Auditory Masked Priming and Lexical Processing in People with Differing Familial Handedness Julia Fisher1, Roeland Hancock1, Thomas G. Bever1

Prior research claims that in early sentence processing right-handers with familial left-handedness (FS+) focus on lexical information, while right-handers without it (FS-) focus on syntax [18]. To determine whether this difference exists in isolated word recognition, we contrasted FS+ and FS- lexical decision using Kouider and Dupoux’s [12] auditory masked priming paradigm.

Auditory masked priming presents subjects with masked primes followed by unmasked targets. Masking is achieved through overlaid noise and prime compression. Using auditory masked priming with synthetic English, Davis et al. [7] found repetition priming for only low neighborhood density words. We used the factor neighborhood density in our study with naturally-spoken English, and further explored auditory masked priming by allowing a small variation in prime-target delay.

We find that FS+ subjects experience priming for both high and low frequency targets while FS- subjects only experience priming for low frequency targets. This suggests that FS+individuals have greater facility with lexical processing than FS- individuals. Additionally, we find priming for both low and high neighborhood density words, suggesting that past results may be due to loss of information in synthetic speech. Finally, the presence of priming shows that the paradigm is robust to small variations in its structure.

Familial Handedness

Past research has demonstrated that there exist language processing and neurological differences between right-handed individuals with (FS+) and without (FS-) left-handed blood relatives. FS- individuals have been shown to be more sensitive than FS+ individuals to the following:

��the difference between main and subordinate clauses ([2] & [6], as cited in [1]) � the difference between active and passive sentences ([5] as cited in [1])��word order [18]

In contrast, FS+ individuals have been shown to do the following:

��read words faster than FS- individuals ([5] as cited in [1]) ��have faster response times in a task involving sentence fragments and probe words/ phrases [18]

Combined, these results indicate that FS- individuals are more sensitive to structural/grammatical information than FS+ individuals, while FS+ individuals are more efficient than FS- individuals at lexical processing. In terms of neurological differences, FS+ individuals were found to suffer from major aphasia less frequently and recover from such aphasia more quickly than FS- individuals when the left-hemisphere was damaged [13]. In an fMRI study of left-hemispheric language dominance, Tzourio-Mazoyer et al. [19] showed that FS+ right-handers with weak manual preference were not left-hemisphere dominant while listening to a story in their native language. Relatedly, Hancock [9] found that a model of the additive genetic effects of handedness correlated with EEG asymmetries.

Because many of the psycholinguistic studies described above involved both lexical and syntactic processing, a question that arises is whether or not FS+ and FS- individuals differ in lexical processing outside of a syntactic context. We investigate this using a relatively new paradigm: auditory masked priming.

1University of Arizona

Auditory Masked Priming

Kouider and Dupoux’s [12] masked priming task is an auditory analogue of Forster and Davis’s [8] visual masked priming paradigm. Like in visual masked priming, auditory masked priming presents subjects with a prime followed by a target. The prime is masked in multiple ways. First, it is sandwiched between auditory “masks” — words that have been reversed, time-compressed, and volume-attenuated. A single forward mask precedes the prime, and multiple backward masks follow it. Additionally, the prime is compressed and volume-attenuated to the same degree as the masks. The target retains its original length and volume and is superimposed immediately after the prime over the backwards masks. The auditory impression is of a loud word spoken over background noise.

Seventy-one native-English-speaking undergraduates at the University of Arizona (34 men, 37 women, mean age = 19.64 years) participated in the experiment. They completed three computer tasks including the lexical decision auditory masked priming task. Presentation of the computer portions of the experiment were done using the Psychophysics Toolbox extensions for Matlab [4], [11], [15]. Following the computer tasks, subjects completed a handedness questionnaire adapted from Oldfield’s [14] handedness inventory, a familial handedness questionnaire [10], and a language history and usage survey.

