Reading as a Perceptual Process

A. Kennedy, D. Heller, J. Pynte, Ralph Radach

• Publisher: North Holland
• Year: 2000
• ISBN:
  9780080515762
• Topic: Psychology
• Subtopic: Experimental Psychology
• Index: Psychology

This book is divided into five sections dealing with various fundamental issues in current research: attention, information processing and eye movement control; the role of phonology in reading; syntax and discourse processing and computational models and simulations. Control and measurement of eye movements form a prominent theme in the book. A full understanding of the where and when of eye movement control is a prerequisite of any complete theory of reading, since it is precisely at this point that perceptual and cognitive processes interact.

Amongst the 'hot topics' included are the relation between parafoveal and foveal visual processing of linguistic information, the role of phonology in fluent reading and the emergence of statistical 'tuning' approaches to sentence parsing.

Also discussed in the book are three attempts to develop quantitative models of reading which represent a significant departure in theory-building and a quantum step in the maturation of reading research.

Much of the work reported in the book was first presented at the 5th European Workshop on Language Comprehension organised in April 1998 which was held at the CNRS Luminy Campus, near Marseilles. All contributions summarise the state-of-the-art in the relevant areas of reading research.

Section 1

Visual Word Processing

Chapter 1

Traces of Print Along the Visual Pathway

Tatjana A. Nazir ISC-CNRS, Lyon

Abstract

Skilled readers can recognize printed words during one single glance and they do so almost equally well for short as for long words. From a visual point of view the absence of a ‘word length effect’ is a rather surprising phenomenon. Given that every additional letter moves outer letters of a word farther into parafoveal vision, the probability that an entire word will be perceived during a fixation should decrease as words become longer. The word length effect which is expected but not observed with skilled readers is, however, noticed in beginning readers. Reading performance of first graders decreases linearly as words become longer and only after several years of reading experience word length ceases to influence performance. Curiously, a similar length effect can be observed after circumscribed lesions in temporo-occipital regions of the left hemisphere in premorbidly literate adults. What is acquired by skilled readers that made them recognize long words as easy as short words and that can be destroyed after localized brain lesions? With a series of examples that demonstrate strong correlations between eye movement pattern and reading performance I will try to argue that reading induces perceptual learning at early stages in the visual pathway. By this token perception of print becomes easy.

The missing word length effect

When a skilled reader is asked to recognize a briefly presented 5-letter word displayed such that the eye is fixating either the first, the second, the third letter of the word and so forth, we observe a systematic variation of performance with the location of the eyes in the word: the word is recognized best when fixated slightly left of center and performance is better with the eyes fixating on the first compared to the last letter in the word (Fig. 1). Visual acuity and differences in the way letters are perceived in the two visual fields are major factors contributing to this variation in performance (Nazir, Jacobs and O’Regan, 1998).

Fig. 1 Proportion of correct word recognition (in a task where participants were required to decide whether a briefly presented letter string was a word or not) as a function of the location of the eye in 5-, 7-, and 9-letters words. Replotted from Nazir (1993).

When the reader is asked to recognize a 7- or a 9-letter word, a similar pattern of result emerges. Independent of word length performance varies with gaze location following an inverted u-shape function, although we notice that the ‘viewing position effect’ (VPE) becomes more pronounced as words become longer. The height of the VPE-curves, however, remains more or less the same. As will be outlined in the following, this latter point bears a theoretical problem.

