The rhythmic hypothesis posits that languages from different rhythm classes can be discriminated at birth, whereas distinctions between languages or dialects within the same rhythm class emerge only later in development. While highly influential, this framework presents a simplified view of early language discrimination: it assumes a tripartite division of languages that ignores the continuous nature of prosodic variation and, by doing so, fails to account for differential discrimination patterns that arise from the acoustic diversity among languages. In a longitudinal optical brain imaging study, we tested 37 infants exposed to Standard Czech from birth on their ability to discriminate the familiar language variant from the unfamiliar Silesian dialect. The results revealed effects in the right prefrontal cortex (PFC) rather than in auditory cortices, which is in line with a recent view of early language discrimination based on top-down processes. At 4 months, PFC oxygenation was higher when listening to Standard Czech than to the Silesian dialect, suggesting infants’ increased attention and preference for the native prosodic patterns stored in the long-term memory. When listening to the Silesian dialect, the level of PFC oxygenation increased within the next 2 months, reaching the level of the familiar variety, indicating a switch in attention. Whereas some studies report left-lateralized activation in early dialect discrimination, we showed right-lateralized processing, probably due to a predominant reliance on intonational contours. Overall, our findings indicate that early discrimination patterns and their perceptual salience cannot be explained solely by rhythmic classification but rather depend on the specific acoustic characteristics of the contrast.