We analyse the two-point correlation function (2PCF) of galaxy groups identified from the 2-degree Field Galaxy Redshift Survey with the halo-based group finder recently developed by Yang et al. With this group catalogue we are able to estimate the 2PCFs for systems ranging from isolated galaxies to rich clusters of galaxies. The real-space correlation length obtained for these systems ranges from ∼4 to ∼15h−1Mpc, respectively. The observed correlation amplitude (and the corresponding bias factor) as a function of group abundance is well reproduced by associating galaxy groups with dark matter haloes in the standard Λ-cold dark matter model. Redshift distortions are clearly detected in the redshift-space correlation function, the degree of which is consistent with the assumption of gravitational clustering and halo bias in the cosmic density field. In agreement with previous studies we find a strong increase of the correlation length with the mean intergroup separation. Although well-determined observationally, we show that current theoretical predictions are not yet accurate enough to allow for stringent constraints on cosmological parameters. Finally, we use our results to explore the power-law nature of the 2PCF of galaxies. We split the 2PCF into one- and two-group terms, equivalent to the one- and two-halo terms in halo occupation models, and show that the power-law form of the 2PCF is broken, when only including galaxies in the more massive systems.