Where does replication collide?

The process by which a cell copies its double-stranded DNA before it divides: the two strands come apart at a moving junction called the replication fork, and each old strand serves as a template for a new matching strand built by the enzyme DNA polymerase, so every daughter molecule keeps one old strand and one new one. Meselson and Stahl's 1958 The replication of DNA in Escherichia coli proved this semiconservative pattern by labelling the strands with a heavy isotope and watching the density of the copies. The referent is the chemical copying of a genome inside a living cell, not the repeating of a whole study to test its claim, the keeping of data copies across many computers, or the repeated units inside one experiment.[1]

Unit

One duplication of a genome, per cell division

Recovery means

DNA polymerase proofreads as it copies, and mismatch repair corrects errors that slip through before the next round

Measured by

Traced originally by density-gradient labelling, and today by sequencing and by assays that track replication forks along the genome

The repeating of a completed study, with fresh data and often new participants, to find out whether its original finding holds up rather than being a fluke of one sample. A direct replication reruns the same procedure as closely as possible, while a conceptual replication tests the same claim with a different design. The Open Science Collaboration's 2015 Estimating the reproducibility of psychological science repeated one hundred published studies and found many effects weaker or absent, a result that became central to the replication crisis. The referent is the repeat of an investigation to check a claim, not a cell copying its DNA, data mirrored across machines, or the repeated units within a single experiment.[2]

Unit

One repeat of a whole study, with its own fresh data

Recovery means

Not applicable

Measured by

Judged by whether the new result reaches significance in the same direction, or by comparing the new effect size against the original

The keeping of several copies of the same data or service on separate computers, so the system stays available and correct even when some machines crash. In state machine replication every copy processes the same commands in the same order and so stays identical to the others, while primary-backup schemes instead route each update through one leader that the backups follow. Schneider's 1990 Implementing fault-tolerant services using the state machine approach set out the canonical account of this method. The referent is redundant copies of data held across networked computers for fault tolerance, not a genome being copied, a published study being repeated, or the repeated units of an experiment.[3]

Unit

One maintained copy of the data or state, held on one node

Recovery means

A failed copy is rebuilt from the surviving ones, restoring both the number of copies and the agreed state

Measured by

Assessed by the number of copies, the consistency guarantee they meet, and how many simultaneous failures the group can survive

The repeated, independent experimental units that each receive the same treatment inside a single study, giving the repetition needed to estimate how much the outcome varies by chance alone. Fisher's 1935 The Design of Experiments named replication, alongside randomization and local control, as a founding principle of good experiments, because it is what yields a valid estimate of error and sharpens the comparison between treatments. The referent is the repeated units within one experiment that measure variability, not the whole study repeated later, a genome copied inside a cell, or data mirrored across machines.[4]

Unit

One independent experimental unit under a given treatment

Recovery means

Not applicable

Measured by

Counted as the number of replicates per treatment, which sets the degrees of freedom available for the error estimate