Jan,

Reviewing the messages Satoru referenced and the code (spm_getSPM.m and spm_list's calls to spm_P) there's no reason that the extent threshold question should be omitted in the case of conjunctions.  In particular, in the case of the inference on the 'Conjunction Null' the theoretical result *is* available, and is just the usual result for a single image.  (To see this, realize that inference under the conjunction null takes place with worst case scenerio possible, where one image is totally null, and all other images conjoined are completely and wildly significant, e.g. T>100 everywhere; in this worst case the null cluster sizes in the conjoined image will be exactly those in the single null statistic image.)

Further, the exent threshold is of course just an arbitrary cut off, and even if applied, doesn't change cluster P-values.  Lastly, when doing a global or partial null conjunction test, the cluster P-values are not available and will simply result in NaN's in the tabular output.

So feel free to find the lines in spm_getSPM.m

%-Extent threshold (disallowed for conjunctions)
%-----------------------------------------------------------------------
if ~isempty(XYZ) & nc == 1

if ~isempty(XYZ)