I am running a standard event related task design assessing response to gains > losses. Each participant has 2 runs. L1 files look fine. At L2, some participants show overly bright filtered_func_data with white backgrounds whereas others show overly dark images and some are normal. 

The error is not systematic and in fact the same participant can have normal data for the third contrast in cope 3 but a bright image for the first contrast in cope1. 

The only thing I'm doing at L2 is combining runs and taking a single group average. I do not have any empty EVs.

My analyses are not throwing any errors and I am using the same model set up for everyone (a model I've used successfully in the past). The raw data look fine as do the L1 data. 

Thanks in advance!

Sarah