In addition to being correlated in space, clusters are also correlated in time. Efremov & Elmegreen (1998) found that the age difference between clusters in the LMC increases with the spatial separation as a power-law, age separation1/2. Elmegreen & Efremov (1996) found the same age-separation correlation for Cepheid variables in the LMC. de la Fuente Marcos & de la Fuente Marcos (2009a) showed that this correlation also applies to clusters in the solar neighborhood. In both cases, the correlation is strongest for young clusters with a separation less than ~ 1 kpc, and it goes away for older clusters. Presumably the young clusters follow the correlated structure that the gas had when the clusters formed. Clusters form faster in regions with higher densities out to a kpc or so, which is probably the ISM Jeans length. This means that small star-forming regions (e.g., cluster cores) come and go during the life of a larger star-forming region (an OB association), and then the larger regions come and go during the life of an even larger region (a star complex). Eventually, the clusters, associations and complexes disperse when they age, taking more random positions after ~ 100 Myr. The correlation is about the same as the size-linewidth relation for molecular clouds (Larson 1981), considering that the ratio of the size to the linewidth is a timescale.
Correlated star formation implies that some clusters should form in pairs. Cluster pairs were discovered in the LMC by Bhatia & Hatzidimitriou (1988) and in the SMC by Hatzidimitriou & Bhatia (1990). An example is the pair NGC 3293 and NGC 3324 near eta Carinae. de la Fuente Marcos & de la Fuente Marcos (2009b) studied these and other pairs. For NGC 3292/3324, the clusters are apparently weakly interacting and the age difference is 4.7 My. NGC 659 and NGC 663 are also weakly interacting and the age difference is 19.1 My. Dieball et al. (2002) determined the distribution function of the number of cluster members per cluster group in the LMC. They found a statistical excess of clusters in pairs compared to the expectation from random groupings.