Macroaggregation is a highly dynamic attribute of soils that is claimed to have a significant impact on their ability to store C and conserve nutrients. A major obstacle to the description and modelling of macroaggregate dynamics, and of the associated processes, is an almost complete ignorance of the real origin of the different types of aggregates found in soils, their turnover times and positions in the soil matrix. We present here a general methodological approach in which the origin of aggregates separated according to visual criteria could be determined by comparing their specific organic matter signatures assessed by Near Infrared Spectrometry (NIRS) to signatures of biogenic structures produced by soil ecosystem engineers (invertebrates and roots) living in the same soil. Macroaggregates and other soil components were separated visually from samples taken at 61 locations regularly distributed across a watershed in Nicaragua and representing crops, pastures, forests, coffee plantations and fallows. Coinertia analyses among soil macroinvertebrate communities and the matrix of soil morphological variables showed highly significant relationships. In Amazonian forest patches and pastures from the state of Pará in Brazil, 75 different types of biogenic structures were collected at the soil surface and on tree trunks, and analysed by the NIRS spectral method. Significant differences among the different types of structures allowed grouping according to their broad phylogenetic origin with large interspecifc differences. In a field experiment conducted at the same site, soils previously under pastures were planted in 16 possible combinations of four plant species, in a fully randomized design replicated three times in different sites. Surface casts of the earthworm species Andiodrilus pachoensis and soil macroaggregates separated by our visual technique had significantly different spectral signatures depending on the location of the plot and the composition of plant cover. However, the comparison of NIRS signatures of soil macroaggregates and casts suggested that Andiodrilus pachoensis was not responsible for the production of the biogenic aggregates that comprised a large proportion on the soil volume in this soil.