Materials

A modified version of a script written by Scott Jackson and Dan Brenner for the computer software PRAAT [3] was used to compress the words and pseudowords to 35% of their original length and create the stimuli. The lexical statistics used to select the targets and primes were taken from the Irvine Phonotactic Online Dictionary [21]. All stimuli were recorded by a female native English speaker in her early twenties who grew up both in Tucson, AZ and in Monterey, CA. She was a student at the University of Arizona at the time of the recording. A brief description of the auditory masked priming stimuli is below:

��240 targets: 120 words, 120 pseudowords��All bisyllabic��Target words chosen to have raw frequency between 10 and 100 occurrences/million��Target pseudowords chosen to have unstressed, unweighted, word-average biphoneme, triphoneme, and positional probabilities at most one standard deviation below the respective means for real words��Half of targets: low neighborhood density (1-3 neighbors)��Half of targets: high neighborhood density (10+ neighbors)��Spoken lexical uniqueness point in rhyme of second syllable or immediately after final phone��Primes chosen based on same criteria as targets��Primes and target pairs created by Python script written by first author with verification of semantic dissimilarity also ensured by first author.

Only right-handed subjects (as determined by strength of hand preference for writing and throwing and strength of foot preference for kicking), subjects that achieved an accuracy rate of at least 80%, and subjects without experimental glitches were included in the analyses. This left 58 subjects (30 women: 19 FS+/11 FS- ; 28 men: 12 FS+/16 FS-). All of the statistical work was done using the R system for statistical computing [16]. Reaction times were measured from the beginning of each target. For each subject, reaction times greater than two standard deviations from the mean were cut to two standard deviations from the mean. Additionally, in order to better approximate a normal distribution, all cut reaction times were transformed using the natural logarithm. In order to satisfy homogeneity of variance (as measured by Levene’s Test of Homogeneity of Variance), words and pseudowords were analyzed separately.

In addition to the factor counterbalanced group, two factors not included in the original design were added to the analysis in order to remove error from the error term: frequency of target (low or high based on an even split of the word targets) and location of the lexical uniqueness point (during the target or immediately following it). Thus, both words and pseudowords were analyzed with a mixed six/seven factor analysis of variance with the following factors:

��familial handedness: FS+/FS-��gender: male/female��neighborhood density: low/high��prime type: repetition/unrelated��frequency (only used for the word analysis): low/high��lexical uniqueness point: during target/following target��counterbalanced group: two levels

Words: Due to a significant interaction among all factors except counterbalanced group (F1(1,50) = 5.11, p < 0.05; F2(1,104) = 5.51, p < 0.05), we split the data by familial handedness and examined the effects of the other factors. For FS+ subjects, there was a significant main effect of prime type (F1(1,27) = 26.38, p < 0.001; F2(1,104) = 24.84, p < 0.001) and a significant interaction among frequency, lexical uniqueness point, and gender (F1(1,27) = 4.83, p < 0.05, F2(1,104) = 5.02, p < 0.05). When investigated, the interaction showed no significant sub-effects. However, the effect of prime type showed that FS+subjects experienced on average 29 ms of priming.

For FS- subjects, there was a significant effect of prime type (F1(1,23) = 13.80, p < 0.01; F2(1,104) = 12.47, p < 0.001) and a marginally significant interaction between prime type and frequency (F1(1,23) = 6.51 p < 0.05; F2(1,104) = 3.76 p = 0.055). An exploration of the interaction revealed that for low frequency targets, there was no effect of prime type (p > 0.05). For high frequency targets, however, prime type was significant (F1(1,25) = 14.86, p < 0.001; F2(1,58) = 8.51, p < 0.01).

These results suggest that right-handed familials only experienced priming for high frequency words while left-handed familials experienced priming for both frequency levels. See Figure 2. In order to truly determine whether or not priming differed between FS+ and FS- subjects for low frequency targets, we calculated a log priming score for each subject and target and conducted an analysis of variance with factors familial handedness and counterbalanced group. Results showed that FS+ subjects experienced marginally greater priming than FS- subjects for low frequency word targets (F1(1,54) = 3.28, p = 0.076; F2(1,58) = 3.19, p = 0.079).

