Parsing the temporal evolution of brain activity in duration discrimination with electroelectro- and magnetomagneto-encephalography Karim N'DIAYE, Micha PFEUTY, Richard RAGOT, Line GARNERO, Viviane POUTHAS Laboratoire de Neurosciences Cognitive & Imagerie Cérébrale, CNRS UPR640-LENA, Paris, France

Introduction Are the cerebral processes involved in time perception specific to the temporal dimension? Or, do they share common neural substrates with other perceptual judgments? If so, do these processes nevertheless differentiate through taskdependent dynamics? Previous functional imaging works have contrasted sub-second timing with other perceptual judgment, in particular pitch [Rao01, Voltz01, Nenadi03, Reiterer05]. While event-related fMRI reach sub-second temporal precision, it is still below the capabilities techniques recording the electromagnetic counterpart of brain activity. Recently, EEG has also been used [Gibbons03] in a task involving time and frequency judgment but analyses were restricted to scalp data. In the present study, we used MEG-EEG co-recordings to investigate the dynamics and the topography of brain activity in a task where duration or frequency information was selectively attended.

Experimental setting

A

Participants Fifteen young healthy adults (5 women, mean age: 24.8 yrs, range: 21-30). All had normal hearing and no reported neurological disorder.

Duration discrimination vs. frequency discrimination At the beginning of each of the six runs, participants were instructed to perform one of the two tasks, i.e. selectively attend one dimension while ignoring the irrelevant one. Each run comprised 64 trials. On each trial, two time intervals were presented separated by a varying delay (ISI=1.0-1.8 sec). One interval was always 600ms (base duration), the other was randomly chosen among: 600, 700, 800 or 900ms. Intervals were marked by 20ms-long pure tones (1000Hz, 3ms on and off ramped, i.e. base pitch). For one of the two intervals, the second tone had a frequency of either 1000, 1020, 1040 or 1060Hz. Duration and pitch factors were totally crossed in a within-subject design. Task sequence was pseudo-randomized across participants.

B

R

At the onset of a visual cue, participants had to designate with a left or right button-press which interval was the longer (resp. ended with a highpitched tone). Delayed motor selection and response was enforced by randomizing response mapping on each trial. Inter-trial time was 2.5-3.5s.

Data recording MEG and EEG were simultaneously recorded at a 625Hz sampling rate (DC-100Hz filter) through a CTF 151-Omega system (151-gradiometers with whole head coverage). EEG potentials were measured from 64 unipolar Ag/AgCl scalp electrodes which positions were digitized using a Polhemus Fast Track system. Vertical and horizontal EOG were also recorded for offline correction of eye blinks. For technical reason, one MEG sensor was disabled at the recording time.

Data processing

Distributed source reconstruction

MEG and EEG data were inspected for rejection of artifactcontaminated trials. Eye-blink related signal was removed from individual channels by subtracting the scaled first component issued by a Principal Component Analysis of each data segment containing a blink.

For the purpose of distributed source analysis, individual anatomical MRI were obtained and processed with BrainVISA software [BrainVISA04] to extract the triangulated meshes of the cortical mantle (grey matter/white matter interface). Co-registration of MEEG and anatomy was assured by using head positioning coils during the MEG session and MR-visible beads at the same locations inside the MR scanner.

Data were then filtered (20Hz zero-phase low-pass filter) and epoched. Trials were averaged separately for each task. Distinct averages were performed for each stimulus category. To increase signal to noise ratio, the first 600 ms of encoding and comparison intervals were averaged for all trials (as they were physically identical up to that point in time). To correct for uneven positioning of participants’ heads in relation to the MEG sensor array, we interpolated individual MEG measurement onto a template sensor array using a high rank estimate.

Only MEG data were used for the distributed source analysis. The BrainStorm Toolbox [BrainStorm04] was employed. L2 minimum-norm estimates were computed for each task and condition with a spherical head forward model. To allow inter-subject averaging, a single anatomy was used.

MEG global field Behavioral results

Comparison

Percentage correct

Participants performed equally well in both tasks. There may be however a plateau effect, as no significant difference was observed between tasks, nor across difficulty levels.

Encoding

Deviation level

Global field power difference were small between the task but larger between conditions (p