Since the description of cortical deafness, it has been known that the supe
rior temporal cortex is bilaterally involved in the initial stages of langu
age auditory perception but the precise anatomical limits and the function
of this area remain debated. Here we reviewed more than 40 recent papers of
positron emission tomography and functional magnetic resonance imaging rel
ated to language auditory perception, and we performed a meta-analysis of t
he localization of the peaks of activation in the Talairach's space. We fou
nd 8 studies reporting word versus non-word listening contrasts with 54 act
ivation peaks in the temporal lobes. These peaks clustered in a bilateral a
nd well-limited area of the temporal superior cortex, which is here operati
onally defined as the speech sensitive auditory cortex. This area is more t
han 4cm long, located in the superior temporal gyrus and the superior tempo
ral sulcus, both anterior and posterior to Heschl's gyrus. It do not includ
e the primary auditory cortex nor the ascending part of the planum temporal
e. The speech sensitive auditory cortex is not activated by pure tones, env
ironmental sounds, or attention directed toward elementary components of a
sound such as intensity, pitch, or duration, and thus has some specificity
for speech signals. The specificity is not perfect, since we found a number
of non-speech auditory stimuli activating the speech sensitive auditory co
rtex. Yet the latter studies always involve auditory perception mechanisms
which are also relevant to speech perception either at the level of primiti
ve auditory scene analysis processes, or at the level of specific schema-ba
sed recognition processes. The dorsal part of the speech sensitive auditory
cortex may be involved in primitive scene analysis processes, whereas dist
ributed activation of this area may contribute to the emergence of a broad
class of "voice" schemas and of more specific "speech schemas/phonetic modu
les" related to different languages. In addition, this area is activated by
language-related lip movement suggesting that a multimodal integration of
the auditory and the visual information relevant in speech perception occur
s at this level. Finally, there is a task-related top-down modulation of th
e pattern of activation of the speech sensitive auditory cortex which may r
eflect the fact that the different parts of this structure are connected to
different down-stream cortical regions involved in the neural processing o
f different types of tasks.