Question on FSL FEAT model design


My group is attempting to analyze the data we have but is unsure of how to proceed using the FSL-FEAT analysis tool.

Our data consists of subjects who each have 4 recordings of them being exposed to 3 continuous different stimuli plus a rest recording. The data has been preprocessed completely.

We are unsure how we are supposed to set up the model, as there is no “event” that would serve as an explanatory variable.

Any help is appreciated; thank you!


Hi Jeonghyun,

FEAT can be used to model more than just event-based designs. You can think of your stimuli as events which have a non-instantaneous duration. Take a look at the EVs page in the FEAT user guide: FEAT/UserGuide - FslWiki

There are also several examples on setting up an analysis with FEAT in the FSL course practicals:


Thank you for the help! I believe we have a sense of how to set up our data thanks to you. Our three stimuli are in separate recordings, would I need to concatenate them into one in order to use FEAT to analyze them?

HI @jhan,

Concatenating data sets is not ideal as it introduces discontinuities into the time series, but may be necessary as a last resort. There is no guarantee that the baseline intensity, nor the amplitudes of the various signal/noise sources within your data is consistent across recordings, so you can’t simply concatenate the raw data together.

It is more common to perform a task-based fMRI analysis from a single recording which contains both rest and task conditions. Are there any rest periods in your stimulus recordings?


Our recordings have no rest periods, as we have a separate recording with the subject in a rest-state, which is why I believed that concatenation would be our only option. Is it possible to have 1 EV that is measured for the entire recording?

Hmm, I’m really not sure of what to suggest - perhaps somebody else here may be able to help you. The problem is that the absolute intensity of a fMRI signal is meaningless - it cannot be meaningfully compared across different scanning sessions; at least I am not familiar with any methods for doing so.

What is meaningful is how the intensity of the signal changes, or varies, throughout the duration of a scanning session. This requires that a single scanning session contains both rest periods, and periods during which your stimuli are being applied, which would allow you to search for changes in intensity/variance which are correlated with your stimuli.

Hi @jhan,

It sounds like naturalistic task analysis paradigms might be more suitable for you, e.g., inter-subject correlation or GLMSingle.


Hi all,

Thank you for many helpful feedback. We will try exploring other paradigms such as ones that Steven mention. I think we will try to move away from FSL for now, as our recordings do not have any contrasting conditions within one recording.

Thank you very much.