Subliminal Emotional Faces Elicit Predominantly Right-Lateralized Amygdala Activation: A Systematic Meta-Analysis of fMRI Studies

Abstract

Prior research suggests that conscious face processing occurs preferentially in right hemisphere occipito-parietal regions. However, less is known about brain regions associated with non-conscious processing of faces, and whether a right-hemispheric dominance persists in line with specific affective responses. We aim to review the neural responses systematically, quantitatively, and qualitatively underlying subliminal face processing. PubMed was searched for Functional Magnetic Resonance Imaging (fMRI) publications assessing subliminal emotional face stimuli up to March 2022. Activation Likelihood Estimation (ALE) meta-analyses and narrative reviews were conducted on all studies that met ALE requirements. Risk of bias was assessed using the AXIS tool. In a meta-analysis of all 22 eligible studies (merging clinical and non-clinical populations, whole brain and region of interest analyses), bilateral amygdala activation was reported in the left (x = -19.2, y = 1.5, z = -17.1) in 59% of studies, and in the right (x = 24.4, y = -1.7, z = -17.4) in 68% of studies. In a second meta-analysis of non-clinical participants only (n = 18), bilateral amygdala was again reported in the left (x = -18, y = 3.9, z = -18.4) and right (x = 22.8, y = -0.9, z = -17.4) in 56% of studies for both clusters. In a final meta-analysis of whole-brain studies only (n=14), bilateral amygdala was also reported in the left (x = -20.2, y = 2.9, z = -17.2) in 64% of studies, and right (x = 24.2, y = -0.7, z = -17.8) in 71% of studies. The findings suggest that non-consciously detected emotional faces may influence amygdala activation, especially right-lateralized (a higher percentage of convergence in studies), which are integral for pre-conscious affect and long-term memory processing.

Keywords: Activation Likelihood Estimation; amygdala; emotional faces; parahippocampal gyrus; subliminal.

Figure 1

PRISMA flow diagram of the relevant steps for the literature search conducted in PubMed (Page et al., 2021). Out of the initial 943 identified studies, 22 fit the inclusion criteria and were included in the meta-analysis and narrative review. During the identification stage, n = 708 studies were removed prior to screening. The n = 238 remaining studies were screened and n = 76 studies were sought for retrieval. From the n = 74 reports assessed for eligibility, n = 52 studies were excluded according to exclusion criteria, leaving n = 22 studies.

Figure 2

Significant ALE cluster maxima of neural activation to subliminal emotional face stimuli surviving FDR correction, p < 0.05 for multiple comparisons, cluster size >100 mm³. Montreal Neurological Institute (MNI) coordinates are given. Meta-analysis 1 peak clusters: left amygdala (x = −19.1, y = 1.5, z = −17.1), right amygdala (x = 24.4, y = −1.7, z = −17.4). Meta-analysis 2 peak clusters: left amygdala (x = −18, y = 3.9, z = −18.4), right amygdala (x = 22.8, y = −0.9, z = −17.4). Meta-analysis 3 peak clusters: left amygdala (x = −20.2, y = 2.9, z = −18.4), right amygdala (x = 22.8, y = −0.9, z = −17.4).

Figure 3

Summary diagrams of the reported clusters of significant brain activation in response to subliminal emotional stimuli in the reviewed fMRI studies. Subliminally presented faces expressed (A) mixed emotional expressions, (B) sadness, (C) anger, (D) fear, (E) surprise, (F) happiness, and (G) disgust. The number within each bar indicates the n of the studies. ACC, anterior cingulate cortex; NAcc, nucleus accumbens.