Ocular and shoot-count defoliation estimation techniques were compared for accuracy and precision at the levels of branches and crown position within trees, whole trees, plots and stands. A shoot-count estimate of defoliation for a mid-crown branch has the best relative accuracy (+/- 7%), while the best ocular estimate for the whole tree by an experienced observer (accuracy +/- 12%). Ocular estimates were biased towards overestimation at low defoliation levels (error 20-30% defoliation), and previous defoliation caused the overestimation of current defoliation by 30-40% on trees that have been severely defoliated. Observers were found to be consistent in their rating, but biased by about +/- 10% : experienced observers were about variance between plots in a stand. It was concluded that ocular estimation of defoliation were about 5-10% more accurate than inexperienced observers. Intertree variance in defoliation was greater than intertree variance or variance between plots in a stand. It was concluded that ocular estimation of defoliation is a viable technique that can give accuracy with limits required for surveys and many research applications, if the influence of observer experience, observer bias, and previous defoliation are recognized, and adjusted for when necessary.