Resale or republication not permitted without written consent of the publisherMar Ecol Prog Ser 317: [237][238][239][240][241][242][243][244] 2006 in field studies has been extremely difficult (Hare & Cowen 1997). For this reason, laboratory experiments have been used to investigate the relative importance of size and age (Leggett & DeBlois 1994). However, the laboratory approach suffers from complications of the artificial constraint of animals within tanks and the focus by researchers on only a portion of the 'predation sequence' (Fuiman 1989), typically predator capture success (Fuiman & Magurran 1994). For logistical reasons, experimental studies generally only examine the interaction between a single species and size class of predator and its prey (Litvak & Leggett 1992, Pepin et al. 1992, Taylor 2003, whereas, in the field, larval fish inhabit an environment that contains numerous species of predators of a great range of sizes. The likelihood of predation for individual members of a cohort will reflect the cumulative effects of all size classes and species of predators within the pelagic environment (Bailey & Houde 1989, Paradis et al. 1996.As the number of predators to which fish are susceptible declines with increasing size (Bailey & Houde 1989), the growth-predation hypothesis predicts that the importance of selective mortality should decline as individuals grow through larval and juvenile phases to become adults. This idea is difficult to test, as it requires representative collections of the same cohort of fish at successive life-history stages, each of which may take months or even years to complete. Furthermore, larvae, juveniles and adults can occupy very different habitats, each entailing specialised gear to attain representative samples. Consequently, no study has been able to examine the relative importance of selective mortality throughout the entire life cycle of a marine fish.Although much of the evidence on the importance of the growth-predation hypothesis has been derived from temperate environments, an increasing number of studies demonstrate that these processes also operate in tropical systems (Searcy & Sponaugle 2001, Bergenius et al. 2002, Vigliola & Meekan 2002, Wilson & Meekan 2002, Meekan et al. 2003, Hoey & McCormick 2004. The apparent ubiquity of sizeselective mortality in fish populations allows predictions of the hypothesis to be tested on tropical fish that are relatively fast growing compared to temperate equivalents. In some of these tropical species, life cycles are completed in only a few months, rather than the years or even decades that are required in temperate environments.The sprat Spratelloides gracilis is a good example of a fast-growing tropical fish. It is an ideal candidate for tests of the predictions of the growth-predation hypothesis, as representative samples of the same cohort can be captured at different stages of a life history that only lasts a maximum of 4 mo (Milton et al. 1991). Here, we use otolith analysis to back-calculate age, size and growth...