The measurement of particle shape by image analysis and light scatteri
ng depends, inter alia, upon projected area. Non-spherical particles e
xhibit a range of values for projected area and perimeter. The distrib
ution functions for these variables, together with their mean values,
are useful as signatures for specific shapes. When the results from in
dividual, random orientations are being used as starting points for fu
rther computation, we investigate how many results are needed to give
good agreement with the actual distribution functions. Choosing the cu
be as an example, it is shown how experimental data may be used to det
ermine size using maximum-likelihood estimators. Adequate accuracy can
be obtained with a modest number of data. The question of whether it
is better to use a random or regular set of directions is investigated
. (C) 1998 Elsevier Science S.A. All rights reserved.