I RECOMMEND THAT YOU READ TABACHNICK AND FIDELL (2001); IT REMAIS ONE OF THE BEST RESOURCES THAT ANSWER ALL YOUR STATISTICAL INQUIRIES IN DETAIL

Your firs step would be to convert all your pitch measurements in hertz to semitones. The semitone is a relative measure: it captures the size of a musical interval. The best-known musical interval is the octave, which a doubling of the frequency (in Hz), i.e. a ratio of 2:1. One octave comprises 12 semitones, which is a ratio of approximatelely 1.06:1.00. It is customary to express the pitch in speech relative to a baseline of 50 Hz (which is the lowest pitch the human - male - voice can produce. The formula to convert an artcitrary frequence to semitones above 50 Hz would then be:

12 * ln (x / 50) / ln2

Here 50 is the baseline (or reference) pitch and ln is the natural logarithm. Any other logarithm would do as well, as log as the same type of logarithm is used in both instances in the formula. Note that ln(x)/ln(2) is an alternative for the logarithm on the base 2 (the 2-log is not a function that is available in most calculators and programming languages, hence the alternative used here).

Note that you may also subdivide the octave into smaller units, e.g. cents (i.e. one 100th part of an octave, as is often used in ethnomusicology) - in which case you replace "12" by "100". If you want to express the semitones relative to another baseline, then replace the value "50" by the alternative baseline frequency.

Once you have converted all your pitch measurements to semitones, your next step is to compute the arithmetic mean of all the semitone measurements. The mean is a necessary in-between step towards computing the standard deviation. Computation of mean and standard deviation can be done by a large number of programs, among which Excel and SPSS but you may also do it by hand. For the mean just add up all your semitone values and divide by the number of measurements. For the standard deviation subtract the mean (as computed in the previous step) from each individual semitone value, and square the difference. Then add up all the squared differences, divide the sum by the number of semitone values minus 1 and take the square root of the result.

It is important that you do the semitone conversion before you compute the mean and the standard deviation.

A practical alternative to all of the above is to ask Praat (www.praat.org) to compute the pitch curve for you, and then query the resulting pitch object, asking praat to output the mean and the standard deviation of the pitch.

Have fun.

Just a quick correction here - cents are one hundredth of a semitone, not of an octave. So there are 1200 cents in an octave.