A power series is derived for 1/T, the reciprocal of the geometrical-acoustics reverberation time of an enclosure. The first term is Sabine's expression and the second term provides a correction to the Sabine value in many practical cases where his value is a good, but not quite good enough, estimate of T. The Eyring formula and other expressions which have been advanced as improvements on the Sabine formula are shown to contain only parts of the second term and hence to be unreliable, yielding in some enclosures a correction of even the wrong sign.