As a mindless activity, I’ve re-visited the Crowley data, which we’re discussing. Among other stereotypes, out of 15 series, Crowley uses 2 bristlecones, Polar Urals, Tornetrask and Dunde. Even so, without any grafting, there’s not much hockey-stick-ness to this dataset. When you parse Crowley, you also see some very odd decisions, which result in lowering MWP levels relative to modern levels. The image here is a bit cluttered – I’ve been looking at too many spaghetti graphs, but bear with me on it.
Figure 1: Crowley unspliced. Black – base case using 15 proxies; dashed – Crowley’s own edit excluding 2 proxies, which results in higher modern to MWP levels; red – replication using freshly collated data; blue – sensitivity with freshly collated data and i) no bristlecones; ii) no Dunde and iii) no Polar Urals in the 11th century.
I commented before on the "odd" decision by Crowley to exclude the Sargasso Sea and Michigan pollen reconstructions from the proxy composite used for further analysis. He used 13 series, rather than 15 series in his composite. He justified this decision as being due to less precise data – then why mention the other 2 series at all other than as a bait-and-switch. It is impossible not to speculate that, if excluding these 2 series had not had the effect of raising modern levels relative to MWP levels, whether Crowley might have decided the other way. Be that as it may, any robust result should be stable to the inclusion of these proxies and I have included them.
A new point is the curious effect of freshly collated data. As noted previously, Crowley does not provide any data citations and "forgot" where he got data. Crowley only provided smoothed and transformed data, so that digital reconciliation to original sources is not possible and considerable detective work is required to sort out candidate versions. I get the impression that Crowley may have digitized some of these series from print publications as some of them are low resolution series.
I think that it is possible that Crowley may have digitized some series, even where there are digital versions available. For example, I’m confident that Crowley’s "central Colorado" series is a version of the Almagre Mt bristlecone pine series. However, while Crowley’s version has a relatively high correlation (0.82) to a smoothed version of an older archived version of this series (co071), there is a definite dilation of the version from the presumed original data. The archived version has higher MWP to modern levels (surprise, surprise).
Some series can be pinned down even with the sketchy information. Crowley’s "White Mountain" series is an old version of Sheep Mountain, the most dominant series in the MBH98 PC1 – so much for "independence". Why would he use an old version rather than the Graybill version? Print digitzation? The Almagre Mt series is also in MBH; the Jasper, Alberta series is used in Jones et al , but not MBH; the fenno series is Briffa’s familiar Tornetrask reconstruction used in MBH and Jones; the Urals series is Briffa’s Urals series used in MBH and Jones; the SE France series is fran010 used in MBH. The Dunde series is used in MBH and Jones, although the grey version used here seems to be sui generis; I’ve used the most up-to-date version in a fresh collation. The GISP2 data is archived. I used the Crowley smoothed versions for C Michigan sediments; Iceland documentary; C England documentary; German C13 and west China tree ring (Dulan), as I was unable to locate archived versions. A Chinese phenological series from Yang  was used instead of the one from Zhu  since Zhu  is apparently misdated according to Zhang .
The net result of the fresh collation is the red version above, which is close to the composite calculated from the email, but with slightly higher MWP-modern levels.
As a sensitivity study, the two bristlecone series, Dunde and the Polar Urals prior to 1100 (with the misdated cores) were excluded. Even on the red version, the 20th century is similar to MWP levels, but without the very series in question, you have a very distinct MWP relative to the modern period.
If you inspect plots of the individual series, it is interesting how few of them have distinctive 20th century values. the strongest hockey stick shape belongs to the Sheep Mountain bristlecone series, which also dominates the MBH North American PC1. The Central Colorado bristlecone series also has elevated 20th century levels (and contributes to the MBH North American PC1). Thompson’s Dunde ice core series here has elevated 20th century values.