Functional imaging data

The main effect of faces (all tasks: fear, happy, neutral) against baseline (all rest blocks) across all 50 participants shows activation in the key emotion processing areas in the brain. These areas include the insula, hippocampus, thalamus, ventromedial and dorsolateral prefrontal cortex, precuneus, the fusiform gyrus, and amygdala (Shin & Liberzon, 2010). These areas were also seen when looking at each emotion in isolation (fear vs baseline, happy vs baseline, neutral vs baseline, see Figure 5.1). The differences in activation in the amygdala during these emotionally valenced blocks were investigated further using ROI analysis.

Figure 5.1 showing A) whole brain activation for main effect of faces across all participants, B) activation of fear against baseline, C) activation of happy against baseline and D) activation of neutral against baseline.

119 5.5.2.2 Amygdala Region of Interest (ROI) analysis

Group level

Across all 50 participants the three-way ANOVA revealed no significant main effect of valence (F(2,87) = 1.31, p=.28, η2p= .03), hemisphere (F(1,49) = .78, p=.38, η2p= .02), or any significant two way interactions (valence x lateralisation (F(2,90) = 0.50, p=.59, η2p= .01); valence x session (F(3,151) = 0.48, p=.70, η2p= .01); lateralisation x session (F(2,82) = 0.83, p=.42, η2p= .02) or three-way interaction (valence x lateralisation x session (F(3,171) =.14, p=.95, η2p=.003 ).

There was however a significant main effect of session (F (2,86) = 6.99, p=.002, η2p= .13). Bonferroni post-hoc tests showed that there was a significant decline in activation from session 1 to 2 (p=.005), and session 1 to 3 (p=.002) but not between session 2 and 3 (p=1; see Figure 5.2).

Figure 5.2 (a) Graph showing significant differences in amygdala activation over time sessions showing a pattern of habituation, (b) graph representing group level amygdala activation to difference valences, (c) graph showing left and right amygdala activation during presentation of face stimuli. Bars represent mean±SE *p<0.05, ** p<0.01, ***p<0.001

a b

120 Gender interaction

Comparison of amygdala activation across emotional blocks revealed that women exhibited significantly higher amygdala activation (x̄ =1.74, SD = 0.90) compared to male participants (x̄ =1.16, SD = 1.18; t (48) = -1.96, p = 0.028, d = 0.49).

Results from the mixed ANOVA with gender as a between subjects’ factor revealed a similar pattern of main effects as the group level interaction. However, it was the interactions with gender that are of interest. There was a trend towards a significant interaction between gender and valence (F (2,96) = 2.84, p=.063, η2_{p = .06). When this was investigated further it was} revealed that there was a significant main effect of valence in the female participants (F (2,56) = .81, p= .005, η2p = 0.17), but not in male participants (F (2,40) =.11, p = .90, η2p = 0.005; see Figure 5.3). Post-hoc Bonferroni tests on the female participants revealed that there is a significant difference between the fear condition and neutral condition (p =.01), a non-significant trend between the fear and happy condition (p=.06) and no significant difference between the happy and neutral condition (p=1) in women.

Figure 5.3. Graph showing the interaction of gender by valence in amygdala activation. As seen in the green bars, there are no significant differences in amygdala activation to different valences in male participants. In contrast there are significant differences in amygdala activation in women (red bars) between fear and neutral, and a trend towards a significant difference between fear and happy conditions. Bars represent mean±SE *p<0.05, ** p<0.01, ***p<0.001.

There were no other significant interactions (gender x hemisphere (F(1,48)=0.03, p= .87, η2 p = 0.01); gender x session (F(2,85)=2.29, p= .11, η2_{p = 0.05); gender x session x hemisphere} (F(2,81)=0.07, p= .90, η2_{p = 0.002);gender x valence x hemisphere (F(2,96)=1.08, p= .34, η}2

121 = 0.02);gender x session x valence (F(3,150)=1.12, p= .35, η2p = 0.02); gender x session x hemisphere x valence (F(4,192)=0.48, p= .75, η2p = 0.01)).

Anxiety interaction

Overall amygdala activation was significantly greater in the high anxiety group (x̄ =1.87± 0.81) than the low anxiety group (x̄ = 1.32± 1.12; t(48) = -1.73, p = 0.05, d = 0.63).

Again the mixed ANOVA with anxiety as a between groups variable revealed a similar pattern of main effects as the group level interaction. In addition, there was a significant three way interaction of session x hemisphere x anxiety group (F(2,82) = 3.87, p = .031, η2_{p= .08).} Post-hoc tests adjusting for multiple comparisons on the three-way interaction revealed that in the low anxiety group both the right and left amygdala showed a significant reduction in activation from session 1 to 2 (Right: p=.005; Left: p= .023) and session 1 to 3 (Right:

p=.003; Left: p< .001) but not session 2 and 3 (Right: p=1; Left: p=.43). In contrast activation levels were stable across sessions for the right and left amygdala in the high anxiety group (session 1 to 2 Right: p=.38, Left: p= .97; session 1 to 3 Right: p=.86, Left: p=.86; session 2 to 3 Right: p=1, Left: p=1), suggesting an absence of habituation within this group (see Figure 5.4).

