关键词: 抑郁, 消极注意偏向, 积极注意偏向, 三水平元分析

Abstract: Background:
Depression is a prevalent and debilitating mental disorder characterized by persistent low mood and diminished interest or pleasure. Cognitive theories posit that attentional bias—the preferential allocation of attention toward mood-congruent information—plays a central role in the onset and maintenance of depressive symptoms. While substantial evidence supports the existence of negative attentional bias in depression, the magnitude, stability, and contextual boundary conditions of this effect remain unclear. Moreover, comparatively limited attention has been devoted to how depressed individuals process positive emotional information.
Previous meta-analyses have provided valuable insights into attentional processes in depression but have typically focused on negative bias alone and employed models that cannot adequately account for dependency among effect sizes. Consequently, estimates of heterogeneity and moderator effects have often been imprecise. Addressing these methodological and conceptual limitations, the present study represents the first comprehensive three-level meta-analysis to synthesize evidence on both negative and positive attentional biases in depression. By modeling within- and between-study variance simultaneously, this work offers a more rigorous and integrated evaluation of attentional asymmetries, providing new insights into the dual mechanisms of cognitive-affective processing in depression.
Aims:
This study aimed to (1) quantify the magnitude and consistency of both negative and positive attentional biases among individuals with elevated depressive symptoms, thereby extending the traditional focus on negative bias alone; (2) identify methodological and demographic moderators—such as task paradigm, stimulus type, emotional valence, age, and sample type—that account for variability in effect sizes; (3) refine the conceptualization of attentional bias in depression by framing it as a bidirectional, context-sensitive construct; and (4) integrate the present findings with prior evidence to advance future theoretical models and empirical research.
Methods
A systematic search was conducted across seven major English and Chinese databases (EBSCO, Web of Science, ScienceDirect, PubMed, CNKI, VIP, and Wanfang) for peer-reviewed studies published between January 1, 2000, and July 8, 2025. Following rigorous screening and double coding, 51 studies examining negative attentional bias (yielding 91 effect sizes) and 33 studies examining positive attentional bias (yielding 56 effect sizes) were included.
A three-level random-effects model, implemented in R using the metafor package, was employed to account for sampling variance, within-study dependence, and between-study heterogeneity. Methodological rigor was ensured through sensitivity analyses, outlier diagnostics, and formal assessments of publication bias. Moderator analyses examined the influence of task paradigm, stimulus type, emotional valence, age, and sample characteristics on variability in effect sizes.
Results:
The meta-analytic findings revealed a significant, moderate negative attentional bias among depressed individuals compared with healthy controls (Hedges’ g = 0.48, 95% CI [0.35, 0.61], p < .001), indicating greater attentional capture by negative or threat-related stimuli. Concurrently, a significant absence of positive attentional bias was observed (Hedges’ g = -0.30, 95% CI [-0.44, -0.18], p < .001), suggesting reduced attentional engagement with rewarding or pleasant stimuli.
Moderator analyses demonstrated that task paradigm significantly influenced the magnitude of both negative (F(4, 84) = 4.49, p = .002) and positive (F(3, 50) = 8.74, p < .001) biases. Specifically, the dot-probe paradigm yielded larger negative bias effects than the emotional Stroop task, whereas the free-viewing eye-tracking paradigm showed greater sensitivity to detecting the absence of positive bias compared with the spatial cueing task. Stimulus type also emerged as a significant moderator for positive bias (F(1, 54) = 8.22, p = .006), with pictorial stimuli producing larger effects than verbal stimuli—a pattern approaching marginal significance for negative bias (p = .057). In contrast, age, sample type (clinical vs. subclinical), and stimulus valence did not significantly moderate either form of bias.
Conclusions and Implications:
This meta-analysis suggests a dual pattern of attentional bias in depression, characterized by heightened attention toward negative information and attenuated engagement with positive stimuli. Such an asymmetrical pattern may reflect an underlying imbalance between enhanced sensitivity to threat and reduced responsiveness to reward, providing a more differentiated account of attentional processing in depression.
Methodologically, the application of a three-level hierarchical meta-analytic framework represents a substantial advance over previous syntheses, allowing a more precise estimation of dependent data structures and moderator effects. The identification of task paradigm and stimulus type as major moderators highlights the importance of methodological design in interpreting attentional bias findings and offers guidance for future experimental research.
From a theoretical standpoint, these findings help redefine attentional bias as a bidirectional, context-dependent process that bridges cognitive and motivational perspectives. Clinically, the results provide preliminary support for dual-track interventions—those aimed not only at attenuating negative attentional capture but also at enhancing engagement with positive or rewarding stimuli. Nonetheless, given that most included studies employed cross-sectional designs, causal interpretations should remain cautious. Future longitudinal, multimodal, and neurocognitive studies are essential to clarify the temporal and neural mechanisms underlying these dual biases in depression.

Key words: depression, negative attention bias, positive attention bias, three-level meta-analysis