Ординатура / Офтальмология / Английские материалы / Eye Movements A Window on Mind and Brain_Van Gompel_2007
.pdf
334
Start of next saccade program
0.8
0.7
Well-located fixations
frequency |
0.6 |
|
|
Mislocated fixations |
|
|
|
All fixations |
|
||
|
|
|
|
||
0.5 |
|
|
|
|
|
0.4 |
|
|
|
|
|
Relative |
|
|
|
|
|
0.3 |
|
|
|
|
|
0.2 |
|
|
|
|
|
|
|
|
|
|
|
|
0.1 |
|
|
|
|
|
0 |
0 |
100 |
200 |
300 |
|
−100 |
||||
Time [ms], Relative to start of current fixation [0 ms]
(a)
R. Engbert et al.
Start of next saccade program
|
1 |
|
|
|
|
frequency |
0.8 |
|
|
|
|
0.6 |
|
|
|
|
|
relative |
|
|
|
|
|
0.4 |
|
|
|
|
|
Cumulative |
|
|
|
|
|
0.2 |
|
|
|
|
|
|
|
|
|
|
|
|
0 |
0 |
100 |
200 |
300 |
|
–100 |
Time [ms], Relative to start of current fixation [0 ms]
(b)
Figure 6. Start of next saccade program, relative to the start of the current fixation for well-located vs mislocated fixations. The 0 value on the x-axis marks the start of the current fixation. Thus, negative values represent saccade programs that were started before the start of the current fixation. (a) Relative frequencies. (b) Cumulative relative frequencies.
the beginning of a fixation is smaller than one (which is the case in SWIFT simulations), the mechanism can only decrease but not increase the mean duration of mislocated fixations.
4.2. Fixation duration IOVP effects in SWIFT
Theoretically, all types of fixations can be mislocated. The explanation of the IOVP effect suggested in Nuthmann et al. (2005) is based on the assumption that mislocated fixations (often) trigger a new saccade program immediately. Thus, we predict and observe an IOVP effect for durations of single fixations, first of multiple fixations, and second of multiple fixations (cf., Vitu et al., 2001). Our empirical estimates, however, were based on all fixations. Consequently, the suggested IOVP mechanism is not able to differentiate between different types of fixations (e.g., single, first, second).
Again, we employed simulations with the SWIFT model to investigate and reproduce quantitative differences between various IOVP functions. The correction mechanism for mislocated fixations, implemented as Principle VI, was able to reproduce the IOVP effect for single fixation durations (Engbert et al., 2005, Fig. 15). However, to reproduce the IOVP effect for the first of multiple fixations as well as the fixation duration trade-off effect for two-fixation cases, the model had to be furnished with Principle (VII): It is assumed that saccade latency is modulated by the amplitude of the intended saccade
Ch. 14: An Iterative Algorithm for the Estimation of Mislocated Fixations |
335 |
Fixation duration [ms]
Empirical data |
Simulated data |
230 |
|
|
|
|
|
|
|
|
|
|
230 |
|
|
|
|
|
|
|
|
|
220 |
|
|
|
|
|
|
|
|
|
|
220 |
|
|
|
|
|
|
|
|
|
210 |
|
|
|
|
|
|
|
|
|
[ms] |
210 |
|
|
|
|
|
|
|
|
|
200 |
|
|
|
|
|
|
|
|
|
200 |
|
|
|
|
|
|
|
|
|
|
190 |
|
|
|
|
|
|
|
|
|
duration |
190 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
180 |
|
|
Word length 8 |
|
|
|
|
180 |
|
|
|
|
|
|
|
|
|
|||
170 |
|
|
|
|
|
Fixation |
170 |
|
|
|
|
|
|
|
|
|
||||
|
|
Word length 5 |
|
|
|
|
|
|
|
|
|
|
|
|
||||||
|
|
Word length 7 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|||||
160 |
|
|
Word length 6 |
|
|
|
|
160 |
|
|
|
|
|
|
|
|
|
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
150 |
|
|
Word length 4 |
|
|
|
|
150 |
|
|
|
|
|
|
|
|
|
|||
140 |
|
|
Word length 3 |
|
|
|
|
140 |
|
|
|
|
|
|
|
|
|
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
130 |
−4 |
−3 |
−2 |
−1 |
0 |
1 |
2 |
3 |
4 |
|
130 |
–4 |
–3 |
–2 |
–1 |
0 |
1 |
2 |
3 |
4 |
−5 |
|
–5 |
||||||||||||||||||
Landing position |
Landing position |
(a) |
(b) |
Figure 7. Mean fixation duration as a function of word length and landing position within a word, based on all fixations except first and last fixations in a sentence. Empirical data (a) vs simulated (b) data.
