The internet affords researchers many opportunities, including continuous data collection and coding, and a large, heterogeneous sample. But it also brings a loss of control. Lab-based results can be replicated online (Birnbaum, 2004; Krantz & Dalal, 2000), but questions remain as to what types of research can be replicated, and where potential limitations may exist. We examined effects of location (laboratory vs. web) on object familiarity ratings, looking closely at variability in dependent measures, including response latencies. Our results suggest web-based data collection is reliable, and we provide suggestions for maximizing this reliability.