In response with a couple of kind replies I've gotten (thank you Jim, Gregg and others whose replies must be over the mail right now...), I refine my question: Is there an algortithm that gives the most reliable X-Y items scale out of an X items scale? The common SPSS output gives the Alpha Cronbach for the deduction of any one item but in social science relations between variables can be complex while the abundance of combinatoric possibilites to compose an X-Y items scale out of an X items scale when the original scale is long is too scary to even begin to check each one seperately. Any idea?