Tracking the mind during reading: The influence of past, present, and future words on fixation durations.

Abstract

Reading requires the orchestration of visual, attentional, language-related, and oculomotor processing constraints. This study replicates previous effects of frequency, predictability, and length of fixated words on fixation durations in natural reading and demonstrates new effects of these variables related to previous and next words. Results are based on fixation durations recorded from 222 persons, each reading 144 sentences. Such evidence for distributed processing of words across fixation durations challenges psycholinguistic immediacy-of-processing and eye-mind assumptions. Most of the time the mind processes several words in parallel at different perceptual and cognitive levels. Eye movements can help to unravel these processes.Keywords: eye movements, fixation duration, gaze, word recognition, reading Reading is a fairly recent cultural invention. The perceptual, attentional, and oculomotor processes enabling this remarkable and complex human skill had been in place for a long time before the first sentence was read. Of course, reading also fundamentally presupposes language, reasoning, and memory processes. If we want to understand how internal processes of the mind and external stimuli play together in the generation of complex action, reading may serve as an ideal sample case, because, despite its complexity, it occurs in settings that are very amenable to experimental control. In addition, the measurement of eye movements yields high-resolution time series that have proven to be very sensitive to factors at all levels of the behavioral and cognitive hierarchy. Most importantly, we already know or can determine basic perceptual, attentional, and oculomotor constraints which any theory of reading and any computational model implementing such a theory at a behavioral microlevel must respect.Looking at the eyes, reading proceeds as an alternating sequence of fixations (lasting 150 to 300 ms) and saccades (30 ms). Information uptake is largely restricted to fixations. For example, fixation durations reliably decrease with the printed frequency of words and with their predictability from prior words of the sentence. Beyond these uncontroversial facts, however, much still needs to be learned about perceptual and attentional processes and properties of words that guide the eyes through a sentence. Starr and Rayner (2001, p. 156) highlighted the following three issues as particularly controversial: "(1) the extent to which eye-movement behavior is affected by low-level oculomotor factors versus higher-level cognitive processes; (2) how much information is extracted from the right of fixations; and (3) whether readers process information from more than one word at a time." Distributed Processing in Fixation Durations 4In this article, we report new empirical results relating to each of these issues. We also present a data-analytic framework within which these issues can be addressed simultaneously and propose a set of theoretical principles which account for a complex set of experimental observations....

Tracking the mind during reading via eye movements: Comments on Kliegl, Nuthmann, and Engbert (2006).

“…However, we hope that we have made clear that we think that many of the claims that they make are inconsistent with a great deal of other data, and that much of the data that are inconsistent with their claims have been obtained in well-controlled experiments. Again, we realize that no experiment is ever perfect, but to the extent that extraneous variables can be controlled, we suspect that they are preferable to the type of regression analysis techniques that are relied upon by Kliegl et al (2006). And, finally, consistent with the prior statement, we strongly urge that claims made from regression analysis techniques should not be accepted until confirmed via controlled experimental techniques.…”

“…With this one additional assumption, the predictability of word n+1 was now positively correlated to the single fixation durations on word n in the long-word condition (r = .062). Although the absolute strength of this predicted relationship is quite modest, it is comparable to that observed by Kliegl et al (2006): a modest correlation (r = .05) between the logit-transformed predictability of word n+1 and single fixation durations on word n 4 . It is also important to note that our manipulation of the word n predictability covariate was a modest one, with the longer (7-13 letter) words being only slightly less predictable than the shorter (2-6 letter) words (.01 vs. .2, respectively).…”

“…In this article, we have questioned a number of the points made by Kliegl et al (2006). While we are critical of some of the claims that they make, we want to again note that we think that they have provided some interesting analyses.…”

“…Let's consider the evidence reported by Kliegl et al (2006) for the distributed nature of lexical processing in eye fixations in reading. There are two findings that they reported which, if true, seem to be the strongest arguments against a serial model of word processing, such as E-Z Reader.…”

“…Another problem with the analyses reported by Kliegl et al (2006) is that all words were included in their analysis. That is, about 25% of their words were 2-3 letters long, and probably consist largely of function words.…”