An exploratory survey of local cattle populations of Gamo Goffa Zone in south-western Ethiopia was conducted between April 2011 and May 2012 to identify and phenotypically characterize cattle populations. Ten focus group discussions with key informants as well as phenotypic data from sample mature animals were used to generate data, including quantitative trait data from 560 animals and qualitative trait data from 867 animals. Findings from focus group discussions revealed that, even if local cattle are found widely distributed throughout the study area, they are not known by any common name or breed type. Results from analysis of variance (ANOVA) on continuous variables showed significant (P < 0.0001) differences between sites. Tukey's multiple mean comparison test showed that each quantitative traits were significant (P < 0.0001) for sites. Chi-square test was also significant (P < 0.0001) for most of the categorical variables per sites. Based on a discriminant analysis, sample populations were classified into their respective sites with overall hitting rate was 63.15 percent for females and 74.89 percent for males. Canonical discriminant (CANDISC) analysis showed quantitative traits and Mahalanobis' distances between sites were significant (P < 0.0001). The stepwise discriminant (STEPDISC) analysis for both populations showed that most variables had significant (P < 0.0001) power in explaining phenotypic variation. These information from focus group discussions and phenotypic variation analyses led to identification of two cattle populations that deemed to be distinct breed types (Gamo highland and lowland). Thus, indigenous cattle population of the study area was not homogenous on their phenotypic features, and further genetic characterization should be done to confirm their genetic distinctiveness.