In order to understand the formation mechanism of the disks around Be stars it is imperative to have a good overview of both the differences and similarities between normal B stars and the Be stars. Here we investigate a previous report that there may be a large population of sub-arcsecond companions to Be stars. We present the first systematic, comparative imaging study of the binary properties of matched samples of B and Be stars observed using the same equipment. We obtained high angular resolution (0.07-0.1 arcsec) K band Adaptive Optics data of 40 B stars and 39 Be stars. The separations that can be probed range from 0.1 to 8 arcsec (corresponding to 20-1000 AU), and magnitude differences up to 10 magnitudes can in principle be covered. We detect 11 binaries out of 37 Be targets (corresponding to a binary fraction of 30 +/- 8%) and 10 binaries out of 36 B targets (29 +/- 8%). Further tests demonstrate that the B and Be binary systems are not only similar in frequency but also remarkably similar in terms of binary separations, flux differences and mass ratios. We find that any hypotheses invoking binary companions as responsible for the formation of a disk need the companions to be closer than 20 AU. Close companions are known to affect the circumstellar disks of Be stars, but as not all Be stars have been found to be close binaries, the data suggest that binarity can not be responsible for the Be phenomenon in all Be stars. Finally, the similarities of the binary parameters themselves also shed light on the Be formation mechanism. They strongly suggest that the initial conditions that gave rise to B and Be star formation must, to all intents and purposes, be similar. This in turn indicates that the Be phenomenon is not the result of a different star formation mechanism.

The origin of the disks around Be stars is still not known. Further progress requires a proper parametrization of their structure, both spatially and kinematically. This is challenging as the disks are very small. Here we assess whether a novel method is capable of providing these data. We obtained spectro astrometry around the Pa beta line of two bright Be stars, alpha Col and zeta Tau, to search for disk signatures. The data, with a pixel to pixel precision of the centroid position of 0.3..0.4 milliarcsecond is the most accurate such data to date. Artefacts at the 0.85 mas level are present in the data, but these are readily identified as they were non-repeatable in our redundant datasets. This does illustrate the need of taking multiple data to avoid spurious detections. The data are compared with simple model simulations of the spectro astrometric signatures due to rotating disks around Be stars. The upper limits we find for the disk radii correspond to disk sizes of a few dozen stellar radii if they rotate Keplerian. This is very close to observationally measured and theoretically expected disk sizes, and this paper therefore demonstrates that spectro-astrometry, of which we present the first such attempt, has the potential to resolve the disks around Be stars.