Oxytocin facilitates top-down and bottom-up attention to emotional faces in a general and temporal-dependent manner

Abstract

Background

Oxytocin (OT) exerts widely modulatory effects on socio-emotional functions in humans, which can be achieved via enhancing the salience of social cues by interacting with the dopaminergic attention system. However, there is a lack of direct evidence for OT modulating attentional processing, with its underlying neural mechanisms remaining to be elucidated.

Methods

In a double-blind, placebo-controlled, between-subject design, 60 healthy male participants were recruited. We combined pharmaco-electroencephalography with two modified tasks (a cue-target visual search [CTVS] task and a face distractor interference [FDI] task) to investigate whether intranasal OT can modulate attentional processing of social cues in top-down versus bottom-up task sets.

Results

In the CTVS task, OT accelerated participants’ response time to target faces, which was paralleled by a larger N170 and stronger theta power, suggesting that OT promoted early top-down attentional processing of social cues. In the FDI task, OT inhibited the distractive effect of task-irrelevant emotional faces in the first half of the task via facilitating top-down attentional control to targets as reflected by enhanced attentional selection (increased N2pc) and more efficient attentional processing (decreased P300). However, in the second half, OT switched from facilitating top-down attentional control to potentiating bottom-up attentional capture by emotional face distractors, as evidenced by OT reducing response accuracy but having no effects on the N2pc and P300.

Conclusions

Our findings not only provide evidence for the role of OT in modulating attentional processing of social cues but also lend support to its therapeutic potential in normalizing such attentional deficits.

Introduction

Oxytocin (OT) is a well-established hypothalamic neuropeptide primarily recognized for its crucial role in maternal behaviors, such as promoting uterine contractions during labor and milk ejection during lactation (Insel, Young, & Wang, 1997; Kendrick, 2000). Beyond these early discovered peripheral functions, OT has been demonstrated a wide range of modulatory effects on social and emotional functions in both animals and humans, including attachment formation, emotional memories, and more complex socio-cognitive behaviors such as interpersonal trust, social memory, and decision-making (Kendrick, Guastella, & Becker, 2018; Ma, Shamay-Tsoory, Han, & Zink, 2016; Quintana et al., 2021; Yao & Kendrick, 2022, 2025).

Several theoretical frameworks have been proposed to interpret OT’s effects, among which the social salience hypothesis is one of the most influential. This hypothesis posits that OT exerts its functional effects by increasing the salience of social cues via interacting with the dopaminergic attention system (Shamay-Tsoory & Abu-Akel, 2016). The important role of OT in attentional processing has also been emphasized by a recent hierarchical ‘SSS’ (survival, security, and sociability) model proposed by Yao and Kendrick (2025). Empirical support can be generally found in previous studies showing intranasal OT increasing gaze fixation especially toward human eye region (Domes et al., 2007; Guastella, Mitchell, & Dadds, 2008), facilitating socially directed gaze following (Le et al., 2021), and improving facial emotion recognition (Fischer-Shofty, Shamay-Tsoory, Harari, & Levkovitz, 2010; Marsh, Yu, Pine, & Blair, 2010). OT has also been found to enhance the detection accuracy of briefly presented emotional faces masked by neutral ones (Schulze et al., 2011). These findings together suggest that OT may influence emotional recognition even at an early stage of visual attentional processing, which may be underpinned by OT modulating amygdala reactivity (Gamer, Zurowski, & Büchel, 2010; Kanat et al., 2015; Kanat, Heinrichs, Schwarzwald, & Domes, 2015). Subsequent studies have provided more direct evidence for OT facilitating early attentional shifting toward emotional faces in spatial cueing and dot-probe paradigms (Domes et al., 2013; Ellenbogen et al., 2012). Using eye-tracking techniques, OT has also been found to impair attentional inhibition to social stimuli in an anti-saccade paradigm, suggestive of OT increasing attention toward social cues in a bottom-up manner (Xu et al., 2019; Zhuang et al., 2022).

