Bernoulli-Euler theory and Bessel functions are used to obtain explicit expressions for the exact dynamic stiffnesses for axial, torsional and flexural vibrations of any beam which is tapered such that A varies as J'"and GJ and I both vary as y(" ' ), where A, GJ and I have their usual meanings; y = (cx:L) + 1; c is a constant such that c > -1; I, is the length of the beam; and x is the distance from onc end of the beam. Numerical checks give better than seven-figure agreement with the stiffnesses obtained by extrapolation from stepped beams with 400 and 500 uniform elements. A procedure is given for calculating the number of natural frequencies exceeded by any trial frequency when the ends of the member are clamped. This enables an existing algorithm to be used to obtain the natural frequencies of structures which contain tapered members. and S =k'8 at x = L) can be substituted in equations (F9) and (Flu) to give i.e. where Equations (F13) and (F17) give F, = K,U, i.e. f(Symmetric K 8 . 8 where K, = D,B,Therefore, one way of finding K, is to use a computer, at each required frequency, to calculate the elements of B, and D, (from equations (F12), (F13), (F16) and (F17)) and then to invert B, and pre-multiply by D,. For large structures the computer is likely to spend much less time performing this arithmetic for all the tapered members in a frame than it spends on the other operations which it has to perform on the frame at the same frequency (see, for example. Reference 31, where the much larger arithmetic effort needed to incorporate stayed columns as members of a space truss proved to be quite a small proportion of the total solution time for the truss). Therefore, this simple procedure would be adequate for many purposes. However, computer time can be saved by using equation (F21), and the expressions for the elements of B, and D, given in equations (F12) and (F16), to obtain algebraic expressions for the elements of K,. Such expressions are particularly useful when only some of the stiffnesses are needed, and are given below. They are not necessarily in the shortest possible form, but arc surprisingly concise, and represent a lot of work by the first author which is too extensive and detailed to be presented. The correctness of the expressions was confirmed by numerical checks which arc discussed in the next section.