Mixed Responses: Why Readers Spend Less Time at Unfavorable Landing Positions

Gary Feng

Full Text: PDF   Paper Package: Feng2009a_1.0 tar.gz PID: 11022/0000-0000-1F0D-1

Abstract


This paper investigates why the average fixation duration tends to decrease from the center to the two ends of a word. Specifically, it examines (a) whether unfavorable landing positions trigger a corrective mechanism, (b) whether the triggering is based on the internal efference copy mechanism, and (c) whether the corrective mechanism is specific to fixations that missed their targeted words. To estimate the mean and proportion of the corrective fixations, a 3-parameter mixture model was fitted to distributions of first fixation duration from two large eye movement databases in studies 1 and 2. Study 3 experimentally created mislocated fixations using a gaze-contingent screen shift paradigm. There is little evidence for the efference copy mechanism and limited support for the mislocated fixations hypothesis. Overall, data suggest a process that terminates fixations sooner than would during normal reading; it is triggered by the visual input during a fixation, and is flexibly engaged at eccentric landing positions and in reading short words. Implications to theories of reading eye movements are discussed.

 

PS:

Study 1 is based on the Dundee English reading corpus, kindly provided by Allen Kennedy. The "dundee.rda" data file contains all the cases, with additional variables derived from the data. Note, the actual fixated words are removed from this dataset as they are not relevant. The full data set is available from Dr. Kennedy. Also see the publication for original references for the Dundee corpus.

PPS:

Studies 2 & 3 are based on a subset of data from Feng et al. (2009). These are the English adult data; see the Method section. Complete dataset available upon request.

 

Feng, G. (2009). Mixed Responses: Why Readers Spend Less Time at Unfavorable Landing Positions. Journal of Eye Movement Research, 3(2):3, 1-26.www.jemr.org/online/3/2/2