Since eyes and facial expressions constitute essential cues of information during social interactions (Cañigueral & Hamilton, 2019; Frith, 2009; Knudsen, 2007; Raymond, 2009), OT’s effects on facilitating attention toward these cues may therefore underpin its actions on higher-order processing or complex behavior, such as cognitive evaluation or social cooperation. This mechanism has been emphasized by the hierarchical ‘SSS’ model, in which enhanced attention can be achieved via a top-down or bottom-up format, possibly via OT interacting with the noradrenergic and dopaminergic systems (Yao & Kendrick, 2025). There is also evidence for OT improving attentional dysfunctions in psychiatric disorders, including autism, schizophrenia, and depression (Bradley et al., 2019; Domes, Normann, & Heinrichs, 2016; Kanat et al., 2017). However, it should be noted that OT’s actions on social attention in previous studies were mainly demonstrated via behavioral measures of response time (RT) or accuracy (RA) (Domes et al., 2007; Marsh, Yu, Pine, & Blair, 2010; Schulze et al., 2011; Xu et al., 2019; Zhuang et al., 2022), which may not specifically reflect its effects on attentional processing but instead could be the result of multiple processes. There is still a lack of direct evidence for OT’s effects on attentional processing of social cues, particularly at the neural level.

Against these backgrounds, the present study combined the electroencephalography (EEG) with two modified visual search tasks: a cue-target visual search (CTVS) task (top-down) and a face distractor interference (FDI) task (bottom-up), which both incorporate emotional faces, to investigate whether intranasal OT modulated attentional processing of social cues in these two contexts. More specifically, in addition to behavioral indices such as RT and RA conventionally used in previous studies, we utilized three event-related potential (ERP) components that are typically associated with attentional processing, including the N170, N2pc, and P300 components as primary neural indices. While the N170 is a face-specific ERP component reflecting very early visual attention processing of face stimuli (Eimer, 2011; Hinojosa, Mercado, & Carretié, 2015), the N2pc is clearly associated with visual selective attention (Kiss, Van Velzen, & Eimer, 2008; Woodman & Luck, 1999) and has been frequently used as a neural index for visual attentional processing of emotional faces (Bola, Paź, Doradzińska, & Nowicka, 2021; Yao, Ding, Qi, & Yang, 2013, 2014). The P300 is a late ERP component related to attentional resource allocation, and its amplitudes can be used as an index of neural processing efficiency (Edwards et al., 2021; Gongora et al., 2021; Polich, 2012). These ERP components, therefore, enabled us to examine OT’s effects on attention-related processing of social stimuli at both early and late stages. Given that neural oscillations, such as theta band power, have also been associated with attention (Fiebelkorn & Kastner, 2019; Karakaş, 2020), we further explored whether OT would impact neural patterns in the frequency domain. Based on studies showing that intranasal OT enhances or biases attention to social cues (Di Simplicio & Harmer, 2016; Ellenbogen, 2018; Hovey et al., 2020; Yao et al., 2018), we hypothesized that in the CTVS task, whereby emotional faces were preset as targets indicated by task cues before visual search, OT would facilitate behavioral performance during visual search of face targets. At the neural level, we expected to observe that OT would increase N170, N2pc amplitudes, and theta power, associated with visual attentional processing (Hinojosa, Mercado, & Carretié, 2015; Karakaş, 2020; Kiss, Van Velzen, & Eimer, 2008), but decrease P300 amplitudes related to neural processing efficiency (Edwards et al., 2021; Gongora et al., 2021). By contrast, in the FDI task, where emotional faces served as task-irrelevant distractors, we hypothesized that OT would be detrimental to the behavioral performance of goal-directed judgment based on previous findings of OT making task-irrelevant emotional faces more distracting (Ellenbogen, Linnen, Cardoso, & Joober, 2013; Xu et al., 2019; Zhuang et al., 2022). Correspondingly, at the neural level, we expected opposite effects of OT to the CTVS task, suggestive of OT increasing attention to those task-irrelevant face distractors.

