I will consider mainly geothermobarometry in medium-grade pelitic rocks, including the garnet-biotite (GB) geothermometer, the Grossular-Al silicate-plagioclase (GASP) geobarometer, and the muscovite-almandine-biotite-sillimanite (MABS) geobarometer. For GB (Holdaway, 2000) experimental data and estimated biotite ΔWTi were used to optimize two exchange parameters and four biotite Margules parameters. Using stepwise linear regression, experimental vs. calculated T were constrained to lie on a line with slope of one and intercept of zero, maximizing r2. The best model involves experiments by Ferry and Spear (1978) and Perchuk and Lavrent’eva (1983), suggesting minimal viAl in the Ferry and Spear product biotite. For GASP (Holdaway, 2001), end-member experimental data do not adequately constrain the equilibrium. I used the GB model above, and allowed the end-member curve to rotate about the best-constrained part of the GASP end-member data. The end-member curve was further constrained with the kyanite-sillimanite (K-S) boundary using published chemical data on 76 pelitic schist samples from 11 localities, rejecting Low-Grs and low-An samples. The Fuhrman and Lindsley (1988) plagioclase model gives the best results. For MABS, work in progress involves 61 samples from the 11 localities which have muscovite analyses. Biotite Margules parameters were based on the GB model and McMullin et al. (1991). The MABS end-member curve was calibrated by comparison of P values determined using trial MABS data and GASP results. The P values for the 61 samples agree well with the K-S boundary, and sillimanite-bearing rocks of west-central Maine all fall in the sillimanite field. Preliminary biotite values are: GAnn = −5149198 − 412.05 T, WAlFe = −14023 + 28.14 T, WAlMg = −259582 + 308.44 T, WTiFe = 124842 − 98.67 T, WTiMg = −186148 + 271.72 T. For geobarometry, the Berman (1988, revised 1992) database was used with adjustable H and S of grossular for GASP and H and S of annite for MABS. The accuracy of currently available databases, activity models and mole fraction models is not adequate for good geothermobarometry, without further refinement. Adjustable parameters tend to compensate for error in activity models, mole fraction models and databases.