The study of xylogenesis or wood formation is a powerful, yet labor intensive monitoring approach to investigate intra-annual tree growth responses to environmental factors. However, it seldom covers more than a few growing seasons, so is in contrast to the much longer lifespan of woody plants and the time scale of many environmental processes. Here we applied a novel retrospective approach to test the long-term (1926-2012) consistency in the timing of onset and ending of cambial activity, and in the maximum cambial cell division rate in two conifer species, European larch and Norway spruce at high-elevation in the Alps. We correlated daily temperature with time series of cell number and lumen area partitioned into intra-annual sectors. For both species, we found a good correspondence (1-10 days offset) between the periods when anatomical traits had significant correlations with temperature in recent decades (1969-2012) and available xylogenesis data (1996-2005), previously collected at the same site. Yet, results for the 1926-1968 period indicate a later onset and earlier ending of the cambial activity by 6-30 days. Conversely, the peak in the correlation between annual cell number and temperature, which should correspond to the peak in secondary growth rate, was quite stable over time, with just a minor advance of 4-5 days in the recent decades. Our analyses on time series of wood anatomical traits proved useful to infer on past long-term changes in xylogenetic phases. Combined with intensive continuous monitoring, our approach will improve the understanding of tree responses to climate variability in both the short- and long-term context.