Prior testing can facilitate subsequent learning, a phenomenon termed the forward testing effect (FTE). We examined a metacognitive account of this effect, which proposes that the FTE occurs because retrieval leads to strategy optimizations during later learning. One prediction of this account is that tests that require less retrieval effort (e.g., multiple-choice relative to cued-recall) should lead to a smaller benefit on new learning. We examined the impact of interpolated multiple-choice or cued-recall testing (relative to no prior testing) on new learning of a four-section STEM text passage. The effect sizes associated with the FTE were numerically, though not significantly larger when the prior tests were cued-recall than multiple-choice, but only when interpolated judgments of learning were not queried. Further, when multiple-choice tests were made more difficult through lure similarity, the FTE was similarly increased. Finally, the FTE was eliminated entirely when participants provided four JOLs after reading each text section. We believe this elimination of the FTE stemmed from an increase in performance for the control participants induced by reactivity from repeated metacognitive queries requiring deep metacognitive reflection. Taken together, these experiments support a metacognitive account of FTE and have important implications for how educators and students should employ retrieval practice and leverage the benefits of metacognitive reflection to improve new learning.