Fix 5.4 Graphs representing amygdala activation change over time showing habituation patterns in the left amygdala (a) and right amygdala (b) amygdala in the high anxiety (orange bars) and low anxiety (purple) groups. Bars represent mean±SE *p<0.05, ** p<0.01, ***p<0.001

No other significant interactions with anxiety were observed (anxiety x hemisphere (F(1,48)=0.98, p= .33, η2p = 0.02); anxiety x session (F(2,83)=2.18, p= .13, η2p = 0.04); anxiety x valence (F(2,86)=0.25, p= .76, η2p = 0.005); anxiety x valence x hemisphere

a

122 (F(2,96)=0.20, p= .82, η2p = 0.004); anxiety x session x valence (F(3,148)=0.83, p= .48, η2p = 0.02);anxiety x session x hemisphere x valence (F(3,166)=0.60, p= .64, η2p = 0.01)).

5.5.2.3 Control Region of Interest: Fusiform Gyrus Group Level

The three-way ANOVA across all 50 participants revealed no significant main effect of valence (F(2,98)=.53, p= .59, η2p = 0.01), but significant main effects of session

(F(2,83)=19.91, p<.001, η2p = 0.29) and hemisphere (F(1,49)=9.08, p= .004, η2p = 0.16). Follow up post-hoc analysis of the main effect of session showed a significant reduction in activation from session 1 to 2 (p<0.001) and session 1 to 3 (p<.001) but not session 2 and 3 (p=1). The main effect of hemisphere was revealed to be caused by greater activation in the right fusiform gyrus (x̄ =2.45± 1.32) compared to the left (x̄=2.10± 1.17; t(49)= -3.01, p=.004, d=0.89) via simple effects analysis controlling for multiple comparisons.

There was also a significant interaction of session by hemisphere (F(2,98)=6.82, p= .002, η2p = 0.12). Following up this interaction revealed significantly greater activation in right

fusiform gyrus compared to left in session 1 t(49)= -3.57, p=.001, d=0.33; Right: x̄ =0.99± 0.48, Left: x̄ =0.83± 0.44) and session 3 (t(49)= -2.68, p=.010, d=0.22; Right: x̄=0.75± 0.49, Left: x̄ =0.65± 0.43). However, this difference did not pass bonferroni correction threshold in session 2 (t(49)= -2.33, p=.024, d=0.19; Right: x̄ =0.71± 0.45, Left: x̄=0.62± 0.41), see Figure 5.5.

123 Fix 5.5. Graph showing mean fusiform activation over sessions showing habituation in the left (solid blue) and right (patterned blue) hemisphere across all participants. Bars represent mean±SE *p<0.05, ** p<0.01, ***p<0.001

No other significant interactions were observed (valence x session (F(3,164)=0.16, p= .94, η2_{p = 0.003); valence x hemisphere (F(2,98)=1.83, p= .17, η}2_{p = 0.04); valence x session x} hemisphere (F(4,196)=.35, p= .85, η2p = 0.01)).

Gender Interaction

In the mixed ANOVA with gender as a between subjects factor there was a similar pattern of main effects as the group level analysis. However, there was only a significant interaction of gender with session (F(2,83)=5.58, p= .008, η2p = 0.10). Following this up revealed that female participants showed a significant decline from session 1 to 2 (p=.001), and 1 to 3 (p<.001) with no differences between session 2 and 3 (p=.523) as seen at group level. However, male participants only demonstrated a significant decline from session 1 to 2 (p<.001), with no other session effects passing significance bonferroni correction threshold (session 1-3 p=.058; session 2-3 p=.102; see Figure 5.6).

124 Figure 5.6. Graph representing mean fusiform activation over time in male (green) and female (red) participants. Bars represent mean±SE *p<0.05, ** p<0.01, ***p<0.001

Gender did not interact with any other factors in the fusiform gyrus (gender x valence (F(2,96)=1.18, p= .31, η2p = 0.02); gender x hemisphere (F(1,48)=1.62, p= .21, η2p = 0.03); gender x valence x session (F(3,160)=.34, p= .82, η2p = 0.01); gender x valence x hemisphere (F(2,96)=0.34, p= .71, η2p = 0.01); gender x session x hemisphere (F(2,96)=0.28, p= .75, η2p = 0.01); gender x valence x session x hemisphere (F(4,192)=0.25, p= .91, η2_{p = 0.01)).}

Anxiety Interaction

Results from the mixed ANOVA with anxiety group as between subjects factors revealed a similar pattern of main effects to the group level analysis. However, there was no significant interaction of anxiety group with any aspect of fusiform gyrus activation (anxiety x valence (F(2,96)=.46, p= .64, η2p = 0.01); anxiety x hemisphere (F(1,48)=.06, p= .81, η2p = 0.001); anxiety x session (F(2,78)=2.62, p= .08, η2p = 0.05); anxiety x valence x session

(F(3,160)=0.87, p= .47, η2p = 0.02); anxiety x valence x hemisphere (F(2,96)=.21, p= .81, η2p = 0.004); anxiety x session x hemisphere (F(2,96)=1.22, p= .30, η2p = 0.03); anxiety x

125