This paper presents a quantative analysis of stress in Savosavo (unclassified), an endangered language spoken on Savo Island, (Solomon Islands). Acoustic analyses comprise the measurements of F0, duration, and intensity for each syllable in a dataset carefully selected from elicited speech from one speaker only, aiming to test the effect of increasing morphological complexity on stress realization in a system that displays some variation. Statistically significant variation is found in all correlates between stressed and unstressed syllables, thus fitting with widely attested manifestations of stress cross-linguistically. Findings were further tested with a re-synthesis tool, to confirm our initial hypotheses. Our results demonstrate that the current annotation scheme is a reliable representation of the data, and that the qTA component embedded in PENTAtrainer is effective in modelling F0 contours, even with less controlled data as input. We will argue for the usefulness of instrumental phonetic investigations in describing lesser-known languages, to enhance our understanding of the characterization of the prosodic systems of the world’s languages.