I was browsing through some Jacoby articles for a different purpose and was reminded of the very interesting references to a relationship between high-latitude tree ring growth and southern exposure to the sun in two of his 1980s articles, which do not get mentioned any more once the GW campaign is on.
Jacoby et al.  discusses a site chronology at Cri Lake, Quebec on the eastern shore of Hudson Bay, which has no 20th century growth trend, concluding that the difference from other northern sites with a 20th century growth trend pertained to its not having a southern exposure:
The reconstruction of summer temperatures for this site contains the expected high-frequency year-to-year changes and also two- significant low-frequency features — an abrupt cooling beginning in 1816 and an absence of trend since about 1880. Many of the high latitude tree-ring chronologies display a long-term warming trend over the past 150 years (Jacoby et al, 1985 and Jacoby and D’Arrigo, 1987.) The chronologies which show a recent long-term warming are obtained almost invariably from sites obtained on flat terrain or southerly facing slopes. We attribute the absence of this trend at Kuujjuarapik to the geomorphology of the tree sampling site. The fact that Cri Lake is a narrow, steep-sided ravine, blocked from low-angle spring and fall insolation seems to have an important impact on the way trees respond to climate. This observation is in agreement with information that we have from white spruce trees sampled at two other sites situated in the same general areas. Similar to Cri Lake, the chronology at Richmond Gulf, another somewhat sheltered site doesn’t show any long-term growth trend. In contrast, the chronology from Castle Peninsula, a south-facing slope presents the strong increased-growth (inferred warming) tend over the past 150 years characteristic of most good high-latitude North American chronologies. We apply the term good to chronologies with strong common variance between trees at both high and low frequencies. The recorded NH air temperatures (Jones and Kelly, 1983) also indicate warming in the average annual temperature for this period, but little trend is present in the summer air temperature series (Wigley et al 1985). The trees at the sampling site apparently respond primarily to summer temperatures. Thus the non-summer warming is not integrated into growth at this site. [Jacoby et al, PPP 1988]
This point was mentioned for what seems to be the only other occasion in Jacoby and D’arrigo  as follows:
Also among the sites that we have collected in the boreal forests, only trees growing on sites with good exposure to the south and hence low-angle sun have strong common low-frequency variance…The eleven chronologies were selected as those with the bet common variance in the red-noise analyses. All sites had good common variance in the white noise analyses. Of the 36 northern boreal forest tree-ring sites we have sampled within the past decade, ten are thus judged to provide the best record in time and space of temperature-influenced tree growth for this region of North America. The one chronology from a lower latitude on the Gaspé peninsula had similarly good common low-frequency variance and is included due to the scarcity of other data in the eastern region.
The relationship between southern exposure for northern spruce sites and a 20th century growth trend sure seems like an interesting relationship. If the relationship was simply between growth and temperature, then you would think that the shadier sites with northern exposure would also show increased growth. In fact, arguably, the relationship would be stronger for them. If the effect is limited to sites with southern exposure, perhaps something else is involved: perhaps there is some relationship between cloudiness and enhanced 20th century growth. Intuitively, I could see how something like this could possibly explain the different responses between sites with and without southern exposure. Whatever the ultimate explanation, the question surely shouldn’t have been dropped.
One of the difficulties in pursuing this is Jacoby’s practice of selective archiving discussed previously here and here. At the NGDC archive , Jacoby archived the results for Castle Peninsula (which has increased 20th century growth), but not for Cri Lake or Richmond Gulf, which don’t. (You can easily confirm this by going to the search function, selecting tree ring, the Canada-Quebec and setting 55N as the southern limit.
Jacoby, G.C., Ileana S. Ivanciu and Linda D. Ulan, A 263-year record of summer temperature for nother n Quebec reconstructed from tree-ring data and evidence of a major climate shift in the early 1800s, PPPP 64, 69-78.
Jacoby, Gordon C. and Rosanne D’Arriog, Reconstructed northern hemisphere annual temperature since 1671 based on high-latitude tree-ring data from North America, Clim Chg 14, 39-59.