You can compare results between blocks with no problem. This is very common practice. SPM adds a regressor for each session (at the design matrix, there is a column with ‘1’ for each scan in a block, and a zero for other blocks), but in effect SPM concatenates sessions. In any multi-scan design, it is typical to combine contrasts from several blocks together.
Let’s assume you want to examine negative > neutral for sessions 1 and 2 (a session being an MRI scan or run).
Well in session 1 you calculate a Beta for neutral and a Beta for negative. Same in session 1. Then you contrast 1 -1 1 -1 for negative > neutral in session 1 and 2.
You are asking the question “is the average Beta value for neutral across sessions greater than the average Beta value for negative across sessions”. If that is a sensible question given the perspective change across blocks is up to you.
Or you can contrast negative trials across sessions (0 1 0 -1). Or the interaction (1 -1 -1 1). The betas are roughly equally weighted across the sessions.
Best of luck,
Colin Hawco, PhD
Neuranalysis Consulting
Neuroimaging analysis and consultation
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On Behalf Of Maxime Résibois
Sent: February-27-17 9:00 AM
To: [log in to unmask]
Subject: Re: [SPM] Mixing blocks and conditions
Dear SPM experts,
Having received no response I report my question (see below).
I’d be happy to be redirected to any literature specifically addressing what can or cannot be analysed/concluded from fMRI designs separating condition manipulation in blocks (with breaks in between) in more details than “it is preferable not to do it”.
Kind Regards,
Maxime
From: Maxime Résibois
Sent: Wednesday 22 February 2017 15:09
To: 'SPM' <[log in to unmask]>
Subject: Mixing blocks and conditions
Dear SPM experts,
In my design, I had two conditions (adoption of a field/observer perspective) that have been mixed with experimental blocks (i.e. participants stayed in the MRI but we had a short break in the middle of the experiment).
Thus, in one block they adopted one perspective, and in the other they adopted the other perspective (order was counterbalanced).
I am interested in how the condition influences the correlates associated with a continuous regressor, and wonder what I can and what I cannot analyse given the design I used to collect my data.
- I can look at the brain correlates of this regressor separately in each block (regressor)
- I can compare within each block negative and neutral trials separately for each condition (contrasts: negative > neutral or neutral > negative)
However, can I
- Compare together the regressors across the two blocks? And why?
- Compare together the results of the contrasts across the two blocks? And why?
Kind Regards,
--
Maxime RésiboisPhD Student (KU Leuven)
Quantitative Psychology and Individual Differences
Tiensestraat 102 bus 3713
3000 LEUVEN
tel. +32 16 37 30 98