1. Phonetic Similarity Effects in Masked Priming Marja-Liisa Mailend 1, Edwin Maas 1, & Kenneth I. Forster 2 1 Department of Speech, Language, and Hearing Sciences 2 Department of Psychology The University of Arizona Background Acknowledgements We are grateful to our participants for their time and patience and we thank the Herbert E. Carter Travel Award and the Department of Cognitive Science at the University of Arizona for supporting this presentation. Contact Author: mailend@email.arizona.edu Introduction  Briefly presented masked primes facilitate word production when primes and targets share onsets (boat-BARN) compared to non-overlapping prime-target pairs (soap- BARN).  This Masked Onset Priming Effect (MOPE) is well-established (Schiller, 2008) but it remains unclear at what stage of language production this effect arises:  letter-to-sound mapping (e.g., Mousikou et al, 2010)  phonological encoding (e.g., Knoshita, 2000)  phonetic encoding (Forster & Davis, 1991) The Present Study  The goal of the present study is to test the hypothesis that masked primes influence the phonetic encoding of targets.  We manipulate the phonetic similarity (segments that share phonetic features vs. segments that do not) between primes and targets.  Logic: if masked primes influence target production at the level of phonetic encoding then we should observe the so called Phonetic Similarity Effect: longer reaction times for prime target pairs that share phonetic features (pill-BILL) compared to those that do not (hill-BILL).  Rationale: The Phonetic Similarity effect is best ascribed to the phonetic encoding stage because (a) it is not observed in phonological priming tasks (Roelofs, 1999), and (b) people with hypokinetic dysarthria – a motor speech disorder that does not affect phonology – show abnormal Phonetic Similarity Effect (Spencer and Rogers, 2005). Task Materials and Design  Targets were 12 monosyllabic words with a single stop consonant in the onset.  12 additional words that did not start with a stop consonant were used as fillers.  Each target was paired with four different primes defining the four conditions: 1.Similar condition –onsets differ in one feature, rhyme is shared Difference in place of articulation (pan-TAN) Difference in voicing (duck-TUCK) 2.Different condition –onsets share no features, rhyme is shared (man-TAN) 3.Onset Overlap condition– same onset and nucleus, different coda (tap-TAN) 4.Control condition – no form overlap (mug-TAN)  Each target appeared once in each condition.  Targets and fillers were presented in randomized order for each participant; each block contained all conditions. Analysis  Dependent variables were reaction time (RT) of correct responses, measured acoustically with the CheckVocal program (Protopapas, 2007), and error rate.  Error data and the reciprocal RT data were analyzed with linear mixed effects models in R (Baayen, 2008; Baayen, Davidson, & Bates, 2008).  Random effects: subjects and items  Fixed effects: the four conditions and trial number Method (continued) References Results (continued) Method Participants  25 undergraduate students from the University of Arizona participated.  Data from 6 students were excluded because English was not their first language and/or they exceeded the error rate criterion of 10%. Figure 1. Timeline of one trial: example from Similar condition: prime and target differ only in the voicing feature of the onset Results Baayen, R. H. (2008). Analyzing linguistic data: A practical introduction to statistics using R. New York: Cambridge University Press. Baayen, R. H., Davidson, D. J., & Bates, D. M. (2008). Mixed-effects modeling with crossed random effects for subjects and items. Journal of Memory and Language, 59, 390-412. Forster, K. I. & Davis, C. (1991). The density constraint on form-priming in the naming task: Interference effects from a masked prime. Journal of Memory and Language, 30, 1–25. Mousikou, P., Coltheart, M., & Saunders, S. (2010). Computational modelling of the masked onset priming effect in reading aloud. European Journal of Cognitive Psychology, 22(5), 725–763. Kinoshita, S. (2000). The left-to-right nature of the masked onset effect in naming. Psychonomic Bulletin and Reviews, 7, 133–141. Protopapas, A. (2007). Check Vocal: A program to facilitate checking the accuracy and response time of vocal responses from DMDX. Behavior Research Methods, 39, 859–862. Roelofs, A. (1999). Phonological segments and features as planning units in speech production. Language and Cognitive Processes, 14(2), 173–200. Spencer, K. A., & Rogers, M. A. (2005). Speech motor programming in hypokinetic and ataxic dysarthria. Brain and Language, 94, 347–366.  Few errors, and no significant differences in error rates between conditions except the Onset Overlap condition had fewer errors than the Similar condition (z=-2.52, p<.05).  Words were named faster in the Onset Overlap condition compared to the Control condition (t=5.86, p<.01). This replicates previous findings (e.g., Forster & Davis, 1991) and validates our experimental paradigm.  Critically, there was also a significant Phonetic Similarity Effect: words in the Similar condition were named slower than words in the Different condition (t=2.43, p<.05).  The Phonetic Similarity Effect was driven primarily by items that differed in place of articulation  place of articulation: 20 ms, t=3.79, p<.01  voicing: 7 ms, t=.13, p>.05 Figure 2. Errors per condition. Figure 3. Reaction times per condition. Figure 4. Similarity effect for items that differ in place (blue) and items that differ in voicing (red). Table 1. Reaction times per condition depicted separately for items that differ in place (blue) and items that differ in voicing (red). Discussion  The Phonetic Similarity Effect suggests that masked primes in the naming paradigm are planned for articulation and the mismatch between prime and target creates interference at the level of phonetic encoding.  Interestingly, the Phonetic Similarity Effect was observed for items that differed in place of articulation but not for items that differed in voicing. Several possible explanations exist for this discrepancy. 1.For items that differ in voicing, it is possible that participants sometimes started to produce the prime but were able to partly correct their motor plan on-line without affecting perceptual accuracy. This may have decreased RT for these trials as well as the mean RT for the group of items that differed in voicing. 2.There may be an interaction between the interference that comes from planning similar items and the mechanical advantage of knowing the place of articulation ahead of time.  A future direction for this research is to examine the role of shared rhymes between the prime and the target in order to investigate whether it is the similarity at the syllable or at the segmental level that produces the Phonetic Similarity Effect in masked priming. Such work is currently underway in our laboratory. Place differenceVoicing difference ##### duck TUCK Time: 479 mask 830 ms prime 50 ms target 830 ms feedback 830 ms