Lysates of P1 from all hosts tested contained at least three morphological variants with respect to head size. These were termed “big” (P1B), “small” (P1S), and “minute” (P1M). Since successive clonings of plaques isolated on many different hosts failed to change the proportions of the variants, we concluded that the production of variants was a function of the P1 genome rather than that of the host. In the electron microscope, the heads appeared to be icosadeltahedra, having face-to-face head diameters of 86 ± 2 nm, 65 ± 2 nm, and 47 ± 2 nm. Assuming the head capsids to be composed of the same protein subunits, these diameters were compatible with T = 16, 9, and 4 with a lattice constant (intercapsomere distance) of 12 to 13 nm. The tails of all variants were morphologically indistinguishable. Each consisted of a hollow tail tube surrounded by a contractile sheath attached to the head by means of a “head-neck connector” which could be a specialized vertex capsomere. In CsCl gradients, a number of bands were observed. One band contained the majority of P1B particles and 99% of the plaque-forming units. Two other bands contained P1S particles whose densities suggested a content of about 40 and 60% of the complete P1B genome. The less dense of these two bands also contained defective P1B particles with a calculated content of only 60% of the complete genome. The P1S particles tested injected their deoxyribonucleic acid (DNA) into host cells and killed them. Genetic markers contained in this band could be rescued by infectious P1B particles, confirming the evidence of Ikeda and Tomizawa that this fraction contains P1 DNA.