Raven roosts are mobile information centres

1. What they did

In winter, young non-breeding common ravens in the forests of Maine wander widely and feed on large carcasses, which are rich but short-lived and scattered. At night they gather in communal roosts. An old idea, the “information centre” hypothesis, suggests that animals that do not know where food is can learn it from those that do, by following them from a shared roost or colony. Until this study, no one had shown convincingly that roosts work this way.

Over three winters in western Maine, the researchers:

• Provided food. They placed carcasses (moose, cattle, horses, deer and meat piles) in the forest and watched ravens find and feed on them, as well as at naturally occurring carcasses and refuse dumps.
• Marked birds. They tagged 357 ravens with coloured, numbered wing tags and fitted 28 with tail-mounted radio transmitters.
• Watched roosts. From observation trees up to 1 km from roosts, they recorded the direction, timing and numbers of ravens arriving in the evening and leaving at dawn, and noted social soaring displays and roost shifts.
• Released naive birds into roosts. They held 29 wild ravens in aviaries so they did not know of current food, then released 14 near a forming roost and 15 controls away from the roost (in space or in time), and checked who found food next morning.
• Released knowledgeable birds at new food. They released 40 birds (15 singly and 5 groups of five) beside new carcasses just before sunset, and followed the radio-tagged ones to see whether they joined roosts and led others back.

2. What they found

Roosts and departures

• All 15 roosts were in dense white pine stands. Each area had a main grove and several others within 2 km.
• Ravens arrived at roosts over up to an hour, from many directions. They left at dawn as tightly synchronized groups, usually all within about 7 minutes, after a few “kaws” and then “honking” calls. In most cases they all left in one direction, and that direction often changed from day to day.

Roosts move with the food

• When ravens kept feeding at the same carcass, they usually stayed at the same roost (52%) or moved to a nearby grove (38%). When they changed feeding sites, they usually moved to a distant roost (74%).
• Radio-tagged birds stayed at a roost for about 4 nights on average before moving 1 to 12.5 km. In 85% of cases they roosted within 1 km of their new feeding site.

Social soaring

• Large groups sometimes soared in long circuits around the roost, covering at least 200 km², and drew in ravens from at least 10 km away.
• Soaring was rare when the roost stayed put (8%), but common when it shifted nearby (49%) or far away (55%). It peaked on the night after a large group first arrived at a carcass, as the birds moved to roost near it.

Recruitment overnight

• Once a carcass was found, the number feeding often more than doubled overnight. In 18 of 24 increases of 10 or more birds, more ravens arrived in the morning than could have known about the food. The 14 significant increases took groups from about 14 knowledgeable birds to about 33.

Release experiments

• All 14 naive birds released into roosts followed roost-mates to food the next morning, some from up to 6.7 km away. Only 4 of 15 control birds found the food that morning, all released less than 2 km from it.
• On nine occasions, a radio-tagged bird joined a roost that was feeding at a known site it had never visited, and every time it followed roost-mates to the food next morning.
• Of the knowledgeable birds released at new food, 10 joined a roost. Four of those returned to their food, and three brought roost-mates with them (1, 3 and 17 days later). In one case, 30 ravens arrived before dawn the morning after a release, flying straight from a roost 2 km away, with the released bird among the first five.
• The same birds could be leaders one day and followers another.

3. What they concluded

• Raven roosts work as information centres: naive birds follow knowledgeable roost-mates to new food the next morning.
• Unlike the stable colonies studied before, these roosts are mobile. Ravens shift roosts to be near newly found carcasses, and most information is exchanged on the night of, or the night before, a roost shift. Once a roost has moved beside the food, little information needs to pass.
• Stable group membership is not needed. Raven roosts are fluid groups of unrelated birds, yet information is still shared, because both leaders and followers gain from forming a group: vagrants can only feed at food guarded by territorial adults once a crowd has gathered.
• No special signal from knowledgeable birds was found. Signs of having fed (soiled feathers, full crops, smell of carrion) were ruled out, because leaders had often not yet fed. “Honking” calls before departure might play a role, or knowledgeable birds may simply leave first and be followed.
• Social soaring displays appear to signal an upcoming roost shift and draw birds from a wide area to a new roost near the next carcass.
• Ravens also find food by hearing other ravens at a carcass, but this seems to draw fewer than five birds from within 2 km.

4. How tracking was or was not used

The study is mainly about behaviour, followed by radio telemetry and direct observation. Tracks and signs were used in a small way: when no ravens were present at a carcass, the researchers approached it to check for signs of feeding, namely raven tracks, droppings and bite marks. Radio-tagged birds were also followed to roosts and feeding sites.

5. Tracker or observer certification

Not reported. The paper does not say who checked carcasses for tracks and signs. A number of volunteers are thanked for helping to observe, capture and care for the ravens.

6. Other technologies, skills or innovations

• Coloured, numbered wing (patagial) tags on 357 birds, and 30-gram radio transmitters on the central tail feathers of 28 birds (and of released birds).
• Experimental releases from cages covered in spruce boughs, opened by a string from a blind 20 m away.
• A formula for estimating how many ravens already knew of a carcass, correcting daily counts for birds that were missed, so that true recruitment from roosts could be tested.
• Circular statistics to describe the directions of roost arrivals and departures.