SpmContrasts - MRC CBU Imaging Wiki
location: SpmContrasts

A SIMPLE EXPLANATION OF CONTRASTS IN SPM

A contrast is can be used to compare different conditions in your study. To use contrasts intelligently, it is very important to understand something of how they work; you may want to start with our SPM statistics tutorial for this.

When you use contrasts with conditions, the conditions that you are interested in take on a positive value, such as 1, and conditions that you want to subtract from these conditions of interest take on a negative value, such as -1. One rule for making a contrast that is applied to conditions is that it always has to equal zero when you add together all the aspects of the contrast applying to the conditions (see Andrew Holmes' statistics chapter in the SPM course notes for an explanation of why this must be so). In SPM96, you will have to input the same number of values as you have conditions in your study (assuming you have no covariates of interest - if you do then - in SPM96 - you will have to add extra numbers in the contrast for these). See the SPM statistics tutorial for more on how contrasts work, and how to use covariates in an analysis.

You can think of your contrasts this way: in PET, for each of the conditions, for each voxel, SPM calculates the mean signal level for that condition. The contrasts tell SPM how to subtract these means, to look at the particular comparison/s you are interested in. For example a contrast of -1 1 0 0 in an analysis with four conditions equates to the mean PET activity, voxel by voxel, of condition 1 subtracted from the mean PET activity of condition 2 (i.e. Condition 2 minus Condition 1).

Here is an example of a set of contrasts using a standard four conditions design:

  • Condition 1: Language task

  • Condition 2: Memory task

  • Condition 3: Motor task

  • Condition 4: Control

Contrast 1: Lang minus Con: 1 0 0 -1 (this contrast is attempting to see what significantly increased areas of regional cerebral blood flow - rCBF - are found with the language task, when the rCBF of the control task is subtracted from it).

N.B. There needs to be a space, or a comma, between each number when you enter it into SPM.

Contrast 2: Motor minus Mem: 0 -1 1 0 (this contrast will show you areas of the brain that have significantly increased activity in the motor conditions, when the memory conditions are subtracted from it. The result will show you brain areas activated more in motor processes than in memory.

Contrast 3: Con minus Motor: 0 0 -1 1 (this contrast will measure areas of the brain that have significantly increased activity in the control task, compared with the motor task - or significantly decreased activity for the motor task compared with the control task).

You can also combine conditions in your contrast, to discover what is common to a set of conditions. For instance:

Contrast 4: (Lang+Mem) minus Con: 1 1 0 -2 (this contrast will measure areas of the brain that have significantly increased activity in the average of the language and memory conditions, compared with the control task. Note that in order for the total to equal zero, the control condition had to be the negative equivalent of the sum of the other conditions. Another way of looking at this contrast is as the sum of the individual condition contrasts of 1 0 0 -1 and 0 1 0 -1.)

Daniel Bor 20/12/99

CbuImaging: SpmContrasts (last edited 2013-03-07 21:22:58 by localhost)