1. Parents often face an investment trade-off between either producing many small or fewer large offspring. When environments vary predictably, the fittest parental solution matches available resources by varying only number of offspring and never optimal individual size. However when mismatches occur often between parental expectations and true resource levels, dynamic models like multifaceted parental investment (MFPI) and parental optimism (PO) both predict offspring size can vary significantly. MFPI is a “realist” strategy: parents assume future environments of average richness. When resources exceed expectations and it is too late to add more offspring, the best-case solution increases investment per individual. Brood size distributions therefore track the degree of mismatch from right-skewed around an optimal size (slight underestimation of resources), to left-skewed around a maximal size (gross underestimation). Conversely, PO is an “optimist” strategy: parents assume maximally good resource futures and match numbers to that situation. Normal or lean years do not affect “core” brood as costs primarily fall on excess “marginal” siblings who die or experience stunted growth (producing left-skewed distributions). 2. Investment patterns supportive of both MFPI and PO models have been observed in nature, but studies that directly manipulate food resources in order to test predictions are lacking. Ant colonies produce many offspring per reproductive cycle, and are amenable to experimental manipulation in ways that can differentiate between MFPI and PO investment strategies. 3. Colonies in a natural population of a harvester ant (Pogonomymex salinus) were protein-supplemented over two years and mature sexual offspring were collected annually prior to their nuptial flight. 4. Several results support either MFPI or PO in terms of patterns in offspring size distributions and how protein differentially affected male and female production. Unpredicted by either model, however, is that supplementation affected distributions more strongly across years than within (e.g., small females are significantly rarer in the year after colonies receive protein). 5. Parental investment strategies in P. salinus vary dynamically across years and conditions. Finding that past conditions can more strongly affect reproductive decisions than current ones, however, is not addressed by models of parental investment.