Decomposing fear perception: A combination of psychophysics and neurometric modeling of fear perception

Abstract

Emotion perception is known to involve multiple operations and waves of analysis, but specific nature of these processes remains poorly understood. Combining psychophysical testing and neurometric analysis of event-related potentials (ERPs) in a fear detection task with parametrically varied fear intensities (N=45), we sought to elucidate key processes in fear perception. Building on psychophysics marking fear perception thresholds, our neurometric model fitting identified several putative operations and stages: four key processes arose in sequence following face presentation - fear-neutral categorization (P1 at 100ms), fear detection (P300 at 320ms), valuation (early subcomponent of the late positive potential/LPP at 400-500ms) and conscious awareness (late subcomponent LPP at 500-600ms). Furthermore, within-subject brain-behavior association suggests that initial emotion categorization was mandatory and detached from behavior whereas valuation and conscious awareness directly impacted behavioral outcome (explaining 17% and 31% of the total variance, respectively). The current study thus reveals the chronometry of fear perception, ascribing psychological meaning to distinct underlying processes. The combination of early categorization and late valuation of fear reconciles conflicting (categorical versus dimensional) emotion accounts, lending support to a hybrid model. Importantly, future research could specifically interrogate these psychological processes in various behaviors and psychopathologies (e.g., anxiety and depression).

Keywords: ERPs; Fear perception; Model fitting; Neurometrics; Psychophysics.

Fig. 1. Experimental paradigm and neurometric hypotheses

(A) Example stimuli with the neutral prototype (at 2% of the neutral-fear morph continuum) and the six fear levels (15%–45% in 6% increments). A total of 686 trials (98 trials per morph level) was randomly presented in four blocks. (B) Participants performed a fear detection task: Following a jittered interval, a face appeared (350 ms) at the center of the screen, to which subjects made a forced choice of “neutral” or “fearful” by button-pressing. (C) The psychological processes in question were modelled according to their psychometric functions. The red sigmoidal curve illustrates the process of fear detection. The yellow quadratic curve illustrates the process of fear-neutral categorization. The grey line illustrates the linear process of fear valuation. Lastly, the blue sigmoidal curve illustrates the process of fear awareness.

Fig. 2. Behavior results

(A) Fear detection rate as a function of fear intensity (indexed by fear morph %). The response pattern fits very tightly to a sigmoid function (solid curve) Estimated based on the sigmoid function, the boxes indicate the thresholds for subthreshold/unconscious (20.01% fear) and suprathreshold/conscious (40.06% fear) fear perception, and the inflection point (30.05% fear). (B) Fear detection reaction time as a function of fear intensity. Error bars indicate individual-mean-adjusted S.E.M. (i.e., S.E.E.).

Fig. 3. ERP results

Global field power (GFP) waveform shows the main ERPs of interest (A). Grand average ERP waveforms at site Oz (marked by 5 black dots) show P1 potentials (B) and at site Pz (marked by 5 black dots) show P3 and early and late LPP potentials (C) for the neutral and 6 fear levels. (D) Neurometric curve fitting (red curves) reveals key processes emerging over time in sequence. Emotion categorization: P1 amplitudes conformed to an upward quadratic function; topographical map—two prototype levels minus two boundary levels: [(Levels 2% + 45%) − (Levels 27% + 33%)]/2. Fear detection: P300 amplitudes conformed to a sigmoid function (primarily, upper half; red solid line); topographical map—asymptote minus neutral levels: [average of Levels 21%–45% − 2% (neutral level)]. We also show a downward quadratic curve in red dotted line, which was nevertheless not confirmed by follow-up level-wise contrasts. Fear valuation: early LPP (400–500 ms) amplitudes conformed to a linear function; topographical map—a linear trend: 3*Level 45% + 2*Level 39% + Level 33% − Level 21% − 2*Level 15% − 3*Level 2). Fear conscious awareness: late LPP (500–600 ms) amplitudes conformed to a sigmoid function (primarily, lower half); topographical map—suprathreshold minus asymptote levels: [(Levels 45% + 39%)/2 − average of Levels 2%–33%]. Vertical dotted lines indicate psychophysical cutoffs for 25%, 50% and 75% fear detection. (E) Brain-behavior association—correlation between fear detection rates and ERP amplitudes. Straw plots indicate no correlation for the P1 component, significant but weak correlation for the P3 component, and significant and strong correlation for the two LPP components. Each black line corresponds to the regression line for an individual subject (between ERP amplitudes and fear detection rates); red lines correspond to regression lines for the group. eLPP = early LPP; lLPP = late LPP; *, p<0.05; ***, p<0.001.