(see Engbert et al., 2005, Appendix B with an incremental model comparison). The joint implementation of both principles is able to reproduce the IOVP effect for all fixations (Figure 7).
5. Discussion
We investigated the prevalence of mislocated fixations during normal reading. First, using an advanced iterative algorithm, we estimated the distributions of mislocated fixations from experimental data. This procedure is a refinement of a previous approach (Nuthmann et al., 2005), which corresponds to the first-order estimation (first iteration step) of the framework presented here. We showed that our algorithm converges after only a few iteration steps and generates self-consistent estimates of the distributions of mislocated and well-located fixations – as a decomposition of the experimentally observed landing distributions. The results indicate that mislocated fixations occur frequently. In our model simulations, about 20% of all fixations turned out to be mislocated. In experimental data, estimates for mislocated fixations as a function of word length range from a few percent for long words up to a third of all fixations for short words (see also Nuthmann et al., 2005).
Second, using numerical simulations of the SWIFT model, we checked the iterative algorithm. From model simulations, we computed the exact distributions of mislocated fixations. Furthermore, the distributions of mislocated fixations were reconstructed from
336 |
R. Engbert et al. |
simulated landing distributions by our algorithm in the same way as for the experimental data. Because the exact and reconstructed distributions of mislocated fixations were in good agreement, we conclude that our iterative algorithm precisely estimates mislocated fixations from landing position data.
Third, we investigated different cases of mislocated fixations occurring in SWIFT simulations. These analyses showed that patterns of mislocated fixations are very specific, with failed skipping being a major source of errors. Finally, we outlined the relation between mislocated fixations and the IOVP effect. We conclude that mislocated fixations play an important role in eye-movement control during reading.
Acknowledgements
We would like to thank Wayne Murray and Keith Rayner for valuable comments, which helped to improve the quality of the manuscript. This research was supported by grants from Deutsche Forschungsgemeinschaft (KL 955/3 and 955/6).
References
Deubel, H., & Schneider, W. X. (1996). Saccade target selection and object recognition: Evidence for a common attentional mechanism. Vision Research, 36, 1827–1837.
Engbert, R., & Kliegl, R. (2001). Mathematical models of eye movements in reading: a possible role for autonomous saccades. Biological Cybernetics, 85, 77–87.
Engbert, R., & Kliegl, R. (2003). Noise-enhanced performance in reading. Neurocomputing, 50, 473–478. Engbert, R., Kliegl, R., & Longtin, A. (2004). Complexity of eye movements in reading. International Journal
of Bifurcation and Chaos, 14, 493–503.
Engbert, R., Longtin, A., & Kliegl, R. (2002). A dynamical model of saccade generation in reading based on spatially distributed lexical processing. Vision Research, 42, 621–636.
Engbert, R., Nuthmann, A., Richter, E., & Kliegl, R. (2005). SWIFT: A dynamical model of saccade generation during reading. Psychological Review, 112, 777–813.
Findlay, J. M., & Walker, R. (1999). A model of saccade generation based on parallel processing and competitive inhibition. Behavioral and Brain Sciences, 22, 661–721.
Kliegl, R., & Engbert, R. (2003). SWIFT explorations. In J. Hyönä, R. Radach, & H. Deubel (Eds.), The mind’s eyes: Cognitive and applied aspects of oculomotor research (pp. 103–117), Oxford: Elsevier.
Laubrock, J., Kliegl, R., & Engbert, R. (2006). SWIFT explorations of age differences in reading eye movements.
Neuroscience and Biobehavioral Reviews, 30, 872–884.
McConkie, G. W., Kerr, P. W., Reddix, M. D., & Zola, D. (1988). Eye movement control during reading: I. The location of initial eye fixations on words. Vision Research, 28, 245–253.
Nuthmann, A., Engbert, R., & Kliegl, R. (2005). Mislocated fixations during reading and the inverted optimal viewing position effect. Vision Research, 45, 2201–2217.
Pollatsek, A., Reichle, E. D., & Rayner, K. (2006). Tests of the E-Z Reader model: Exploring the interface between cognition and eye-movement control. Cognitive Psychology, 52, 1–56.
Radach, R., & Heller, D. (2000). Spatial and temporal aspects of eye movement control. In A. Kennedy, R. Radach, D. Heller, & J. Pynte (Eds.), Reading as a perceptual process (pp. 165–191). Oxford: Elsevier
Rayner, K. (1979). Eye guidance in reading: Fixation locations within words. Perception, 8, 21–30.
Rayner, K., Warren, T., Juhasz, B. J., & Liversedge, S. P. (2004). The effect of plausibility on eye movements in reading. Journal of Experimental Psychology: Learning, Memory and Cognition, 30, 1290–1301.
