Rio is a California Sea Lion (Zalophus californianus). She was born in captivity at Marine World in Northern California, and due to insufficient maternal care from her biological mother, she was transferred to the Long Marine Laboratory at UC Santa Cruz when she was just a few days old. There, she was raised by a human surrogate mother, in the context of an imprinting study. She’s got a pretty good life. She gets all the fish she could want, and has 3 other pinniped (seals, sea lions, otters) friends at the lab. There’s Burnyce, a female Northern elephant seal, Sprouts, a male harbor seal, and Odin, a male Southern sea otter. Rio likes long walks on the beach, and prefers males who weigh at least 800 pounds.
Burnyce (right), Sprouts (left).
As Rio grew up, she participated in a lot of behavioral experiments of learning and memory, and other cognitive tasks. Two of Rio’s human friends, Colleen Kastak and Ronald J. Schusterman, from the Pinniped Lab were interested in Rio’s long term memory capacity. In particular, they were interested in her ability to remember abstract concepts in the long term. Because Rio spent her whole life participating in research at the laboratory, she was in a unique position to help address the question.
Conceptual equivalence learning – once thought to be unique to humans – is now thought to be the highest level of abstraction attainable by non-human animals. Conceptual learning involves learning the general nature of entire groups of stimuli, as opposed to learning about the explicit relationships that exist between individual stimuli, responses, and outcomes. It is advantageous for animals (human or non-human) to be able to use concept- or rule-governed problem-solving because it allows the individual to solve novel problems without having had prior experience with that the specific problem in question. Conceptual learning is demonstrated when a participant provides the proper response to a new problem based on general rules already learned.
There are three types of conceptual categories: perceptual, relational, or associative:
(1) Perceptual concepts are categories like “flowers,” “birds,” or “telephones.”
(2) Relational concepts are based on abstractions like “fabric,” or “food.”
(3) Associative concepts are stimuli that have common associations with other stimuli. For example, stimulus A and stimulus C may become related to each other because they are both associated with the same stimulus B. Although they are not perceptually similar, they are treated similarly.
In 1999, Rio learned two conceptual categories of visual stimuli. Each category contained 10 trained stimuli. After she demonstrated categorization, she was not exposed to the stimuli, testing apparatus, or procedure for 361 days (approximately one year). Would she remember the categorization after a year had passed?
Originally, she was trained on a two-choice simple discrimination repeated-reversal procedure. What does that mean, exactly?
two choice: for each trial, she was presented with a forced-choice task with only two options
simple discrimination: she was shown one target stimulus, and had to discriminate between the two alternative choices (sometimes the goal was to choose the one from the same category, or from the different category)
repeated-reversal: correct choices were rewarded with a fish and a high-pitched tone that meant “yes”; incorrect choices were not rewarded and were met with a low-pitched tone that meant “no.” After a certain amount of correct trials in which “correct” meant choosing the stimulus from the same category, the rule was changed. Now, “correct” meant choosing the stimulus from the different category. After one trial in which the previously “correct” option was chosen by Rio and NOT rewarded, could Rio alter her behavior based on the new rule, and begin choosing the now-”correct”/previously-”incorrect” alternative from the different category?
First, she was shown the target item in the center (in this case, a “5″). Then she heard a bell that indicated it was time to make the choice, at which time the two choices were offered. Rio chose the item belonging to the same category (numbers, not letters), and was rewarded with a fish and a high-pitch tone. Training continued until her performance was maintained at perfect levels.
A year passed. It was October, 2000. During the previous year, Rio did not see any of the stimuli she had been trained on, nor the testing apparatus, and did not participate in any similar type of concept learning experiment. The apparatus was re-introduced, as well as the stimuli and testing paradigm. Would she remember?
As in the original training paradigm, she was presented with a letter or number in the center, and had to choose either the letter or number on the sides depending on the rule. Each time Rio chose correctly on at least ten consecutive trials, the rule was changed, and the reward pattern was reversed. Sometimes this happened after 10 trials, but sometimes after more than 10 trials – this ensured that she couldn’t simply anticipate when the rule would change.
How did she do? The final ten reversals from 1999 were compared with the first 10 reversals in 2000.
If Rio’s performance depended simply on rote memorization, then she should have made errors on each of the trials following a reversal. However, if her performance depended on remembering the conceptual categories, then she should have realized following a reversal trial, that a change in reinforcement applied to one member of a category would predict the same change in reinforcement for all other members of the same category.
In fact, her performance was well above chance, even a full year later. Not only was she able to represent conceptual categories in her mind, but she remembered those categories after a whole year!
But…what about after TEN years?! This time instead of using associative concepts (in which the stimuli are arbitrarily associated because of common reinforcement patterns), perceptual concepts were used (“trees,” “people,” “flowers”).
When Rio first learned this task in 1991, she was six years old; she was 16 years old when she was retested in 2001. In the intervening 10 years, she did not participate in the same task, or see the same stimuli. A total of 154 stimulus pairs were used in the re-test: 84 of them had been used in 1991 in the original learning task (“identity” pairs), 70 of them had been seen in other experiments, but never in this type of task (“arbitrary” pairs), and 42 were completely new (“novel” pairs).
How did she do? Did she remember the rules of the game after 10 years? Could she apply the rule to the novel stimuli that she had never seen before?
These studies show that a California sea lion can remember abstract problem-solving strategies for at least one year, and up to ten years. This suggests that sea lions have the potential to remember and apply previously established rules when faced with new environmental (or experimental) problems.
In nature, sea lions have been observed remembering the locations of their birthplaces, and breeding and feeding locations. The identities of other individuals are remembered for long periods of time as well. Young pinnipeds recognize their mothers’ vocalizations for up to four years after their youth; male Stellar sea lions appear to categorize their rivals as “neighbors” or as unknown “intruders,” and they appear to remember those designations over multiple breeding seasons. It is clear that sea lions rely on long-term memory for a large variety of social and ecological problems, up to ten years (a significant portion of an individual’s lifetime, since average life expectancy for this species is ~25 years).
Cool, but who cares about sea lions?
Well, people who study sea lions care about sea lions. But I’ve tried to convince you that studying the minds of animals can help us better understand the human mind. By finding evidence for this in pinnipeds, it is possible that the capacity for generalized problem-solving evolved much earlier than we previously thought. By investigating the features and limitations of conceptual behavior in animals in controlled laboratory environments, we will be in a better position to understand the architecture and evolution of generalized problem-solving in the human mind.
Bonus: Want to see videos of Rio, Burnyce, Sprouts, and Odin participating in various experiments?
Reichmuth Kastak C, & Schusterman RJ (2002). Long-term memory for concepts in a California sea lion (Zalophus californianus). Animal cognition, 5 (4), 225-32. PMID: 12461600