1—In a number of papers dealing with the stability of fluid motion, RAYLEIGH employed a certain method, which we may refer to as the “characteristic-value” method. For some problems this method gives results in agreement with observation. For example, it establishes that a heterogeneous inviscid liquid at rest under gravity is stable if the density decreases steadily as we pass upward; it establishes that an inviscid liquid rotating between concentric circular cylinders is stable if, and only if, the square of the circulation increases steadily as we pass outward. This result was stated by RAYLEIGH, and its validity appears to be confirmed by the experiments of TAYLOR, but a simple <
d
2
u
0
/dy
2
retains the same sign throughout the liquid, u
0
being the velocity in the steady motion and
y
the distance from one of the planes. This result is deduced from the fact that mathematical proof by the characteristic value method was not given. I have recently supplied such a proof, extending the problem to include a heterogeneous liquid. But when the method is applied to some other problems, the situation is not so satisfactory. Among the results to which Rayleigh was led is the following. If an inviscid liquid flows between parallel planes, the motion is stable if the characteristic values of a parameter in a certain differential equation cannot be complex, the implication being that they are therefore real. Rayleigh further claimed that the method established the stability of a uniform shearing motion, for which
d
2
u
0
/dy
2
=0. KELVIN and LOVE criticized the method, and a review of the situation in 1907 was given by ORR. In spite of the fact that its general validity remains obscure, the characteristic-value method has been widely employed. It is not the purpose of the present paper to attempt to justify or to discredit the characteristic-value method in general. The paper deals only with the simplest of all stability problems, that of an inviscid liquid flowing between fixed parallel planes. In §2 the method is discussed in some detail and in §3 an argument is developed to show that Rayleigh’s criterion for stability, mentioned above, cannot be legitimately deduced by his method. He proved that complex characteristic values are impossible, and I now prove that real characteristic values are also impossible. The conclusion to be drawn is that the characteristic-value method is not applicable to this case.