The molecular clock is a figurative term for a technique that uses the mutation rate of biomolecules to deduce the time in prehistory when two or more life forms diverged. The biomolecular data used for such calculations are usually nucleotide sequences for DNA, RNA, or amino acid sequences for proteins.

The notion of the existence of a so-called "molecular clock" was first attributed to Émile Zuckerkandl and Linus Pauling who, in 1962, noticed that the number of amino acid differences in hemoglobin between different lineages changes roughly linearly with time, as estimated from fossil evidence. They generalized this observation to assert that the rate of evolutionary change of any specified protein was approximately constant over time and over different lineages (known as the molecular clock hypothesis).

The genetic equidistance phenomenon was first noted in 1963 by Emanuel Margoliash, who wrote: "It appears that the number of residue differences between cytochrome c of any two species is mostly conditioned by the time elapsed since the lines of evolution leading to these two species originally diverged. If this is correct, the cytochrome c of all mammals should be equally different from the cytochrome c of all birds. Since fish diverges from the main stem of vertebrate evolution earlier than either birds or mammals, the cytochrome c of both mammals and birds should be equally different from the cytochrome c of fish. Similarly, all vertebrate cytochrome c should be equally different from the yeast protein." For example, the difference between the cytochrome c of a carp and a frog, turtle, chicken, rabbit, and horse is a very constant 13% to 14%. Similarly, the difference between the cytochrome c of a bacterium and yeast, wheat, moth, tuna, pigeon, and horse ranges from 64% to 69%. Together with the work of Emile Zuckerkandl and Linus Pauling, the genetic equidistance result led directly to the formal postulation of the molecular clock hypothesis in the early 1960s.

Similarly, Vincent Sarich and Allan Wilson in 1967 demonstrated that molecular differences among modern primates in albumin proteins showed that approximately constant rates of change had occurred in all the lineages they assessed. The basic logic of their analysis involved recognizing that if one species lineage had evolved more quickly than a sister species lineage since their common ancestor, then the molecular differences between an outgroup (more distantly related) species and the faster-evolving species should be larger (since more molecular changes would have accumulated on that lineage) than the molecular differences between the outgroup species and the slower-evolving species. This method is known as the relative rate test. Sarich and Wilson's paper reported, for example, that human (Homo sapiens) and chimpanzee (Pan troglodytes) albumin immunological cross-reactions suggested they were about equally different from Ceboidea (New World Monkey) species (within experimental error). This meant that they had both accumulated approximately equal changes in albumin since their shared common ancestor. This pattern was also found for all the primate comparisons they tested. When calibrated with the few well-documented fossil branch points (such as the lack of primate fossils of modern aspect found before the K-T boundary), this led Sarich and Wilson to argue that the human-chimp divergence probably occurred only about 4–6 million years ago.

The observation of a clock-like rate of molecular change was originally purely phenomenological. Later, the work of Motoo Kimura developed the neutral theory of molecular evolution, which predicted a molecular clock. Let there be N individuals, and to keep this calculation simple, let the individuals be haploid (i.e. have one copy of each gene). Let the rate of neutral mutations (i.e. mutations with no effect on fitness) in a new individual be ${\displaystyle \mu }$. The probability that this new mutation will become fixed in the population is then 1/N, since each copy of the gene is as good as any other. Every generation, each individual can have new mutations, so there are ${\displaystyle \mu }$N new neutral mutations in the population as a whole. That means that each generation, ${\displaystyle \mu }$ new neutral mutations will become fixed. If most changes seen during molecular evolution are neutral, then fixations in a population will accumulate at a clock-rate that is equal to the rate of neutral mutations in an individual.

To use molecular clocks to estimate divergence times, molecular clocks need to be "calibrated". This is because molecular data alone does not contain any information on absolute times. For viral phylogenetics and ancient DNA studies—two areas of evolutionary biology where it is possible to sample sequences over an evolutionary timescale—the dates of the intermediate samples can be used to calibrate the molecular clock. However, most phylogenies require that the molecular clock be calibrated using independent evidence about dates, such as the fossil record. There are two general methods for calibrating the molecular clock using fossils: node calibration and tip calibration.

Sometimes referred to as node dating, node calibration is a method for time-scaling phylogenetic trees by specifying time constraints for one or more nodes in the tree. Early methods of clock calibration only used a single fossil constraint (e.g. non-parametric rate smoothing), but newer methods (BEAST and r8s) allow for the use of multiple fossils to calibrate molecular clocks. The oldest fossil of a clade is used to constrain the minimum possible age for the node representing the most recent common ancestor of the clade. However, due to incomplete fossil preservation and other factors, clades are typically older than their oldest fossils. In order to account for this, nodes are allowed to be older than the minimum constraint in node calibration analyses. However, determining how much older the node is allowed to be is challenging. There are a number of strategies for deriving the maximum bound for the age of a clade including those based on birth-death models, fossil stratigraphic distribution analyses, or taphonomic controls. Alternatively, instead of a maximum and a minimum, a probability density can be used to represent the uncertainty about the age of the clade. These calibration densities can take the shape of standard probability densities (e.g. normal, lognormal, exponential, gamma) that can be used to express the uncertainty associated with divergence time estimates. Determining the shape and parameters of the probability distribution is not trivial, but there are methods that use not only the oldest fossil but a larger sample of the fossil record of clades to estimate calibration densities empirically. Studies have shown that increasing the number of fossil constraints increases the accuracy of divergence time estimation.

Sometimes referred to as tip dating, tip calibration is a method of molecular clock calibration in which fossils are treated as taxa and placed on the tips of the tree. This is achieved by creating a matrix that includes a molecular dataset for the extant taxa along with a morphological dataset for both the extinct and the extant taxa. Unlike node calibration, this method reconstructs the tree topology and places the fossils simultaneously. Molecular and morphological models work together simultaneously, allowing morphology to inform the placement of fossils. Tip calibration makes use of all relevant fossil taxa during clock calibration, rather than relying on only the oldest fossil of each clade. This method does not rely on the interpretation of negative evidence to infer maximum clade ages.