Correlated Errors in Hipparcos Parallaxes toward the Pleiades and the Hyades

We show that the errors in the Hipparcos parallaxes toward the Pleiades and the Hyades open clusters are spatially correlated over angular scales of 2°-3°, with an amplitude of up to 2 mas. This correlation is stronger than expected based on the analysis of the Hipparcos catalog. We predict the parallaxes of individual cluster members, π_pm, from their Hipparcos proper motions, assuming that all the cluster members move with the same space velocity. We compare these parallaxes with their Hipparcos parallaxes, π_Hip, and find that there are significant spatial correlations in the latter quantity. We derive a distance modulus to the Pleiades of 5.58 ± 0.18 mag from the gradient in the radial velocities of the Pleiades members in the direction parallel to the proper motion of the cluster. This value, derived using a geometric method, agrees very well with the distance modulus of 5.60 ± 0.04 mag determined using the main-sequence fitting technique, compared with the value of 5.33 ± 0.06 mag inferred from the average of the Hipparcos parallaxes of the Pleiades members. We show that the difference between the main-sequence fitting distance and the Hipparcos parallax distance can arise from spatially correlated errors in the Hipparcos parallaxes of individual Pleiades members. Although the Hipparcos parallax errors toward the Hyades are spatially correlated in a manner similar to those of the Pleiades, the center of the Hyades is located on a node of this spatial structure. Therefore, the parallax errors cancel out when the average distance is estimated, leading to a mean Hyades distance modulus that agrees with the pre-Hipparcos value. We speculate that these spatial correlations are also responsible for the discrepant distances that are inferred using the mean Hipparcos parallaxes to some open clusters, although an agreement between the mean Hipparcos parallax distance and the main-sequence fitting distance to other clusters does not necessarily preclude spatially correlated Hipparcos parallax errors. Finally, we note that our conclusions are based on a purely geometric method and do not rely on any models of stellar isochrones.