Phonology by itself: Masked phonological priming effects with and without orthographic overlap

Abstract

The activation of the phonological code plays a central role in visual word recognition. However, it is still unclear how this activation is integrated within this mainly bottom-up process. In the present masked priming study we combined Greek and Spanish, two languages with common phonemes and partially overlapping graphemes to investigate this issue. Greek-Spanish bilinguals performed lexical decisions on Greek and Spanish targets, briefly preceded by either phonologically related or orthographically and phonologically related prime words of the nontarget language. Results revealed significant bidirectional cross-script masked phonological priming effects which disappeared under the influence of nearly overlapping orthographic representations. This pattern of effects suggests that there is fast and automatic language nonselective activation of the phonological code during the initial stages of visual word recognition but that this is clearly dependent on the orthographic properties of the input stimulus. The implications of our findings are discussed within the framework of current models of monolingual and bilingual visual word recognition.

Keywords:

• Bilingualism
• Cross-script
• Masked priming
• Phonology

Acknowledgements

Maria Dimitropoulou was the recipient of a postgraduate grant from the Government of the Canary Islands (BOC 241, 02/12/2008). This research has been partially supported by Grants SEJ2006-09238/PSIC, CONSOLIDER-INGENIO 2010 (CSD2008-00048), and SEMA (PSI 2009-08889) from the Spanish Ministry of Science and Innovation.

The authors would like to thank the Society of Hispanic Studies in Athens, “Miguel de Cervantes”, for the collaboration in the recruitment of participants. Thanks are also due to Joana Cholin, Rob Hartsuiker, and Wouter Duyck for their technical support and suggestions. The authors also express their gratitude to Janet van Hell, to Ana Schwartz, and to two anonymous reviewers for helpful comments on an earlier draft.

Notes

¹In order to easily identify the overlapping segments (orthographic and/or phonological) across primes and targets, these are presented in underlined format throughout the paper.

²The Greek-Spanish language combination would have also been ideal to study orthographic effects without the influence of phonology. Unfortunately, this is not possible since the group of Greek letters overlapping visually to Roman ones but with distinct phonological representations is too small to create a sufficiently large list of experimental items (see Table 1).

³Out of the 3.5 (72%) phonemes common across related primes and targets of the P + O– condition, 0.4 (9%) were also visually similar. Word items containing these letters had to be included in the experimental material due to the limited amount of items.

⁴All the van Orden graphemic similarity scores (van Orden, 1987) were calculated in an online application developed by R. Hartsuiker (http://users.ugent.be/~rhartsui/Applet1.html). The procedure followed to calculate the van Orden score of graphemic similarity for each prime–target pair was as follows: First, 20 students of the University of La Laguna rated the visual similarity between the Greek and the Roman versions of the letters. Those Greek letters that were rated as visually indistinguishable from Roman letters were then replaced in the Greek words by their Roman version. Then, the graphemic similarity scores for each Spanish word was calculated by comparing them to their exact repetitions (e.g., the similarity of fibra [fibre] and fibra). Following this, the graphemic similarity between each Spanish word and its paired Greek related and unrelated word was calculated (e.g., fibra-f?tpo [bud]). Finally, the van Orden score for each prime–target pair was obtained dividing the first value by the second.

⁵We thank Janet van Hell for suggesting these analyses.

⁶Note that in the effect obtained by Dijkstra et al. (1999) for interlingual homographs with shared orthophonological representations there was a trend towards inhibition, also evident in our study for the P + O+ condition. This trend reached significance in a further combined analysis of Experiments 1 and 2 in which target language was included as a factor, F1(1, 27) = 1.56, MSE = 1006, p > .22; F2(1, 118) = 4.95, MSE = 1246, p < .05. However, considering that this result was only significant in the analysis by items, we will not draw further conclusions at this regard.