Auditory Masked Priming Discussion

In terms of the auditory masked priming paradigm, we found that the task is robust to minor variations in timing. Even though the time between the end of the prime and the beginning of the target varied, we still found repetition priming for words. In contrast to Davis et al. [7], however, we did not find a priming difference between word targets of differing neighborhood densities. Rather, both neighborhood density ranges experienced priming; there were no interactions indicating differences in the amount of priming. See Figure 3. Synthetic speech is unlikely to contain the full set of cues available in natural speech. It is possible that information loss due to synthetic speech made processing of high neighborhood density words more challenging, causing them not to experience priming in [7]. To fully understand the interplay between the paradigm and speech type, more research is needed along the line of Schluter [17], who examined the differences in auditory masked priming results for synthetic and natural speech.

1. Bever, T.G., C. Carrithers, W. Cowart, and D.J. Townsend. 1989. Language processing and familial handedness. In: Galaburda, Al., Editor., 1989. From neurons to reading, MIT Press, Cambridge, MA. 331-360.2. Bever, T.G., C. Carrithers, and D. Townsend. 1989. Sensitivity to clause structure as a function of familial handedness. University of Rochester, Cognitive Sciences Technical Report no. 43.3. Boersma, P. 2001. “Praat, a system for doing phonetics by computer,” GLOT. 5. 341-345.4. Brainard, D. H. 1997. The Psychophysics Toolbox. Spatial Vision. 10. 433-436.5. Carrithers, C. 1988. Canonical sentence structure and psych-ergative verbs. Journal of Psycholinguist Research. 6. Cowart, W. 1988. Familial sinistrality and syntactic processing. In J. M. Williams and C. J. Long, eds., Cognitive approaches to neuropsychology. 273-286. New York: Plenum.7. Davis, Chris, Jeesun Kim, and Angelo Barbaro. 2010. Masked speech priming: Neighborhood size matters (L). Journal of the Acoustical Society of America. 127. 2110-2113.8. Forster, K. and C. Davis. 1984. Repetition priming and frequency attenuation in lexical access. Journal of Experimental Psychology: Learning, Memory, and Cognition. 10. 680-698.9. Hancock, R. 2012. Bayesian estimates of genetic handedness predict oscillatory brain activity. Presented at the 14th Annual Meeting of the International Behavioural and Neural Genetics Society. May 15-19. Boulder, Colorado.10. Hancock, R. & Bever, T.G. 2009. Familial Handedness Pedigree Form. University of Arizona, Tucson, AZ.11. Kleiner, M., D. Brainard, and D. Pelli. 2007. “What’s new in Psychtoolbox-3?” Perception 36 ECVP Abstract Supplement. 12. Kouider, Sid and Emmanuel Dupoux. 2005. Subliminal Speech Priming. Psychological Science. 16. 617-625.13. Luria, A.R. 1947. Traumatic aphasia: Its syndrome, psychopathology, and treatment (Russian). Moscow: Academy of Medical Sciences. Translation, The Hague: Mouton, 1970.14. Oldfield, R.C. 1971. The Assessment and Analysis of Handedness: The Edinburgh Inventory. Neuropsychologia. 9. 97-113.15. Pelli, D.G. 1997. The VideoToolbox Software for visual psychophysics: Transforming numbers into movies. Spatial Vision. 10. 437-442.16. R Development Core Team. 2009. R: A language and environment for statistical computing (ver. 2.14.1) [Software]. Vienna, Austria: R Foundation for Statistical Computing.17. Schluter, K. In prep. Sorry, Crystal, I think we should start hearing other people: The non- equivalence of synthetic and natural speech in subliminal speech priming. Ms., University of Arizona. 18. Townsend, D. J., C. Carrithers, and T. G. Bever. 2001. Familial Handedness and Access to Words, Meaning, and Syntax during Sentence Comprehension. Brain and Language. 78. 308-331.19. Tzourio-Mazoyer, N., L. Petit, A. Razafimandimby, F. Crivello, L. Zago, G. Jobard, M. Joliot, E. Mellet, and B. Mazoyer. 2010. Left Hemisphere Lateralization for Language in Right-Handers Is Controlled in Part by Familial Sinistrality, Manual Preference Strength, and Head Size. The Journal of Neuroscience. 30. 13314-13318.20. Ussishkin, A., A. Wedel, K. Schluter, and C. Dawson. In prep. Overcoming the Orthographic Confound in Semitic: Supraliminal and Subliminal Root and Pattern Priming in Maltese. Ms., University of Arizona.21. Vaden, K.I., Halpin, H.R., Hickok, G.S. 2009. Irvine Phonotactic Online Dictionary, Version 2.0. [Data file]. Available from http://www.iphod.com.

