Comets formed as icy planetesimals in the outer solar nebula. Drag forces due to non-Keplerian rotation of the nebular gas induced relative motions of solid particles. They grew by collisional coagulation, as gravitational instability was inhibited by shear-induced turbulence and the dispersion of radial velocities. The drag-induced radial velocity was size-dependent but did not vary monotonically; the peak velocity was reached by objects about one meter in size, and decreased at larger sizes. This behavior had interesting consequences for the structure of accreting bodies. Bodies of comparable size had low relative velocities and rarely collided; more typically bodies grew by sweeping up others much smaller than themselves. Numerical modeling of this growth process suggests that meter-sized bodies primarily accreted sub-mm aggregate particles that impacted at relatively high speeds (tens of m/sec). Larger bodies in the range of 10 m to kilometers tended to accrete objects with masses about 2 or 3 orders of magnitude smaller than their own mass. During this stage of growth impact velocities decreased with size, reaching a minimum of order 10 cm/sec. For bodies larger than a few km, impact velocities increased as gravitational interactions became more important than drag forces. This growth process would have tended to produce fairly uniform compact structure at sub-meter scale, with macroscopic voids and fractal-like structure at larger sizes. The components were assembled most gently at sizes of tens to hundreds of meters, and comets are most likely to preserve primordial structure on this scale. The greatest uncertainty in relating potentially observable structure to the formation process is our lack of knowledge of the impact strength of primitive icy bodies.