This paper presents a study on the stability and repeatability of Hidden Markov Modeling based probability outputs across several different dynamic handwritten signature features. This study compares such values extracted from signature samples compiled within a single data collection session (i.e. between intra-session) and across several data collection sessions (i.e. between inter-sessions). The primary aim of this study is to investigate the indentifying capability of local online signature Hidden Markov Modeling based probability outputs which have implications on the accuracy of biometrics signature verification system which utilize similar HMM approach. This paper reports on an analysis results carried out on the online genuine signature counterparts of Sigma database - a compilation of over 6000 genuine signature samples that were gathered over a series of data collection sessions.