presented by Tobias Scheer (UCA, BCL) Lavinia Slabu (UCA, BCL) Project members

1. From Mind to Brain Deuxième Rencontre C@UCA 14-15 June Fréjus presented by Tobias Scheer (UCA, BCL) LaviniaSlabu (UCA, BCL) Project members Maureen Clerc (UCA, INRIA) Théodore Papadopoulo (UCA, INRIA) Demian Wassermann (UCA, INRIA) Fabien Mathy (UCA, BCL) Fanny Meunier (UCA, BCL) Alexander Chabot (doctorant, UCA, BCL)

2. overall goal • the project seeks to • identifycorrelates of phonologicalprocesses in the brain • isolatephonologicaleffects, as opposed to other possible origins • there are alternations in sound patterns • k-s • electri[k] – electri[s]-ity • how are theyproduced by grammar (Mind) and in the Brain? • linguistsdon't know. There are a number of analytic options. • an alternation maybe the result of • two distinct lexical entries ==> twowordsstored • allomorphy ==> two stems and an affixstored • phonetics ==> • spell-out (phoneticinterpretation) ==> translation phonophonetics • differentcomputational system: ==> morpho-phonology • analogy ==> looks at otherexistingwords

3. phonologists compare apples and pears • a theoryis about a set of phenomena to beexplained. • the theory of phonologyis about the set of phonologicalphenomena, i.e. about those alternations thatinvolvephonological computation. • despitesubstantial effort over decades, thereis no agreement: • nobodyknowswhatthis set looks like • in actual practice whether or not a given alternation isconsideredinvolvingphonologyislargely a matter or personal, theoretical and intuitive inclination. today phonology is a blindman's business

4. phonologists compare apples and pears • When you are unable to distinguish stone from plastic • currently phonologists are in the position of, say, geologists who aim to make a theory of the characteristics of stone, but are unable to distinguish stone from plastic. • They thus collect samples on which they build their theory, some of which contain 10% of plastic, others 30%, still others 60% and so on. • Unsurprisingly enough, competing theories built on these wildly varying sets of empirical material then significantly diverge – • not because of the theorizing itself but because of the plastic.

5. EEG will tell them phonologists don't know what a phonological process is EEG will tell them Goal for any alternation in any language, to build a machine that is able to tell whether it does or does not involve phonological computation. ==> build a phonology-detector

6. Sahin et al. (2009) • Sahin et al. (2009) • intra-cranial EEG (deep electrodes) • 3 patients implanted for medical evaluation (epilepsy) Sahin, Ned T., Steven Pinker, Sidney S. Cash, Donald Schomer & Eric Halgren 2009. Sequential Processing of Lexical, Grammatical, and Phonological Information Within Broca's Area. Science 326: 445-449.

7. Fig. 2 (A) Main results: sequential processing of lexical, grammatical, and phonological information in overlapping circuits. N T Sahin et al. Science 2009;326:445-449 Published by AAAS

8. adaptation as a phonology-detector A (n unintended) sideeffect of Sahin et al.: theirexpirmental setup allows us to isolatephonologicalactivity in the brain

9. work program of the project • replication of Sahin et al. • stimulus set: English plural formation • deep electrodes  surface electrodes • source localization (MRI anatomical pre-scan) • evaluation of the following (phonologically contentious) alternations: • velar softening (electri[k] - electri[s]ity) [English] • trisyllabicshortening (Chr[aj]st - chr[ɪ]stian) [English] • floatingconsonants (par[t]ir- il part) [French] • schwa-ɛ (app[ə]ler - il app[ɛ]lle) [French] • crazy rules (phenomenon to be determined) [language to be determ.] • 3. for each alternation • sameprotocol and stimuli, twomodalities • neuro-physiological (EEG, MRI pre-scan) • behavioural (reaction time)

10. language data: English plural formation • English plural • plural marker -s in spelling • in fact [-z]: tree – tree-[z] • but devoiced -z  -s when stem ends in voiceless consonant • rock - rock-[s] • no devoicing with stem-final • voiced consonants: job – job-[z] • sonorants: school – school-[z] • special behaviour of stem-final sibilants (s, z, ʃ, ʒ, ʧ, ʤ): epenthesis • fish - fiʃ- ɨ -[z] • page - paʤ- ɨ -[z] Two phonological processes involved: progressive devoicing -z  -s epenthesis

