Watson and Crick's model for the structure of DNA suggested that replication of the strands of DNA could be accomplished using the rules of base pairing, but did not specify the exact mechanism. Three general mechanisms were considered (Figure 4.13):

1. Conservative–From one parental double-stranded DNA, two "daughter" double-stranded DNAs are made. One contains two new strands and the other contains both of its original strands.

2. Semi-conservative–From one parental double-stranded DNA, two daughter double-stranded DNAs are made. Each daughter DNA contain one parental DNA strand and one newly made strand. (Figure 4.12)

3. Dispersive–From one parental double-stranded DNA, two daughter double-stranded DNAs are made. Each strand in the daughter molecules contains portions of old and newly synthesized material.

In 1958, Matthew Meselson and Franklin Stahl determined which mechanism was actually used (Figure 4.14) when they:

1. Grew cells for many generations in medium containing ¹⁵N (a heavy isotope of nitrogen). This caused double-stranded DNA in these cells to be denser (1.724 g/mL) than normal double stranded DNA (1.710 g/mL) because the heavier nitrogen atoms were incorporated into the bases of the DNA.

2. The cells were then shifted to a medium containing the normal ¹⁴N for one generation. This would allow only one round of DNA replication to take place.

3. The density of the double-stranded DNA obtained from cells after the shift to ¹⁴N for one generation was midway between the density of normal and heavy DNA.

4. When cells were allowed to grow for two generations in ¹⁴N medium, equal amounts of double-stranded DNA with a medium density and normal density were obtained.

5. Comparing the density of single-stranded DNA from these cells (by separating the strands at pH 12) revealed that cells grown in ¹⁴N for one generation contained equal amounts of light and heavy strands.

These observations are all consistent with the semi-conservative model of DNA replication.

Observations 3 and 4, on the other hand, were not consistent with the conservative model of replication and observation 5 was inconsistent with the dispersive model of replication.