We present a new 23-site network of white spruce ring-width chronologies near boreal treeline in Old Crow Flats, Yukon Territory, Canada. Most chronologies span the last 300 years and some reach the mid-16th century. The chronologies exhibit coherent growth patterns before the 1930s. However, since the 1930s, they diverge in trend and exhibit one of two contrasting, but well-replicated patterns we call Group 1 and Group 2. Over the instrumental period (1930–2007) Group 1 sites were inversely correlated with previous-year July temperatures while Group 2 sites were positively correlated with growth-year June temperatures. At the broader northwestern North America (NWNA) scale, we find that the Group 1 and Group 2 patterns are common to a number of white spruce chronologies, which we call NWNA 1 and NWNA 2 chronologies. The NWNA 1 and NWNA 2 chronologies also share a single coherent growth pattern prior to their divergence (ca. 1950s). Comparison of the NWNA 1/NWNA 2 chronologies against gridded 20th-century temperatures for NWNA and reconstructed northern hemisphere summer temperatures (ad 1300–2000) indicates that all sites responded positively to temperature prior to the mid-20th century (at least back to ad 1300), but that some changed to a negative response (NWNA 1) while others maintained a positive response (NWNA 2). The spatial extent of divergence implies a large-scale forcing. As the divergence appears to be restricted to the 20th century, we suggest that the temperature response shift represents a moisture stress caused by an anomalously warm, dry 20th-century climate in NWNA, as indicated by paleoclimatic records. However, because some sites do not diverge and are located within a few kilometres of divergent sites, we speculate that site-level factors have been important in determining the susceptibility of sites to the large-scale drivers of divergence.