if the scores are normally distributed then one can use mean + - a given SD to create two groups. The median split is also a most frequently use procedure. But it seems that you are interested in only using mild and moderate anxiety groups and not the severe one. Thus, cut-off based on SD or percentile may be better.

Further, the choice also depends on the N, if N is large dichotomization on the basis of SD may be advised for small N median or quartile split may be a better choice.

As far as the published cut-off score is concerned they are always sample specific and influenced by a host of contextual factors. Thus, such cut-off should be preferred only when they have been behaviorally validated across different samples and sociocultural contexts.

You could also search for articles that examined the anxiety levels of people with an anxiety disorder in comparison with a healthy control group by using the STAI and then compare your own results of participants on the STAI with the result from these articles.

The STAI-T is not a clinical instrument as it does not assess impairment and thus no clinical cutoff levels exist.

I am also interested in finding out the optimal cut off score for the STAI-State. Has anyone perhaps come up with an answer?

I know there is such a cut off score for some other similar self-report instruments, such as the BDI-II.

Thanks in advance.

I do not recommend using the STAI, Form Y, in any way other than instructed, because that is how it was studied and shown to be valid. I especially recommend against reducing the state and trait scores down to a yes/no result unless you run a pre-study to validate the reduced calculation under reasonable assumptions of validity. Failure to do this renders your results scientifically invalid, in my opinion.

I am looking at an article from the NIH which suggests 39-40 to " detect clinically significant symtoms for the scale" Other studies have suggested 54-55 for older adults.