An investigation into the perceptual and cognitive factors affecting word recognition during normal reading

The present thesis examines the effects of a range of factors on the processing of written language. The present thesis principally uses eye movement recording technology while participants read short passages of text. Factors known to influence written language processing range from lower-level perceptual constraints to higher-level discourse contingencies. Examples of lower-level to higher-level variables are, respectively, intraword orthographic constraints, such as word-initial letter constraint (WILC) – how many other words share the same three initial letters of a given word; lexical level word frequency – how often a word occurs in written language; and extraword contextual predictability – how likely a word is to occur given the discourse up to the position of the word in the passage. The present thesis not only investigates the main effects of these factors, but also studies the simultaneous effects that these factors have on written language processing. Information acquired from the right of current fixation location – parafoveal preview – is essential for reading to proceed at a normal rate. Preview is typically studied using gaze-contingent display change paradigms – non-foveal text is obscured or manipulated, and effects on eye movement behaviour recorded. The present thesis studies an additional method of measuring the effects of preview, without manipulating the text displayed: launch distance – how far readers’ prior fixation is from a given word, before foveal processing of that word. Visual acuity diminishes as retinal eccentricity increases. The present thesis examines the how the effects of the above factors, and any interactions between them, are modulated by launch distance. Standard effects of frequency and predictability were found across all studies. Lower-frequency words (LF) were processed with greater difficulty than higher-frequency words (HF); low-predictability words (LP) were processed with greater difficulty than (HP) words. Consistent with prior research (Rayner, Ashby, Pollatsek, Reichle, 2004), Experiment 1 found additive effects of frequency and predictability on eye movement behaviour. However, further investigation revealed that when preview was highest (i.e., Near launch distances), frequency and predictability exerted an interactive effect. Experiment 2a further investigated the simultaneous effects of frequency and predictability, addressing methodological concerns about Experiment 1. Principally, that HP contexts in Experiment 1 were medium-predictability (MP), potentially obscuring any interaction, as the acquisition of parafoveal information is influenced by the frequency and predictability of the parafoveal word. Comparing very low-predictability (VLP) items to very high-predictability (VHP) items, the interactive pattern of effects observed in the Near launch distance condition of Experiment 1 was replicated in the global analyses of Experiment 2a. In Experiment 2b, comparisons of HF and LF words in VLP and specifically-designed MP items yielded an additive pattern of effects, consistent with Experiment 1. Furthermore, conditionalised analyses of these items by launch distance showed an interactive pattern of effects, but only at Near launch distances. Conditionalised analyses of HF and LF words in VLP and VHP materials from Experiment 2a revealed an interactive pattern of frequency and predictability effects at both Near and Middle launch distances. It is argued that frequency and predictability can interact under two distinct conditions, but both manners are dependent on preview. When HF and LF words are presented in MP contexts, a high level of preview must be provided by a Near launch distance for an interaction to be observed; when HF and LF words are presented in VHP contexts, sufficient information can be extracted at further launch distances, generating an interactive pattern of effects in global analyses. Experiment 3 examines whether fixation durations are inflated prior to skipping a word in text. An overall non-significant effect of word skipping on prior fixation durations was observed. However, this result was somewhat misleading – inflated fixation durations prior to skipping were observed, but only when to-be-skipped words were either HF or HP; indeed, the largest mean inflation prior to skipping was observed when the to-be-skipped word was both HF and HP. These results suggest that when readers are able to extract most information about parafoveal words (e.g., when those words are HF or HP), fixation durations prior to skipping these words are inflated. It is tentatively suggested that these effects reflect a longer accumulation of information from parafoveal to-be-skipped word. These effects are consistent with models of eye movement control permitting parallel processing of written information, as opposed to a strictly serial approach. Experiments 4a and 4b tested the effects of WILC. Experiment 4a employed a lexical decision task, examining the separate and combined effects of WILC and frequency. LF words were responded to less quickly than HF words. Words with low WILC (LC words; e.g., “clown” shares its initial trigram “clo” with many words) were processed more quickly than words with high WILC (HC words; e.g., “dwarf” shares its initial trigram “dwa” with few words). It is suggested that LC words in a lexical decision task are responded to quickly as their initial trigram is shared by a large number of viable words, facilitating a “word” response. The initial trigrams of HC words are not shared by many other words, potentially hindering a “word” response. Experiment 4b re-tests the role of WILC on eye movement behaviour during reading, based on an earlier study by Lima and Inhoff (1985). Unlike Lima and Inhoff’s study, the frequency and predictability (known to influence the extraction of parafoveal information) of LC and HC target words was also manipulated. In contrast to the findings of Lima and Inhoff (but, consistent with their original prediction), HC words were found to exhibit a processing advantage over LC words in measures of eye movement behaviour reflecting early, lexical processing. Further analyses based on launch distances from, and landing positions within target words suggested that the pattern of effects observed may be due to the accumulation of WILC information from the parafovea. The present thesis finds that word frequency and contextual predictability can yield interactive effects on processing, but that any possible interaction is dependent on acquisition of parafoveal information. Evidence of inflated fixation durations prior to word skipping were observed, but these effects are modulated by the characteristics of the parafoveal to-be-skipped word. Initial letters of words have a substantial effect on processing, but this effect is task-dependent. In lexical decision, activation of “wordness” is advantageous, and LC words exhibit an advantage over HC words. In natural reading, information is available from sentential context and the parafovea, and HC words carry an advantage over LC words. The present thesis argues for the use of launch distance as a metric for measuring preview benefit, albeit in a necessarily post-hoc fashion. Reliable effects of launch distance were found across all experiments where it was examined as a factor – eventual fixation time on a word increases as the distance of prior fixation from beginning of that word increases. Launch distance was also shown to influence the effects of a range of factors which influence written language processing, and the interactions between these variables.EThOS - Electronic Theses Online ServiceGBUnited Kingdo