In a dynamic allocation setup, we experimentally study agents' choices under the First-In-First-Out (FIFO) and Last-In-First-Out (LIFO) queuing protocols. We find that agents are nearly rational under FIFO but tend to be overly selective under LIFO. In the model that anchors our experiment, we show that the magnitude of such an excessively selective bias reduces the welfare performance gap between FIFO and LIFO. Since strategic complementarities can reinforce overly selective behavior, we show