Economy of effort, a popular notion in contemporary speech research, predicts that dynamic extremes such as the maximum speed of articulatory movement are avoided as much as possible and that approaching the dynamic extremes is necessary only when there is a need to enhance linguistic contrast, as in the case of stress or clear speech. Empirical data, however, do not always support these predictions. In the present study, we considered an alternative principle: maximum rate of information, which assumes that speech dynamics are ultimately driven by the pressure to transmit information as quickly and accurately as possible. For empirical data, we asked speakers of American English to produce repetitive syllable sequences such as wawawawawa as fast as possible by imitating recordings of the same sequences that had been artificially accelerated and to produce meaningful sentences containing the same syllables at normal and fast speaking rates. Analysis of formant trajectories shows that dynamic extremes in meaningful speech sometimes even exceeded those in the nonsense syllable sequences but that this happened more often in unstressed syllables than in stressed syllables. We then used a target approximation model based on a mass-spring system of varying orders to simulate the formant kinematics. The results show that the kind of formant kinematics found in the present study and in previous studies can only be generated by a dynamical system operating with maximal muscular force under strong time pressure and that the dynamics of this operation may hold the solution to the long-standing enigma of greater stiffness in unstressed than in stressed syllables. We conclude, therefore, that maximum rate of information can coherently explain both current and previous empirical data and could therefore be a fundamental principle of motor control in speech production.