A B S T R A C T :Regression analysis of high-precision structural and chemical data on trioctahedral and dioctahedral micas yielded interrelationships between unit-cell parameters, chemical composition and structural details. Regression equations relating b and csinfl parameters of micas to composition were used for estimating composition from cell data in order to analyse P-T conditions of rock formation. Algorithms for computing atomic coordinates for 2M1, 3T and IM dioctahedral micas having either centrosymmetric or non-centrosymmetric layers and 1M trioctahedral micas are presented. Deviations of computed atomic coordinates from experimental values are, on average, 0-002 A for octahedral cations and 0.005-0.010 ,~ for other atoms. Discrepancies between calculated and experimental individual interatomic distances seldom exceed 0.01 A. Computed atomic coordinates were used to calculate X-ray diffraction patterns for glauconite and illite. Results indicate a close fit between the calculated and experimental patterns. The local structure around an octahedral cation of interest can be determined.Diverse isomorphous cation substitutions depending on bulk chemistry and rock formation conditions lead to considerable variations in phyllosilicate structural features. Direct diffraction structure analysis, however, is rarely done, not only because of the expense and time involved in the procedure but also because suitable single crystals are difficult to find. Furthermore, phyllosilicates are often clay-sized, poorly crystallized and impure, thereby making interpretation of diffraction and spectroscopy data problematic. Various simulation procedures to overcome this difficulty should involve modelling of the average structure and/or local structural fragments. On the other hand, even direct chemical analysis of each component in a polymineral sample is not always feasible. It is of interest, therefore, to develop methods to allow detailed structural and crystal-chemical information to be deduced from easily obtainable data, such as bulk chemical analysis and X-ray diffraction (XRD) powder patterns.Two aspects of this problem are developed in this paper. First, cell parameters and chemical composition of dioctahedral micas are examined for interrelationships. This is useful for estimating compositions of polycomponent samples for which direct chemical analysis is laborious or expensive, and for evaluating the degree of transformation of lowand medium-grade metamorphic rocks. Second, structural features are predicted and atomic coordinates are calculated, using a specially devised computer program, from the data on chemical composition and cell parameters of dioctahedral and trioctahedral micas having disordered cation distribution. Such calculations are useful to interpret infrared (IR), M6ssbauer and EXAFS spectroscopic data for monomineral samples, to simulate XRD patterns, and to aid in electrostatic energy calculations and estimations of O-H vector 9 1993 The Mineralogical Society