Ch. 14: An Iterative Algorithm for the Estimation of Mislocated Fixations |
337 |
Reichle, E. D., Rayner, K., & Pollatsek, A. (2003). The E-Z Reader model of eye movement control in reading: Comparisons to other models. Behavioral and Brain Sciences, 26, 445–526.
Richter, E. M., Engbert, R., & Kliegl, R. (2006). Current advances in SWIFT. Cognitive Systems Research, 7, 23–33.
Vitu, F., McConkie, G. W., Kerr, P., & O’Regan, J. K. (2001). Fixation location effects on fixation durations during reading: an inverted optimal viewing position effect. Vision Research, 41, 3513–3533.
This page intentionally left blank
PART 5
EYE MOVEMENTS AND READING
Edited by
ROBIN L. HILL
This page intentionally left blank
Chapter 15
EYE MOVEMENTS IN READING WORDS AND SENTENCES
CHARLES CLIFTON, Jr., ADRIAN STAUB, and KEITH RAYNER
University of Massachusetts, Amherst, USA
Eye Movements: A Window on Mind and Brain
Edited by R. P. G. van Gompel, M. H. Fischer, W. S. Murray and R. L. Hill Copyright © 2007 by Elsevier Ltd. All rights reserved.
342 |
C. Clifton, Jr., et al. |
Abstract
Word recognition processes seem to be reflected quite straightforwardly in the eye movement record. In contrast, eye movements seem to reflect sentence comprehension processes in a more varied fashion. We briefly review the major word identification factors that affect eye movements and describe the role these eye movement phenomena have played in developing theories of eye movements in reading. We tabulate and summarize 100 reports of how syntactic, semantic, pragmatic, and world-knowledge factors affect eye movements during reading in an initial attempt to identify order in how different types of challenges to comprehension are reflected in eye movements.
Ch. 15: Eye Movements in Reading Words and Sentences |
343 |
Readers move their eyes through a text in order to acquire information about its content. Measurements of the duration and location of the fixations they make have taught researchers a great deal about how people acquire information from the printed text, how they represent it, and how they integrate it in the course of understanding a text (see Rayner, 1978, 1998, for extensive overviews). Much of the systematic variance in fixation duration and location can be attributed to processes of recognizing the individual words in the text. Understanding of the relation between word recognition and eye movements has progressed to the point where several formal and implemented models of eye movements exist. Many of these models are described in detail, as well as compared and evaluated, by Reichle, Rayner, and Pollatsek (2003; more recent descriptions of new or updated models can be found in Engbert, Nuthmann, Richter, & Kliegl, 2005; Feng, 2006; McDonald, Carpenter, & Shillcock, 2005; Pollatsek, Reichle, & Rayner, 2006; Reichle, Pollatsek, & Rayner, 2006; Reilly & Radach, 2006; Yang, 2006). In our opinion, the most successful models are those that link the word recognition process to the time when an eye moves from one fixation to the next and the target of the saccade that accomplishes this movement. Our favored model, the E-Z Reader model (Pollatsek et al., 2006; Rayner, Ashby, Pollatsek, & Reichle, 2004; Reichle et al., 1998; Reichle et al., 1999), predicts a large proportion of the variance in eye movement measures on the basis of variables whose effect on word recognition has been independently established.
Despite their success, word recognition-based models of eye movement control do not yet provide fully satisfactory answers about all aspects of eye movements during reading. In the E-Z Reader model, a distinction made between two phases of recognizing a word (which are assumed to control different aspects of programming a saccade and the shifting of attention) has been criticized as being not fully compelling (see the replies to Reichle et al., 2003). No model fully specifies the nature of the mental representations of words (e.g., their orthographic or phonological or morphological content) nor does any model fully specify how information that specifies these different representations is acquired foveally vs parafoveally. No model fully specifies how the sequence in which orthographic symbols that appear in a printed word is mentally represented. And, even though it has been clear at least since Frazier and Rayner (1982; Rayner et al., 1983) that higher-level factors such as syntactic parsing and semantic integration can influence fixation durations and eye movements, no existing model adequately accounts for their effects.
In the first section of this chapter, we will briefly review some of the well-understood effects of word recognition on eye movements and comment on the extensions of these effects that are discussed in the chapters that appear in the Eye Movements and Reading part of this volume. In the next part, we go on to analyze the effects of syntactic, semantic, and pragmatic factors on eye movements, and discuss one basis of the difficulty of modeling, namely the apparently variable way that these factors find expression in eye movements. We begin this section with a discussion of one case study of how different measurements of eye movements can provide very different pictures of how some highlevel factors influence reading and language comprehension (Clifton, 2003). We continue with an extensive survey of published articles that investigate the effects of high-level