Distance sampling along a line transect is used commonly for monitoring changes of birds' abundance at sea. A critical yet rarely tested assumption of line-transect-sampling theory is that all birds along the transect line (i.e., directly in front of the boat) are detected or that probability of detecting a bird on the line can be estimated. As part of a long-term research and monitoring program for the Kittlitz's Murrelet (Brachyramphus brevirostris), we tested the assumption of complete detection of murrelets on the water along a transect line directly in front of a moving boat. Following standard survey procedures, we approached groups of murrelets (n = 57) at sea and recorded their distance, response (diving or flying), and duration of response. Flying murrelets (n = 27) were easily detected, but diving birds (n = 30) were more difficult to detect because of the duration of their dive. The probability that a bird dove and remained underwater long enough to avoid detection was low because birds that dove more than 150 m from the boat surfaced before the boat passed whereas birds that “waited” to dive near the boat were easily detected prior to diving. The greatest probability of nondetection was for birds diving at 55 m (diving long enough for the boat to pass) but was only 0.032 ± 0.007 (P SE). These experiments quantifying detection probability along the transect line could be applied to any species surveyed from a boat.