The purpose of the present work is (1) to study the statistical relevance of twin-twin junctions and (2) to study statistically the influence of twin-twin junctions and microstructure on nucleation and growth of twins in h.c.p. materials. A new automated twin recognition technique has been developed and is used to extract statistics from EBSD scans of high purity clock-rolled zirconium specimens loaded along the through-thickness and one of the in-plane directions. The technique allows for recognition of tensile and compressive twin systems within each individual grain. The ten possible twin-twin junction types that may occur in Zr between first generation twins are introduced as well as their associated frequencies in cases of through-thickness and in-plane compression. The present study shows that twin-twin junctions between twins belonging to the most active twinning modes are statistically relevant. It is also shown that twin-twin junctions hinder twin growth. In agreement with previous studies, it is found that irrespective of the loading direction and twin mode, both grain size and crystallographic orientation largely influence the propensity of grains for twin activation. However, the study suggests large differences in nucleation and growth mechanisms for each twinning mode.