Even though Broca's area has been associated with speech and language processing since the 19th century, the exact role that it plays is still a matter of debate. Recent models on the neuroanatomical substrates of language have assigned Broca's area to different processes: syllabification (Indefrey and Levelt 2004), articulatory code storage (Hickok and Poeppel 2004) and verbal working memory (Chein and Fiez 2001; Chein et al. 2002). The subject of this doctoral dissertation, is to examine language production and disambiguate the role of Broca's area. This issue was addressed in a series of functional magnetic resonance imaging studies (fMRI) involving speech production, where the phonological properties of pseudowords were manipulated in a way that differentiated between syllabification and articulatory code generation. The load on verbal working memory was also changed. The behaviour of Broca's area was then examined in response to these manipulations to determine the dependence of the observed results on the different levels of processing and verbal working memory. The results from the present studies suggest that the dorsal premotor cortex has a consistent role in articulatory code generation irrespective of verbal working memory demands. In contrast, Broca's area, specifically Brodmann area 44, showed a main effect of phonetic encoding only during delayed response tasks. Interestingly, area BA44 was also found to be functionally segregated between the dorsal and ventral part. The dorsal part was sensitive to articulatory and phonological load, such as stimulus length. The ventral part on the other hand was sensitive to sub-lexical stimulus properties, but only during delayed response trials. These findings suggest that BA44 is not a homogeneous region, but it is divided into a dorsal premotor and a ventral prefrontal part. These results add another dimension of complexity to the study of Broca's area, its functional segregation and its role in language production.