Materials and methods

Participants and treatment

Sixty healthy male participants (mean age = 21.07 years, standard deviation = 2.14) were recruited for the present study. This sample size was adequate to achieve a power > 80% (effect size = 0.25, α = 0.05) for a mixed analysis of variance (ANOVA) based on a priori power analysis using the G*Power v.3.1 toolbox (Faul, Erdfelder, Buchner, & Lang, 2009). To control for potential confounding effects, participants completed questionnaires of mood states and personality traits before treatment. Participants self-administered 24 international units of OT or PLC intranasally following a standardized protocol (Guastella et al., 2013) and tasks started 45 minutes post-treatment (Figure 1a). All participants were provided with written informed consent before study involvement, and all procedures conformed to the latest version of the Declaration of Helsinki and were approved by the ethical committee of University of Electronic Science and Technology of China. The study was preregistered on clinicaltrials.gov (NCT04413786). Details of participant recruitment/exclusion, questionnaires, and treatment procedure were reported in the Supplementary Material.

Figure 1.

(a) Experimental protocol. (b) Timeline of the cue-target visual search task (top-down). (c) Timeline of the face distractor interference task (bottom-up). Faces used in the present study are from the Taiwanese Facial Expression Image Database (Chen & Yen, 2007) and are masked following terms of use. Icons were obtained from flaticon.com under the free license with attribution.

Experimental tasks

The present study consisted of a CTVS (top-down) and an FDI (bottom-up) attention task modified based on previous studies (Delchau et al., 2020; Moradi, Mehrinejad, Ghadiri, & Rezaei, 2017), with the task order being counterbalanced across participants. In the CTVS task (Figure 1b), participants were instructed to judge the location (left or right) of an emotional face target on a screen in a low load condition of two faces versus a high load condition of four faces. Visual search of the target face was therefore goal-directed and engaged top-down attentional processing. The target face appeared at a random location, with left or right sides being counterbalanced. In the FDI task (Figure 1c), participants were clearly instructed to judge the direction (upward or downward) of a ‘U’ shape target overlaid on the nose of a neutral face on the stimulus screen as quickly and accurately as possible. Of note, the emotional face used in each trial always appeared at the contralateral side of the ‘U’ target. Participants were explicitly required to judge the direction of the ‘U’ shape, and emotional faces were thus used as task-irrelevant distractors. This created a situation in which emotional face distractors competed with ongoing goal-directed attention to the ‘U’ target, thereby allowing measurement of how OT modulated stimulus-driven attentional capture, normally indicated indirectly by measuring its interference effects on top-down task performance (e.g. Belkaid, Cuperlier, & Gaussier, 2017; Moradi, Mehrinejad, Ghadiri, & Rezaei, 2017). There were also two load conditions similar to the CTVS task (see Supplementary Material for a more detailed description of both tasks). Of note, although attentional sets were manipulated in a top-down and bottom-up manner separately in these two tasks, they may be treated more as attentional tendencies rather than strictly categorical distinctions. For example, even when attention is guided in a top-down manner, salient features of emotional faces could still capture attention in a bottom-up manner at an early stage.

EEG data acquisition and analyses

The EEG data were recorded using a 64-channel ActiCap system and preprocessed using the EEGLAB 14.1.1 toolbox (Delorme & Makeig, 2004). Different ERP components (the N170, N2pc, and P300) and theta band power related to attentional processing at different stages were calculated and averaged across trials for each condition in each participant for subsequent analyses (see Supplementary Material for detailed description of EEG data analyses).

Statistical analyses

Independent t-tests or repeated-measures ANOVAs were conducted to analyze questionnaire scores, behavioral data, or EEG data, respectively. The Greenhouse–Geisser correction was employed, whereby assumptions of sphericity were violated, and the Bonferroni correction was applied for post-hoc multiple comparisons. For the N170, given that we examined changes in its amplitudes at two hemispheres, a Bonferroni correction was also applied to account for the number of hemispheres. Correlations between personality traits, behavioral responses, and neural signals were analyzed using Pearson or Spearman correlations, depending on data distribution, with correlation group differences being tested using Fisher’s z-transformation test (see Supplementary Material for details).

Results

Demographics and questionnaires

Independent t-tests on personality traits revealed no significant group differences (see Supplementary Table S1). There were also no significant group differences in both pre- and posttreatment measures of positive and negative mood (see Supplementary Table S2).

