This thesis explores the relationship between fixation location and the locus of attention during reading. Early theories of eye movement control during reading suggested that a very tight coupling exists between the two (Just & Carpenter, 1980); however, it has since been shown that dissociations do exist. Whether these dissociations necessarily implicate parallel lexical processing, or whether they can be accommodated for within a serial-sequential framework is explored in a series of experiments. Experiment 1 tested whether parallel lexical processing is, at the very least, psychologically plausible. Two horizontally aligned letter strings were presented simultaneously on a screen, the task being to decide whether they were physically identical or not. Even when presentation duration should have been short enough to prohibit the strictly serial processing of each word in turn, the results show clear lexical effects: high frequency words were responded to faster and with fewer errors than low frequency words. Effects of lexicality and orthography were also found. These results suggest that the two words had been processed at a lexical level in an overlapping fashion. Experiments 2 and 3 investigated the nature and range of word n+2 preview effects. In Experiment 2, word n+1 was either a determiner or 3-letter alternative higher frequency word; in Experiment 3, word n+1 was either a 4- or a 6-letter high frequency word. A gaze contingent display change technique was employed, where prior to passing an invisible boundary located immediately after word n, one, the other, neither or both of words n+1 and n+2 received a nonword preview. In addition to showing orthographic parafoveal-on-foveal effects stemming from word n+1, there was also evidence that word n+2 preview influenced targeting decisions on words n and n+1. Word n+2 preview effects are also found on word n+2 and in the spillover region. These effects were most wide ranging when word n+1 length was an average of 5- compared to 3-letters. Higher-level plausibility preview effects were explored in Experiments 4-6, again using a gaze contingent display change technique. In Experiment 4 word n+1 received either an identical preview, a different but plausible one, or an anomalous, or nonword preview. Critically, an effect of plausibility arose on word n+1, with anomalous previews receiving longer inspection times than alternative plausible previews. Experiments 5 and 6 investigated the range over which these effects might occur, testing for a plausibility preview effect on word n+2. Results showed numerical, but not statistical evidence for a plausibility-related preview effect on word n+2. There were, however, clear orthographic word n+2 preview effects. Finally, Experiment 7 experimentally tested the immediate oculomotor response to a mislocated fixation, using a text shift paradigm to simulate saccadic error and measuring the effect on lexical processing. Critically, this experiment showed that a quick error correction strategy appears to be engaged following a simulated saccadic undershoot, rather than a stay and process response. This suggests that a mislocated fixation account coupled with a stay and process response is unlikely to provide a viable explanation for lexical parafoveal-on-foveal effects. Overall, it is suggested that current instantiations of both serial (e.g., Reichle, Warren & McConnell, 2009) and parallel (e.g., Schad & Engbert, 2012) models of eye movement control during reading appear to fail to capture major aspects of these patterns of results. The results do, however, appear to fit most parsimoniously with a perspective on eye movement control that allows for multiple words to be processed in an overlapping fashion.

To investigate eye-movement control in reading, the present thesis examined three phenomena related to the eyes’ landing position within words, (1) the optimal viewing position (OVP), (2) the preferred viewing location (PVL), and (3) the Fixation-Duration Inverted-Optimal Viewing Position (IOVP) Effect. Based on a corpus-analytical approach (Exp. 1), the influence of variables word length, launch site distance, and word frequency was systematically explored. In addition, five experimental manipulations were conducted. <br><br> First, word center was identified as the OVP, that is the position within a word where refixation probability is minimal. With increasing launch site distance, however, the OVP was found to move towards the word beginning. Several possible causes of refixations were discussed. The issue of refixation saccade programming was extensively investigated, suggesting that pre-planned and directly controlled refixation saccades coexist. Second, PVL curves, that is landing position distributions, show that the eyes are systematically deviated from the OVP, due to visuomotor constraints. By far the largest influence on