The biological carbon pump is a critical component for uptake of carbon by the oceans. Most of this is mediated by gravitational sinking of particles and it is generally assumed that there is a positive relationship between sinking velocities and sizes of particles. Due to the difficulties inherent in measuring the sinking velocity of untouched and undamaged particles, the majority of studies have been based on artificial solid particles or laboratory generated marine snow formed from homogenous material, e.g. phytoplankton cultures. Here, we present results from a newly developed optical method that measures sizes and sinking velocities of undisturbed in situ aggregates in the mesopelagic zone. The measurements were done at depths between 90 and 530 m throughout day and night. In total we measured 55 image sequences of aggregate sizes and sinking velocities, resulting in measurements of 1060 individual aggregates. Only 10 sequences showed significant correlations between aggregate size and sinking velocity. Furthermore, only 8 of the 10 significant correlations between size and settling were positive. Despite similar optical appearance, similar sized aggregates had different sinking velocities, suggesting that aggregates formed in situ are heterogeneous with different compositions, structure and densities. This complicates estimates of sinking velocity from aggregate size only and one has to be careful when estimating downward particle flux based on in situ size-distribution and abundance of aggregates and theoretical relationships between size and sinking velocity. Hence, this study forces the conclusion that estimates of downward particle flux which use particle size distributions must also include information about aggregate composition, compactness and density from in situ optics or direct particle and aggregate sampling.