Behavioral results of the CTVS task

A repeated-measures ANOVA on RT revealed a significant main effect of face (F(1.58, 82.20) = 76.99, p < 0.001, η² _p = 0.60), with participants responding fastest to happy faces but slowest to angry faces (Supplementary Figure S1a). The main effect of load was significant (F(1, 52) = 1337.54, p < 0.001, η² _p = 0.96), as reflected by a faster RT under the low compared to the high load condition. Importantly, there was a significant main effect of treatment (F(1, 52) = 7.37, p = 0.009, η² _p = 0.12), with participants responding faster in the OT than in the PLC group (Figure 2a). Interaction between treatment and load was also significant (F(1, 52) = 4.68, p = 0.035, η² _p = 0.08). Post-hoc analyses indicated that although OT accelerated RT under both load conditions, OT’s effects were stronger under the high (p = 0.009) relative to the low load (p = 0.014; Figure 2b). For RA, there was a significant main effect of face (F(2, 104) = 13.80, p < 0.001, η² _p = 0.21), with the highest accuracy for happy faces and the lowest one for fearful faces (see Supplementary Figure S1b). The main effect of load was also significant (F(1, 52) = 11.44, p = 0.001, η² _p = 0.18), with a higher accuracy under the low compared to the high load. There were no other significant main or interaction effects (ps ≥ 0.212).

Figure 2.

(a) Response time for judging the location of target face stimuli in the oxytocin (OT) and placebo (PLC) groups across conditions in the cue-target visual search task. (b) Response time for judging the location of the target face stimuli under the low and high load conditions in the two groups. (c) N170 components across conditions and corresponding topographical maps following OT and PLC treatments. (d) Correlations between N170 amplitudes across conditions and alexithymia scores in the OT and PLC groups. (e) Theta band (4–7.5 Hz) power changes at frontal electrodes following OT and PLC treatments across conditions (*p < 0.05, **p < 0.01). Error bars indicate the standard error of the mean.

Neural results of the CTVS task

On the neural level, the N170, N2pc, and P300 components were analyzed to examine whether OT’s effects on social stimuli were exerted via modulating early or late attentional processing. A repeated-measures ANOVA on N170 amplitudes in the left hemisphere revealed a significant main effect of treatment (F(1, 52) = 4.75, p = 0.034, η² _p = 0.08), which became marginal after Bonferroni correction for the number of hemispheres (p = 0.068), with a larger N170 following OT compared to PLC treatment (Figure 2c). The main effect of load was also significant (F(1, 52) = 34.33, p < 0.001, η² _p = 0.40), with larger N170 amplitudes in the high relative to the low load condition. However, no other significant effects were observed (ps ≥ 0.289). While larger N170 amplitudes were also found following OT treatment in the right hemisphere (OT: −1.57 ± 2.74 versus PLC: −0.56 ± 3.08 μv), the difference did not reach significance (p = 0.207). Pearson correlation analyses found a positive correlation between N170 amplitudes in the left hemisphere across conditions and alexithymia scores in the PLC group (r = 0.523, p = 0.005) but a marginal negative correlation in the OT group (r = −0.344, p = 0.079; Figure 2d). The Fisher z-transformation test showed a significant correlation difference between groups (Fisher z-score = −3.253, p = 0.001).

For the N2pc, the main effect of face was significant (F(2, 104) = 3.34, p = 0.039, η² _p = 0.06), but post-hoc analyses showed no significant difference in pairwise comparisons (ps ≥ 0.091; Supplementary Figure S2). The main effect of load was significant (F(1, 52) = 9.25, p = 0.004, η² _p = 0.15), with larger N2pc amplitudes under the low relative to the high load. No significant main effect of treatment or interactions was found (ps ≥ 0.487). For the P300, the main effect of load was significant (F(1, 52) = 27.38, p < 0.001, η² _p = 0.35), with larger P300 amplitudes under the low load than the high load. However, no other significant effects were found (ps ≥ 0.109).

