The plagioclase loop width is defined as AnS – AnL where S and L are solid and liquid, respectively, and An is the mole fraction (or per cent) An/(An + Ab). It is measured from the plagioclase composition and an estimate of the liquid composition based on standard assumptions rooted in the norm recalculation routine. A survey of a large, new database allows study of the 1 atm experimental literature involving 819 experiments. The approach to equilibrium is tested empirically by the loop width and its transform, the linear partitioning function KD. This function is the exchange coefficient formed as the intercept at pure AnS from the array over all XS of the partition coefficient D when defined as (
) where X is the mole fraction. D and KD are thus limited to lie between zero and 1·0. The results show a 1 atm loop width ranging from 0·2 to 68 mol % An with associated KD = 0·03–0·98. These extrema represent loops both far too small and far too large to represent stable equilibrium between plagioclase and liquid. Some of the higher values lie near the persodic limit because the An component of the liquid is scarce, and the liquid composition is Ab-rich. The loops from Ne-normative liquids span the range from 40 to 90 mol % An in the crystal with loops 2–50 mol % wide. A baseline for loop width is derived from the precise synthetic study of the system Di–An–Ab at 1 atm and an array of high-pressure data that extrapolates to the same value of KD = 0·26 at 1 atm. Many of the 1 atm experimental data fall far outside any reasonable uncertainty limits on the baseline. Causes of departure include crystal disequilibrium that can lead to errors in both directions, and compositional problems. Comparison of the values of KD with liquid compositions in silica, oxygen fugacity, temperature, kinetics, and sodium reveals the common presence of sodium deficiency. A crystallinity study shows a definite decrease in sodium content with liquid evolution, whereas the sodium content should increase or at least remain nearly constant with evolution of the melt. A selection of self-consistent 1 atm data from a few laboratories shows a close correspondence to the standard distribution of loop widths and indeed most literature data lie within the limited range of KD = 0·26–0·4. Experiments known to be hydrous show very low values of KD and corresponding large loop widths, as previously understood from literature data. When data limited to near-baseline KD values are plotted against temperature, they yield expected thermal slopes and concave-up curves of temperature plotted against plagioclase composition. Literature data show a positive correlation of KD with dry pressure, and a separate database from 2 to 15 kbar shows that more than half the data comport with this pressure effect. A different group ranges down to KD as low as 0·2, with increasing loop widths. Some of these are alkalic and hence rich in the Ab component of the liquid. Others were obtained in Fe containers that control the oxygen fugacity to low values and convert the fayalite component of the liquid to iron, oxygen, and silica. This reaction increases the activity of silica and in turn the activity of Ab in the melt, thus widening the loop. Thermodynamic studies concerned with plagioclase–melt equilibria will profit from an awareness of these effects.