The reflex motion of a star induced by a planetary companion is too small to detect by photographic astrometry .
The apparent discovery in the 1960s of planetary systems around certain nearby stars , in particular Barnard ’ s star , turned out to be spurious .
Conventional stellar radial velocities determined from photographic spectra at that time were also too inaccurate to detect the expected reflex velocity changes .
In the late 1970s and early 1980s , the introduction of solid-state , signal-generating detectors and absorption cells to impose wavelength fiducials directly on the starlight , reduced radial velocity errors to the point where such a search became feasible .
Beginning in 1980 , our team from UBC introduced an absorption cell of hydrogen fluoride gas in front of the CFHT coudé spectrograph and , for 12 years , monitored the radial velocities of some 29 solar-type stars .
Since it was assumed that extra-solar planets would most likely resemble Jupiter in mass and orbit , we were awarded only three or four two-night observing runs each year .
Our survey highlighted three potential planet hosting stars , \gamma Cep ( K1 IV ) , \beta Gem ( K0 III ) , and \epsilon Eri ( K2 V ) .
The putative planets all resembled Jovian systems with periods and masses of : 2.5 yr and 1.4 M _ { J } , 1.6 yr and 2.6 M _ { J } , and 6.9 yr and 0.9 M _ { J } , respectively .
All three were subsequently confirmed from more extensive data by the Texas group led by Cochran and Hatzes who also derived the currently accepted orbital elements .

None of these three systems is simple .
All 5 giant stars and the supergiant in our survey proved to be intrinsic velocity variables .
When we first drew attention to a possible planetary companion to \gamma Cep in 1988 it was classified as a giant , and there was the possibility that its radial velocity variations and those of \beta Gem ( K0 III ) were intrinsic to the stars .
A further complication for \gamma Cep was the presence of an unseen secondary star in an orbit with a period initially estimated at some 30 yr .
The implication was that the planetary orbit might not be stable , and a Jovian planet surviving so close to a giant then seemed improbable .
Later observations by others showed the stellar binary period was closer to 67 yr , the primary was only a sub-giant and a weak , apparently synchronous chromospheric variation disappeared .
Chromospheric activity was considered important because \kappa ^ { 1 } Cet , one of our program stars , showed a significant correlation of its radial velocity curve with chromospheric activity .

\epsilon Eri is a young , magnetically active star with spots making it a noisy target for radial velocities .
While the signature of a highly elliptical orbit ( e = 0.6 ) has persisted for more than 3 planetary orbits , some feel that even more extensive coverage is needed to confirm the identification despite an apparent complementary astrometric acceleration detected with the Hubble Space Telescope .

We confined our initial analyses of the program stars to looking for circular orbits .
In retrospect , it appears that some 10 % of our sample did in fact have Jovian planetary companions in orbits with periods of years .