In the time-frequency analysis, a repeated-measures ANOVA on theta band power showed a significant main effect of treatment (F(1, 52) = 4.90, p = 0.031, η² _p = 0.09), with a stronger theta power in the OT compared to the PLC group (Figure 2e). The main effect of load was also significant (F(1, 52) = 6.16, p = 0.016, η² _p = 0.11), with a higher theta power in the high load condition compared to the low load condition. However, neither the interaction nor the main effect of face was significant (ps ≥ 0.133).

Behavioral results of the FDI task

A repeated-measures ANOVA on RT showed that while the main effect of treatment was not significant (F(1, 52) = 1.45, p = 0.234, η² _p = 0.03), the main effects of face (F(3, 156) = 3.43, p = 0.019, η² _p = 0.06) and load (F(1, 52) = 475.83, p < 0.001, η² _p = 0.90) were significant, with a faster RT in judging the ‘U’ direction presented together with neutral relative to angry faces (p = 0.040) and under the low relative to the high load. There were no significant interactions for RT (ps ≥ 0.053). A similar repeated-measures ANOVA on RA revealed a significant main effect of treatment (F(1, 52) = 4.36, p = 0.042, η² _p = 0.08), as reflected by a lower accuracy in judging the direction of the target ‘U’ in the OT compared to the PLC group (Figure 3a). However, no other significant main effects or interactions were found (ps ≥ 0.161).

Figure 3.

(a) Choice accuracy for judging the direction of the target “U” following OT and PLC treatments across conditions in the face distractor interference task. (b) N2pc amplitudes at the electrodes of PO7 and PO8 following OT and PLC treatments across conditions. (c) P300 amplitudes at the electrode CPz and topographical maps following OT and PLC treatments across conditions (*p < 0.05). Error bars indicate the standard error of the mean.

Neural results of the FDI task

For the N170, a significant main effect of load was found at electrodes in both the left (F(1, 52) = 64.03, p < 0.001, η² _p = 0.55) and right hemispheres (F(1, 52) = 59.29, p < 0.001, η² _p = 0.53), with larger N170 amplitudes in the high relative to low load. However, no other significant main effects or interactions were found (ps ≥ 0.217). For the N2pc, there was a marginal main effect of treatment (F(1, 52) = 4.02, p = 0.050, η² _p = 0.07), with a larger N2pc following OT compared to PLC treatment (Figure 3b). The main effects of face (F(3, 156) = 7.20, p < 0.001, η² _p = 0.12) and load (F(1, 52) = 5.68, p = 0.021, η² _p = 0.10) were also significant, with larger N2pc amplitudes when judging the ‘U’ direction presented together with neutral faces compared to happy (p = 0.001), angry (p < 0.001), and fearful faces (p = 0.039; Supplementary Figure S3), and under the low than the high load. No significant interactions were found (ps ≥ 0.314). Furthermore, the main effect of treatment on P300 amplitudes was also marginal (F(1, 52) = 3.64, p = 0.062, η² _p = 0.07), with P300 amplitudes being lower following OT compared to PLC treatment (Figure 3c). There was also a marginal main effect of load (F(1, 52) = 3.67, p = 0.061, η² _p = 0.07), with larger P300 amplitudes under the low load compared to the high load. No significant main effect of face or interactions was found (ps ≥ 0.683). For the theta band power, there was a significant main effect of load (F(1, 52) = 38.16, p < 0.001, η² _p = 0.42), with a stronger theta power under the high load compared to the low load. However, no other significant main effects or interactions were found (ps ≥ 0.078).

