The knowledge of tree age is important for understanding tree growth and forest dynamics. It may be estimated by 'direct' methods involving growth ring counts, or by 'indirect' methods involving field measurements of growth rates. Direct methods are considered more accurate, but it is not clear if they are appropriate for all species, notably from the humid tropics. In this paper we assess the occurrence of annual growth rings and their utility for age estimation in three tropical tree species, Acrocarpus fraxinifolius, Dalbergia latifolia (Fabaceae) and Syzygium cumini (Myrtaceae), growing in traditional shade coffee plantations of the southern Western Ghats, India. These species previously were described as having "indistinct or absent" growth rings. We used anatomical studies, field measurements and computational methods to characterise growth rings and assess similarities between directly and indirectly estimated tree ages. Our study revealed that annual growth rings were characterised by different sets of anatomical features per species and were most distinct in the fast-growing deciduous A. fraxinifolius. Growth rates measured in the field showed annual periodicity in all three species, and reflected annual rainfall-drought cycles in D. latifolia and S. cumini. Direct age estimates were most similar to indirect estimates in D. latifolia, and least so in S. cumini. The results of direct age estimation by counting rings are consistent with them being annual in nature in tropical species with distinct and reliable annual growth ring formation. However, for species with poorly defined growth rings, indirect age estimation methods might be more useful.