During the breeding season male humpback whales sing long, complex songs that are thought to be sexual displays3. Songs consist of a series of themes, progressing in a predictable order, that may repeat for several hours4. We used a small observation vessel to find singing humpbacks and conduct focal sampling5, recording behaviour before, during and after playback. (Strictly speaking, we have evaluated the additional impact that LFA sounds have on a singing whale that is already being followed.)

We recorded the vocal behaviour of each focal singer continuously for several hours using a towed, calibrated hydrophone array6. When the whale was at the surface, observers sampled visible behaviour. Photographs of fluke and dorsal fin features confirmed the whale's identity throughout each follow7. At least two songs were recorded before the observation vessel requested the US Navy R/V Cory Chouest to transmit ten (in one case four) 42-s LFA signals at 6-min intervals. The sonar was broadcast at less than full strength, and no focal singer was exposed to a signal louder than 150 dB (with respect to 1 μPa).

Sixteen singers were followed during 18 playbacks. In nine follows, the whale sang continuously throughout the playback; in four the singer stopped when he joined other whales (typical of normal social interaction); and in five the singer stopped, presumably in response to the playback. We recorded at least one complete song in all conditions from six individuals, and pooled the songs of each of the two individuals that were subjects in two experiments. For these six whales, we measured the duration and theme structure of song spectrograms, comparing song duration in the three conditions using analysis of variance8.

On average, humpback whales' songs were 29% longer during LFA playbacks (Fig. 1) — a particularly strong result, given the low power of the test and small sample size9. Song duration returned to normal after exposure, suggesting that this response has a limited duration. There was little difference in the likelihood of an aberrant theme transition across exposure conditions (χ2=3.273, P=0.195), indicating that long songs resulted from longer themes within a normal song structure. Across the six singers, maximum received level of the sonar at the whale did not correlate positively with either the increase in mean song duration from pre-exposure to exposure condition (r=−0.90) or with the subsequent decrease from exposure to post-exposure condition ( r=−0.63).

Figure 1: Duration of songs (±s.e.m.) produced by humpbacks before, during and after exposure to LFA sonar transmissions.
figure 1

(Bold, filled diamonds, mean of all six singers; other symbols, individual singers). The maximum received level of the sonar at the whale ranged from 130 to 150 dB re 1 μPa. Songs were grouped in the exposure condition if a sonar transmission occurred at any point during the song. The average number of songs per singer in the pre-exposure, exposure and post-exposure conditions was 3.2, 4.7 and 3.8, respectively. Differences were assessed using a mixed-model analysis of variance treating exposure condition as a fixed factor, whale identity as a random factor, and each song duration as an independent observation. The effect of exposure condition on song duration was statistically significant at P=0.047 (F2,10=4.200, power=0.50, n=6).

We suggest that humpbacks sang longer songs during LFA sonar transmissions to compensate for acoustic interference. Our study shows that it is possible to measure the behavioural responses of individual whales in controlled experiments at sea.