Influence of prime lexicality, frequency, and pronounceability on the masked onset priming effect

Abstract

The present study investigates the origins of the masked onset priming effect (MOPE). There are two alternative interpretations that account for most of the evidence reported on the MOPE, so far. The speech planning account (SP) identifies the locus of the MOPE in the preparation of the speech response. In contrast, the dual-route theory proposes that the effect arises as a result of the processing of the prime by the nonlexical route. In a series of masked onset priming word naming experiments we test the validity of these accounts by manipulating the primes' frequency, their lexical status, and pronounceability. We found consistent MOPEs of similar magnitude with high- and low-frequency prime words as well as with pronounceable nonwords. Contrarily, when primes consisted of unpronounceable consonantal strings the effect disappeared, suggesting that pronounceability of the prime is a prerequisite for the emergence of the MOPE. These results are in accordance with the predictions of the SP account. The pattern of effects obtained in the present study further defines the origins of the MOPE.

Keywords:

• Masked onset priming
• Word naming
• Frequency
• Lexicality
• Pronounceability

Acknowledgments

This research has been partially supported by Grants SEJ2006–09238/PSIC and CONSOLIDER-INGENIO 2010 (CSD2008–00048) from the Spanish Ministry of Science and Innovation. The authors are grateful to Mark Seidenberg for providing information regarding his ongoing work. Thanks are also due to Johannes Ziegler and Conrad Perry for valuable comments on the operation of the CDP+ model. This manuscript would not have been possible without the helpful comments by Sachiko Kinoshita, Betty Mousikou, and an anonymous reviewer on earlier drafts.

Notes

¹ In a recent experiment, Mousikou et al. (2009) tested once more the dual-route's proposal that only the first letter/phoneme of masked primes is effectively processed in a masked onset priming experiment with nonword prime–target pairs. In contrast to what the theory predicted, the authors found shorter naming latencies when the primes shared the first two letters/phonemes with the targets than when they shared only the first letter/phoneme. However, the larger benefit (4 and 5 ms in the subject and the item analysis, respectively) for the two-letter/phoneme overlap appeared only for two thirds of the participants. The authors concluded that the reason for this unexpected difference was the sometimes faster operation of the nonlexical route of this subgroup of participants (see also Thompson, Connelly, Fletcher-Flinn, & Hodson, 2009, for evidence in the same line).

² The DRC model has successfully simulated MOPEs with word primes of varying frequencies with different sets of materials (see Mousikou et al., in press-a; Mousikou et al., in press-b). However, further information regarding the frequency of the specific primes that gave rise to the effects reported in these simulations has not been provided. The only simulation in which a subanalysis of the primes as a function of the MOPEs observed was the one carried out with the materials used by Forster and Davis (1991, Experiment 1). As mentioned, in this simulation the DRC did not produce a significant MOPE due to the inclusion of some very-high-frequency primes (Mousikou et al., in press-a).

³ Despite the fact that the MOPE for nonword primes was only significant in the analysis by participants, it should be clearly noted that the critical interaction between type of prime and relatedness was not significant in either the participant or the item analysis (both Fs  <  1.9, both ps > .28), and therefore there was no statistical need for the pairwise comparisons (i.e., these tests are only reported in order to show the consistency of the data).

⁴ Acknowledging the fact that the lexical activation of masked primes is problematic for the DRC model, Mousikou et al. (in press-a) have recently stated they are currently trying to minimize it. By reducing the lexical activation of masked primes, the DRC model could potentially simulate the significant MOPE with high-frequency prime words. Nevertheless, if such a modification were to be implemented, it would be unclear how the model could simulate existing evidence from other masked priming effects, reported in the reading aloud and the lexical decision literature, which have been shown to be modulated by the prime's frequency or lexical status (e.g., Andrews, 1996; Duñabeitia et al., 2009a; Duñabeitia et al., 2009b).

⁵ The possible reasons for the discrepancy across the results obtained in the MOPE studies reported by Kinoshita (2000), Mousikou et al. (in press-b), and Schiller (2004) have not been clearly identified. Mousikou and colleagues proposed as a potential reason the different methods used to measure the responses to the targets. Kinoshita and Schiller used voice keys, while Mousikou and colleagues used the CheckVocal software to hand-mark the responses of their participants. It should be also mentioned that Schiller's onset-related primes consisted of one or two letters followed by a string of percentage signs (e.g., b%%%%%%–BROEDER, br%%%%%– BROEDER), while both Kinoshita and Mousikou et al. used full letter primes. This difference could have affected the obtained pattern of effects in various ways, since the processing of symbols is fundamentally different to the processing of letters (for further discussion of this point see the Results and Discussion section of Experiment 2 and the General Discussion).