Abstract: The purpose in this study was to explore whether aggressive information could interact with attention by reducing the extent of inhibition of return(IOR) effect and whether the discrepancies between implicit and explicit self-esteem(SE) could modulate IOR effect under attentional biases toward aggressive information. A standard cue-target task was adopted to examine whether the attentional bias was stemmed from attention capture or attentional disengagement. Participates were asked to respond to Rosenberg’s(1965) 10-item Self-Esteem Scale and completed the implicit association task(IAT). According to the results of the tests, the fragile HSE group(a combination of high explicit and low implicit SE) and secure HSE group(a combination of high explicit and high implicit SE) were selected. The Cue-target task was applied in two experiments to measure IOR. Two types of words was served as target in Experiment 1 but as cue in Experiment 2.Participates were asked to respond to the location of the target(the word in Experiment 1 and a black asterisk in Experiment 2) . In Experiment 1, mean RTs subjected to a 2(group: fragile HSE vs. secure HSE) ×2 SOA(500,and 1000ms)×2(Target Valence: aggressive words vs. neutral words) ×2 and (Cueing validity：valid vs. invalid) mixed analysis of variance(ANOVA) revealed a significant main effect on target valence[F（1,66）=5.11,p=.027,η2=.072],reaction times for aggressive words(317.50±7.12ms) were significantly faster than that for neutral words(319.48±7.05ms).And the main effect of cueing validity was significant [F（1,66）=34.85，p<.001，η2=.346], slower RTs on valid(324.32±6.81ms) than on invalid trials(312.65±7.46ms) showing a significant IOR effect emerged. Importantly, the interaction between target valence and cueing validity reached significance [F(1,66)=11.94,p=.001,η2=.153], which was not modified by the discrepancies between implicit and explicit self-esteem. Further analysis indicated that for both groups, the reaction times for aggressive targets(321.99±6.82ms) on valid location was significantly faster than that for neutral targets(326.66±6.84ms) on valid location, which might reflect facilitated engagement towards aggressive information in the source of attentional capture. on the results of the amounts of IOR showed that there was a significant main effect for target valence[F（1,66）=11.94，p=.001，η2=.153], smaller IOR extent on aggressive words(-8.98±2.07ms) than that on neutral words(-14.36±2.18ms). The results of Experiment 1 suggested that the reduction of IOR effect with aggressive targets reflect facilitate engagement towards aggressive information for both groups on the role of attentional capture. In Experiment 2, mean RTs were subjected to a 2(group: fragile HSE vs. secure HSE) ×2 SOA(500,and 1000ms)×2(Target Valence: aggressive words vs. neutral words) ×2 and (Cueing validity：valid vs. invalid) mixed-factors ANOVA. There was a significant main effect of cueing validity, F（1,64）=114.57，p<.001，η2=.642, overall, the IOR was found for both cue types in overall, and response times to valid cueing condition(311.71±4.06ms) was significantly faster than invalid cueing(295.92±3.74ms), indicating significant IOR effect. Importantly, there was a significant three-way interaction of cue valence×cueing×group [F（1,64）=8.29，p=.005，η2=.115]. The simple effect analysis showed that the reaction time for aggressive words on invalid cueing condition(301.33±5.58ms) was significantly slower than that for neutral words on invalid cueing condition(295.84±5.11ms) in fragile HSE group; the reaction time for aggressive words on invalid cueing condition(291.25±5.58ms) was significantly shorter than that for neutral words on invalid cueing(295.27±5.11ms) in secure HSE group. It showed that attentional bias in fragile HSE might reflect the difficulty in disengaging from aggressive information. As for the tests on the amounts of IOR, there was a significant interaction between group and cue valence[F（1,64）=8.29，p=.005，η2=.115].Further analysis showed that, compared to secure HSE group(-18.74±2.33ms), the effect of IOR on aggressive words in fragile HSE group(-11.94±2.33ms) was much smaller; For fragile HSE group, the effect of IOR on aggressive words (-11.94±2.33)was smaller than that on neutral words(-16.73±2.26ms). The results of Experiment 2 suggested that reduction of the IOR effect with aggressive cues reflect the reluctance of fragile HSE group to disengage attention from aggressive information. The findings in this study discussed with regard to the mechanisms responsible for IOR. The attentional biases in fragile HSE are the cause of facilitated engagement