Acoustic input vs. phonological representationWhat you hear:ry o: h o: m i m i g a a tt e ry o: h o: d e o t o o k i k uGoal: recover string of phonemic symbolsProblems:• Segmentation• Categorization• High vowels are devoiced in Japanese when between voiceless obstruentsKazanina et al. (2006): perception ofphonemic versus allophonic contrastsVery similar patterns at the phonetic surface in Russian and Korean:Kazanina et al. (2006): perception ofphonemic versus allophonic contrastsHowever, different surface distributions:»Kazanina et al. (2006): perception ofphonemic versus allophonic contrastsAnd different underlying ( lexical) representations However, different surface distributions:Kazanina et al. (2006): perception ofphonemic versus allophonic contrastsIn other words (picture):phonemic contrastallophonic contrast/T/ [d] in between voiced sounds[t] elsewhereExperiment• Russian and Korean participants• stimuli: [ta]-[da] continuum– one for each language, constructed from a naturalrecording of [ta] by a native speaker• tasks:– Russians: identification– Koreans: naturalness rating– all: AX discriminationResultsidentification /ratingResultsidentification /ratingdiscriminationExperiment• MEG recordings during passive listening• oddball paradigm(AAAABAAAABAAAABAA…)• two conditions:– linguistic (da vs. ta)– non-linguistic (1 vs. 1.2 kHz tones)Magnetoencephalography (MEG)• a non-invasive neurophysiological ...
High vowels are devoiced in Japanese when between voiceless obstruents
o o k i
k
u
Kazanina et al. (2006): perception of phonemic versus allophonic contrasts
Very similar patterns at the phonetic surface in Russian and Korean:
Kazanina phonemic
However, different surface distributions:
et al. (2006): perception of versus allophonic contrasts
Kazanina phonemic
And different underlying (»lexical) representations
However, different surface distributions:
et al. (2006): perception of versus allophonic contrasts
Kazanina et al. (2006): perception of phonemic versus allophonic contrasts
In other words (picture):
/T/[d] in between voiced sounds [t] elsewhere
phonemic contrast
allophonic contrast
• •
•
Experiment
Russian and Korean participants stimuli: [ta]-[da] continuum one for each language, constructed from a natural recording of [ta] by a native speaker tasks: Russians: identification Koreans: naturalness rating all: AX discrimination
identification / rating
Results
identification / rating
discrimination
Results
•
•
•
Experiment
MEG recordings during passive listening
oddball paradigm (AAAABAAAABAAAABAA)
two conditions:
linguistic (da vs. ta)
non-linguistic (1 vs. 1.2 kHz tones)
• •
•
•
Magnetoencephalography (MEG)
a non-invasive neurophysiological technique measures the magnetic fields generated by neuronal activity of the brain; hence like EEG it provides a direct measure of brain function (cf. fMRI, which reflects brainmetabolismand only secondarily brain function) very high temporal resolution (like EEG); events with time scales on the order of milliseconds can be resolved excellent spatial resolution (like fMRI); sources can be localized with millimeter precision