Abstract: In multi-line text reading, longer saccades are required to bring the reader’s fixation point from the end of the previous line to the beginning of the next line, which is called a return sweep. The return-sweep interval usually ranges from approximately 4~6 characters at the end of the previous line to approximately 5~8 characters at the beginning of the next line, with an end-to-end line distance of approximately 30~70 characters.
Return sweeps are essential for reading multi-line texts, which are common in natural reading. Research on return sweeps has mainly focused on the return-sweep launch site, target selection, accuracy, and the fixation adjacent to the return sweep. Studies have drawn five main conclusions: (1) Compared to those with high spelling and reading capacity, individuals with low spelling and reading capacity have a return-sweep launch site closer to the end of the line. (2) Readers can generally anticipate a return-sweep target based on the global typographic properties of the text, since selection of the return-sweep target is affected by the salience of the line-initial word, the return-sweep launch site, the font size, the line length, and individual differences among readers, as well as the combination of the length of the first word of the line and the return-sweep launch site, the combination of font size and the return-sweep launch site, and the combined influence of the font size, the line length, and the return-sweep launch site. (3) The accuracy of the return sweep is affected by the line length, the degree of deviation between the expected target and actual return-sweep landing position, the font size, the return-sweep launch site, salience of the line-initial word, and individual differences. (4) The fixations of the return sweep are divided into line-final and line-initial fixations, which are further divided into accurate line-initial and undersweep fixations. When durations between these three fixations are compared, the fixation duration for the line not adjacent to the return sweep is longer than for the line-final fixation and shorter than for the accurate line-initial fixation. Individual variations are also present in these durations. The duration of undersweep fixations is usually shorter than that for the line not adjacent to the return sweep. (5) During the undersweep fixation, a reader’s attention can temporarily focus on the fixation word instead of the first word of the line. Both adults and children can extract fixation words and information from the left side of the fixation words during the undersweep fixation duration to create a seamless transition in reading.
However, many aspects remain unclear. The influence of parafoveal information on word processing and saccadic target selection is not well understood across multi-line text and requires further study. For in-line saccades, a reader can obtain parafoveal prevision of the next word and select a saccadic target. However, during the return sweep, the longer interval and line distance places the first word of the next line beyond the parafoveal preview range, and the reader cannot preview the first word of the next line before the return sweep to select the saccadic target. Furthermore, the E-Z reader and SWIFT models are currently only used to explain single-line text reading. There is no model to explain the return sweep, which must be included for a reading model to simulate natural reading. Based on the above research, we propose that future research should aim to improve the existing reading model to accommodate the return sweep, examine the return sweep in other languages, compare the return sweep in first- and second-language reading, explore whether the line-final fixation durations are affected by the lexical properties of end-of-line fixation words, and investigate the influence of the lack of preview on lexical processing in the return-sweep process to understand the “preview benefit” in the boundary paradigm.

Key words: reading, return sweep, eye movement

摘要： 回扫是读者的注视点从文本上一行行末到下一行行首的眼动,在多行文本阅读中非常重要且必不可少。目前,回扫的相关研究较少,且主要集中于回扫起跳位置、回扫目标选择、回扫的精确性、与回扫相邻的注视四个方面。本文先简单介绍回扫研究中常用的眼动指标,再详细介绍回扫的四个方面的研究,最后提出未来的研究设想：关注其他文字与文字系统的回扫、考察二语阅读中的回扫、探究行末注视是否受行末注视词词汇特征的影响、探讨回扫过程中预视的缺失对词汇加工的影响以帮助我们理解边界范式中的“预视效益”,完善现有模型以适应回扫。

关键词: 阅读, 回扫, 眼动