First, the T1 performance was analyzed. The mean accuracy of T1 discrimination across all conditions was 92.7 ± 1.2% (M ± SE, the same below). A three-way ANOVA with factors of the fear-expectation (high, low), expression (fear, neutral) and lag (lag3, lag8) was conducted on T1 accuracy. The main effect of lag was significant [F(1, 29) = 4.95, p = 0.034, η2 p = 0.15, BF10 = 1.21], with higher T1 accuracy at the lag3 condition (93.4 ± 1.1%) than at the lag8 condition (92 ± 1.3%). Other main effects [fear-expectation: F(1, 29) = 0.79, p = 0.38, BF01 = 5.2; expression: F(1, 29) = 0.798, p = 0.38, BF01 = 4.99] and interactions [fear-expectation × expression: F(1, 29) = 2.02, p = 0.17, BF01 = 2.53; fear-expectation × lag: F(1, 29) = 0.59, p = 0.45, BF01 = 3.74; expression × lag: F(1, 29) = 0.32, p = 0.58, BF01 = 3.56; fear-expectation × expression × lag: F(1, 29) = 1.38, p = 0.25, BF01 = 1.97] were not significant.

The mean accuracy of T2 discrimination across all conditions was 68.1 ± 3.1%. An ANOVA with the fear-expectation (high, low), expression (fear, neutral) and lag (lag3, lag8) as factors on T2 discrimination was performed (Fig. 2). The main effect of fear-expectation was significant [F(1, 29) = 8.05, p = 0.008, η2 p = 0.22, BF10 = 3.22]. T2 performance was significantly better in the low fear-expectation condition (69.7 ± 3.1%) than that in the high fear-expectation condition (66.5 ± 3.2%). The main effect of expression was significant [F(1, 29) = 4.71, p = 0.04, η2 p = 0.14, BF10 = 1.96], with higher T2 accuracy for fearful T2 (69.6 ± 2.8%) than for neutral T2 (66.7 ± 3.5%). Besides, the interaction between the fear-expectation and lag was significant [F(1, 29) = 13.96, p = 0.001, η2 p = 0.33, BF10 = 16.8]. Simple effect analysis revealed that there was no significant difference in mean accuracy between high fear-expectation condition (69.3 ± 3.1%) and low fear-expectation condition (68.7 ± 3.2%) in the lag8 condition (p = 0.605), but the accuracy in high fear-expectation condition (63.7 ± 3.7%) was lower than that in the low fear-expectation condition (70.8 ± 3.3%) in the lag3 condition (p < 0.001). More importantly, the interaction of the fear-expectation, expression and lag was significant [F(1, 29) = 5.32, p = 0.03, η2 p = 0.16, BF10 = 0.56]. Further simple effect analysis revealed that in the high fear-expectation condition, there was a significant expression effect in the lag 8 condition only, with significantly higher accuracy of fearful faces than neutral faces [F(1, 29) = 4.92, p = 0.04, η2 p = 0.15, BF10 = 1.73; fear: 71.2 ± 2.7%; neutral: 67.5 ± 1.6%]. In addition, in the low fear-expectation condition, the expression effect was significant in the lag 3 condition only, with significantly higher accuracy of fearful faces than neutral faces [F(1, 29) = 7.4, p = 0.01, η2 p = 0.2, BF10 = 4.16; fear: 74.0 ± 3.3%; neutral: 67.6 ± 3.8%]. The main effect of lag [F(1, 29) = 0.78, p = 0.38, BF01 = 2.53] and other interactions [fear-expectation × expression: F(1, 29) = 1.29, p = 0.27, BF01 = 3.46; expression × lag: F(1, 29) = 0.24, p = 0.63, BF01 = 4.28] were not significant Fig. 3.

Behavioral results of Experiment 1. Mean percentage of the correct T2 identification for fearful and neutral faces in the high fear-expectation condition A and in the low fear-expectation condition B, depicted separately for the lag3 and lag8 condition

ERP data of the correct discrimination for the T1 and T2 in Experiment 2. Grand-average ERPs for fearful (High Fear-ex Fearful, magenta line) and neutral faces (High Fear-ex Neutral, black line) in the high fear-expectation condition, and fearful (Low Fear-ex Fearful, red line) and neutral faces (Low Fear-ex Neutral, blue line) in the low fear-expectation condition at the Fz, Cz, and Pz A; the difference waveforms and 95% confidence interval of ERPs generated by fearful faces minus neutral faces for the high fear-expectation condition (black line) and the low fear-expectation condition (blue line) at the Fz, Cz, and Pz A. The scalp topographies of difference waves between fearful faces and neutral faces in the high and low fear-expectation conditions at 180–220 ms and 470–570 ms B. The bar graphs showing the average amplitude of the four conditions (the fear-expectation × expression) for VPP and P300 C

