Averaging data
The acquisition software can average your data in real time and show the results as the experiment is running. The online averaged data will also be saved in a separate file. This option isn't as useful as you would think, as the HPI coils interfere with the results and makes them almost unusable. Therefore it is much better to average you data offline, using the saved raw data. However, to do so using the Elekta software requires you to have specified all necessary details at the time of the recording. Changing or adding averaging details is difficult. For that reason it is highly recommended to fully specify all averaging information before running any experiment, and checking that everything is working correctly by using the online averager, even if its output is irrelevant.
HPI measurements
Before any recording can be made the locations of the HPI coils have to be measured by the MEG machine and compared to the locations measured by the 3D digitiser. This process requires the machine to be well tuned, as any noise at this point will result in unreliable locations. Therefore, this can be seen as a good test for the whole setup.
When the acquisition software doesn't recommend to accept the HPI locations, it is usually best to have a look at the raw sensor signal and see if any of the sensors is behaving badly. Any problem with the HPI has to be solved before you can proceed with the recording, as the data will be virtually useless with out decent HPI.
Using continuous HPI
If using continuous head position indication (cHPI), acquire a minimum of 20 seconds of data before turning on cHPI (ie before first stimulus). Otherwise MaxFilter has problems detecting bad channels (info from Matti Kajola).
Generic Recommendations to Experimenters
There are several goals in the MEG Data Collection.
Get the participant to perform the task.
Get the participant comfortable. An uncomfortable participant creates lots of motion and muscle artifacts.
Train the participant to avoid artifacts (e.g. eye muscle artifacts)
These last two points are extremely important. They have a great influence on the quality of your data. The person can be performing the task very well and the data can be garbage. Also, stressing that you want them to be comfortable will help them relax. It is this type of encouragement that is needed to build a good relationship with your participant. If they are uncomfortable during any part of the experiment, it will affect whether they return to complete the study and/or are willing to participate in future experiments.
It's kind of difficult to write about this, but you need to be very nice and casual, while maintaining professionalism. Be sure to speak clearly and calmly. Stand with your feet firmly planted on the ground. Look the participant in the eye. Don't fidget with your hands. These things will make the participant feel that you are comfortable and in control. A participant wants to feel safe.