A dolphin hunts with a shell, its calf picks it up and does the same three minutes later

A fish takes refuge in a large empty shell lying on the bottom. A dolphin works its rostrum into the opening, lifts the shell out of the water, shakes it, and the fish drops into its mouth. This behaviour is called shelling, it had never been observed anywhere other than at Shark Bay, in Western Australia, and a note published open access on 8 August 2026 in Marine Mammal Science documents it five thousand kilometres away.

Exactly what was observed

Alexis Levengood and colleagues at the University of the Sunshine Coast have been running regular boat-based observation campaigns since 2022 in the Great Sandy Marine Park, on the Queensland coast, on Butchulla country. Two episodes were recorded in 2025, involving three different dolphins.

6 July 2025, 7.50 a.m.	an adult female, Stevie Nicks, positions a Melo amphora shell at the surface with her rostrum and lifts it entirely out of the water over three surfacings, then lets it go. She is alone, the nearest dolphin is 510 m away
1st October 2025, 9.01 a.m.	a mother, Maserati, surfaces with a shell beneath a drone. A fish escapes from it, she lets the shell go and gives chase
1st October 2025, 9.04 a.m.	three minutes later, her calf, Bentley, surfaces with that same shell, at exactly the same spot, its rostrum worked into the opening as its mother’s was, and strikes it against the surface
Depths	6.1 m and 5.4 m, a glassy sea in both cases
Duration of the observations	9 minutes and 29 minutes

Two details are worth noting. The body and tail of a fish, probably a lizardfish of the genus Synodus, can be seen emerging from the opening of the shell while the calf lifts it. And in neither of the two episodes were the researchers able to confirm that the manoeuvre had succeeded: rain interrupted the first follow, turbidity masked the end of the second.

The detail that changes the reading

The shorthand now in circulation, “the calf imitates its mother”, says more than the filmed sequence does, and that sequence has to be set out exactly.

The mother did not pull it off: the fish escaped, she abandoned the shell to give chase, and she never came back for it. Three minutes later, the calf picks up this object left on the spot and handles it as it has seen done.

It then let go of it, surfaced without it, then took it up again some thirty seconds later before releasing it after seven seconds and moving away. The authors note that it did not resume its infant position against its mother’s flank until the sequence was over.

What can be seen, then, is a young animal attempting a technique it has just watched fail, with the tool its mother left behind, and not managing it. That is more modest than a successful transmission, and more interesting.

The behaviour is not new in Hervey Bay

One element of the note deserves emphasis, because it runs counter to the idea of a sudden discovery. Two earlier sightings appear in the paper, made from a tourist watching boat in the bay, with photographs in support: one in 2013, the other in 2019, in the same area.

They are placed on record thanks to the permanent presence of a photographer on board, and two of the seven authors of the note belong to that watching operation. The behaviour has therefore existed in this bay for at least twelve years; what was missing was a publication.

There was also, from 2011, an anecdotal mention of shelling in south-east Queensland, with no image or data to go with it. The note turns that report into a documented fact. It is a good example of what non-academic field observation produces, and our page on how to take action lists the schemes that exist.

What was known about Shark Bay

Comparing the two sites means recalling the earlier state of knowledge, which is more nuanced than is usually said.

At Shark Bay, shelling has been observed in fewer than 2% of the population, a figure the authors of the time judged to be an underestimate, the manoeuvre lasting only a few seconds. The same population also practises sponging, a far more demanding technique that we covered in a separate article, and which involves Indo-Pacific bottlenose dolphins.

Only one study has modelled the way shelling spreads, in 2020. It concluded that diffusion through the social network was by far the best supported explanation, with a ratio of 10 to more than 300 in favour of social transmission rather than individual learning or a genetic basis. And it placed that transmission outside the mother-calf bond, between peers or from unrelated adults.

One detail of that same study has drawn little comment and now takes on relief: the behaviour appeared only in individuals carrying 2 haplotypes out of 8, in which learning appeared to be about fifty times faster. In other words, a maternal lead was already present in the data, without being taken up.

