Perhaps Berner would prefer you reference his actual article in Nature rather than the staff writers' "Cliff's Note's" proxy.
Nature 446, 530 (2007). doi:10.1038/nature05699
Authors: Dana L. Royer, Robert A. Berner & Jeffrey Park
A firm understanding of the relationship between atmospheric carbon dioxide concentration and temperature is critical for interpreting past climate change and for predicting future climate change1. A recent synthesis2 suggests that the increase in global-mean surface temperature in response to a doubling of the atmospheric carbon dioxide concentration, termed 'climate sensitivity', is between 1.5 and 6.2 °C (595 per cent likelihood range), but some evidence is inconsistent with this range1, 2, 3, 4, 5. Moreover, most estimates of climate sensitivity are based on records of climate change over the past few decades to thousands of years, when carbon dioxide concentrations and global temperatures were similar to or lower than today1, 6, so such calculations tend to underestimate the magnitude of large climate-change events7 and may not be applicable to climate change under warmer conditions in the future. Here we estimate long-term equilibrium climate sensitivity by modelling carbon dioxide concentrations over the past 420 million years and comparing our calculations with a proxy record. Our estimates are broadly consistent with estimates based on short-term climate records, and indicate that a weak radiative forcing by carbon dioxide is highly unlikely on multi-million-year timescales. We conclude that a climate sensitivity greater than 1.5 °C has probably been a robust feature of the Earth's climate system over the past 420 million years, regardless of temporal scaling.
Looking at the Yale news release for the article [ http://www.yale.edu/opa/newsr/07-03-28-02.all.html ], we find an interesting statement in regards the lack of validation for temperature in the paleo record they used in respect to their simulations of CO2, to come to their conclusions:
"Led by Dana L. Royer, assistant professor of Earth and Environmental Sciences at Wesleyan University, who did his graduate work in geology at Yale, the collaboration simulated 10,000 variations in the carbon-cycle processes such as the sensitivity of plant growth to extra CO2 in the atmosphere. They evaluated these variations for a range of atmospheric warming conditions, using the agreement with the geologic data to determine the most likely warming scenarios. The model-estimated atmospheric CO2 variations were tested against data from ancient rocks.
Other proxy measurements of soil, rock and fossils provided estimates of CO2 over the past 420 million years. Calculation of the climate sensitivity in this way did not require independent estimates of temperature. It incorporated information from times when the Earth was substantially warmer and colder than today, and reflects the sensitivity of the carbon-cycle balance over millions of years."
Seems they only took empirical measures of CO2 proxies with respect to geological age to compare against their simulations of what CO2 ought to be according to their implicit assumptions of with no validation of actual magnitude of temperature variation from the geological record.
As can be observed from the results represented in #22 above, raw climate sensitivity based on temperature ranges throughout the paleo-record do not correlate well at all with CO2 variation.
Per the statement from the abstact "but some evidence is inconsistent with this range1, 2, 3, 4, 5." "some evidence is inconsistent" could easily be seen as a prime candidate for understatement of the decade.
You need a subscription to read articles in Nature. That's why I don't do that.