mean and standard deviation of the Gaussian PVL curve was exhibited by launch site distance. Third, it was investigated how fixation durations vary as a function of landing position. The IOVP effect was replicated: Fixations located at word center are longer than those falling near the edges of a word. The effect of word frequency and/or launch site distance on the IOVP function mainly consisted in a vertical displacement of the curve. The Fixation-Duration IOVP effect is intriguing because word center (the OVP) would appear to be the best place to fixate and process a word. A critical part of the current work was devoted to investigate the origin of the effect. It was suggested that the IOVP effect arises as a consequence of mislocated fixations, i.e. fixations on unintended words, which are caused by saccadic errors. An algorithm for estimating the proportion of mislocated fixations from empirical data was developed, based on extrapolations of landing position distributions beyond word boundaries. As a new central theoretical claim it was suggested that a new saccade program is started immediately if the intended target word is missed. On average, this will lead to decreased durations for mislocated fixations. Because mislocated fixations were shown to be most prevalent at the beginning and end of words, the proposed mechanism generated the inverted U-shape for fixation durations when computed as a function of landing position. The proposed mechanism for generating the effect is generally compatible with both oculomotor and cognitive models of eye-movement control in reading. / Um Blickbewegungen beim Lesen zu untersuchen, wurden in der vorliegenden Dissertation drei Phänomene in Bezug auf die Landeposition des Auges innerhalb des Wortes betrachtet, (1) die optimale Blickposition (OVP), (2) die präferierte Blickposition (PVL) und (3) der Invertierte Optimale Blickpositionseffekt für Fixationsdauern (IOVP). In einem corpus-analytischen Ansatz (Exp. 1) wurde systematisch untersucht, wie die Variablen Wortlänge, Sakkadenstartdistanz und Wortfrequenz die Parameter der OVP-, PVL-, bzw. IOVP-Funktion beeinflussen. Des weiteren wurden fünf experimentelle Manipulationen durchgeführt. <br><br> Erstens, die Wortmitte wurde als OVP identifiziert, operationalisiert als die Buchstabenposition innerhalb eines Wort, an der die Refixationswahrscheinlichkeit minimal ist. Mit zunehmender Sakkadenstartdistanz verschob sich die OVP jedoch in Richtung Wortanfang. Verschiedene in Betracht kommende ursächliche Faktoren für Refixationen wurden diskutiert. Des weiteren wurden Fragen zur Programmierung von Refixationen untersucht, wobei die Ergebnisse nahe legen, dass vorgeplante und unmittelbar gesteuerte Refixationssakkaden koexistieren. Zweitens, Landepositionsverteilungen (PVL-Kurven) zeigen, dass die Augen systematisch von der OVP abweichen, was im Wesentlichen auf visuomotorische Faktoren zurückzuführen ist. Mittelwert und Standardverteilung der normalverteilten PVL-Kurven wurden v.a. von der Sakkadenstartdistanz beeinflusst. Als dritter Schwerpunkt wurde untersucht, wie Fixationsdauern als Funktion der Landeposition variieren. Der Invertierte Optimale Blickpositionseffekt wurde repliziert: In der Wortmitte lokalisierte Fixationen sind länger als solche, die sich an den Worträndern befinden. Der Effekt von Wortfrequenz bzw. Sakkadenstartdistanz auf die IOVP-Funktion zeigte sich im Wesentlichen in einer vertikalen Verschiebung der Kurve. Der Befund eines invertierten OVP-Effektes für Fixationsdauern ist kontraintuitiv, denn die Wortmitte (OVP) wird als optimaler Ort betrachtet, um ein Wort zu fixieren und zu verarbeiten. Ein wesentlicher Beitrag der vorliegenden Arbeit bestand darin, ursächliche Faktoren für den IOVP-Effekt zu identifizieren. Es wurde vorgeschlagen, dass der Effekt auf sog. fehlplazierte Fixationen, d.h. Fixationen auf nicht-intendierten Wörtern, zurückzuführen ist. Fehlplazierte Fixationen werden durch okulomotorische Fehler in der Sakkadenprogrammierung verursacht. Es wurde ein Algorithmus entwickelt, um den Anteil fehlplazierter Fixationen aus empirischen Lesedaten abzuschätzen, basierend auf Extrapolationen von Landepositionsverteilungen über die Wortgrenzen hinweg. Als zentrale theoretische Annahme wurde formuliert, dass ein neues potentiell korrigierendes Sakkadenprogramm unverzüglich gestartet wird, wenn das intendierte Zielwort verfehlt wurde. Dadurch verringert sich die mittlere Dauer von fehlplazierten Fixationen. Da fehlplazierte Fixationen am häufigsten am Wortanfang und am Wortende auftreten, generierte der vorgeschlagene Mechanismus die invertierte U-Form für Fixationsdauern als Funktion der Landeposition. Der Mechanismus, der – gemäß der hier entwickelten Argumentation – dem IOVP-Effekt zugrunde liegt, ist prinzipiell sowohl mit okulomotorischen als auch mit kognitiven Theorien der Blicksteuerung beim Lesen vereinbar.