Vertical or horizontal, and why it matters

A socially learned behaviour can circulate in two ways, and they do not have the same consequences.

Vertically, from mother to calf, during the years of dependence. The technique then follows the maternal lineages, spreads slowly and persists for a long time. That is the mode of transmission of sponging.

Horizontally, between individuals of the same age, or obliquely from unrelated adults. The technique can then cross a group in a few seasons, like a fashion, and disappear just as fast.

The observation of 1^st October is the first compatible with the vertical route, and the authors liken it to a strategy described elsewhere under the name “do as mother does”. They recall that the prolonged dependence of the young dolphin, several years in near-permanent contact with its mother, offers countless occasions for this, as our page on lactation in cetaceans sets out.

Three readings remain possible

No video can decide between what follows, and the authors do not claim otherwise.

1. Observational learning. The calf watched its mother do it and reproduced the handling. This is the reading argued for in the note.

2. Local enhancement. This is the most serious competing hypothesis, and the detail of the sequence gives it weight. By handling and then abandoning the shell, the mother drew her calf’s attention to a particular object, in a particular place. It can then pick it up without copying anything of the action. This mechanism, well documented in other animals, produces exactly the same picture.

3. Exploratory handling. Young dolphins carry, push and turn over everything they find, seaweed, shells, debris. A calf that picks up a floating shell demonstrates nothing on its own.

Deciding between them would require following who practises the behaviour, who associates with whom, and checking that the technique spreads along social ties rather than at random. That means dozens of observations over years, not two.

An unexpected sex bias

The three animals concerned are two adult females and a juvenile of unknown sex. On so small a number, that proves nothing, but the authors note it because the Shark Bay study had found no sex bias for shelling.

The explanation put forward is worth knowing, because it holds for all feeding specialisations. Males devote a large part of their time to courting females and, in some populations, to maintaining alliances, time-consuming activities that sit badly with a demanding technique tied to particular places. Females, for whom the energy cost of reproduction is considerable, invest more in feeding. Our pages on social cohesion and cetacean diets set out these trade-offs.

The limits, as the authors set them out

They are stated without evasion in the note, and they have to be restated just as plainly.

The sample is two events. The authors write explicitly that the small size of the sample calls for caution and that their results do not demonstrate vertical transmission.

No manoeuvre was confirmed as successful. Not for the lone female, nor for the mother, nor for the calf. What is observed is a technique attempted, not a technique that feeds.

The actual frequency remains unknown. The authors consider the behaviour probably under-reported in the bay, for want of survey effort and because it lasts only a few seconds.

The identification of the species remains generic. The note speaks of bottlenose dolphins without settling the question of the subspecies, unlike the Shark Bay work, which deals explicitly with Tursiops aduncus.

An innovation that appeared twice

The conclusion of the authors is more precise than the summary that has been made of it. They do not say that the behaviour spread from one coast to the other: they see in it an argument in favour of an independent innovation, which appeared separately in the west and in the east of the continent.

This reading is consistent with a material constraint that is rarely noted. Shelling requires a shell large enough for a fish to take refuge in it and for a dolphin to be able to lift it, which restricts the technique to a few very large gastropods. A population perfectly capable of inventing the action will never do so if nothing of the kind is lying about on the bottom.

The authors propose, finally, to reclassify shelling. Long described as opportunistic, it would in fact meet the criteria of a feeding specialisation: it involves only a fraction of the population, it is repeated over time by the same individuals, and it requires knowing where to find the shells and how to handle them. Our pages on the brain and cognition, the cultural clans of sperm whales and our video on tool use by dolphins place this file within the body of traditions documented in cetaceans, alongside the killer whales that groom one another with seaweed.

Source: Levengood A. L., Hume G. V., Buckman J. M., Steel I., Toohey B. E., Lynch P. and Smith C., “Observations of Shelling Behavior and Potential Evidence of Vertical Social Learning by Bottlenose Dolphins in the Great Sandy Marine Park, Queensland Australia“, Marine Mammal Science, 42(4), e70252, 8 August 2026, open access.