Interpolation bug July 2011

we have detected a bug in EEGLAB that pertains to reading interpolated data when processing multiple subjects in EEGLAB "STUDY". Channel order and location proved to be inaccurate.

This bug applies to you if:

• you used EEGLAB STUDY to process multiple subjects
• and you removed some channels from some datasets
• and you interpolated channels when computing ERP, spectrum, ERSP and ITC

This bug does not applies to you if:

• you did not use STUDY within EEGLAB
• or you used STUDY to process ICA components and ICA clusters.
• you did not reject any channels in any subject (even if you rejected a single channel in a single subject and interpolated it, the data channels from ALL subject might be corrupted when the data is being read)
• you did not interpolate missing channels
• you are using any version of EEGLAB 8. The bug only appeared in EEGLAB 9.

Consequences of the bug:

• Correspondence between channels was lost between subjects and channel location were also wrongly assigned
• Effect that might have been consistent accros subjects were not visible
• Average scalp topographies might have looked scrambled
• The good news is that by replotting your data (there is no need to recompute anything), effects that were previously not visible because of the bug may now appear

What you need to do:

• download the latest version of EEGLAB 10 or EEGLAB 9 at http://sccn.ucsd.edu/eeglab
• you do not need to recompute any measure since the bug pertain to reading interpolated data

We are apologizing for the inconvenience. We are trying to make EEGLAB as bug free as possible and this relies on both our unit testing framework (http://sccn.ucsd.edu/wiki/) and your feedback and bug submission at http://sccn.ucsd.edu/eeglab/bugzilla. This specific bug was reported independently by two users: Agatha Lenartowicz and Claire Braboszcz.

A. Delorme - July 25th, 2011