13 Years of Timing of PSR B1259-63

This paper summarizes the results of 13 years of timing observations of a unique binary pulsar, PSR B1259$-$63, which has a massive B2e star companion. The data span encompasses four complete orbits and includes the periastron passages in 1990, 1994, 1997 and 2000. Changes in dispersion measure occurring around the 1994, 1997 and 2000 periastrons are measured and accounted for in the timing analysis. There is good evidence for a small glitch in the pulsar period in 1997 August, not long after the 1997 periastron, and a significant frequency second derivative indicating timing noise. We find that spin-orbit coupling with secular changes in periastron longitude and projected semi-major axis ($x$) cannot account for the observed period variations over the whole data set. While fitting the data fairly well, changes in pulsar period parameters at each periastron seem ruled out both by X-ray observations and by the large apparent changes in pulsar frequency derivative. Essentially all of the systematic period variations are accounted for by a model consisting of the 1997 August glitch and step changes in $x$ at each periastron. These changes must be due to changes in the orbit inclination, but we can find no plausible mechanism to account for them. It is possible that timing noise may mask the actual changes in orbital parameters at each periastron, but the good fit to the data of the $x$ step-change model suggests that short-term timing noise is not significant.