towards aggressive information (attentional capture) and the difficulty in disengaging attention from aggressive information. The sensitivity on aggressive stimuli is probably the potential mechanism on aggressive behavior. Further researches should focus on the aggression intervention from a cognitive perspective.In Experiment 1,mean RTs were subjected to a 2×2×2×2 mixed analysis of variance(ANOVA) with group(fragile HSE vs. secure HSE) as the between-participants factor and SOA(500,and 1000ms),Cue Valence(aggressive words vs. neutral words) and Cueing(valid vs. invalid) as within-participants factor.IOR was found for both two cue types,a finding confirmed by the significant main effect of cueing,F（1,64）=114.57，p=0.000，η2=.642.Overall,response times to valid cueing condition(311.71±4.06ms) was significantly faster than invalid cueing(295.92±3.74ms) condition,which indicates significant IOR effect.Importantly,there was a significant three-way interaction of cue valence×cueing×group[F（1,64）=8.29，p=0.005，η2=.115]. The simple and simple effects analysis showed that the reaction time for aggressive words on invalid cueing condition(301.33±5.58ms) was significantly longer than that for neutral words on invalid cueing condition(295.84±5.11ms) in fragile HSE group;the reaction time for aggressive words on invalid cueing condition(291.25±5.58ms) was significantly shorter than that for neutral words on invalid cueing(295.27±5.11ms) in secure HSE group. It means that attentional bias in fragile HSE may reflect a difficulty in disengaging from aggressive information.As for the tests on the amounts of IOR, there was significant interaction effect between group and cue valence[F（1,64）=8.29，p=0.005，η2=.115].Further analysis showed that, compared to secure HSE group(-18.74±2.33ms),the effect of IOR on aggressive words in fragile HSE group(-11.94±2.33ms) was much smaller;For fragile HSE group,the effect of IOR on aggressive words (-11.94±2.33)was smaller than that on neutral words(-16.73±2.26ms). The results of Experiment 1 tell us that reduction of the IOR effect with aggressive cues is thought to reflect the reluctance of fragile HSE group to disengage attention from aggressive information. In Experiment 2,mean RTs were subjected to a 2×2×2×2 mixed analysis of variance(ANOVA) with group(fragile HSE vs. secure HSE) as the between-participants factor and SOA(500,and 1000ms),Target Valence(aggressive words vs. neutral words) and Cueing(valid vs. invalid) as within-participants factor.The analysis revealed a significant main effect on target valence[F（1,66）=5.11,p=0.027,η2=.072],the reaction time for aggressive words(317.50±7.12ms) was significantly faster than that for neutral words(319.48±7.05ms).There was also a main effect of cueing[F（1,66）=34.85，p=0.000，η2=.346],with participants being slower on valid(324.32±6.81ms) than on invalid trials(312.65±7.46ms),thus showing a siginificant IOR effect.Importantly,the interaction between target valence and cueing reached significance[F(1,66)=11.94,p=0.001,η2=.153],which was not modified by the discrepancies between implicit and explicit self-esteem.A further analysis indicaed that for both groups,the reaction time for aggressive targets(321.99±6.82ms) on valid location was significantly faster than that for neutral targets(326.66±6.84ms) on valid location,which may reflect facilitated engagement towards aggressive information(attentional capture).As for the tests on the amounts of IOR, there was a significant main effect for target valence[F（1,66）=11.94，p=0.001，η2=.153],the effect of IOR on aggressive words(-8.98±2.07ms) was much smaller than that on neutral words(-14.36±2.18ms). The results of Experiment 2 tell us that reduction of the IOR effect with aggressive targets is thought to reflect facilitated engagement towards aggressive information for both groups(attentional capture). These findings are discussed with regard to the mechanisms responsible for IOR.The attentional biases in fragile HSE are due to facilitated engagement towards aggressive information(attentional capture) and a difficulty in disengaging attention from aggressive information.The sensitivity on aggressive stimuli is probably the potential mechanism on aggressive behavior.Further research will focus on the aggression intervention from a cognitive perspective.

Key words: Fragile high self-esteem, College students, Aggression, Inhibition of return

摘要： 为考察攻击性信息对脆弱型高自尊大学生返回抑制的影响机制，本研究采用线索靶子范式，分别将攻击性及中性词汇作为靶子（实验1）和线索（实验2）呈现。实验1发现，脆弱型和安全型高自尊组对攻击性信息均存在注意捕获，从而使攻击性信息的返回抑制量降低。实验2发现，脆弱型高自尊组对攻击性信息存在注意解除困难，从而使攻击性信息的返回抑制量降低。脆弱型高自尊大学生对攻击性信息存在过多的注意维持，降低了视觉搜索效率，这可是其好发攻击性行为的潜在认知加工机制。

关键词: 脆弱型高自尊, 大学生, 攻击, 返回抑制