The divide-and-conquer approach to the study of human cognition has succeeded in focusing researchers' efforts on behavioral phenomena that fall under well-defined categories such as attention, perception, language, memory, emotion, and motor control. The result has been the development of coherent bodies of work in each area replete with successful explanatory theories and a rich collection of paradigms, tasks, and analytic techniques. There has been a renewed in recent years in combining and integrating ideas across these domains, as well as in incorporating neuroscientific data, as a way to build more powerful and general models of cognition. Here we review the history and current state of integration between verbal short-term memory (VeSTM) and language, two domains of study that have significant areas of overlap but have not been fully integrated. We review evidence from cognitive neuroscience that has generally shown VeSTM to greatly depend on the network of brain regions that are known to form the core sensory-motor basis of human language. Whereas classic psychological models of VeSTM posit the existence of dedicated short-term storage buffers, we suggest that temporary verbal memory emerges from the coordinated interplay of a fronto-temporal sensory-motor circuit that has evolved to support the perception and production of speech. Phonological rehearsal in the service of temporary maintenance is achieved by feedforward and feedback pathways connecting the auditory- and motor-speech systems via a sensorymotor interface component situated in the Sylvian-parietal-temporal region (Spt). Reciprocal connectivity between the frontal and temporal speech systems enables the kind of "round-tripping" of dual speech codes long hypothesized by cognitive models such as Baddeley and Hitch's "phonological loop".