The original work with the paradigm showed that it can produce natural-speech word repetition priming in French (without subject awareness) when the prime is compressed to 35% of its original length [12]. Ussishkin et al. [20] found form priming in Maltese when prime-target pairs shared a consonantal root. Davis et al.’s [7] work on synthetic English showed that the paradigm can be sensitive to neighborhood density (only low neighborhood density word targets (not high) achieved priming.

We conduct a basic auditory masked priming experiment with repetition and unrelated primes. We follow Davis et al. [7] in using two neighborhood density groups in order to maximize the range of environments in which FS+ and FS- individuals might differ. We differ from previous auditory masked priming work in aligning the end of each target with the end of the backwards masks (see Figure 1). This produces differing amounts of time between the end of the prime and the beginning of the target. While this difference is subtle, it allows us to test the robustness of the paradigm in a novel way.

Figure 1. Visual depiction of an auditory masked priming stimulus item

Pseudowords: For the pseudoword analysis, one subject (male, FS-) had to be removed from the subject analysis and three targets (one low neighborhood density, two high neighborhood density) from the item analysis in order to ensure that the analysis contained no empty cells. The only significant effect was lexical uniqueness point (F1(1,49) = 132.41, p < 0.001; F2(1,109) = 14.50, p < 0.001). However, because the experiment was not designed to have equal numbers of targets with the two lexical uniqueness point levels, only 10 of the 120 pseudoword targets had lexical uniqueness points that fell after the target, and only seven of those were included in the pseudoword analysis. Thus, the above result, while intriguing, is likely an artifact of stimuli imbalance.

Figure 2. Error bars represent one standard deviation above and below the mean.

Figure 3. Error bars represent one standard deviationabove and below the mean.

RESULTSABSTRACT

INTRODUCTION

DISCUSSION

� FS+ and FS- individuals differ in isolated lexical processing. Specifically, FS+ individuals show priming for both high and low frequency word targets while FS- individuals only show priming for high frequency targets.

� Kouider and Dupoux's [12] auditory masked priming paradigm is robust to minor variations in stimulus timing.

� In contrast to Davis et al. [7], we find priming for both low and high neighborhood density word targets, indicating that the auditory masked priming paradigm is sensitive to speech type.

KEY FINDINGS

METHODS

REFERENCES

Familial Handedness Discussion

In terms of familial handedness, results revealed that there are frequency-modulated differences in isolated word processing between FS+ and FS- individuals. FS+ individuals experienced significant priming to both high and low frequency word targets. In contrast, FS- individuals only experienced priming to high frequency word targets, implying that they were not able to process low frequency repetition primes quickly enough to facilitate target response. These results are consistent with the prior findings that FS+ individuals are more efficient than FS- individuals in processing lexical information.

ACKNOWLEDGMENTS

We would like to thank Devon Dale, Nicholas Denisuk, Kimberly Golisch, and Vanessa Nguyen for for the vast amount of time they spent creating stimuli, running subjects, and processing data. We would also like to thank Kevin Schluter for his constant help with the auditory masked priming paradigm and Kenneth Forster, Adam Ussishkin, and Andrew Wedel for their thoughtful advice.

This work was supported by the National Science Foundation Graduate Research Fellowship Program.