11. language data: English plural formation phonological processes involved

12. stimuli set #1: C-type voiced (20) voiceless (20) sibilants (20) sonorants (20) frequ. = lemma frequency COCA (per million)

13. stimuli set #2: frequency & length long & frequ. (20) long & rare (20) short & frequ. (20) short & rare (20) frequ. = lemma frequency COCA (per million)

14. Participants • 9 participants • language • English native speakers from USA/Canada • born or have grown up in the USA/Canada • undisrupted until at least age 10 • no other language spoken from birth) • stayed no more than 5 years outside of the USA/Canada recently • medical condition • healthy • right-handed • no speech or hearing disorders • no metallic implants (MRI) • normal or corrected-to-normal vision • no previous history of either mental or neurological disease • 4 female, 5 male • 20-38 years old

15. Pilot study • 64-channels, ANT neuro EEG, with the average of activity at all electrodes as reference and AFz as ground; • Online band-pass filter: 0.05 to 200 Hz ; • Sampling rate 2048; • Bipolar electrodes for HEOG and VEOG; • Faraday cage booth; Visual stimuli: • 80 unique words in 3 conditions repeated 3 times in 3 blocks, randomized Analysis: 0.05-100 Hz. Artifacts exceeding 100 μV were rejected.Epochs [-0.1, 1.2]s were averaged. Means of repeated-measures ANOVA on the peak amplitude, peak latency and mean amplitude (in a 4ms time window corresponding to the 3 peaks, N2, P3, N4) were calculated separately, including the factor Type (Null-Inflect, Overt-Inflect and Read).

16. Experimental Design Trials 1403-1603 in steps of 100 ms] Target Word Inflectional Cue This is + word 253 ms 1103 ms 603 ms + Cue epoch (1706 ms) 1503 ms (after Sahin et al., 2009)

17. Results Grand average ERP of all conditions and standard variation electrode F7 Sahin et al.'s 320ms (morpho-syntax) Sahin et al.'s 450ms (phonology) Sahin et al.'s 200ms (lexical access)

18. Results Overall replication of Sahin et al. • expected: • N2 (200ms): Read = Null = Overt • P3 (320ms): Read vs. Null & Overt • N4 (450ms): Read & Null vs. Overt

19. Results #1: all C-types mixed AF4 P3 * N4 * N2 not sign.

20. Results #1: all C-types mixed – zoom on N4 AF4 T-test ~ p=0,096 T-test * p=0,027 not the orderexpected

21. Results Replication by different C-types • expected at N4 (450ms): • voiced Cs: Read = Null = Overt (no extra phonological activity) • sonorant Cs: Read = Null = Overt (no extra phonological activity) • sibilant Cs: Read & Null vs. Overt (extra phonological activity) • voiceless Cs: Read & Null vs. Overt (extra phonological activity)

22. Results #2: onlyvoiced consonants AF3 N4 not sign.

23. Results #3: onlysonorants Fp1 N4 not sign.

24. Results #4: only sibilants N4 (corresp.) * Pz

25. Results #5: onlyvoiceless consonants AF3 N4 T-test * Overt vs. Read T-test Overt vs. Null: not sign.

26. Results Sahin et al.: effect of word frequency and word length • 450ms: • no effect of frequency (?) • long higher amplitude than rare • 200ms: • rare higher amplitude than common • no effect of length

27. Results Replication word frequency • expected: • N2 (200ms): rare higher amplitude than frequent • N4 (450ms): rare = frequent

28. Results #6: wordfrequency F5 N4 not sign. N2 not sign.

29. Results Replication word length • expected: • N2 (200ms): long = short • N4 (450ms): long higher amplitude than short

30. Results #7: wordlength F7 N2 not sign. N4 ~ p=0.085

31. Thank you for your attention

32. Results #4: only sibilants F7