We split the sample to “blinkers” (who do show an AB) and “non-blinkers” (who do not show an AB) based on whether participants showed an AB effect in the neutral condition or not. AB magnitude was calculated as the percentage of decrement in T2 performance (given that T1 was accurately identified) relative to T1 performance at the lag3 condition according to the following formula: (T1lag3-T2|T1lag3)/T1lag3 × 100%. According to the criteria of blinkers and non-blinkers proposed by Martens and Valchev [25], six participants with an AB magnitude of 10% or less were classified as non-blinkers (mean = − 3.03%), another twenty-four participants as blinkers (mean = 36.12%). For blinkers, an ANOVA with the fear-expectation (high, low), expression (fear, neutral) and lag (lag3, lag8) as factors on T2 discrimination was performed. Still, we observed a similar pattern of results. The main effects of the fear-expectation [F(1, 23) = 11.93, p = 0.002, η2 p = 0.34, BF10 = 5.04] and the expression [F(1, 23) = 10.15, p = 0.004, η2 p = 0.31, BF10 = 19.93] were significant. The interaction between the fear-expectation and lag was significant, F(1, 23) = 15.69, p = 0.001, η2 p = 0.41, BF01 = 27.17. And the interaction of the fear-expectation, expression and lag was significant [F(1, 23) = 6.06, p = 0.022, η2 p = 0.21, BF10 = 1.19]. Further simple effect analysis revealed that in the high fear-expectation condition, there was a significant expression effect in the lag 8 condition only, with significantly higher accuracy of fearful faces than neutral faces (fear: 68.6 ± 3.1%; neutral: 63.3 ± 4.0%, p = 0.01]. In addition, in the low fear-expectation condition, the expression effect was significant in the lag 3 condition only, with significantly higher accuracy of fearful faces than neutral faces (fear: 71.7 ± 3.7%; neutral: 62.7 ± 4.1%, p = 0.001]. The main effect of lag [F(1, 23) = 1.28, p = 0.27, BF01 = 1.19] and other interactions [fear-expectation × expression: F(1, 23) = 1.22, p = 0.28, BF01 = 3.89; expression × lag: F(1, 23) = 0.85, p = 0.37, BF01 = 4.32] were not significant.

The mean accuracy of T1 across all conditions was 96.7 ± 0.6%. A three-way ANOVA with factors of the fear-expectation (high, low) and expression (fear, neutral) was conducted on T1 accuracy. The main effects of the fear-expectation [F(1, 26) = 1.64, p = 0.21, BF01 = 1.19] and the expression [F(1, 26) = 2.31, p = 0.14, BF01 = 3.27], as well as the interaction between the expression and the fear-expectation [F(1, 26) = 0.01, p = 0.99, BF01 = 3.69] were not significant.

The mean accuracy of T2 across all conditions was 78.9 ± 2.7%. A two-way ANOVA with factors of the fear-expectation (high, low) and expression (fear, neutral) performed on T2 discrimination did not reveal significant main effects of the fear-expectation [F(1, 26) = 2.99, p = 0.096, BF01 = 1.02] and the expression [F(1, 26) = 0.37, p = 0.546, BF01 = 4.06]. The interaction effect between the fear-expectation and the expression was also not significant [F(1, 26) = 0.64, p = 0.43, BF01 = 3.14].

According the AB magnitude, seven participants with an AB magnitude of 10% or less were classified as non-blinkers (mean = 4.94%) and another twenty participants as blinkers (mean = 22.89%). Again, a two-way ANOVA with factors of the fear-expectation (high, low) and the expression (fear, neutral) was performed on T2 discrimination for blinkers did not reveal significant main effects of the fear-expectation [F(1, 19) = 3.09, p = 0.095, BF01 = 0.92] and the expression [F(1, 19) = 0.26, p = 0.618, BF01 = 3.49], as well as the interaction effect [F(1, 19) = 0.32, p = 0.578, BF01 = 3.53].

There was a significant main effect of the expression [F(1, 26) = 4.4, p = 0.046, η2 p = 0.15, BF10 = 0.66], with larger VPP amplitudes elicited by fearful faces (3.409 μV) than neutral faces (2.911 μV) (Fig. 3). No further main effect of the fear-expectation [F(1, 26) = 0.109, p = 0.744, BF01 = 4.593] or interaction was significant [F(1, 26) = 0.002, p = 0.968, BF01 = 3.599].

There was no significant main effect of the fear-expectation [F(1, 26) = 2.798, p = 0.106, BF01 = 0.7] or the expression [F(1, 26) = 1.985, p = 0.171, BF01 = 2.16]. However, the interaction between the fear-expectation and the expression was significant [F(1, 26) = 8.147, p = 0.008, η2 p = 0.24, BF10 = 0.972] (Fig. 3). A simple effect analysis revealed that there was no significant difference between the fearful faces (2.798 μV) and neutral faces (2.948 μV) in the high fear-expectation condition (p = 0.758, BF01 = 3.594). However, fearful faces (4.227 μV) elicited larger P300 amplitudes than neutral faces (3.076 μV) in the low fear-expectation condition (p = 0.003, BF10 = 11.936).

We analysis these ERP results only for blinkers. For the VPP, the main effects of the fear-expectation [F(1, 19) = 0.004, p = 0.951, BF01 = 4.35] and the expression [F(1, 19) = 4.01, p = 0.06, BF01 = 1.64] were not significant; and the interaction between the fear-expectation and the expression was also not significant [F(1, 19) = 0.156, p = 0.697, BF01 = 3.11]. Again, we observed similar results of P300 in blinkers. The interaction between the fear-expectation and the expression was significant [F(1, 19) = 10.219, p = 0.005, η2 p = 0.35, BF10 = 1.51]. A simple effect analysis revealed that there was no significant difference between the fearful faces (2.91 μV) and neutral faces (3.28 μV) in the high fear-expectation condition (p = 0.549, BF01 = 2.8). However, fearful faces (4.96 μV) elicited larger P300 amplitudes than neutral faces (3.58 μV) in the low fear-expectation condition (p = 0.006, BF10 = 7.2). The main effect of the fear-expectation was significant [F(1, 19) = 5.023, p = 0.037, BF10 = 4.31], while the main effect of the expression were not significant [F(1, 19) = 1.239, p = 0.28, BF01 = 2.42].