Seeing the world through non rose-colored glasses: anxiety and the amygdala response to blended expressions

Abstract

Anxious individuals have a greater tendency to categorize faces with ambiguous emotional expressions as fearful (Richards et al., 2002). These behavioral findings might reflect anxiety-related biases in stimulus representation within the human amygdala. Here, we used functional magnetic resonance imaging (fMRI) together with a continuous adaptation design to investigate the representation of faces from three expression continua (surprise-fear, sadness-fear, and surprise-sadness) within the amygdala and other brain regions implicated in face processing. Fifty-four healthy adult participants completed a face expression categorization task. Nineteen of these participants also viewed the same expressions presented using type 1 index 1 sequences while fMRI data were acquired. Behavioral analyses revealed an anxiety-related categorization bias in the surprise-fear continuum alone. Here, elevated anxiety was associated with a more rapid transition from surprise to fear responses as a function of percentage fear in the face presented, leading to increased fear categorizations for faces with a mid-way blend of surprise and fear. fMRI analyses revealed that high trait anxious participants also showed greater representational similarity, as indexed by greater adaptation of the Blood Oxygenation Level Dependent (BOLD) signal, between 50/50 surprise/fear expression blends and faces from the fear end of the surprise-fear continuum in both the right amygdala and right fusiform face area (FFA). No equivalent biases were observed for the other expression continua. These findings suggest that anxiety-related biases in the processing of expressions intermediate between surprise and fear may be linked to differential representation of these stimuli in the amygdala and FFA. The absence of anxiety-related biases for the sad-fear continuum might reflect intermediate expressions from the surprise-fear continuum being most ambiguous in threat-relevance.

Keywords: ambiguity; amygdala; anxiety; expression; fMRI adaptation; face processing; fear; representation.

Figure 1

Task stimuli and design. (A) Three emotional expression continua were created by morphing between sad and fearful, surprised and sad, and surprised and fearful expressions, using six 16.7% morph steps. These continua were created for both a male and a female actor, faces were taken from the Pictures of Facial Affect set (Ekman and Friesen, 1976). The actors' neutral expressions were also used as stimuli, but not included in the continuum construction. For the main fMRI analysis, using a categorical model, faces were labeled according to the end of the continuum to which they belonged (“A” or “B,” e.g., “surprise” or “fear”) with a third label being given to “50/50” morphs that comprised equal amounts of each end expression. (B) Within the fMRI task, participants were shown faces with expressions from one continuum at a time, presented together with neutral faces and double length null trials in a pseudo-random fashion using a “type 1 index 1 sequence” (Figure S1). Participants were asked to respond by key press to presentation of the neutral face. (C) In the behavioral task, an equivalent type 1 index 1 presentation sequence was used but participants were asked to respond to all faces except the neutral trials, categorizing the faces as showing “mainly” expression “A” or expression “B” for a given continuum (e.g., “mainly surprise” or “mainly fear” for the surprise-fear continuum, as shown in the example.)

Figure 2

Key transitions coded by adaptation regressors in the categorical model. The stimulus presented on each trial “n” was labeled using both direct effect and adaptation regressors. Direct effect regressors code stimulus type. For the categorical model, all faces with 67%+ expression “A” were coded as “end A” (e.g., “surprise,” shown in green), those with 67%+ expression “B” were coded as “end B” (e.g., “fear,” shown in pink), and the 50/50 morph was coded using a third separate regressor. Adaptation regressors describe the difference between the stimulus on trial n and the stimulus on trial n − 1. These are illustrated here for the surprise-fear continuum. “Surprise <> fear”: transitions between any face from the surprise (green, “A”) end of the continuum and any face from the fear (pink, “B”) end of the continuum, where either face type can occur at trial n − 1 or trial n (note, all adaptation effects are coded symmetrically). (ii) “50/50 <> fear”: transitions between a 50/50 surprise/fear morph and a fear (“B”) end face, (iii) “50/50 <> surprise”: transitions between a 50/50 surprise/fear morph and a surprise (“A”) end face. Adaptation regressor values for the additional graded (linear) adaptation model can be found in Figure S2.

Figure 3

Behavioral results. Logistic regression fits across subjects for each of the expression continua: (A) sad-fear; (B) surprise-sad; (C) surprise-fear. Data points indicate the mean proportion of responses across subjects (±SEM) made for each expression morph with reference to the right-ward (“B”) end of the continuum—i.e., proportion of fear responses for the sad-fear and surprise-fear continua, proportion of sad responses for the surprise-sad continuum. The solid line represents the logistic regression fit to the group data. (D) A mixed effects logistic regression model revealed an interaction of trait anxiety x morph content (% fear in morph) for the surprise-fear continuum, β = 0.056, p = 0.0248. To illustrate the effect of anxiety upon categorization responses as a function of morph level, the logistic regression fit is presented for trait anxiety levels ±two standard deviations from the mean (this corresponds to scores of 20 and 59, respectively, the mean STAI trait score being 39.5, SD = 9.75). These fits are plotted using solid lines. Red = high anxious. Green = low anxious. The derivatives of these fitted functions (which gives the rate of change in responses from surprise to fear) are presented using dotted lines. Data points (triangles) for individuals with trait anxiety scores in the top tertile (red) and bottom tertile (green) of the group are also shown. It can be seen that heightened anxiety is associated with a more rapid transition from surprise to fear responses with this difference emerging for morphs with 33–50% fear content (see the derivative plots) and leading to more fear responses being made by high anxious individuals until high levels of fear content (83–100%) are reached, at which point both high and low trait anxious individuals predominantly make fear judgments (see the main fits).

