Seed yield data from 16 genotypes (G) of chickpea (Cicer arietinum L.), tested in 49 transcontinental locations (L) were analyzed using residual maximum likelihood (REML) procedure and pattern analysis methods of classification and ordination to gein an understanding of the structure of Genotype x Environment (GE) interactions. The genotypes included the two major types of chickpea-desi and kabuli drawn from two main areas for chickpea adaptation, the Indian subcontinent and the Mediterranean region. Using hierarchical classification, the 16x49 GL mean data matrix was reduced to 8x9 (G-group x L-group) matrix capturing 50% of GL interaction sums of squares (SS). The genotypes fell into three discernible groups according to their geographic proximity. The location grouping suggested the presence of two mega-environments. Sub grouping of locations showed that generally the tropical (23°) locations grouped together and separate from sub-tropical locations. The performance plots of G-groups indicated that two low yielding genotypes, Giza and Harigantars, contril?uted to large variation from one L-group to another. In general the medium-duration G-group (K 850, L 550, ICCC 8 and ICCC 4) from India was consistently better in most L-groups. In contrast, the G-group comprising Rabat (Morocco) and ILC 482 (Turkey), generally performed well in higher latitudes. The principal component biplot results were essentially similar to those obtained from the performance plots. It appears that development of early to medium-duration varieties may increase ahd stabilize seed yields. Based on genotype grouping, separate breeding programs for subtropical, tropical and the Mediterranean regions appear appropriate. This study showed that breeding for kabuli type and relatively bod seed might not necessarily affect varietal productivity adversely. Implications of these findings for chickpea improvement are discussed.