Ao is a Tibeto-Burman language spoken in Nagaland, India. It is an under-studied language and hence, there are not many resources available for the language. Ao is a tone language and it is known to have three lexical tones, namely, high, mid and low. Three distinct dialects, namely, Chungli, Mongsen and Changki are reported for the language and depending on the dialect, in a polysyllabic word, tones may be arranged differently for even the same words. In this work, an acoustic study is conducted on the three tones of the two dialects of Ao, namely, Changki and Mongsen. Tone distributions in trisyllabic words are considered, expecting dialect specific tone assignment patterns. The tonal distributions unique to each dialect are identified through native speakers' judgment and later through acoustic measurements for a subset of words. Based on the analysis, the feasibility of performing automatic dialect discrimination is explored. Initially, a semi-automatic method, using Praat derived values, is implemented for dialect discrimination. An automatic method is then proposed using zero-frequency filtering (ZFF) approach for estimating pitch values. The results of the study demonstrate the potential of tone features based dialect discrimination. © 2017 IEEE.