White (1977) suggests several potential indicators of grain boundary sliding. These are: constant small grain size, equidimensional grain shape, random lattice orientation, bimodal alignment of grain boundaries, presence of grain boundary bubbles, and presence of high dislocation density adjacent to grain boundary irregularities and triple points. In addition to these, Beere (1978) and Fliervoet et al. (1997) showed that with a dominant grain boundary sliding, there is no crystallographic relation between adjacent grains due to grain rotation. Since the present technique of the experiment described in this paper does not allow the analysis of the microstructure on TEM/SEM scale, the last two indicators will not be discussed here.

In experiment TO-105, the average grain area at the end of the deformation is about 1.76 x 10^-2 mm², which is larger than that of experiment TO-110 (about 1.23 x 10^-2 mm²). Since the grain boundary sliding is more extensive in experiment TO-105, smaller grain size does not necessarily favor grain boundary sliding.

In all experiments where grain boundary sliding appears to be extensive with the development of grain boundary openings, the average grain axial ratio is less than 1.4 throughout the deformation. This value is not much lower than that of experiment TO-110 (about 1.5 from shear strain = 0.9 onwards) where intragranular plastic deformation is more important. Also in another experiment, not described here (experiment TO-109), the average grain axial ratio is lower throughout the deformation (1.1 - 1.3) and intragranular plastic deformation is predominant. Therefore equidimensional grain shape may not be a good indicator of grain boundary sliding, particularly if grain-shape adjustment by grain boundary migration is effective (Rubie, 1990).

Random lattice orientation may not be a reliable indicator of grain boundary sliding either since sample TO-105 shows a strong lattice preferred orientation (Fig. 8). It has been suggested that some lattice preferred orientation could develop when grain boundary sliding is accommodated by intracrystalline slip (Edington et al., 1976; Etheridge and Wilkie, 1979; Schmid et al., 1987). Furthermore, Rutter at al. (1994) showed that in a deformation with dominant grain boundary sliding, there is only a weak development of lattice preferred orientation at 50% strain of fine-grained calcite aggregates, whereas there is a strong lattice preferred orientation at more than 600% strain, similar to that of intracrystalline plastic flow regime.

Bimodal alignment of grain boundaries has been considered as a strong evidence of grain boundary sliding (Raj and Ashby, 1971; Singh et al., 1973; White, 1977; Schmid et al., 1987; Drury and Humphreys, 1988). In simple shear deformation the first maximum is known to be parallel to the shear zone boundary with the second maximum being at about 70^o to the shear zone boundary (Fig. 16a, Schmid et al., 1987; Drury and Humphreys, 1988). In pure shear deformation, two maxima are known to be symmetric with respect to the shortening direction with an angle of about 45^o (Fig. 16d, Singh et al., 1973; White, 1977; Drury and Humphreys, 1988). Although the grain boundary orientation in experiment TO-105 shows two maxima perpendicular to each other, with one maximum 25^o off from the bulk shear direction (Fig. 16c), other experiments do not generate a preferred bimodal alignment of grain boundaries (Figs. 16b, e and f).

The preferred orientation of grain boundary openings is seen to be a reliable criterion of grain boundary sliding in experiments described here (Fig. 15). The disappearance of grain boundary openings during static readjustment of the microstructure after deformation (Fig. 14) suggests that it may not be a valuable indicator of grain boundary sliding in naturally deformed rocks. However, if grain boundaries are found to carry remnants of former grain boundary openings such as an array of bubbles or voids and of second phase inclusions in observation under electron microscopy (White and White, 1981; Behrmann, 1985; Fliervoet et al., 1997), and if these grain boundaries have a well-defined preferred orientation (Behrmann and Mainprice, 1987), these will be a strong indicator of the former existence of grain boundary openings and grain boundary sliding. Preferred orientation of grain boundary bands with a slightly different chemical composition from the grain interior (Hall, 1984) might be another possible indicator of grain boundary opening and sliding.

In summary, any single indicator alone of all above, except preferred orientation of remnants of grain boundary openings, does not satisfactorily imply grain boundary sliding. The main reason for this is probably that accommodation mechanisms for grain boundary sliding will easily erase the signatures of grain boundary sliding.

Acknowledgements
I thank my teacher, Prof. W.D. Means, for his continuous support and patience while I was at Albany.  I also thank Youngdo Park for his help with making movies used in this paper.  The manuscript was improved by comments from Paul Bons and Mark Jessell.  The publication of this paper was partially supported by Korea Science and Engineering Foundation grant 981-0401-005-2.