The small black fish swim around the tank; gracefully gliding back and forth with little discernible pattern.
Then, barely audible over the watery whirr of the marine laboratory, a low buzz sounds.
And a few seconds later, the scene is transformed as the fish crowd into a small area, corralling themselves into a tightly packed ball.
They are awaiting a reward.
"Now we drop some food into the feeding tube," says Scott Lindell, who is running the experiment.
"And voila, they are satiated and happy."
Mr Lindell is the director of the scientific aquaculture programme at the Marine Biological Laboratory (MBL) in Woods Hole, Massachusetts. And for the last few years, he has been training fish.
It takes about four weeks to train a black sea bass
It's a Pavlovian response, he explains.
Just as 19th Century scientist Ivan Pavlov conditioned dogs to associate the sound of a bell with the prospect of food, Mr Lindell's black sea bass associate a 280Hz tone emitted from an underwater speaker with food, and respond by gathering in a partitioned "feeding zone".
"It takes about three or four weeks to train them," says Mr Lindell.
But why train a fish?
Currently, offshore fish farms use cages or nets to enclose fish, where they are regularly fed and then harvested.
But Mr Lindell believes "acoustic fish ranching" could offer another option.
He explains: "We are trying to train fish so that we don't have to use cages to confine them, so that they can forage for natural food as well as the food that we offer them.
"It would keep them in a general area, and you could eventually recapture them and take them to market.
"This could lower the cost of fish farming."
To test this, last year Mr Lindell moved his experiments out to the ocean - along with about 4,200 tagged black sea bass.
But not everything went to plan.
The black sea bass were trained up once they were in the aquadome
Mr Lindell says: "We stocked the fish into what we call an aquadome: basically a hemisphere anchored to the sea floor - it's about 32ft across and 16ft high (10m by 5m)."
At first, the scientists covered the dome with a tight mesh, locking the fish inside while they underwent their rigorous sound-training regime with the help of an underwater speaker. This took about four weeks.
"At that point, we took some of the 1in-mesh (3cm) off the cage, and replaced it with a 3in-mesh (8cm)," Mr Lindell continues.
"And that allowed some of our smaller fish to swim out, and it kept most of the predators from getting into the cage."
Bluefish proved to be deadly predators
But there was one exception.
At first, the fish began to forage outside of the aquadome, moving in and out at the prompt of the sound, just as the researchers had hoped.
"But then we start seeing these bluefish circling our cage. And these are notorious for being ravenous and ruthless hunters," he says.
"Very frustratingly, we went back day after day to find these fish still showing up at the cage, and we couldn't for the life of us call the black sea bass back.
The tags helped the researchers to identify their bass
"They were scared to death - we went diving, and we could see them amongst the rocks, but nothing was going to make them run that gauntlet between the rocks and the cage when it would put their lives at risk."
And the fish had good reason to be scared.
When the team caught one of the bluefish and slit open its belly, they discovered 12 tiny tags - the fish that they had been attached to had already been digested.
Mr Lindell adds: "It goes to show that we can provide all sorts of positive enforcement, but it only takes a little bit of negative enforcement to undo it."
But the underwater massacre has not put Mr Lindell off.
He explains: "It is the same problem that sheep farmers have - if you don't have fencing, then they need to have some kind of protection against predators. So with sheep, people have sheep dogs.
"We need to find the equivalent in terms of what would protect our 'herd of fish' if they were not protected by the confines of the cage."
Could robotic fish research help with the fish ranching experiments
And a shark might just be the answer, he says. Something like mako shark that has a taste for bluefish.
He says: "I can't imagine us training a mako shark the way you would train a killer whale.
"But it could be as simple as having a fibreglass, mako shark look-alike that is mechanically driven, that could swim around and around the cage, keeping the bluefish away."
This summer, Mr Lindell and his team returned to the spot where the aquadome experiment had taken place.
They wanted to see if any of his black sea bass had survived the onslaught of the bluefish and if they still associated the sound with the prospect of a tasty reward.
He says: "We submerged our speaker and produced a noise a couple of times.
"But we only had about 4ft (1m) of visibility, and to the best of our knowledge we did not see any black sea bass streaming back a year later as I'd hope. Although the odds of those fish being in that spot a year later would have been small."
We are now asking what fish species in what geographical areas could be more suitable for this application
Despite the problems he experienced during last year's experiment, Mr Lindell still believes trained fish could benefit fish farms.
And as well as looking into predator defence, he is also looking at the feasibility of training different fish and employing acoustic ranching in different locations.
An experiment that he carried out this summer revealed that a Pavlovian response could be prompted in wild striped bass - and he believes other fish can also be trained.
He says: "We are now asking what fish species in what geographical areas could be more suitable for this application.
"It might be that this makes more sense in reservoirs, for instance, which have a more limited flora and fauna - you wouldn't have to worry so much about intrusion of predators."
He adds: "I think it definitively has some possibilities in terms of fisheries management or aquaculture but it has to be the right place and we obviously have to avoid potential problems with predators."
This page is best viewed in an up-to-date web browser with style sheets (CSS) enabled. While you will be able to view the content of this page in your current browser, you will not be able to get the full visual experience. Please consider upgrading your browser software or enabling style sheets (CSS) if you are able to do so.