Although ATR harmony has been studied in a number of languages, very little acoustic phonetic work has been conducted on ATR harmony systems in Nilo-Saharan languages (Guion et al. 2004). What is particularly intriguing about the Kube dialect of the Nilo-Saharan language Zaghawa is the asymmetry between underlying contrast and surface realization in its vowel inventory. The [+ATR] mid vowels [e], [o], are always found in the context of a [+ATR] high vowel [i], [u], and never alone. While 9 vowels are phonetically present (i.e., [i], [ɪ], [e], [ɛ], [u], [ʊ], [o], [ɔ] and [a]), only 7, excluding [e], [o], are underlyingly contrastive, which is a less common inventory (Anonby 2007, Casali 2008). This paper reports on an acoustic study of [ATR] vowels in the Kube dialect. The results indicate that while the acoustic difference between [+/-ATR] harmonic pairs is in line with previous findings of the first formant (F1) being the most robust cue, the [+ATR] allophonic mid vowels [e], [o] and [-ATR] high vowels [ɪ], [ʊ] have significant F1 differences as well, which has not always been the case in previous instrumental studies of ATR harmony languages.