Andy/Kent -
Just to extend some of Eric's comments to multiple event-types...
The optimal SOA depends on which effect (contrast) is of interest.
If you have two event-types, randomly ordered every SOA, and you are
interested in a differential [1 -1] contrast, you want the SOA to be
as short as possible (but not below 2s, where nonlinearities are likely
prevail, as Eric states). This is similar to a [1 0] contrast in the
stochastic design that Eric describes. If however you are interested
in the common effect (a [1 1] contrast), you want a longer ISI of
approximately 20s (depending on your bandpass filter). A particularly
clever design is the one used by Dale & Buckner, with three event-types,
one of which is a null event (a "fixation" trial; another example of
a stochastic design). With these events randomly ordered, you can be
sensitive to both common and differential effects at short SOAs.
Josephs & Henson (in press) illustrate these results via simulation.
Best wishes
Rik
-------------------------------- 8-{)} ----------------------------------
DR R HENSON INTERNET [log in to unmask]
Wellcome Department of
Cognitive Neurology TEL (work1) +44 171 833 7483
12 Queen Square TEL (work2) +44 171 391 1138
London, WC1N 3BG FAX +44 171 813 1420
URL: http://www.psychol.ucl.ac.uk/rik.henson/index.html
-------------------------------------------------------------------------
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|