Phonological and orthographic overlap effects in fast and masked priming

Abstract

We investigated how orthographic and phonological information is activated during reading, using a fast priming task, and during single-word recognition, using masked priming. Specifically, different types of overlap between prime and target were contrasted: high orthographic and high phonological overlap (track–crack), high orthographic and low phonological overlap (bear–gear), or low orthographic and high phonological overlap (fruit–chute). In addition, we examined whether (orthographic) beginning overlap (swoop–swoon) yielded the same priming pattern as end (rhyme) overlap (track–crack). Prime durations were 32 and 50 ms in the fast priming version and 50 ms in the masked priming version, and mode of presentation (prime and target in lower case) was identical. The fast priming experiment showed facilitatory priming effects when both orthography and phonology overlapped, with no apparent differences between beginning and end overlap pairs. Facilitation was also found when prime and target only overlapped orthographically. In contrast, the masked priming experiment showed inhibition for both types of end overlap pairs (with and without phonological overlap) and no difference for begin overlap items. When prime and target only shared principally phonological information, facilitation was only found with a long prime duration in the fast priming experiment, while no differences were found in the masked priming version. These contrasting results suggest that fast priming and masked priming do not necessarily tap into the same type of processing.

Keywords:

• Eye movements
• Fast priming
• Visual word recognition
• Orthography
• Phonology

This research was conducted when the first author was a sabbatical visitor at the University of California, San Diego. We are grateful to Timothy Slattery for his help in programming Experiment 1 and his help with analysing the data, to Andrew Olson and Bernhard Angele for their help with the statistical analyses, to Juneka Begum and Kelly Tipton for their help running Experiment 2, and to Ken Forster and Steve Lupker for their helpful comments on an earlier draft.This research was supported by Grant HD26765 to the third author. The second author was supported by a FQRSC (125964) Postdoctoral Fellowship.

Notes

1 Confirmation that our priming effects are very much in line with previous research comes from a comparison with previous research: The priming effects reported in Experiment 1 are comparable to those reported by Rayner et al. (1995) and slightly smaller than those found by H. W. Lee et al. (1999) and Y. A. Lee et al. (1999).

2 One P+O− item pair (crock–stalk) was mistakenly classified as rhyming and therefore deleted from all analyses, together with the data from the control nonword prime.

3 We also examined whether the P+O− condition showed an effect in other eye movement measures. To this end, we checked whether there was a main effect of priming overall, an interaction between prime and duration, or a priming effect at each duration interval separately for the following measures (see Pickering, Frisson, McElree, & Traxler, 2004): percentage of first-pass regressions out, regression-path duration, percentage of regressions in, second-pass duration, and total gaze duration. None of the comparisons approached significance (all ps > .22), again emphasizing the need to treat the single fixation finding for this condition cautiously.

4 There were also near significant main effects of duration for some of the different types, due to longer fixation times in general for the longer prime duration. We do not discuss this further.

5 In R, this means that instead of a model such as F1*F2*F3 + (1+F1:F2:F3|ppt) + (1+F1:F2:F3|item), we used F1*F2*F3 + (1|ppt) + (1|ppt:F1:F2:F3) + (1|item) + (1|item:F1:F2:F3).

6 Note that these fixation times are somewhat longer in the present experiment than those observed in normal reading experiments, which is due to the absence of a parafoveal preview, though in line with other fast priming experiments (e.g., H. W. Lee et al., 1999; Y. A. Lee et al., 1999).

7 One could argue that the P+O− primes also exhibit higher orthographic overlap with the targets than the nonword primes. While this is indeed the case for most primes (e.g., the prime–target pair greet–bleat have two letters in common, while nuxin–bleat do not), research by Humphreys, Evett, and Quinlan (1990) and Humphreys, Evett, Quinlan, and Besner (1987) has shown that orthographic priming with primes that only have 2 out of 4, or 2 out of 5, letters overlapping is comparable to having a row of xs as prime. We also calculated whether the relative Levenshtein distance between the prime and target (Levenshtein distance divided by the number of letters), which measures degree of orthographic overlap, was a significant predictor of the reading times. This was not the case, all ps > .10.

8 Note that this prime duration is longer than the prime durations used in homophone fast priming experiments, in which no homophone priming has been found for durations longer than 36 ms.