Biogenic reefs are an important component of aquatic ecosystems where they enhance biodiversity. These reefs are often established by dense aggregations of a single taxa and understanding the fundamental principles of biogenic reef formation is needed for their conservation and restoration. We tested whether substratum type and density affected the aggregation behaviour of two important biogenic-reef forming species, the horse mussel, Modiolus modiolus (Linnaeus, 1758), and the blue mussel, Mytilus edulis (Linnaeus, 1758). First, we tested for effects of substratum type on mussel movement and aggregation behaviour by manipulating substrata available to mussels in mesocosms (three treatments: no sediment added, sediment added, sediment and shells added). Both mussel species moved less in treatments with sediment and with both sediment and shells present than when no sediment or shells were added and the percentage of these mussels that aggregated (clumps of two or more individuals) was lower when shells were present compared to treatments without shells, however, fewer M. modiolus attached to shells than M. edulis. There was no effect of different substratum type on patch complexity of either mussel species. In addition, motivated by our interest in the restoration of M. modiolus, we also tested experimentally whether the aggregation behaviour of M. modiolus was density-dependent. M. modiolus moved a similar distance in three density treatments (100, 200 and 300 mussels m), however, their aggregation rate appeared to be greater when mussel density was higher, suggesting that the encounter rate of individuals is an important factor for aggregation. M. modiolus also formed aggregations with a higher fractal dimension in the high and medium density treatments compared to lower density, suggesting that at higher density this increased patch complexity could further facilitate increased recruitment with the enhanced habitat available for settlement. These findings add to the growing evidence showing that adding dead shells to substratum to encourage M. modiolus restoration is not likely to be effective. Our findings suggest that mussel density is a more important driver for patch and subsequently reef formation. Moreover, two seemingly functionally similar mussel species showed some differences in their behaviour (e.g. attachment to shells). This highlights the importance of considering the specific ecology of a target species, such as M. modiolus, when designing restoration methods because information garnered from experimentation on an ostensibly similar species (e.g. M. edulis) may not be appropriate.