Conditional branching/contingencies

[Don.Rojas at UCHSC.edu]

I am currently planning an experiment where simple same/different pitch
discriminations are made.  I need the experiment to become harder or easier
depending on subject performance.  For example, a subject will start by
making easy discriminations between a 500 Hz tone (reference) and another
tone (comparison) that is 50 percent different in frequency.  After 30
trials (half of which have reference/comparison pairs of the same
frequency), the accuracy of that block of trials is used to determine
whether to increase or decrease the difficulty of the comparison for the
next block of 30 trials.  The end point would be a block of trials that
subjects can discriminate the smallest percent difference between
frequencies at 85% accuracy.  I suppose that I can use the summation object
to keep running track of accuracy to make a decision after each block of
trials.  I'm having difficulty imagining how to organize the tones in terms
of eprime objects to best do the conditional branching after each trial.
Anyone got any ideas?

Don Rojas
U. of Colorado Health Sciences Center