You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

 
Poster B47, Wednesday, November 8, 3:00 – 4:15 pm, Harborview and Loch Raven Ballrooms

Asymmetrical MMNs to socially-marked biological sounds: a potential challenge to the phoneme underspecification hypothesis

Roberto Petrosino1, Diogo Almeida2, Andrea Calabrese1, Jon Sprouse1;1University of Connecticut, 2New York University - Abu Dhabi

Mismatch negativities (MMNs) occur when two sounds are presented in an oddball paradigm with one sound played frequently as standard, and the other played infrequently as the deviant. Prima facie, swapping the roles of two sounds shouldn’t affect the MMN; however, there is a growing literature demonstrating asymmetrical MMN effects (aMMNs) when the two sounds are phonemes: the MMN is larger with one role ordering than the reverse. MMNs have long been used to probe abstract categories in speech perception (Näätänen et al. 1997, Phillips et al. 2000), as such these aMMNs have been interpreted as evidence that MMNs are sensitive to asymmetries in the abstract representation of phonemes, specifically whether each phoneme in the pair is fully specified for articulatory features in the lexicon, or underspecified for one or more features (e.g., Eulitz & Lahiri, 2004, Scharinger et al. 2012, Schluter et al. 2017). Our goal in this project is to better understand aMMNs so as to better interpret the strength of this evidence. We have three concrete questions: (i) Do aMMNs arise for other biological sounds?, (ii) Are aMMNs sensitive to other higher-order abstract categories, such as social markedness?, and (iii) Can lower-level acoustic differences explain the directionality of aMMNs? To answer these questions, we looked for aMMNs to three biological sounds – [flatulence], [cough] and [sniff] – that are (i) relatively frequent, (ii) differ in social markedness, and (iii) differ in spectral and temporal acoustic properties. We also included a linguistic pair – [s] and [z] – to confirm that we can observe aMMNs in general. We created ten tokens of each condition, 250ms long, processed using the best practices of the aMMN literature. We tested 24 right-handed participants in an auditory oddball paradigm with six blocks (with the order randomized for each participant): [s]-[z], [flatulence]-[cough], [cough]-[sniff], in each of the two role assignments. We calculated MMNs by comparing the deviant of a sound to the standard of the same sound from a different block (the identity MMN). For the linguistic condition, the MMN response to [z] was larger than the MMN to [s], as expected. For the biological sounds, the MMN to [flatulence] was larger than the MMN to [cough]. There was no significant difference between [cough] and [sniff]. First, our results suggest that aMMNs are not specific to language. This suggests that aMMNs could in principle be used to explore other types of abstract representations. Second, the larger MMN to [flatulence] suggests that aMMNs may be sensitive to the abstract representations of social markedness. Third, spectral MFCC-based analyses suggest that lower-level acoustic properties cannot explain this aMMN, as it places [cough] and [sniff] as most distinct from one another, and [flatulence] in between the two (contrary to our results). This new dimension to aMMNs raises a new question for existing aMMNs to linguistic sounds: Is the aMMN due to underspecification as previously argued, or due to a different abstract property like phonological markedness? Future studies dissociating underspecification and markedness will be necessary to answer that.

Topic Area: Perception: Auditory

Back to Poster Schedule