The Peruvian anchovy or anchoveta (Engraulis ringens) supports the highest worldwide fishery landings and varies in space and time over many scales. Here we present the first comprehensive sub-mesocale study of anchoveta distribution in relation to the environment. During November 2004, we conducted a behavioural ecology survey off central Peru and used a series of observational and sampling tools including SST and CO2 sensors, Niskin bottles, CTD probes, zooplankton sampling, stomach content analysis, echo-sounder, multibeam sonar, and bird observations. The sub-mesoscale survey areas were chosen from mesoscale acoustic surveys. A routine coast-wide (∼2000 km) acoustic survey performed just after the sub-mesoscale surveys, provided information at an even larger population scale. The availability of nearly concurrent sub-mesoscale, mesoscale and coast-wide information on anchoveta distribution allowed for a unique multi-scale synthesis. At the sub-mesoscale (100s m to km) physical processes (internal waves and frontogenesis) concentrated plankton into patches and determined anchoveta spatial distribution. At the mesoscale (10s km) location relative to the zone of active upwelling (and age of the upwelled water) and the depth of the oxycline had strong impacts on the anchoveta. Finally, over 100s km the size of the productive area, as defined by the upwelled cold coastal waters, was the determining factor. We propose a conceptual view of the relative importance of social behaviour and environmental (biotic and abiotic) processes on the spatial distribution of anchoveta. Our ecological space has two y-axis; one based on self-organization (social behaviour), and the other based on the environmental processes. At scales from the individual (10s cm), to the nucleus (m), social behaviour (e.g. the need to school) drives spatial organization. At scales larger than the school, environmental forces are the main driver of fish distribution. The conceptual ecosystem models presented in this paper may provide the final links needed to develop accurate forecasts of the spatial distribution of anchoveta over multiple scales. © 2008 Elsevier Ltd. All rights reserved.