Stem-initial accent has been identified in a number of NW Bantu languages (Hyman 1989). In Eton (A70; Van de Velde 2008), stem-initial accent is manifested phonotactically (half of the consonant phonemes restricted to the stem-initial position C1), phonetically (C1 consonants are longer than consonants in other positions), phonologically (C1 not subject to lenition rules that occur elsewhere), as well as tonologically (only accented syllables can host two underlying tones). Consonant lengthening as a significant stress correlate is typologically rare and even rarer as the primary phonetic parameter in the realization of stress. In order to evaluate the hypothesis that the stem-initial accent in NW Bantu is realised only or mainly through consonant length, we recorded questionnaires for seven NW Bantu languages (Bapuku, Basaa, Eton, Fang, Kota, Kwasio, Bagyeli), with a particular focus on different regional varieties of A70 languages. The acoustic analysis of the data confirms the role of consonant lengthening as a major phonetic correlate of stem-initial accent, as illustrated in Fig. 1 for Eton, where the second [m] is stem-initial and is almost twice as long as the other two [m]. The only language in our sample that lacks stem-initial accent is Bapuku, which has penultimate vowel length and which lacks the typical NW Bantu morphosyntactic characteristics. Interestingly, within A70, relative length of C1 appears to decline as one moves further south. Furthermore, the analysis of the data suggests that the exaggeration of the duration of consonants rather than vowels in a certain position in a word, such as the stem-initial position C1, is in origin an utterance-level prosodic/intonational phenomenon marking a particular emphasis on a given element within the utterance, which we refer to as C-emphasis prosody.