Temporal-dependent effects of OT in the FDI task

Given that the FDI task involved attentional control over responding to targets while inhibiting interference from emotional face distractors, which can be strengthened by trial repetition or attenuated by fatigue with the progression of the task, we therefore first conducted an exploratory analysis by dividing the task into two equal halves to examine whether OT’s effects particularly these marginal ones on the N2pc and P300 varied between the two halves. In the first half, the significant main effect of treatment on overall RA disappeared (F(1, 52) = 1.63, p = 0.208, η² _p = 0.03; Figure 4a), but we found significant main effects of treatment on N2pc (F(1, 52) = 7.29, p = 0.009, η² _p = 0.12; Figure 4c) and P300 (F(1, 52) = 4.24, p = 0.044, η² _p = 0.08; Figure 4e), with larger N2pc but lower P300 amplitudes following OT compared to PLC treatment. In the second half, results showed a significant main effect of treatment on RA (F(1, 52) = 6.79, p = 0.012, η² _p = 0.12), with a lower RA in the OT compared to the PLC group (Figure 4b). However, no significant effects related to treatment were observed at the neural level in the second half (all ps ≥ 0.094). To determine whether these effects were derived from participants benefiting from trial repetition or getting fatigue with the progression of the task, we compared their behavioral performance between the two halves and found a significant main effect of timepoint on both RA (F(1, 52) = 7.97, p = 0.007, η² _p = 0.13) and RT (F(1, 52) = 49.00, p < 0.001, η² _p = 0.49), with participants performing better in the second half compared with those in the first, suggesting the possibility of trial repetition improving task performance rather than fatigue attenuating it. Further exploratory analyses showed that this improvement for RA was mainly driven by changes in the PLC (p = 0.005) but not in the OT group (p = 0.279, Supplementary Figure S4b). A complete version of the results of these analyses was reported in Supplementary Results. We also conducted similar analyses for the CTVS task and found similar behavioral and neural patterns between the two halves (see Supplementary Figure S5).

Figure 4.

Choice accuracy for judging the direction of the target “U” following OT and PLC treatments across conditions in the first half (a) and the second half (b), respectively, of the FDI task. N2pc amplitudes at the electrodes of PO7 and PO8 following OT and PLC treatments across conditions in the first half (c) and the second half (d). P300 amplitudes at the electrode CPz and topographical maps following OT and PLC treatments across conditions in the first half (e) and the second half (f) (*p < 0.05, **p < 0.01, n.s. = not significant). Error bars indicate the standard error of the mean.

Discussion

The present study combined pharmaco-EEG with two modified tasks to investigate whether intranasal OT can modulate attentional processing of social cues in top-down and bottom-up task sets and its underlying neural mechanisms. Results showed that in the CTVS task, OT significantly enhanced participants’ RTs in judging the location of target face stimuli, independent of face expressions, which was accompanied by a larger N170 and stronger theta power following OT treatment. In the FDI task, OT’s effects on potentiating bottom-up attentional capture by emotional face distractors were modulated by the time course of the task. These findings provide evidence on how OT modulates attention to social cues in top-down versus bottom-up task sets.

In the CTVS task, on the behavioral level, consistent with previous studies (Burnham, 2010; Wadlinger & Isaacowitz, 2008), participants responded more quickly and accurately to happy faces or in the low load condition. More importantly, we found that OT significantly accelerated participants’ RT to target faces independent of face expressions and load conditions, suggesting a general effect of OT on facilitating attentional processing to social stimuli, which is in accordance with previous findings using a dot-probe paradigm and a facial emotion recognition paradigm (Domes et al., 2013; Hubble et al., 2017). By contrast, we did not find OT’s effects on RA, which might be due to judging the position of target faces being not very challenging, as reflected by an averaged RA higher than 97%, and RT might be more sensitive in this context. In partial support of this argument, we found a similar but stronger facilitatory effect of OT on RT under the high load compared to the low load condition, suggestive of a modulatory effect of cognitive load on OT’s effects on top-down attentional processing of social cues.

