The role of age of acquisition and word frequency in reading: Evidence from eye fixation durations

Abstract

Over the past several decades, many researchers have examined how a word's age of acquisition (AoA) contributes to word recognition. Most of these studies, however, have used word-in-isolation experiments. At the same time, many studies have utilized eye tracking techniques to investigate the word frequency effect during reading. The present experiments sought to tie these two types of research together by investigating the influence of AoA and frequency on word processing. Specifically, eye movements were recorded as participants read sentences. The first experiment orthogonally manipulated frequency and AoA. In Experiment 2, participants read sentences that contained target words that varied in AoA but were controlled on various measures of word frequency. The same participants also read sentences containing high and low frequency words that were controlled on AoA. Results from these two experiments converged to demonstrate that both frequency and AoA affect eye fixation durations during sentence reading.

Acknowledgments

This research is based on a master's thesis by the first author at the University of Massachusetts. She was supported by a pre-doctoral fellowship on Grant MH16745. The research was also supported by grants HD17246 and HD26765. We thank the other members of the master's thesis committee, Chuck Clifton, Marvin Daehler, and Alexander Pollatsek, for their helpful comments. We also thank Bob Johnston, Viv Moore, and an anonymous reviewer for their helpful comments on an earlier draft. Portions of the data were presented at the 2001 meeting of the European Conference on Eye Movements and at the 2002 meeting of the Psychonomic Society.

Notes

¹AoA ratings were also collected from University of Massachusetts undergraduates for these words after the experiment, during the time when norms were being collected for the Experiment 2. The rated norms correlated .834 with the Gilhooly and Logie (1980) AoA estimates.

²The group factor was not a variable of interest in this experiment. Therefore, the results of this variable will not be discussed in the results section. The only significant effects of the grouping variable were as follows: There was three-way interaction between word frequency, AoA, and group if first fixation duration that was significant in the participants analysis, F1(2, 62) = 3.92, p = .025, but not in the items analysis, F2(2, 60) = 1.56, p=.219; a main effect of group in gaze duration that was significant in the participants analysis, F1(2, 31) = 5.26, p= .008, but not in the items analysis (p>.1); and a main effect of group in total fixation time that was significant in the participants analysis, F1(2, 31) = 4.64, p=.013, but was not in the items analysis (p>.3). Also, when WFG was added as a covariate in the analysis, there was a significant main effect of group on gaze duration, F2(2, 59) = 3.52, p = .036. One possible reason for these group effects is due to the fact that although items within the groups were controlled on many variables, there was a significant difference between the groups in word length, F(2, 69) = 9.92, p < .001, and understandability, F(2, 69) = 4.88, p = .01.

³Stimuli were selected for this analysis on the basis of their frequency. The range of the high frequency group was 43–204 per million. Three items in the high frequency condition had a frequency of 43. All three items were included in the analysis, thus resulting in an unequal number of items in each condition.

⁴This subset of items was not controlled on familiarity, imageability, or concreteness.

⁵Experiment 2 is an analysis of a subset of the stimuli used by Juhasz and Rayner (2003) in a multiple regression experiment designed to explore the nature of a set of five intercorrelated variables on eye fixation times. Data from 36 of the original 72 nouns used by Juhasz and Rayner were used in the reported analysis.

⁶The Celex frequencies (Baayen et al., 1995) used in this experiment are written frequencies that have been standardized to represent frequency out of 1 million words, as to be on the same scale and therefore comparable to Francis and Kucera's (1982) frequency.

⁷Items analyses are not reported for Experiment 2 for multiple reasons. First of all, the number of stimuli in each condition is not very large. This is due to the fact that the stimuli were chosen for a set of 72 items so as to be controlled on as many variables as possible. This limited the number of possible stimuli. Also, unlike in Experiment 1, each target noun appeared in a different sentence frame. While the sentences were as controlled as possible on variables such as target word predictability and target word goodness-of-fit, there is unavoidable extra noise added to the items' analyses.