Much of our knowledge is transmitted socially rather than through firsthand experience. Even our memories depend on recollections of those around us. Surprisingly, when people recall memories with others, they do not reach the potential number of items they could have recalled alone. This phenomenon is called collaborative inhibition. Recently, Luhmann and Rajaram (2015) analyzed the dynamics of collaborative inhibition at scale with an agent-based model, extrapolating from previous small-scale laboratory experiments. We tested their model against human data collected in a large-scale experiment and found that participants demonstrate non-monotonicities not evident in these predictions. We next analyzed memory transmission beyond directly interacting agents by placing agents into networks. Contrary to model predictions, we observed high similarity only within directly interacting pairs. By comparing behavior to model predictions in large-scale experiments, we reveal unexpected results that motivate future work in elucidating the algorithms underlying collaborative memory.