An attention cascade model is proposed to account for attentional blinks in rapid serial visual presentation (RSVP) of stimuli. Data were collected using single characters in a single RSVP stream at 10Hz [Shih, S., & Reeves, A. (2007). Attentional capture in rapid serial visual presentation. Spatial Vision, 20(4), 301-315], and single words, in both single and dual RSVP streams at 19Hz [Potter, M. C., Staub, A., & O'Connor, D. H. (2002). The time course of competition for attention: Attention is initially labile. Journal of Experimental Psychology: Human Perception and Performance, 28(5), 1149-1162]. The model adopts similar architecture of the cognitive accounts of attentional blinks and employs computational details from theories of attention gating. The model has elaborated working memory and attention control mechanism. Both bottom-up and top-down salience are explicit in the model. Quantitative fits are good and the model parameters have plausible values. The model handles stimulus competition, lag 1 sparing, intrusion errors, and magnitude of the dip; it also accounts for commonly observed effects such as stimulus similarities (local and global), target+1 blank, and stimulus salience.

中文翻译：

---

注意级联模型和注意眨眼

提出了一种注意级联模型来解释刺激的快速串行视觉呈现 (RSVP) 中的注意眨眼。在 10Hz 的单个 RSVP 流中使用单个字符收集数据 [Shih, S., & Reeves, A. (2007)。快速串行视觉呈现中的注意力捕获。Spatial Vision, 20(4), 301-315] 和单字，在 19Hz 的单和双 RSVP 流中 [Potter, MC, Staub, A., & O'Connor, DH (2002)。注意力竞争的时间过程：注意力最初是不稳定的。实验心理学杂志：人类感知与表现，28（5），1149-1162]。该模型采用了注意力眨眼的认知帐户的类似架构，并采用了注意力门控理论的计算细节。该模型阐述了工作记忆和注意力控制机制。自下而上和自上而下的显着性在模型中都是明确的。定量拟合良好，模型参数具有合理的值。该模型处理刺激竞争、滞后 1 保留、入侵错误和下降幅度；它还解释了常见的影响，例如刺激相似性（局部和全局）、目标+1 空白和刺激显着性。