On the neural level, a larger N170 was observed following OT relative to PLC treatment. Given that the N170 is an early ERP component specifically associated with face processing (Eimer, 2011; Schindler et al., 2021), increased N170 amplitude therefore suggests that OT promotes the early processing of emotional faces. Consistent with this, previous studies have shown that intranasal OT increases N170 amplitudes when mothers view happy and sad faces of both infants and adults (Peltola, Strathearn, & Puura, 2018) or to fearful faces in a social salience attribution task (Santiago et al., 2024). It is noteworthy that this effect became marginal after multiple comparison corrections; inference related to this finding should be made with caution. Interestingly, positive correlations between N170 amplitudes and alexithymia scores were observed in the PLC group, suggesting that individuals with higher alexithymia scores may be more insensitive to facial stimuli at the very early stage of face processing. Given the close association between alexithymia and autism (Kinnaird, Stewart, & Tchanturia, 2019; Poquérusse, Pastore, Dellantonio, & Esposito, 2018), to some extent, this is consistent with the finding that individuals with autism spectrum disorder exhibit a less pronounced N170 in response to faces (Webb et al., 2023). Of note, OT reversed this positive correlation to a trend of negative correlation, which was mainly driven by OT increasing N170 amplitudes. OT may therefore have the potential to improve deficient early facial processing in alexithymia. Although there is a lack of direct evidence for OT modulating the N170 in alexithymia, decreased plasma OT concentrations have been found to correlate with more severe alexithymia (Baskaran et al., 2017; Schmelkin et al., 2017). OT has also been reported to improve the performance of high-trait alexithymia individuals in the Reading the Mind in the Eyes test (Luminet, Grynberg, Ruzette, & Mikolajczak, 2011).

For the time-frequency analysis, OT was found to increase theta power when judging the location of target faces independent of face expressions and load conditions. Given that theta oscillation is associated with visual attention (Fiebelkorn & Kastner, 2019; Karakaş, 2020), increases in theta power induced by OT, therefore, may reflect enhanced attentional processing toward emotional faces. This is consistent with the improved performance observed at the behavioral level. However, we did not observe any significant effects of OT on either N2pc or P300 amplitudes. Given that larger N2pc and smaller P300 amplitudes are indicative of enhanced attentional selection (Kiss, Van Velzen, & Eimer, 2008; Woodman & Luck, 1999) and more efficient attentional processing (Edwards et al., 2021; Gongora et al., 2021) at early and late stages respectively, these findings suggest an absence of OT effects on modulating these aspects of attentional processing in the CTVS task. By contrast, the early face-specific N170 component might be a more sensitive biomarker for OT’s effects on attentional processing of emotional faces in a top-down task set.

In the FDI task, OT was found to decrease RA across conditions at the behavioral level. In this task, participants were instructed to discriminate the direction of the ‘U’ overlaid on a neutral face while emotional faces were presented contralaterally as salient distractors. The observed detrimental effect of OT on RA may thus indicate that OT increases the salience of emotional face distractors, which consequently enhances their attentional capture ability in a bottom-up format but attenuates participants’ performance in judging the ‘U’ direction. This aligns with the social salience hypothesis of OT (Shamay-Tsoory & Abu-Akel, 2016) and a previous study reporting that OT impaired the ability to inhibit task-irrelevant emotional faces (Ellenbogen, Linnen, Cardoso, & Joober, 2013). However, on the neural level, OT marginally increased N2pc but decreased P300 amplitudes in response to judging the ‘U’ direction, suggesting a trend of OT enhancing attentional selection and processing efficiency of the target (Edwards et al., 2021; Gongora et al., 2021; Kiss, Van Velzen, & Eimer, 2008), which seems paradoxical given that OT reduced RA at the behavioral level.

