The Role of Spaces

Several studies have manipulated spacing in spaced alphabetic writing. The reading rate of people accustomed to words segmented by space slows by approximately 35% when spaces are removed from the text. The average fixation duration when reading normal spaced text is around 220 ms and increases to around 380 ms when spaces between words are absent (Rayner et al. 1998; Spragins et al. 1976).

Furthermore, when spaces are removed the PVL of readers of spaced writing systems shifts towards the beginning of the word (Rayner et al. 1998; Rayner and Pollatsek 1996). This suggests that readers of spaced writing systems can identify word beginnings in unspaced text and adapt their oculomotor strategies accordingly. These studies demonstrate that while the absence of spaces for readers of usually-spaced texts does not make reading impossible, it changes the PVL to the presumably less efficient word-initial position.

On the basis of the shift in PVL when readers of spaced writing systems read unspaced text, it might be expected that readers of inherently unspaced scripts would also show PVLs near word beginnings. Based on the same reasoning, it is also possible that if spaces are inserted between words they may perform better, as measured by shorter viewing times and a shift in PVL from word initial towards word center.

In the case of Thai, Kasisopa et al. (2013) investigated eye movements of readers reading normal, unspaced Thai text. Surprisingly, they found that the PVL for Thai was near the centre of the word, similar to that for spaced English. However, the PVL curve was flatter than that found for spaced writing systems. This indicated that the eye was indeed targeting the word centre, but with less certainty about where the actual centre was located when compared with, say, English.

What cues might Thai readers be using achieve this? Kasisopa et al. (2013) proposed that the relative frequency of consonants in word-initial and word-final positions could be acting as a word-boundary cue and found some evidence to support this hypothesis.

The effect of spaces between words has also been investigated in Chinese. Bai et al. (2008) presented Chinese text in four conditions - normal, unspaced text, and three spacing conditions: appropriately at word boundaries, appropriately between each character; or inappropriately between characters, thus creating non-words. The results showed that spaces between words neither hindered nor facilitated reading Chinese. However, spacing between each character and spacing that gave rise to non-words caused readers to take longer to read this type of text suggesting that it disrupted word processing.

Whether a PVL exists for Chinese has been the subject of some debate. In one of the first studies of eye movements of Chinese readers, Yang and McConkie (1999) found a more-or-less flat distribution of landing sites. However, in a more controlled study, Yan, Kliegl, Richter, Nuthmann, and Shu (2010) found clear evidence of a PVL curve, but for a subset of fixations in which readers were fixating quite close to the target word and therefore had a clearer view it.

Overall, therefore, there is reasonable evidence that readers of both Thai and Chinese, where possible, target word centres and demonstrate a PVL curve similar to that found for spaced scripts. The question remains therefore, as to the precise mechanism that allows these readers to estimate the location of word centres.