In 1993, six Canadian populations of Atlantic cod (Gadus morhua) had collapsed to the point where a moratorium was declared on fishing. It has been argued that the collapses were caused by poor recruitment of cod to the fishery. Yet we are unable to detect a difference between the recruitment of year classes that should have contributed most to the spawning stock at the time of the collapse and recruitment levels in earlier years. A power analysis shows that we would have almost certainly detected an overall reduction of recruitment of 20%. There are considerable differences in the abundance trends as determined by research surveys and reconstructed from the commercial catch at age data (called “virtual population analysis” [VPA]) for each stock. VPA-based abundances consistently depict lower recruitment levels than do survey-based estimates in recent years. More important is the observation that from the early 1980s the VPA-based trend shows a decline where none is apparent in the survey-based trend. One explanation of these differences would be an increase in discarding of young fish as fishing mortality increased. We test the hypothesis that the mortality for young cod is unrelated to the fully recruited fishing mortality. This hypothesis is rejected; in each of the six stocks, high juvenile mortality was associated with high adult mortality. This is consistent with the discarding hypothesis. We suggest that age-specific abundance trends estimated from research surveys and VPA should be compared for all stocks where such data exist, and that high priority should be given to the measurement of discarding levels and the extent to which catch misreporting is related to changes in fishing mortality.