To reconcile these contradictory findings, we further divided the task into two phases and found that, in the first half of the task, OT significantly increased N2pc and decreased P300 amplitudes, but its effect on overall RA disappeared. Increased N2pc, but decreased P300 amplitudes, indicate that OT facilitates attentional selection (Kiss, Van Velzen, & Eimer, 2008; Woodman & Luck, 1999) and processing efficiency (Edwards et al., 2021; Gongora et al., 2021) to the target ‘U’, which could have counteracted the distractive effect of task-irrelevant emotional faces and thus contributed to the null effect on RA. By contrast, in the second half, the OT group exhibited significantly reduced RA, suggesting that the overall decreased RA was mainly derived from the second half. Consistently at the neural level, OT failed to enhance attentional processing of the target ‘U’, as reflected by the absence of OT’s effects on the N2pc and P300 components, which may be caused by a more distractive effect of emotional face distractors at the contralateral side, consistent with the decreased RA in the OT group. To examine whether these temporal-dependent effects were derived from trial repetition or fatigue with the progression of the task, exploratory analyses comparing behavioral performance between the two halves were conducted and showed a significant improvement from the first half to the second, thereby supporting the possibility of trial repetition improving task performance instead of fatigue attenuating it. Further analyses revealed that the RA improvement was mainly driven by changes in the PLC group, and therefore, decreased RA in the OT group suggests that OT eliminated the improvement induced by trial repetition rather than decreasing it. These findings together suggest a temporal-dependent effect of OT such that it supports judging the target ‘U’ when participants were not adept at performing the task in the first half, but fails to do so when participants get practiced in the second half. OT may help engage more in responding to targets via more active top-down attentional control before participants became practiced, and therefore leave less attentional resources for task-irrelevant distractors. However, after practice, more attentional resources were available, which allowed for attentional capture in a bottom-up manner, strengthened by OT, increasing the salience of emotional distractors. Although there is no direct evidence supporting this time-dependent effect of OT on attentional allocation, OT has been proposed to modulate human behavior via initially facilitating attention to salient social and other important stimuli (Yao & Kendrick, 2025). Consistently, OT has been shown to reduce top-down attentional control by increasing bottom-up attentional capture of social cues (Xu et al., 2019; Zhuang et al., 2022) and to alter attentional biases toward salient social cues, including emotional faces (Olivia & Sarah, 2025; Shoenfelt et al., 2024). The present design of our study is unable to clarify the specific mechanism underlying this temporal-dependent effect of OT.

There are several limitations in the present study. First, only males were recruited, which limits the generalizability of our findings to females. Given the demonstrated gender-specific effects of OT in both animal models and human studies (Caldwell, 2018; Procyshyn, Dupertuys, & Bartz, 2024), future studies should include both genders to examine potential gender differences of OT’s effects on social attention. There is also evidence for a gender difference in attentional processing of social cues (Boyle, Johnson, & Ellenbogen, 2022; Domes et al., 2007, 2010), which should be taken into consideration when developing OT-based therapy for future translational application. Second, although our further analyses justified a temporal-dependent effect of OT between the two halves of the task, the present design of our study did not allow us to clarify the specific mechanism. Third, given increasing evidence demonstrating that OT’s effects are not necessarily social-specific (Eckstein et al., 2015; Yao & Kendrick, 2025; Zhou et al., 2024), it is beyond the scope of the present study to determine whether OT’s effects on attention are social-specific or not. We therefore did not include a nonsocial control condition and mainly focused on OT’s effects on attentional processing of social cues (Domes et al., 2007; Marsh, Yu, Pine, & Blair, 2010; Schulze et al., 2011), which is more relevant to clinical translation in mental disorders characterized by social attentional deficits (e.g. autism and alexithymia). However, it should still be noted that our current findings provide evidence only for OT modulating social attention but do not exclude possibilities in the nonsocial context.

In conclusion, the present study provides both behavioral and electrophysiological evidence for OT modulating attentional processing of social cues in top-down versus bottom-up task sets. In the CTVS task, OT accelerated RT and induced a larger N170 and stronger theta power in response to emotional targets, suggesting that OT promoted early top-down attentional processing of social cues. In the FDI task, while OT had no effect on RA but increased N2pc and decreased P300 amplitudes in the first half of the task, it reduced RA but had no effect on N2pc and P300 components in the second half. OT, therefore, may promote top-down attentional control over judging the target ‘U’ to counteract the distractive effect from emotional face distractors in the first half, but failed to do so in the second one. Together, the present study not only provides evidence for the role of OT in modulating attentional processing of social cues and its neural mechanisms but also lends support to the therapeutic potential of OT in mental disorders characterized by attentional deficits, particularly to social cues such as autism and alexithymia.

Supporting information

Zhou et al. supplementary material

Zhou et al. supplementary material

Author contribution

S.Y. and M.Z. designed the study. M.Z., Y.Z., and Z.Z. conducted the experiment and collected the data. M.Z., X.W., and Q.Z. performed the data analysis. M.Z. and S.Y. wrote the manuscript draft. S.Y. and K.M.K critically revised the manuscript draft.

Funding statement

This study was supported by the National Natural Science Foundation of China (grant number: 32471139).

Competing interests

The authors declare none.

Ethical standard

The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.