Vocabulary learning in a Yorkshire terrier: slow mapping of spoken words

Abstract

Rapid vocabulary learning in children has been attributed to "fast mapping", with new words often claimed to be learned through a single presentation. As reported in 2004 in Science a border collie (Rico) not only learned to identify more than 200 words, but fast mapped the new words, remembering meanings after just one presentation. Our research tests the fast mapping interpretation of the Science paper based on Rico's results, while extending the demonstration of large vocabulary recognition to a lap dog. We tested a Yorkshire terrier (Bailey) with the same procedures as Rico, illustrating that Bailey accurately retrieved randomly selected toys from a set of 117 on voice command of the owner. Second we tested her retrieval based on two additional voices, one male, one female, with different accents that had never been involved in her training, again showing she was capable of recognition by voice command. Third, we did both exclusion-based training of new items (toys she had never seen before with names she had never heard before) embedded in a set of known items, with subsequent retention tests designed as in the Rico experiment. After Bailey succeeded on exclusion and retention tests, a crucial evaluation of true mapping tested items previously successfully retrieved in exclusion and retention, but now pitted against each other in a two-choice task. Bailey failed on the true mapping task repeatedly, illustrating that the claim of fast mapping in Rico had not been proven, because no true mapping task had ever been conducted with him. It appears that the task called retention in the Rico study only demonstrated success in retrieval by a process of extended exclusion.

Figure 1. Experimental Setup.

Formal trials began with the owner and the experimenter seated on the floor in the entryway out of sight from the location of the toys to be retrieved. A video camera (with no attendant) was focused on the toys, which were arrayed in front of the fireplace. After a command was given, the dog would retrieve a toy and return to the entryway. A trial was deemed correct if she had the correct toy in her mouth as soon as she came into view of the owner on return.

Figure 2. Results of Word Learning Phase I.

The results in the blue and pink segments of the figure and summarized in the green segment, illustrate that Bailey was able to retrieve, at a statistically significant level, the novel objects (Dora the Explorer and Triceratops) both in the Training Exclusion and Retention tests. In the yellow segments, the data illustrate that Bailey failed utterly on the subsequent two-choice tasks that pitted the two novel items against each other. Even after four sessions, no progress had been made on the two-choice test. Before each Training Exclusion and Retention test, the owner had conducted a short set of Informal Training trials with either Dora the Explorer or Triceratops in front of the fireplace (note the time spent on the trials in minutes and seconds is recorded in the figure plus the number of retrievals for that particular Informal Training segment). Thereafter, during each session Training Exclusion and Retention tests were conducted by the primary experimenter in the formal testing setting as in Figure 1 with the novel item that had been the target during the Informal Training segment. The formal tests required two or three retrievals, one of which always targeted Dora the Explorer or Triceratops, and the others of which targeted other known items in the test set. The test set (for both Training Exclusion and Retention tests) always included known items, and in Retention also included four completely novel items that had never before been included in any test set. The columns labeled “Correct” indicate Bailey's performance only on the trials where Dora the Explorer or Triceratops was asked for. Bailey was virtually always correct (36 of 37 trials) when known items were requested. A two-choice task with ten trials pitting Dora the Explorer and Triceratops against each other was conducted at the end of each of the four sessions. Bailey did not exceed chance performance on the two-choice task in any of those four sessions. For additional details see text.

Figure 3. Results of Word Learning Phase II: for details see text.

In Phase II, Bailey's owner spent considerable additional time training (as indicated in minutes of training and number of retrievals) on Dora the Explorer and Triceratops (without any experimenter present), and eventually the accumulated evidence from formal two-choice testing by the primary experimenter showed that Bailey had learned the mappings. The pink segments of the figure show the data on the training and the yellow segments show the results of the ten trials of two-choice testing targeting Dora the Explorer or Triceratops during each segment of testing. In green is the summary of the tests. Note that in the third and fourth segments, the test sets actually included additional known items (decoy toys) that were included to help maintain the dog's attention, and Bailey was correct each time one of those was requested. The data reported in the figure, however, only concern the trials for the target items. For additional details see text.