When using waver with the -tstim option and giving the event times from a file, but not specifying the a:b syntax for event duration, does waver assume that the event duration is actually 1 TR as specified in -dt?

Not exactly. As the output of waver -help tries to explain, if an input to -tstim is a pure number, then the result is a impulse response to an infinitely rapid event at that time. If the input is of the form a:b, then the result is the sum of a bunch of impulse responses spread out over the time from a to b. These will not quite be the same.

You can see this by doing the following 2 commands:

waver -dt 1.0 -tstim 5:6 -GAM -numout 20 | 1dplot -stdin &
waver -dt 1.0 -tstim 5 -GAM -numout 20 | 1dplot -stdin &

The graphs are similar but not identical. In particular, you'll see that the first one (with the stimulus spread out over 1 second) is slightly delayed from the second one. This is as it should be.

For practical purposes, though, you'd never be able to tell these two responses apart with FMRI data. So I recommend that you do a pair of plots like the above for the cases that you care about, and you'll probably find that you won't care about the tiny differences.

bob cox