The timing of the commencement of photosynthesis (P*) in spring is an important determinant of growing-season length and thus of the productivity of boreal forests. Although controlled experiments have shed light on environmental mechanisms triggering release from photoinhibition after winter, quantitative research for trees growing naturally in the field is scarce. In this study, we investigated the environmental cues initiating the spring recovery of boreal coniferous forest ecosystems under field conditions. We used meteorological data and above-canopy eddy covariance measurements of the net ecosystem CO2 exchange (NEE) from five field stations located in northern and southern Finland, northern and southern Sweden, and central Siberia. The within- and intersite variability for P* was large, 30–60 days. Of the different climate variables examined, air temperature emerged as the best predictor for P* in spring. We also found that ‘soil thaw’, defined as the time when near-surface soil temperature rapidly increases above 0°C, is not a useful criterion for P*. In one case, photosynthesis commenced 1.5 months before soil temperatures increased significantly above 0°C. At most sites, we were able to determine a threshold for air-temperature-related variables, the exceeding of which was required for P*. A 5-day running-average temperature (T5) produced the best predictions, but a developmental-stage model (S) utilizing a modified temperature sum concept also worked well. But for both T5 and S, the threshold values varied from site to site, perhaps reflecting genetic differences among the stands or climate-induced differences in the physiological state of trees in late winter/early spring. Only at the warmest site, in southern Sweden, could we obtain no threshold values for T5 or S that could predict P* reliably. This suggests that although air temperature appears to be a good predictor for P* at high latitudes, there may be no unifying ecophysiological relationship applicable across the entire boreal zone.