<p>Proxy records from climate archives provide evidence about past climate changes, but the recorded signal is affected by non-climate related effects as well as time uncertainty. As proxy based climate reconstructions are frequently used to test climate models and to quantitatively infer past climate, we need to improve our understanding of the proxy records’ signal content as well as the uncertainties involved. In this study, we empirically estimate signal-to-noise ratios (SNRs) of temperature proxy records used in global compilations of the mid to late Holocene. This is achieved through a comparison of proxy time series from close-by sites of three compilations and model time series data at the proxy sites from two transient Holocene climate model simulations. In all comparisons, we found the mean correlations of the proxy time series on centennial to millennial time scales to be rather low (R < 0.2), even for nearby sites, which resulted in low SNR estimates. The estimated SNRs depend on the assumed time uncertainty of the proxy records, the time scale analysed, and the model simulation used. Using the correlation structure of the ECHAM5/MPI-OM simulation, the estimated SNRs on centennial time scales the SNRs ranged from 0.05 – assuming no time uncertainty – to 0.5, for a time uncertainty of 400y. On millennial time scales, the estimated SNRs were generally higher. Use of the TraCE-21ka model simulation correlation structure resulted generally in lower SNR estimates than for ECHAM5/MPI-OM. As the number of available high-resolution proxy records continues to grow, a more detailed analysis of the signal content of specific proxy types should become feasible in the near future. The estimated low signal content of Holocene temperature records should caution against over-interpretation of these kinds of datasets until further studies are able to facilitate a better characterisation of the signal content in paleoclimate records.</p>