How do goats "read" 2D-images of familiar and unfamiliar conspecifics?

Abstract

To study individual recognition in animals, discrimination tasks are often conducted by presenting 2D images of real conspecifics. However, animals may discriminate the images merely as visual stimulus combinations without establishing referential relationships to the individuals depicted. In the current study, we investigated whether goats are able to discriminate photos of familiar and unfamiliar conspecifics, whether they not only process the photos as visual stimuli, but also understand them as virtual copies of real conspecifics and whether they grasp the concept of familiarity. Using a computer-controlled learning device, in three tests, goats of two experimental groups (A and B) had to discriminate portrait (Te1), profile (Te2) or headless body photos (Te3) of conspecifics. Tests were presented as 4-choice tasks, with one photo from Group A (rewarded) plus three photos from Group B (distractors). That is, the rewarded photo was familiar to Group A, but unfamiliar to Group B. Finally, in a reversal test (Te4) we reversed this principle. The goats learned the discriminations in Te1 to Te3 within two (Te1 and Te2) and three training days (Te3), respectively, and they needed between 91 [CL (66, 126)] and 174 [CL (126, 241)] trials to reach the learning criterion, with no statistically significant differences between the groups. In Te4, in contrast, the animals took 403 [Group A; CL (291, 557)] and 385 [Group B; CL (286, 519)] trials, respectively, to learn the task. The lack of spontaneous preferences for the photo of the familiar conspecific in the pretests of Te1 to Te3 in Group A, as well as the lack of differences in the number of trials to learn the discriminations between both groups, do not at first glance suggest that the goats established a correspondence between real conspecifics and their 2D representations. However, the higher number of trials in Te4 suggests that both groups formed the learning rule of choosing either the known (Group A) or the unknown goat (Group B) over the course of Te1 to Te3 and then failed after the rule was reversed, providing evidence that goats can associate 2D photos of conspecifics with real animals.

Keywords: domestic ungulates; face recognition; individual recognition; reversal learning; visual discrimination.

Figure 1

Lateral view of a goat inside the compartment with the learning device: 1 = entrance (only one goat could enter at a time to avoid observational learning by pen mates), 2 = light beam to indicate when a goat entered/left the device, 3 = collar with RFID transponder for individual identification at the device, 4 = yoke to put the head through plus integrated RFID antenna, 5 = water bowl for reward delivery, and 6 = touchscreen for stimuli presentation.

Figure 2

Visual four-choice discrimination problems for training (Tr) and for the tests (Te). The rewarded photo (S⁺) within each test is placed in the upper left corner in this example. After each trial, the photos switched positions on the monitor. The numbering of the photos is given in Tr1. In Te1 to Te3, the rewarded photo was a goat from Group A (S⁺), while the three other photos were goats from Group B (S⁻). In Te4, this principle was reversed. It featured one photo from Group B (S⁺) and three photos from Group A (S⁻).

Figure 3

Estimated mean percentage of choice (%Choice, LSM ± CL) of the different photos in Group A and Group B for PT1 to PT4.

Figure 4

Learning curves for the six days of discrimination training in Te1–Te4 for Group A and Group B. The boxplots show the distribution of the data with the 25th, 50th (median), and 75th percentiles as coloured boxes, the 90th percentile as whiskers and black dots as outliers.

Figure 5

Estimated mean number of trials (TtC) to reach the learning criterion in Te1–Te4 for Group A and Group B. The boxplots show the distribution of the data with the 25th, 50th (median), and 75th percentiles as coloured boxes and the 90th percentile as whiskers. LSMs (±CL) are indicated in gray.