High-precision compositional analyses of 54 whole-rock samples from the ad 1783 Laki lava flow field show small but statistically significant variations in trace and major element concentrations and ratios. Strong linear correlations exist between major and trace element concentrations, and variations in incompatible element ratios, such as Zr/Y, are modest. Point-counting results indicate that the lava contains an average of 12 vol. % phenocrysts, with plagioclase, clinopyroxene and olivine present in relative volumetric proportions of 57:32:11. Whole-rock compositions vary linearly with the total mass fraction of phenocrysts in the samples, such that samples with the lowest concentrations of incompatible trace elements have the highest proportion of phenocrysts. On first inspection, such correlations might be interpreted to arise from variable crystal accumulation into the carrier liquid within the magma. However, simple models of crystal accumulation fail to match the relationships between whole-rock composition and phenocryst content. Instead, the phenocrysts must have formed the solid part of a magmatic mush, with the mush liquid being more evolved than the carrier liquid. This mush was entrained into the carrier liquid prior to eruption, with incomplete mixing of the mush into the carrier liquid allowing for the preservation of whole-rock compositional variation. A mathematical description of the mass balance involved in mush mixing is developed to constrain the properties of the mush. Although there is a trade-off between estimates of mush liquid composition and mush porosity, independent constraints on mush liquid composition from phenocryst compositions are used to estimate an average mush porosity of 46-65%. The success of the binary mixing fits to whole-rock compositions indicates that the mean compositions of the mush and the carrier liquid cannot have changed substantially during the eruption. However, more detailed observations reveal that on average the mush proportion was higher during the later stages of the eruption, and this coincides with the presence of primitive high Mg# olivine and clinopyroxene and anorthitic plagioclase primocrysts in the mush. The key observations cannot be accounted for by a model of in situ evolution of mush liquid in the cooling margins of a magma chamber. Instead, the juxtaposition of evolved mush liquid with primitive phenocrysts that is required to generate the mush may perhaps occur as a result of compositional convection at the chamber roof, or alternatively by the partitioning of phenocrysts into more viscous magma during the mixing of primitive basalt and evolved melt in the chamber. It is likely that many porphyritic basaltic eruptions carry disaggregated mush and it is straightforward to apply the methods described in this study to other eruptions, allowing for future improvements in the characterization of the properties of mushes in basaltic magma chambers.