1. Sahin, Pinker, Cash, Schomer, & Halgren (2009) Sequential processing of lexical, grammatical, and phonological information within Broca’s area

2. Broca’s Area Implicated in several kinds of processes –Language Phonology Morphology Syntax –More generally Sequential ordering? (Near motor face and hand regions)

3. Brodmann Areas Defined histologically Broca’s Area = 44 & 45

4. Depth Electrodes - Used in surgical patients only - Seizures not sufficiently controlled by meds - Surgery to excise tissue causing seizures - Main purpose to help localize seizure focus - Contacts every 5mm at different depths along a single needle - Electrode diameter = 1 mm - Record from adjacent pairs (= bipolar) - Less noisy - Polarity reversals when cross sources - Allows better localization of seizure source - Record Local Field Potentials (LFP)

5. Participants (Patients) PatientSexAgeSeizure Onset Age fMRI done? Completed runs (of 9) AF4114Yes6 BF5118No9 CF385Yes7 Inclusion criteria: - No clinical language or cognitive impairments - Right handed - Seizures thought to have a focal source not in language areas - Electrode placements through language areas From supporting online material www.sciencemag.org/cgi/content/full/326/5951/445/DCI

6. Electrode Placement

7. Electrodes A & B During Surgery

8. Depths 1-6 along one probe Electrode Locations Determined with Structural MRI

9. Experimental Design

10. Evidence of Relatively Normal Processing in Broca’s Area in These Patients Same paradigm & stimuli used with normal participants & these same patients in an fMRI study

11. Relationship of fMRI Loci of Activity & Depth Probe Locations

12. Grand Means Across All Conditions Triphasic response - Notice polarity reversals across channels - Means local generators

13. Regional Specificity of Triphasic LFPs

14. Response to Cue vs Target in Broca’s Area Response to cue not reliably modulated by experimental condition

15. Evidence 1 st Peak Reflects Lexical Processing - Amplitude reliably bigger for rarer words - But not reliably affected by word length - Peak time consistent w/ other estimates of word ID time

16. Evidence 2 nd Peak Reflects Inflectional Processing Amplitude equally large for Null- & Overt-Inflection What else do 2 “inflection” conditions have in common?

17. No Differences for Inflecting Nouns vs Verbs

18. Evidence 3 rd Peak Reflects Phonological Processing Amplitude larger in Overt-Inflection

19. Additional Evidence 3 rd Peak Reflects Phonological Processing Amplitude larger for longer words

20. Consistency of 1 st peak of triphasic response across participants From supporting online material www.sciencemag.org/cgi/content/full/326/5951/445/DCI

21. Consistency of 2 nd peak of triphasic response across paticipants - Notice differences in - Amplitude scales - Electrode channels - Polarity From supporting online material www.sciencemag.org/cgi/content/full/326/5951/445/DCI

22. Consistency of 3 rd peak of triphasic response across participants - Notice different channels than previous figure From supporting online material www.sciencemag.org/cgi/content/full/326/5951/445/DCI

23. Localizing Generators (Equivalent Current Dipole [ECD] Modeling)

24. Conclusions Processing a word for production –(in these hybrid comp/prod tasks) –Consists of sequentially ordered processes –Originating in spatially close but distinct regions in Broca’s area Argue that this provides better support for –Language production models that have discrete non- interactive stages for different components of production e.g. Levelt, Roelofs, & Meyer –Over models that incorporate interaction & feedback e.g., Dell; Goldrick

25. Conclusions Stronger conclusion than warranted by data –Online response from Goldrick, Dell, Kroll, & Rapp “…full discreteness claim is not required by the Sahin et al. data and conflicts with an extensive body of research that has already demonstrated its inadequacy. … To account for the inadequacy of both fully discrete and fully interactive theories, language production models must allow for interaction while incorporating critical constraints on the degree of coupling between processing components. Such architectures respect the sequentiality of processing different types of information (in line with Sahin et al.’s results) without sacrificing the ability to account for interactive effects. Rather than viewing results such as those of Sahin et al.’s through a Boolean lens – supporting either fully interactive or fully discrete models – a more profitable approach would acknowledge the continuum of possibilities between these extremes.”