It is possible that a given lot of ammunition fired at the same aim point by two different tanks will produce center of (shot) impacts, COIs, that differ in magnitude by several times the lot's dispersion. It is difficult to discern what fraction of this variation is due to barrel differences, alone, since changing tubes alters both the mounting conditions and the occasion. A means has been devised to 'change tubes' without altering the mount, or the occasion, by merely changing the shape of a given tube within the same mount. This is accomplished by localized control of the barrel's axial thermal expansion, implemented through a series of temperature-controlled heating pads adhered to the outer barrel wall. Using this technique, it was found that a simple, yet very common, bow-shaped curvature to the right vs. left, for example, produced a significant shift in the COl, the magnitude of which was on the order of the ammunition dispersion. In particular, inducing a right bow moved the muzzle to the left, as expected, but had the opposite, and perhaps unexpected, effect on the shot impacts--moving them to the right. Conversely, forming a left bow moved the muzzle to the right and the shot impacts to the left. An attempt to gain an understanding of the observed COl-centerline dependence through modeling the effects of projectile-barrel interaction (using the gun dynamics code Rascal) did not succeed. It appears that a more comprehensive model is needed before a predicative capability is possible.