Predictions in the visual domain have been shown to modulate conscious access. Yet, little is known about how predictions may do so and to what extent they need to be consciously implemented to be effective. To address this, we administered an attentional blink (AB) task in which target 1 (T1) identity predicted target 2 (T2) identity, while participants rated their perceptual awareness of validly versus invalidly predicted T2s (Experiment 1 & 2) or reported T2 identity (Experiment 3). Critically, we tested the effects of conscious and non-conscious predictions, after seen and unseen T1s, on T2 visibility. We found that valid predictions increased subjective visibility reports and discrimination of T2s, but only when predictions were generated by a consciously accessed T1, irrespective of the timing at which the effects were measured (short vs. longs lags). These results further our understanding of the intricate relationship between predictive processing and consciousness.

视觉领域的预测已被证明可以调节有意识的访问。然而，人们对预测如何做到这一点以及它们需要在多大程度上有意识地实施才能有效知之甚少。为了解决这个问题，我们执行了一项注意眨眼 (AB) 任务，其中目标 1 (T1) 身份预测目标 2 (T2) 身份，而参与者评估他们对有效和无效预测的 T2（实验 1 和 2）或报告的 T2 的感知意识身份（实验 3）。至关重要的是，我们测试了有意识和无意识预测在可见和不可见的 T1 之后对 T2 可见性的影响。我们发现，有效的预测增加了主观可见性报告和 T2 的辨别力，但仅当预测是由有意识地访问的 T1 生成时，无论测量效果的时间如何（短期 vs. 长期滞后）。这些结果进一步加深了我们对预测处理和意识之间错综复杂关系的理解。