Abstract: Food caching birds hide food and recover the caches when supplies are less abundant. There is, however, a risk to this strategy because the caches are susceptible to pilfering by others. Corvids use a number of different strategies to reduce possible cache theft. Scrub-jays with previous experience of pilfering other's caches cached worms in two visuospatially distinct caching trays either in private or in the presence of a conspecific. When these storers had cached in private, they subsequently observed both trays out of reach of a conspecific. When these storers had cached in the presence of a conspecific, they subsequently watched the observer pilfering from one of the trays while the other tray was placed in full view, but out of reach. The storers were then allowed to recover the remaining caches 3 h later. Jays cached more worms when they were observed during caching. At the time of recovery, they re-cached more than if they had cached in private, selectively re-caching outside of the trays in sites unbeknown to potential thieves. In addition, after a single pilfering trial, the jays switched their recovery strategy from predominantly checking their caches (i.e. returning to a cache site to see whether the food remained there) to predominantly eating them. Re-caching remained constant across the three trials. These results suggest that scrub-jays use flexible, cognitive caching and recovery strategies to aid in reducing potential future pilfering of caches by conspecifics.

Abstract: An adult California sea lion ( Zalophus californianus) with extensive experience in performing discrimination learning tasks was tested to evaluate her long-term memory for two previously learned concepts. An associative concept, that of equivalence classification, was retested after a retention interval of approximately 1 year. The sea lion had originally shown emergent equivalence classification with nonsimilarity-based classes of stimuli in a simple discrimination repeated-reversal procedure as well as in a matching-to-sample procedure. The 1-year memory test revealed no decrement in classification performance in either procedure. A relational concept, that of generalized identity matching, was retested after approximately 10 years. The sea lion had originally received trial-and-error exemplar training with identity matching-to-sample problems prior to transferring the concept to novel stimulus configurations. In the 10-year memory test, the sea lion immediately and reliably applied the previously established identity concept to familiar and novel sets of matching problems. These are the first reports of long-term conceptual memory in a nonprimate species. The experimental findings are consistent with a variety of observations of sea lions in natural settings, which indicate that natal sites, feeding areas, and individuals may be remembered over long periods of time.

Abstract: Single-trial learning and long-term memory of “what” and “who” information were examined in an adult gorilla (Gorilla gorilla gorilla). We presented the gorilla with a to-be-remembered food item at the time of study. In Experiment 1, following a retention interval of either approximately 7 min or 24 h, the gorilla responded with one of five cards, each corresponding to a particular food. The gorilla was accurate on 70% of the short retention-interval trials and on 82% of the long retention-interval trials. In Experiment 2, the food stimulus was provided by one of two experimenters, each of whom was represented by a card. The gorilla identified the food (55% of the time) and the experimenter (82% of the time) on the short retention-interval trials. On the long retention-interval trials, the gorilla was accurate for the food (73%) and for the person (87%). The results are interpreted in light of theories of episodic memory.

Abstract: Two mazes were used to test the learning ability and memory of horses, and changes in these abilities. Testing was done on four occasions. On Occasion 1, the horses were run through Maze A until they had reached the criterion of three consecutive correct runs. A week later (Occasion 2), they were retested in Maze A to the same criterion as a measure of memory. On Occasion 3,2 months later, the horses were run through Mazes A and B until they reached the criterion. Occasion 4 took place 1 week later when they were run through Mazes A and B. An estimation of changes in ability to learn came from a comparison of results from Occasions 1 and 3. Similarly, changes in ability to remember came from a comparison of results from Occasions 2 and 4. Nine horses with a variable amount of riding training were the subjects. All horses were able to learn the maze, but the ability varied among horses. There was no obvious correlation between quality of handling of the horses and learning ability. Once the horses had learned the maze, they remembered it perfectly on subsequent occasions. There were changes in the memory and learning ability of the horses, but no clear explanation for this could be found.

Abstract: Delayed-responses have been traditionally employed to investigate the temporal characteristics of animals“ ability to represent and recall objects that have disappeared. In the typical condition, the animal, usually a mammal, observes the experimenter hiding an interesting goal (e.g. some food) in a certain location. A delayed-response task (DRT) was administered to 4 female Esperia pony (2 years old) coming from a free-range breed (Frosinone, Italy) and to 7 female Amiata donkeys (4.2±2 years old) coming from a conservation stock (University of Pisa, Italy). The DRT's apparatus was located in a square fence. A single ”U-shaped“ screen (330x160x140 cm) made by wood shavings blocks was positioned in the centre of the fence. A gap (40x50 cm) on the ground was in the middle of the central side of the U-shaped-screen and served to make the food-attractor disappear. The food-attractor consisted in cereal flakes and fresh grass for ponies and cereal flakes for donkeys. A bucket full of food was placed on a dolly tied on a rope which could be pulled by an experimenter. In a preliminary training each animal was allowed to eat food from the bucket and, while the animal was eating, the dolly was gently pulled away from the animal, and beyond the screen through the gap. The subjects needed to move around of the screen in order to retrieve the food. As a reinforcement, they were allowed to eat some food from the bucket once behind the screen. From trial to trial, the bucket was presented farther and farther (starting with a distance of 1 m in front of the screen to reach 7 m). Therefore subjects were tested in the DRT requiring them to rejoin the bucket with the goal-food disappearing behind the screen as in the preliminary training but following a 10 s delay. For the DRT, the bucket was placed 7 m in front of the screen, 3 m away from the animal's starting area. Then the dolly was pulled away from the animal. Ten seconds after the disappearance of the dolly behind the screen the animal was released from the starting area. The DRT ended when the subject had reached the attractor behind the screen on 3 consecutive trials. Results showed that all animals were able to rejoin the food behind the screen after 10 s delay. The mean time of the delayed-response (mean±sd, in s) in the ponies (1st: 19.8±8; 2nd: 10.8±2.2; 3rd: 12.8±2.8) and in the donkeys (1st: 28.4±10; 2nd: 26.9±13; 3rd: 24.3±16.6) showed a trend to decrease from first trial to third. These preliminary results suggest that like other mammals our ponies and donkeys can maintain a working memory trace of the location where biologically attractive objects have been seen to disappear. In conclusion, this study paves the way to set up a viable model system for the investigation of the more sophisticated aspects of Equids” cognitive abilities such as working memory.