The observed fraction of pulsars with interpulses , their period distribution and the observed pulse width versus pulse period correlation is shown to be inconsistent with a model in which the angle \alpha between the magnetic axis and the rotation axis is random .
This conclusion appears to be unavoidable , even when non-circular beams are considered .
Allowing the magnetic axis to align from a random distribution at birth with a timescale of \sim 7 \times 10 ^ { 7 } years can , however , explain those observations well .
The timescale derived is consistent with that obtained via independent methods .
The probability that a pulsar beam intersects the line of sight is a function of the angle \alpha and therefore beam evolution has important consequences for evolutionary models and for estimations of the total number of neutron stars .
The validity of the standard formula for the spin-down rate , which is independent of \alpha appears to be questionable .