A new micromechanical technique for experimental determination of fiber- matrix interfacial properties is presented. This technique consists in tensile loading of the fiber, with a matrix droplet on it, at both ends, accompanied by continuous direct observation of interfacial crack propagation. In comparison with the well-known microbond test, the new method has two important advantages. First, crack propagation is stable for any embedded fiber length and any relation between adhesion and friction at the interface. Second compliance of the test equipment does not affect the results, and specimens with long free fiber ends can be successfully tested. A similar result can be reached using the pull-out or microbond test with an ＇infinite＇ (very long) embedded fiber length. An algorithm for separate determination of the interfacial adhesion and friction from experimental relationships between the crack length and applied load is described. The new test was employed to determine the interfacial parameters for composites of glass fibers with polypropylene, polystyrene, and polycarbonate. For each fiber-polymer system investigated, the following parameters were calculated: ultimate interfacial shear strength; critical energy release rate for crack propagation; and adhesional pressure. Our approach to the estimation of the work of adhesion, W-A, from micromechanical tests, based on the concept of adhesional pressure, allowed us to calculate the W-A values for several thermoplastic matrix-glass fiber pairs and to obtain values consistent with previous estimations made according to other approaches.