Recent theoretical studies on the population dynamic consequences of cannibalism have focused on mechanisms behind the emergence of large cannibals (giants) in size-structured populations. Theoretically, giants emerge when a strong recruiting cohort imposes competition induced mortality on stunted adults, but also provides a profitable resource for a few adults that accelerate in growth and reach giant sizes. Here the effects of a recruitment pulse on the individual and population level in an allopatric Arctic char population have been studied over a 5-year period and these results were contrasted with theoretical model predictions for the conditions necessary for the emergence of cannibalistic giants. The recruitment pulse had negative effects on invertebrate resource abundance, and the decrease in body condition and increase in mortality of adult char suggested that strong intercohort competition took place. The frequency of cannibalism increased and a few char accelerated in growth and reached ‘giant’ sizes. The main discrepancy between model predictions and field data was the apparently small effect the recruited cohort had on resources and adult char performance during their first summer. Instead, the effects became pronounced when the cohort was 1 year old. This mismatch between model predictions and field observations was suggested to be due to the low per capita fecundity in char and the restricted nearshore habitat use in young-of-the-year (YOY) char. This study provides empirical evidence that the emergence of giants is associated with the breakthrough of a strong recruiting cohort and also suggests that the claimed stable char populations with large cannibals may instead be populations with dynamic size structure that results in intermittent breakthroughs of recruitment pulses, providing the conditions necessary for char to enter the cannibalistic niche. The data suggest that increased recruit survival through restricted habitat use may destabilize dynamics and cause the emergence of giants. However, they also suggest that this does not necessarily develop into populations with bi-modal size structure in populations with low per capita fecundity and size- and density-dependent habitat use of recruiting cohorts.