An efficient temperature and humidity retrieval algorithm for radiometric measurements at high spectral resolution is introduced and applied to climatological profiles. The algorithm is developed for analyzing Infrared Atmospheric Sounding Interferometer (IASI) data of the European weather satellite METOP-1 (launch scheduled 2005) for climatological purposes but is also applicable for other purposes and to other similar data. The algorithm's core features are a channel selection methodology followed by a linearized optimal estimation. The key concept of the former is that a small subset (5--10%) of all available IASI channels (~8000) is selected based on maximizing a suitable information content measure at each retrieval level. This enables efficiency and robustness of the retrieval algorithm and curtails the high redundancy in the measurements. In addition to profile and error covariance estimates optimal estimation furnishes various sensitivity functions of which we used weighting functions for quantifying the utility of measurement channels and averaging kernel functions for assessing the resolution of retrieved profiles. Results based on simulated IASI spectra computed from a set of standard climatological profiles and a realistic radiometric noise model demonstrate, for clear air, the capabilities of high spectral resolution measurements for improving temperature and humidity soundings compared to current operational sensors. In the troposphere (below ~200 hPa), retrieved profiles exhibit temperature errors of <1 K and specific humidity errors of <10% at most heights, associated with a vertical resolution of ~1.5--2 km. Promising performance was found in the upper troposphere (500--200 hPa), where about five independent reliable values of temperature and humidity are available indicating the high potential of the IASI sensor for monitoring climatic changes in upper tropospheric moisture. Tests on the sensitivity of retrieved profiles to the quality of a priori profiles showed weak sensitivity of temperature but significant sensitivity of humidity. The results provide a solid basis and clear guidance for improvements of the presented algorithm for reliable large-scale application on cloud-free spectra.