Figure 4

Adaptation effects in the amygdala for the surprise-fear continuum. (A) Activation was extracted and averaged from left and right amygdala ROIs, shown here. These ROIs were specified using the MNI Automated Anatomical Labeling Template (Tzourio-Mazoyer et al., 2002). (B) Group level amygdala adaptation effects (mean and S.E.) for trial-to-trial transitions, as revealed by the categorical model. Significant adaptation of the right amygdala response (decrease in BOLD signal) was observed for transitions between 50/50 surprise/fear morphs and faces from the surprise end of the surprise-fear continuum (“50/50 <> surprise,” see Figure 2) *Group-level one sample t-test against zero, t₍₁₈₎ = −2.37, p <0.05. For transitions between 50/50 surprise/fear morphs and faces from the fear end of the surprise-fear continuum (“50/50 <> fear”), there was higher inter-individual variability and the group level adaptation effect did not reach significance. There was slight but non-significant release from adaptation for transitions between faces from one end of the continuum and the other (“surprise <> fear”). (C) “Adaptation bias” refers to the extent to which adaptation was greater for transitions between 50/50 morphs and fear end faces than between 50/50 morphs and surprise end faces (i.e., “50/50 <> fear”-“50/50 <> surprise”). This was significantly modulated by anxiety, r₍₁₇₎ = −0.44, p < 0.05. High trait anxious individuals showed greater adaptation (i.e., decrease) of the right amygdala BOLD response when 50/50 surprise/fear faces followed or preceded faces from the fear (vs. surprise) end of the continuum, this effect reversing in low trait anxious individuals.

Figure 5

Additional regions of interest from the face processing network. (A) Right OFA as functionally defined by activity to neutral faces vs. baseline, using the graded model (statistical map thresholded at t > 13; activation peak: x, y, z = 40, −86, −4). (B) Right FFA as functionally defined by activity to any emotional face including emotional faces after nulls (statistical map thresholded at t > 7; activation peak: x, y, z: 40, −48, −18). (C) Right STS as functionally defined by activity to any emotional face including emotional faces after nulls (statistical map thresholded at t > 5.25; activation peak: x, y, z: 56, −38, 12). Note. These contrasts were orthogonal to our analyses of interest, as reported within the Results section. Right lateralized ROIs for OFA, FFA, and STS were mirrored across the sagittal plane to create corresponding left lateralized ROIs (not shown).

Figure 6

Contrasting adaptation effects in OFA and FFA: results from the categorical model. (A) The BOLD response for both right and left OFA (averaged across each ROI) showed significant release from adaptation not only for transitions from one end of a continuum to the other (“between end” transitions), right OFA: t₍₁₈₎ = 8.21, p < 0.0001, left OFA: t₍₁₈₎ = 5.94, p < 0.0001, but also for transitions between 50/50 morphs and expressions from either end of the continua (“50/50 <> end” transitions), right OFA: t₍₁₈₎ = 5.61, p < 0.0001, left OFA: t₍₁₈₎ = 3.72, p < 0.005, and for transitions between expressions within a given end of a continuum (“within end” transitions), right OFA: t₍₁₈₎ = 4.21, p < 0.001, left OFA: t₍₁₈₎ = 2.94, p < 0.01. This is consistent with the OFA representing physical differences in expressions even when they are perceived similarly. (B) Both right and left FFA showed significant release from adaptation of the BOLD response for “between end” transitions, right FFA: t₍₁₈₎ = 3.85, p < 0.005, left FFA: t₍₁₈₎ = 3.69, p = 0.005. For “50/50 <> end” transitions, release from adaptation only reached significance in right FFA, right FFA: t₍₁₈₎ = 2.73, p < 0.05, left FFA: t₍₁₈₎ = 1.73, p = 0.10. Further, no significant release from adaptation was observed for “within end” transitions, ps > 0.5. These findings suggest that representation of expressions in OFA may be more graded and less categorical in nature than that within FFA. Bars show group mean (±S.E.) One sample t-tests against 0, ^*p < 0.05, ^**p < 0.01, ^***p < 0.001.

Figure 7

Differences in linear adaptation effects across regions: results from the graded model. Bilateral OFA and FFA showed significant linear release from adaptation of the BOLD response as a function of percentage change in expression (“morph steps”) across continua, right OFA: t₍₁₈₎ = 4.29, p < 0.0005, left OFA: t₍₁₈₎ = 3.61, p < 0.005, right FFA: t₍₁₈₎ = 4.28, p < 0.0005, left FFA: t₍₁₈₎ = 4.10, p < 0.001. In contrast, linear adaptation effects were not significant in either right or left STS or right or left amygdala (ps > 0.1). “amy” = Amygdala. Bars show group mean (±S.E.) One sample t-tests against 0, ^**p < 0.01, ^***p < 0.001.