The aim of the present thesis is to outline, how psychology in general and occupational psychology in particular can benefit from automatic biosignal analysis. Progress in this field within the last decade bears chances to widen the horizon of commonly employed statistics and use an interdisciplinary approach to gain new insights of typical occupational issues with the help of human state computing. Advances within psychological methodology are necessary, because recent approaches are neither sufficient to measure relevant human states in a suitable way nor show feasibility regarding every day usage. Hence, a way of obtaining more convenient assessments is presented. This thesis demonstrates how the assessment of leadership relevant states, fatigue and stress based on voice, mouse and head movement is facilitated with natural data gathered while acting in typical tasks. Advantageous is, that measurements take place during work automatically with the help of non-intrusive devices resulting in a minimally resource consuming way of testing. Obtaining proper features of recordings is assigned a key factor for state computing. Hence, the theoretical focus is set on the derivation of suitable features. The novelty of the presented approach lies within its generalizability. It is demonstrated, how features from several sources can be transferred to both different human states and biosignals. For that purpose, nonlinear dynamics as well as Wavelet based and signal-specific features have been extracted in addition to commonly examined temporal and spectral features and functionals in order to generate optimized prediction models. Results prove a wide ranged feasibility of the approach starting to close the gap resulting from drawbacks of current methods. Nonetheless, it is still only the beginning of emancipating promising statistical methods in today’s psychology with possibilities reaching far beyond occupational matters. Future work mainly consists of improving data corpora, prediction algorithms and adapting features to several signals as well as states for optimal results. Moreover, long-time validations within companies were helpful to further prove the added value of human state computing to recent approaches.