Speakers of intonation languages use bundles of cues to express prosodic prominence. This work contributes further evidence for the multi-dimensionality of prosodic prominence in German reporting articulatory (3D EMA) and acoustic recordings from 27 speakers. In particular, we show that speakers use specific categorical and continuous modifications of the laryngeal system (tonal onglide) as well as continuous modifications of the supra-laryngeal system (lip aperture and tongue body position) to mark focus structure prosodically. These modifications are found between unaccented and accented but also within the group of accented words, revealing that speakers use prosodic modulations to directly encode prominence. On the basis of these findings we develop a dynamical model of prosodic patterns that is able to capture the manipulations as the modulation of an attractor landscape that is shaped by the different prosodic dimensions involved.