01The field. Stellar polarimetry from McCall (1990), re-cross-matched to Gaia DR3, gives a cluster-distance field at PA ≈15°. An independent Planck 353 GHz dust-emission measurement agrees to within 2° — the direction is secure.
02Depth separation. Because every polarization sightline now has a Gaia parallax, foreground dust (≈130°) can be separated from the M8-distance field. The two point in genuinely different directions. The intermediate distance bins hold only two or three stars each, so this is best read as two orientation regimes with a transition near ~900 pc, not as a smooth rotation.
03The motions. Peculiar velocities of 980 members show anisotropic expansion (preferentially in declination) with rotation comparable to the expansion — a dynamically active, feedback-shaped cluster.
04The tempting result. With rotation removed, the strain axis lies 28.2° from B, against 45° expected by chance, and the axis is tightly determined (166.8° ± 2.3°). It reads as alignment. It is not: one axis compared with one other axis lands this close or closer with probability p = 0.31, however precisely it is measured. No improvement in astrometry could rescue it.
05The other result, and its control. Individual stars really do move across B, strongly so (⟨cos 2Δθ⟩ = −0.285). But their velocities fill an ellipse 1.83:1 with its major axis at 90° ± 2° — nearly across B — and that shape alone predicts −0.254, or 89 per cent of the effect. The cluster is also elongated at ~110°. So the perpendicular preference is a statement about the shape of the velocity distribution, not about the field. Nothing here says the field is irrelevant to how M8 formed; it says these data cannot demonstrate it. At b = −1.3° the field also lies only 14° from the Galactic plane, so field and Galaxy cannot be told apart on this line of sight.
Stars are shown at their true relative positions and scale over the cluster (0.25° ≈ 5.7 pc; scale bar); motions are amplified for visibility and are not plate-solved to individual image counterparts. All position angles and the expansion/rotation rates are measured values from the analysis.