Speech has been used as a biomarker for the binary classification of multiple diseases, with promising results. However these speech affecting diseases often co-exist in the same individual and produce similar manifestations in the speech signal. Thus we propose to characterize normative speech using reference intervals for interpretable speech features (acoustic and linguistic), as a first step towards the adoption of speech analysis for multidisease screening in health applications. We discuss the impact of demographics and speech tasks. Finally, we compare the reference intervals with subjects suffering from Parkinson's disease, Alzheimer's disease and depression.