Technical Comments

Comment on “Number-space mapping in the newborn chick resembles humans’ mental number line”

See allHide authors and affiliations

Science  26 Jun 2015:
Vol. 348, Issue 6242, pp. 1438
DOI: 10.1126/science.aaa8577


Rugani et al. (Reports, 30 January 3015, p. 534) tested 3-day-old domestic chicks using an innovative experimental setup and demonstrate the presence of the mental number line. We raise concerns regarding this conclusion by highlighting the possible loopholes in the experimental design and the data analysis procedures. We further suggest auxiliary experiments that can substantiate the authors’ claim.

Empirical evidence supports the possibility that humans represent numbers along a mental number line (MNL) in ascending order from left to right (1). The origin of this MNL, however, is contentious. Interspecific comparison of numerical competence among humans and nonhuman animals suggests a continuous and nonverbal representation of numbers (2). Rugani et al. (3) tested the existence of the MNL in 3-day-old domestic chicks. An initial training phase involving a reward association with a certain number was followed by a test phase in which the same subjects are challenged by a number smaller or larger than the one used for training. The authors concluded that the subjects associated smaller numbers with the left space and larger numbers with the right space during the test phase. Additionally, this bias was displayed according to the relative magnitude of the challenge and was not an absolute bias toward a certain number.

Experiment 1 tested 15 subjects with numbers “2” and “8” after training with “5,” and experiment 2 tested 12 subjects with numbers “8” and “32” after training with “20.” In both experiments, half of the subjects were tested for the smaller number first, and the other half were tested for the larger number first (five trials each). The results are shown to be independent of the order of testing; thus, the authors consider each of the 2 × 5 = 10 trials to be independent.

There are confounds that possibly violate the assumption of the 10 trials to be independent.

Several studies have demonstrated that young chicks show lateral biases at spatial tasks (46). Any such bias is likely to be exaggerated as an artifact when the same subject is used repeatedly. Simply put, 70% of 15 subjects × 5 trials—that is, ~53 out of 75—is significant (binomial test with hypothetical probability of success = 0.5; P < 0.001), but 70% of 15 subjects × 1 trial—that is, ~11 out of 15—is not (binomial test, P = 0.119). One can argue that if such bias exists, it will manifest in the second set of trials as well, with the subject showing bias inconsistent with the MNL. However, once the chicks associate the smaller number 2 with the left space (which is not unlikely after five consecutive trials), just changing the stimulus to 8 can result in a move toward the right space, possibly because of the discrepancy in the association and the novelty factor (6). The same can also be true for the group of subjects challenged with the larger number stimulus 8 in the first place. Thus, the assumption of the 10 trials being independent is objectionable.

Moreover, the data on chicks choosing either the left or the right space follows the binomial distribution; the use of the Mann-Whitney U test is likely to be erroneous. Although according to the Mann-Whitney U test, the two data sets: (6, 6, 6, 6, 6, 6, 6, 6, 6, 6) and (5, 5, 5, 5, 5, 5, 5, 5, 5, 5) are different (U = 0; df = 19; P < 0.001), they are not different according to a binomial test (binomial test: 60/100; P = 0.060). For the same reason, studies on manual asymmetries in nonhuman primates use binomial z scores to examine individual-level biases; the z test is then used to examine population-level biases (7).

The concerns we have raised suggest that the evidence for the presence of the MNL in chicks might be insufficient currently.

In addition to the existing set of experiments, we would like to suggest a parallel set of experiments involving a blank panel. Apart from substantiating the current findings, these experiments will give a clue to the presence of zero in the MNL. Train the chicks with a blank panel and then test them with number panels; in another experiment, train the chicks with a number panel and then test them with blank panels. If one assumes the absence of the concept of zero in chicks, there should be no bias in either experiment, which would support the findings of Rugani et al. (3). However, a strong bias in these experiments will substantiate the authors’ claim, with a startling proposition that chicks have some parallel to zero in the MNL.


View Abstract

Navigate This Article