionally, clusters may cross anatomical boundaries and thus the spatial resolution that may be considered is lower than the voxel-based correction methods.
The Equalized time model (Study 1)
The response times (RTs) of each condition (GO, GC, EO, EC) were compared in order to test the effect of mean RT on the interaction analysis. A 2X2 repeated-measure ANOVA was conducted (Task:G,E; Condition:O,C), which revealed a significant differences in Task [F(1,17)=30.04,p<0.001], Condition [F(1,17)=10.94,p<0.01, and the interaction effect [F(1,17)=20.99,p<0.001]. Further paired t-test [t(17)=5.51, p<0.001] revealed that the Generation task had longer RTs (M=7.70, SD=2.99) than the evaluation task (M=4.16, SD=2.99). The paired t-test [t(17)=3.31, p<0.01] also revealed that the original condition (M=6.55, SD=2.35) had longer RTs than the characteristic condition (M=5.32, SD=1.16). Finally, Bonferroni corrected (p<.05) paired t-tests revealed that GC (M=6.26, SD=1.95) was significantly longer than the EC condition (M=4.38, SD=0.77) and the EO condition (M=3.96, SD=0.6), and significantly shorter than the GO condition (M=9.15, SD=4.33). The EC condition was significantly longer than EO condition, and shorter than the GO condition.
In order to make the RTs in the four conditions more similar, we created a new protocol, taking the shortest condition (EO) as our standard and correcting the three other conditions to resemble this standard. To this end, we shortened trials in the GO condition that were longer than 4.4sec to make them equal to 4.4sec. Trials in the GC condition longer than 4.6sec were shortened to equal 4.6 sec, and trials in the EC condition longer than 4sec were shortened to equal 4sec.
We then shortened the RTs (as described in Section 2.1.6 Contrast Analysis), and conducted a new 2X2 repeated-measure ANOVA (Task:G,E; Condition:O,C), which revealed non-significant differences in Task [F(1,17)=1.95,p=.18], Condition [F(1,17)=1.42,p=.25], and the interaction effect [F(1,17)=1.68,p=.21]. The new means for the four conditions were: GO=4.18sec, GC=4.19sec, EC=4.1sec, and EO=3.96sec.
The new RTs were used to create a new model. The results of the new interaction analysis [p<0.001, cluster size>474 mm3] (see Figure 1, Table I) were very similar to the results obtained from the former model (see Figure 2).
We therefore believe this analysis confirms that the initial analysis was not affected by the time differences between the four conditions and that it represents the neural networks related to the generation and evaluation systems.
Table I. Interaction between task (G, E) and condition (O, C) – Equal time model
Regions / BAPeak/center-of-gravity / Voxel of peak activation (x,y,z) / F-value
(1,17) / Cluster size (mm3) / GO-GC
T(17) / EO-EC
T(17)
Medial Frontal Gyrus / 6 / -9 / -4 / 58 / 78.79 / 5861 / -4.73*** / 2.61 m
Middle Frontal Gyrus / 6 / -42 / 5 / 52 / 22.06 / 923 / -3.83** / 2.72m
Inferior Frontal Gyrus / 45 / -57 / 20 / 19 / 32.32 / 3463 / -4.65*** / 1.97
Transverse Temporal Gyrus / 41 / -51 / -16 / 10 / 27.86 / 997 / -4.56*** / 1.62
Postcentral Gyrus / 3 / -57 / -10 / 47 / 27.05 / 668 / -4.58* / 1.09
***p<0.0005,**p<0.005,*p<0.025,Xm=marginal (0.025<p<0.05), corrected for multiple comparisons; G=generation task, E=evaluation task, O=originality condition (generation or evaluation), C=characteristics condition (generation or evaluation).
Figure 1. Equal time model:
Figure 2. Variable time model:
AUT scoring (Study 2)
The sample was later extended to include responses from 110 participants. A proposed idea was awarded two points (2) if less than 3% of the participants in the pretest had thought of this idea. One point (1) was awarded for ideas thought of by 3%-5% of pretest participants, and no points (0) were awarded for ideas thought of by more than 5% of pretest participants. Accordingly, originality was measured as the average number of accumulated points awarded for each item.
Motor threshold procedure (Study 2)
The resting motor threshold (rMT) was defined as the lowest stimulus intensity capable of eliciting at least five motor responses in the relaxed APB muscle with amplitude of at least 50 µV in a series of ten consecutive trials of single-pulse TMS. aMT was measured during a voluntary isometric contraction of the contralateral APB with a force level of about 20% of maximal EMG. The aMT was defined as the minimum stimulus intensity required to produce motor responses > 100 µV in five consecutive single-pulse TMS trials.
The average location of the left motor area was X=32, Y=-28, Z=68 (Talairach space), and the average intensity of the motor single-pulse stimulation was 45.46±4.5. The average location of the right motor area was X=-35, Y=-26, Z=67, and the average intensity of the motor single-pulse stimulation was 45.91±4.21.