In IPCC 1995 [SAR] – An Extended Excerpt, I quoted the following key statements from IPCC 1995:
Alpine glacier advance and retreat chronologies (Wigley and Kelly, 1990) suggest that in at least alpine areas, global 20th century temperatures may be warmer than any century since 1000 AD and perhaps as warm as any extended period (of several centuries) in the past 10,000 years. Crowley and Kim (1995) estimate the variability of global mean temperature on century time scales over the past millennium as less than +- 0.5 deg C. [my italic]
Then I wondered:
These last two lines are the ones that jettison the issue of attribution of past climate change. I’ll try to get to comments on the two sources quoted here. It would be nice if they were relying on sources that were not so closely associated with the campaign. I wonder how strong Wigley and Kelly, 1990 and Crowley and Kim, 1995 really are, especially relative to some of the recent work from Joerin and Nicolussi noted up recently on this site.
Wigley and Kelly, 1990 did not support the claims attributed to them at all, as discussed here. Here’s the rest of the story on Crowley and Kim, 1995, which proves to be no better than a secondary and perhaps tertiary source. I have now consulted Crowley and Kim 1995 and will post up a pdf if there is any demand. This is a model study based on an Energy Balance Model and does not contain any discussion of proxies. It compares projected temperature increases under this model to past temperatures. The main lifting is done in their Figure 3, shown below with the original caption. The historical part of this graph is interesting since it shows the striking and long-term decline in global temperature over the past 50 million years with temperatures in the most "recent" period (on say a 100,000 year step) being the lowest in the entire period (perhaps also in the 500 million year period), Figure 3. Comparison of future greenhouse projections against the geologic record. Curves on the right represent estimates of global temperature change in the past from the oxygen isotope record (see Fig. 4); green represents differences from the Holocene core tops and magenta represents differences from the observed global average temperature. Crossbars indicate fitting points for calibration of oxygen isotope curve in terms of global temperature (See text). Labeled scale on right-hand side represents calculated values for peak warming on left-hand side of figure. Curves on the left represent estimates from restricted and unrestricted CO2 scenarios, utilizing the standard IPCC range and best guess for sensitivities equivalent to CO2 doubling. "Error bars" represents a generous estimate of the range of natural variability based on records of the last 1000 years [Crowley and North, 1991]. [my bold – see the scale denoted +- 0.5 degree C in the bottom left hand corner of the figure.] The comment in the caption to Figure 3 is made in passing once more in the running text as follows:
The projected increase in global average temperatures for all scenarios also greatly exceeds the past record of climate variations over the last millennium [Crowley and North 1991] (Paleoclimatology, Oxford University Press, 339 pp). A generous estimate of that natural variability for that time is +- 0.5 degree C.
So Crowley and Kim  is not a primary source for the estimate of "variability of global mean temperature on century time scales over the past millennium"; it is a secondary source relying on a textbook, which is probably not a primary source either. One would have hoped for a little more from the numerous review iterations and thousands of scientists involved in an IPCC assessment report. Anyway on to Crowley and North, 1991 to see what it says.