The distribution of sunspot decay rates is studied using the Greenwich Photoheliographic Results (GPR) for a total of approximately hundred years between 1874 and 1976. The decay rates are seen to be lognormally distributed. The discrepancies between the decay rates given in the past by different authors are shown to originate as a consequence of this asymmetric distribution. It is pointed out that the extended tails shown by the lognormal distributions are associated to spots decaying much faster than suggested by Bumba's (1963) work. A cycle by cycle analysis of the lognormal distributions associated with each sunspot group type and for single spots is presented. The differences between the nine solar cycles involved are studied. As a remarkable property of the decay process, we show that it happens at a nearly constant total to umbral area ratio. This property holds for decaying spots which are still large enough to show a penumbra. We have studied the suitability of a decay law with the instantaneous decay rate proportional to the length of the spot boundary. This law predicts a parabolic decay pattern with some specific characteristics. No definite conclusion in favour of this law is reached, but it is suggested that a linear decay is as weakly supported by the GPR data as a peripheral one. On the other hand, weak non-linearities are seen in the decay of isolated spots with a clear tendency to produce a convex pattern in the area vs. time diagram. The implication is that sunspot decay is braked as time proceeds.