Dear Susanne,
|Is it true that when using short SOAs I have to choose 'Interaction
|among trials (Volterra )' - YES, because otherwise SPM does not acount
|for short presentation rates at all ?
No, you don't have to choose this option routinely. The Volterra option
allows modelling of the influence that evoked activity at some previous
time has on activity at a later time - in other words, a (non-linear)
interaction between successive trials in this context. At very short SOAs
(~ <1s) this interaction becomes important; the evoked response to a
current trial is signficantly reduced by the presence of an immediately
preceding (<1s) trial (see Friston 1998 MRM 39;41-). But at slightly
longer SOAs, as Karl was suggesting, evoked responses to succesive trials
add linearly and you need not model interactions explicitly with the
Volterra kernels.
|How short can the SOA reasonably be ?
Karl may have some comments on this, but our empirical data suggested that
non-linear interactions between successive trials became important at SOAs
less than 1-1.5 seconds. So for everyday use, an SOA above this range
should work fine without the need for modelling Volterra interactions
explicitly.
best,
Geraint
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|