Effects of load on the time course of attentional engagement, disengagement, and orienting in reading

Abstract

We examined how the frequency of the fixated word influences the spatiotemporal distribution of covert attention during reading. Participants discriminated gaze-contingent probes that occurred with different spatial and temporal offsets from randomly chosen fixation points during reading. We found that attention was initially focused at fixation and that subsequent defocusing was slower when the fixated word was lower in frequency. Later in a fixation, attention oriented more towards the next saccadic target for high- than for low-frequency words. These results constitute the first report of the time course of the effect of load on attentional engagement and orienting in reading. They are discussed in the context of serial and parallel models of reading.

Keywords:

• Attention
• Load; Reading
• Time course
• Word frequency
• Engagement
• Disengagement
• Orienting

Acknowledgments

The research reported here was funded by a grant from the Leverhulme Trust (F/07 605/L) to Karina J. Linnell and Martin H. Fischer. We would like to thank Keith Rayner, Heiner Deubel, Sandy Pollatsek, Ben Tatler, and our anonymous reviewers for their insightful comments.

Notes

¹ A pilot experiment in which the same stimuli were used but participants were only asked to read (there was no probe) produced an average saccade length of 7.6 characters. Therefore a +7 probe would have been a good choice for a right probe because it would have coincided with the average location of the next fixation. However, we also wanted to assess the effects of word boundaries and therefore to have some right probes falling on the critical words (rather than the next word). We increased the chances of a right probe occurring on the critical word by choosing a +6 rather than a +7 probe. Given that (a) readers prefer to fixate somewhere between the middle and the beginning of words (Vitu, O'Regan, & Mittau, 1990), and (b) the average word length was 6.17 characters (standard deviation = 1.20) for low-frequency words and 6.24 characters (standard deviation = 1.46) for high-frequency words, probes occurring 7 characters to the right of fixation would almost always have fallen on the next word; by choosing a +6 probe, 20.3% of right probes in Experiment 1 and 30.6% of the right probes in Experiment 2 fell on the critical word (15.3% of the low-frequency and 15.2% of the high-frequency words were 8, 9, or 10 characters long).

² Frequencies of critical words were measured using the written portion of the British National Corpus (BNC), a 100-million-word balanced corpus of British English. They were measured on raw word forms rather than lemmas and were normalized to words per million. According to the BNC counts, high-frequency words had an average frequency of 102.1, and low-frequency words an average frequency of 4.2 per million. The maximum frequency of the low-frequency items was 15.4, and the minimum frequency of the high-frequency items was 23.8 per million.

³ Of the 480 sentences we used, 246 came from previous research containing sentences with manipulated frequency and controlled predictability (in particular Drieghe, Desmet, & Brysbaert, 2007; Frisson, Rayner, & Pickering, 2005; Hand, Miellet, O'Donnell, & Sereno, 2010). We thank these authors for making their materials available. As all these studies used predictable critical words, we created low-predictable versions by substituting words that were not the same as (or close synonyms of) the predictable words used. For some of the sentences from the existing literature it was also necessary to change the sentences due to the need to shorten sentences to fit on our 77-character display. An additional 234 sentences were devised and tested in a word completion study with 20 different native English speaking participants to ensure a predictability of less than .05. In this control experiment, each sentence was shown up to the point of (but excluding) the critical word, and participants completed the most likely next word.

⁴ On average, the first fixation on the critical word occurred at 2.8 characters (SEM = 0.1).

⁵ On average, the first fixation on the critical word occurred at 2.6 characters (SEM = 0.1).

⁶ In a separate analysis, we excluded all trials in which a +6 probe occurred on the critical word. A 2 (word frequency) × 3 (spatial offset) × 3 (temporal offset) ANOVA again showed an interaction between word frequency and spatial offset of the probe, F(2, 42) = 3.37, p = .04, η² = .012; breaking down this interaction for different spatial offsets, there was a marginal effect of word frequency for right probes (p = .07) but not for central or left probes.

⁷ In the accepted trials in Experiment 2, the average length of the fixated word when a right probe occurred on the fixated word was 7.05 characters (SEM = 0.07) whereas it was 5.8 characters (SEM = 0.04) when a right probe occurred on the next word.

⁸ One might argue that the probe was encoded into a preattentive memory store—that is, iconic memory—and that attention then operated over that iconic representation (e.g., Lachter & Durgin, 1999). One would therefore conclude that our paradigm did not reveal the spatiotemporal distribution of attention—we thank an anonymous reviewer for pointing out this possibility. To address this, first, we should note that visual information about the probe was processed at higher levels than preattentive memory stages, as indicated by correct identification performance (as opposed to mere detection). Secondly, if attention only operated over a prelexical iconic memory, any effects of load should be equal for left and right probe locations. In other words, the same effects of word frequency should occur 6 characters to the left or right of fixation. However, this was not observed: Load effects were asymmetrically distributed in both experiments.