ABSTRACT. We used correlation and path coefficient analysis based on an ontogenetic approach to develop selection criteria in barley (Hordeum vulgare L.) for an early production system in Ethiopia. A total of 100 genotypes using 10 x 10-simple lattices with two replications were used to perform the experiment at Ambo and Asasa. The combined analysis of the measured traits showed significant differences among genotypes for all traits. A positive correlation was observed between grain yield and spike/m 2 , kernel number/spike, and 1000-kernel weight. The path analysis showed that spike/m 2 , 1000-kernel weight, and kernel number per spike had significant positive direct effects on grain yield, which shows that these traits can be used as selection criteria to improve grain yield. The significant positive correlation of spike/ m 2 , 1000-kernel weight, and grain-filling period and the positive direct effect on grain yield indicated the potential of these traits as indirect selection criteria to improve grain yield in the early production system in Ethiopia. This study also showed that the path coefficient analysis based on an ontogenetic model was efficient and produced results that can be interpreted clearly.