We compared the mean number of responses to the familiar test stimuli (original melody) to the mean number of responses to the unfamiliar test stimuli (modified versions) across the three tests. A Repeated-Measures ANOVA with the within factors Familiarity (familiar, unfamiliar) and Test (Pitch, Tempo, Timbre) revealed an effect of Familiarity (F(1,39) = 11.325, p = 0.002, η2 = 0.225) and Test (F(1.57,61.11) = 3.604, p = 0.044, η2 = 0.085), as well as an interaction between them (F(2,78) = 12.155, p < 0.001, η2 = 0.238). The Greenhouse–Geisser correction for violations of sphericity was used. Post-hoc pairwise comparisons with the Bonferroni alpha correction revealed that the animals’ responses did not differ between the familiar melody (M = 21.83, SD = 12.70) and the versions with transposed octaves (Pitch test; M = 22.43, SD = 13.35; MD = − 0.60, p = 0.640); nor between the familiar melody (M = 21.95, SD = 11.71) and the versions with novel tempi (Tempo test; M = 21.23, SD = 9.38; MD = 0.73, p = 0.584). In contrast, the animals produced more responses after the familiar melody (M = 20.28, SD = 11.25) than after the versions played with new instruments (Timbre test; M = 12.88, SD = 9.41; MD = 7.40, p < 0.001; see Fig. 3). There were no significant differences between the responses to familiar test stimuli across test sessions (all p = 1), but the responses to unfamiliar test stimuli were smaller for the Timbre test compared to the Pitch (MD = − 9.55, p = 0.007) and to the Tempo (MD = − 8.35, p = 0.001) tests. To explore if there were any changes across sessions for each test, we conducted separate repeated-measures ANOVA for the Pitch, the Tempo and the Timbre tests. For the Pitch Test, we observed a significant effect of Session (F(2, 37) = 19.91, p < 0.001, η2 = 0.469) on the number of responses to test stimuli but no significant effects of Familiarity (F(1, 37) = 0.249, p = 0.621, η2 = 6.228e−4) or the interaction between them (F(2,37) = 0.52, p = 0.519, η2 = 0.003). Similarly, for the Tempo test, results show a significant effect of Session (F(2, 37) = 3.487, p = 0.041, η2 = 0.134), but no significant effects of Familiarity (F(1, 37) = 0.267, p = 0.609, η2 = 0.001) or interaction between them (F(2,37) = 0.257, p = 0.775, η2 = 0.002). Thus, rats produced more nose-poking responses from session 1 to session 3, but crucially, this increase in responses did not have any effect in the pattern of responses that they were producing after familiar and unfamiliar stimuli. In contrast, for the Timbre test, results show a significant effect of Familiarity (F(1, 37) = 40.54, p = < 0.001, η2 = 0.116), as the animals consistently responded more to the familiar than to the unfamiliar stimuli, but no significant effects of Session (F(2, 37) = 2.862, p = 0.070, η2 = 0.104) or interaction between them (F(2,37) = 0.812, p = 0.452, η2 = 0.005). We also explored if there were any effects from the order in which the tests were presented. We observed an effect of Order (F(2, 78) = 25.943, p < 0.001, η2 = 0.305), as the animals presented with the Pitch test first produced more responses than the animals in the other groups. We also observed an effect of Familiarity (F(1, 39) = 11.941, p = 0.001, η2 = 0.019), but crucially, the interaction between factors was not significant (F(2, 78) = 0.596, p = 0.553, η2 = 0.002).
To further explore if the lack of significant differences in responses that we observed in the Pitch and Tempo tests was a failure of the animals to discriminate between test items, or rather a true null result in which the data is actually providing information about the null hypothesis, we ran Bayesian paired sample t tests (JASP Team, 2020). The data, in fact, provide no support for the alternative hypothesis neither for the Pitch test (BF10 = 0.189), nor for the Tempo test (BF10 = 0.197), but it does provide moderate support for the null hypothesis for both of them (Pitch [BF01 = 5.279, error % 9.153e−6]; and Tempo [BF01 = 5.081, error % 8.807e−6]). This suggests that the animals are generalizing their responses from the familiar stimuli to the novel ones. So, it is not that the animals are behaving randomly, but rather, they are responding to the novel test stimuli as if it is functionally equivalent to the familiar stimuli. For the Timbre test, a Bayesian paired sample t-test confirm strong evidence in favor of the alternative hypothesis (BF10 = 95,059.227, error % 2.729e−10).
We also explored any possible bias towards high or low frequencies, fast or slow tempo, or the timbre of the instruments by comparing the responses to the two kinds of unfamiliar stimuli we used during each test. We did not find differences between the higher frequency (M = 11.05, SD = 7.32) and the lower-frequency (M = 11.38, SD = 7.31) test items (t(39) − 0.34, p = 0.733, 95% CI [− 2.24, 1.59]) in the Pitch test; between the faster tempo (M = 9.85, SD = 6.19) and the slower tempo (M = 11.38, SD = 5.32) test items (t(39) − 1.44, p = 0.159, 95% CI [− 3.67, 0.62]) in the Tempo test; nor between the violin (M = 6.15, SD = 4.37) and the piccolo (M = 6.73, SD = 6.03) test items (t(39) − 0.77, p = 0.447, 95% CI [− 2.10, 0.94]) in the Timbre test (see Fig. 4; individual data are reported in the supplementary material accompanying this article).