If perception is constructed, what keeps perception from becoming mere hallucination unlinked to world events? The visual system has evolved two strategies to anchor itself and correct its errors. One involves completing missing information on the basis of knowledge about what most likely exists in the scene. For example, the visual system fills in information only in cases where it might be responsible for the data loss. The other strategy involves exploiting the physical stability of the environment as a reference frame with respect to which the eyes and body can move.

Lehar provides useful insights into spatially extended phenomenology that may have major consequences for neuroscience. However, Lehar's biological naturalism leads to counterintuitive conclusions, and he does not give an accurate account of preceding and competing work. This commentary compares Lehar's analysis with that of Velmans, which addresses similar issues but draws opposite conclusions. Lehar argues that the phenomenal world is in the brain and concludes that the physical skull is beyond the phenomenal world. Velmans argues that the brain is in the phenomenal world and concludes that the physical skull is where it seems to be.

Steven Lehar allows too much to his direct realist opponent in using the word “subjective” of the sensory field per se. The latter retains its nonconceptual, nonmental nature even when explored by perceptual judgement. He also needs to stress the evolutionary value of perceptual differences between person and person, a move that enables one to undermine the direct realist's superstitious certainty about the singular object.

To avoid endless and futile debate, critics of an alternative paradigmatic hypothesis cannot simply state their own paradigmatic assumptions as if they were plain fact while dismissing those of the opposition as self-evidently absurd, because it is exactly those initial assumptions that are brought into question by the paradigmatic proposal. Perceived incredibility is no valid ground for rejection of a paradigm whose alternatives are at least equally incredible, and arguably more so.

The E-Z Reader model (Reichle et al. 1998; 1999) provides a theoretical framework for understanding how word identification, visual processing, attention, and oculomotor control jointly determine when and where the eyes move during reading. In this article, we first review what is known about eye movements during reading. Then we provide an updated version of the model (E-Z Reader 7) and describe how it accounts for basic findings about eye movement control in reading. We then review several alternative models of eye movement control in reading, discussing both their core assumptions and their theoretical scope. On the basis of this discussion, we conclude that E-Z Reader provides the most comprehensive account of eye movement control during reading. Finally, we provide a brief overview of what is known about the neural systems that support the various components of reading, and suggest how the cognitive constructs of our model might map onto this neural architecture.

E-Z Reader's account of the interaction between oculomotor and cognitive processes depends critically on distinguishing between early and late stages of lexical processing, because this distinction allows saccadic programming to be decoupled from shifts of attention. Precisely specifying the nature of this distinction has important implications both for current models of lexical retrieval and for the development of E-Z Reader 8.

Evidence is presented for the notion that selection-for-perception and selection-for-action progress in parallel to become tightly coupled at the saccade target before the execution of the movement. Such a conception might be incorporated in the E-Z Reader model of eye-movement control in reading.

Reichle et al. claim to successfully simulate a frequency effect of 60% on skipping rate in human data, whereas the original article reports an effect of only 4%. We suspect that the deviation is attributable to the length of the words in the different conditions, which implies that E-Z Reader is wrong in its conception of eye guidance between words.

We discuss five basic principles of E-Z Reader in terms of their potential for models of eye-movement control in object and scene perception. We identify several obstacles which may hinder the extrapolation of the E-Z Reader principles to nonreading tasks, yet find that sufficient similarities remain to justify using E-Z Reader as a guide for modeling eye-movement control in object and scene perception.

This commentary focuses on the limitations of the E-Z Reader model in its attempt to explain refixation saccades in reading. Listing factors that influence probability of refixating leads the model to assume two sorts of refixations. However, taking into account data on the metrics of refixation saccades allows us to propose an alternative explanation for empirical observations reported in the literature.

Computational models such as E-Z Reader and SWIFT are ideal theoretical tools to test quantitatively our current understanding of eye-movement control in reading. Here we present a mathematical analysis of word skipping in the E-Z Reader model by semianalytic methods, to highlight the differences in current modeling approaches. In E-Z Reader, the word identification system must outperform the oculomotor system to induce word skipping. In SWIFT, there is competition among words to be selected as a saccade target. We conclude that it is the question of competitors in the “game” of word skipping that must be solved in eye movement research.

Parameters in E-Z Reader models are estimated on the basis of a simple data set consisting of 30 means. Because of heavy aggregation, the data have a severe problem of multicolinearity and are unable to adequately constrain parameter values. This could give the model more power than the empirical data warrant. Future models should exploit the richness of eye movement data and avoid excessive aggregation.

This commentary analyses the quantitative parameters of Reichle et al.'s model, using estimates when explicit information is not provided. The analysis highlights certain features that appear to be necessary to make the model work and ends by noting a possible problem concerning the variability associated with oculomotor programming.

Reichle et al. show that saccades in reading are controlled by linguistic processing. The authors' Figure 13 shows the parietal and frontal eye fields as parts of a neural implementation. This commentary presents data from dyslexics performing nonreading saccade tasks. The dyslexics exhibit deficits in antisaccade control. Improvement of the deficits is achieved in 85% of the cases and results in advantages in learning how to read.

The authors' review of alternative models for reading is of great value in identifying issues and progress in the field. More emphasis should be given to distinguishing between models that offer an explanation for behavior and those that merely simulate experimental data. An analysis of a model's discrete structure can allow for comparisons of models based upon their inherent dimensionality and explanatory power.

A central component in the E-Z Reader model is a two-stage word processing mechanism made responsible for both the triggering of eye movements and sequential shifts of attention. We point to problems with both the verbal description of this mechanism and its computational implementation in the simulation. As an alternative, we consider the use of a connectionist processing module in combination with a more indirect form of cognitive eye-movement control.

We argue that although E-Z Reader does a good job in simulating many basic facts related to readers' eye movements, two phenomena appear to pose a challenge to the model. The first has to do with word length mediating the way compound words are identified; the second concerns the effects of initial fixation position in a word on eye behavior.

The E-Z Reader model assumes that the parafoveal selection for fixation and the subsequent selection for attention allocation encompass the same spatially distinct letter cluster. Recent data suggest, however, that an individual letter sequence is selected for fixation and that more than one letter sequence can be selected for attention allocation (processing).

E-Z Reader is a highly successful model of eye-movement control, employing the notion of a serial-sequential attentional spotlight switched from word to word. Evidence of parallel processing of words in text calls this notion into question. Modifications to the model to accommodate this evidence are possible but will not address the fundamental objection that reading should not be seen as “surrogate listening.”

We question the assumption of serial attention shifts and the assumption that saccade programs are initiated or canceled only after stage one of word identification. Evidence: (1) Fixation durations prior to skipped words are not consistently higher compared to those prior to nonskipped words. (2) Attentional modulation of microsaccade rate might occur after early visual processing. Saccades are probably triggered by attentional selection.