Hi David,
You can have AFNI spit out a time series. First, you have to identify the "mask." Is this a cluster in a certain dataset? As an example I'll make some assumptions that may or may not be true of your data. Let's say the cluster is a patch of activation from a GLT or maybe an ANOVA, in the 10th sub-brik of dataset
"anova_output+orig".
Typically one would use the 3dmerge command with the -1clust_order option on the specific sub-brik containing the cluster you want as the mask. Let's say you've decided that your cluster must have a min radius of 3mm and min size of 10 voxels (270microL).
3dmerge -prefix anova_cluster_order -1clust_order 3 270 'anova_output+orig[10]'
This command will generate a new dataset named "anova_cluster_order+orig" and replace the activation values in each cluster with a number that indicates the relative size of a cluster in the specific sub-brik, 1 being the largest cluster, 2 the next largest, etc. Then by viewing the output dataset in AFNI you can determine what "number" your cluster is. This cluster, let's say it's cluster #4, becomes the mask for use with the command 3dmaskave.
The 3dmaskave command then uses the cluster ordered dataset as a mask and by using the -mindex flag is further restricted to only those voxels with a value of 4 (your amygdala cluster). The command would look something like this...
3dmaskave -quiet -mask anova_cluster_order+orig -mindex 4 4 \
timeseries_data+orig > timeseries_text_file
timeseries_text_file will then contain one entry for each time point in the timeseries data with the average of all the voxels in the cluster #4 of the mask dataset.
Good luck,
-jim