Given the linear increase of the eye’s minimum angle of resolution over the central 10° of the visual field (e.g., Olzak and Thomas, 1986), the probability of identifying single letters drops linearly as the distance from fixation increases. By multiplying these individual letter probabilities we can determine the theoretical probability that an entire letter string will be identified while the eye is fixating a certain location within the string. Table 1 illustrates the idea. In this example, the probability of identifying a directly fixated letter in a string of five letters is assumed to be 1. Given the drop of acuity, the probability of identifying neighboring letters in the string will decrease by a constant value. In the present case this drop-off rate is arbitrarily set to 0.03 (the exact value depends on visual display conditions). Thus, the probability of recognizing the letter immediately to the right of the fixated letter is (1 – 0.03) = 0.97, for the following letters it is 0.94, 0.91, and so forth. For reasons that will be discussed later, this drop-off rate is stronger when letters are displayed in the left visual field by a ratio of 1:1.8 (Nazir, O’Regan and Jacobs, 1991; see as well Bouma, 1973; Bouma and Legein, 1977; Hagenzieker, van der Heijden and Hagenaar, 1990; Nazir, Heller and Sussmann, 1992). Hence, when the probability to recognize a target letter drops from 1 at the center of gaze to [1 – 0.03] = 0.97 at a given eccentricity in the right visual field, it drops to [1 – (0.03 x 1.8)] = 0.946 when the letter is presented at the same eccentricity in the left visual field. The theoretical probability of recognizing the entire letter string is given by multiplying these individual letter probabilities and is plotted in the last column of Table 1. Fig. 2 displays these data together with empirical data from Fig. 1, demonstrating strong agreements between prediction and observation (for an alternative model see Clark and O’Regan, 1999).

Table.1

Theoretical probability of recognizing a 5-letter string as a function of fixation location

Position of the fixated letter in the string	Recognition probabilities of individual letters, 5-letter string					Probability of recognizing the entire string
	1	2	3	4	5
1	1	0.97	0.94	0.91	0.88	0.73
2	0.94	1	0.97	0.94	0.91	0.78
3	0.89	0.94	1	0.97	0.94	0.76
4	0.84	0.89	0.94	1	0.97	0.68
5	0.78	0.84	0.89	0.94	1	0.55

Note that the probability of recognizing the letter at fixation is set to 1. The drop-off rate is 0.03 going rightwards and (1.8 x 0.03) going leftwards. The probability of recognizing the entire letter string is calculated by multiplying individual letter recognition probabilities.

Fig. 2 Theoretical (left panel) and empirical word recognition data (right panel) as a function of fixation location in 5-letter words. Note that the shapes of the VPE-curves are almost identical. The difference in height of these curves comes from lexical factors. Words with different frequency of occurrence show a shift of the offset of the VPE-curve, with higher-frequency words being above those of lower frequency.

Fig. 3 displays theoretical VPE-curves for strings up to 9-letters length, calculated as in Table 1. Compatible with the empirical results (Fig. 1) the theoretical VPE becomes more pronounced as the length of the string increases. The height of these theoretical curves, however, drops systematically with every additional letter in the string. Thus, if word recognition were letter-based we should find signs of upgrading reading costs as words become longer¹. It is worth noting that the absence of a word length effect in skilled reading is typically interpreted as indicating parallel letter processing during word recognition (LaBerge and Samuels, 1974; but for a different view see Just and Carpenter, 1987). Given the limits of acuity it is obvious, however, that performance should deteriorate with word length despite the fact that letters are processed in parallel. Hence, the missing word length effect in skilled reading is a puzzle that remains to be solved.

Fig. 3 Theoretical VPE-curves for 5- to 9-letters words.

Conditions where word length does affect reading

Beginning readers and impaired readers

In contrast to skilled readers, systematic word length effects are noticed in beginning readers (Fig. 4). First graders recognize short words better than longer words independent of where in the word they are fixating. This length effect diminishes slowly as reading experience increases but disappears only after several years of reading experience (Aghababian and Nazir, in press). What takes place with experience is an adjustment of the height of the VPE-curves without apparent changes in their form. In other words, with increasing experience something about (long) words is learned that overcomes limits of acuity. Interestingly, children who do not develop normal reading skills (i.e., developmental dyslexia) persist in showing this length effect even after several years of instruction (Aghababian and Nazir, in prep.).

Fig. 4 VPE-curves for first graders (left panel) and adults (right panel) for words of different len...

Table of contents

• Cover image
• Title page
• Table of Contents
• Copyright page
• Reading as a Perceptual Process?
• Contributors
• Section 1: Visual Word Processing
• Section 2: Attention, Information Processing and Eye Movement Control
• Section 3: Phonology in Reading
• Section 4: Syntax and Discourse Processing
• Section 5: Models and Simulations
• Author Index
• Subject Index