• Keine Ergebnisse gefunden

2.3. Results

2.3.1. Song learning performances in the tutored groups

Sample spectrograms of the selected juvenile and tutor songs are shown in Figure 2.

11 to Figure 2. 18. The distributions of the similarity scores of each juvenile in the

50

tutored groups are presented below the song spectrograms. The song similarities of each bird and the group means are summarized in Table 2. 1.

51

Similarity to tutor song of bird k8g8 (410 LONG) Similarity score, median ± MAD: 18.00 ± 4.40

Figure 2. 11: Similarity of 20 juvenile songs of bird k8g8 (410 LONG) compared with 20 tutor songs.

52

Top, Song spectrograms of typical juvenile songs, with the tutor song spectrogram at the bottom for reference. Both the juvenile and tutor songs were selected from the recording in the last day (i.e., when the juvenile bird was 59 dph). Color is used to represent sound amplitude with blue indicating low and red indicating high values. Some juvenile and tutor song motifs are outlined with red rectangles, to serve as examples of the songs compared.

Bottom left, Histogram with a Gaussian fit of the resulting 400 similarity scores (20 juvenile songs vs. 20 tutor songs).

Bottom right, Heatmap of the similarity scores. The horizontal axis represents the 20 juvenile songs, and the vertical axis represents the 20 tutor songs. Color is used to represent the value of the similarity scores. For reference, see the color bar on the right.

* Figure 2. 11 – Figure 2. 18 and Figure 3. 2 – Figure 3. 9 are all plotted with the same settings and styles. The scale of the axes and the color codes are the same for the song spectrograms, histograms, and heatmaps.

53

Similarity to tutor song of bird k20g10 (801 LONG) Similarity score, median ± MAD: 20.00 ± 5.47

Figure 2. 12: Similarity of 20 juvenile songs of bird k20g10 (801 LONG) compared with 20 tutor songs. See Figure 2. 11 legend for more information.

54

Similarity to tutor song of bird k1g10 (901 SHORT) Similarity score, median ± MAD: 19.00 ± 6.02

Figure 2. 13: Similarity of 20 juvenile songs of bird k1g10 (901 SHORT) compared with 20 tutor songs. See Figure 2. 11 legend for more information.

55

Similarity to tutor song of bird p18g10 (909 SHORT) Similarity score, median ± MAD: 28.00 ± 5.83

Figure 2. 14: Similarity of 20 juvenile songs of bird p18g10 (909 SHORT) compared with 20 tutor songs. See Figure 2. 11 legend for more information.

56

Similarity to tutor song of bird g14o5 (1014 SHORT) Similarity score, median ± MAD: 18.00 ± 5.06

Figure 2. 15: Similarity of 20 juvenile songs of bird g14o5 (1014 SHORT) compared with 20 tutor songs. See Figure 2. 11 legend for more information.

57

Similarity to tutor song of bird g14o4 (1227 LONG) Similarity score, median ± MAD: 31.00 ± 5.92

Figure 2. 16: Similarity of 20 juvenile songs of bird g14o4 (1227 LONG) compared with 20 tutor songs. See Figure 2. 11 legend for more information.

58

Similarity to tutor song of bird g15o2 (1503 LONG) Similarity score, median ± MAD: 40.00 ± 8.93

Figure 2. 17: Similarity of 20 juvenile songs of bird g15o2 (1503 LONG) compared with 20 tutor songs. See Figure 2. 11 legend for more information.

59

Similarity to tutor song of bird g15o3 (1618 SHORT) Similarity score, median ± MAD: 27.00 ± 3.76

Figure 2. 18: Similarity of 20 juvenile songs of bird g15o3 (1618 SHORT) compared with 20 tutor songs. See Figure 2. 11 legend for more information.

60

Juvenile songs in the LONG group had developed distinguishable syllable and motif structures at 59 dph, but not in the SHORT group. In the SHORT group, for example bird g14o5 (sample 1014) still sang songs that were quite plastic and that had no clear or nearly stereotyped syllable structures. These were expected since the birds in the SHORT group were only tutored for one day after 58 days of song isolation.

For every individual bird, the resulting similarity scores formed a single cluster without noticeable outliers, and roughly fitted to the Gaussian distribution judged by eye. However comparing among individual birds the data distributed differently, with some are broader (e.g.: g15o2), some are narrower (e.g.: g15o3), some skewed to the left (e.g.: g14o4), and some skewed to the right (e.g.: g14o5). Therefore as mentioned in the method section 2.2.1, I decided to take the median values of the data to represent each bird in order to ignore the different distributions. The selected juvenile songs are represented on the horizontal axis, and the selected tutor songs are represented on the vertical axis. Therefore, inconsistent selections of the juvenile or tutor songs would result in distinctive vertical or horizontal lines, respectively.

Several slightly visible vertical lines can be seen in the heatmaps of birds k8g8, k20g10, and k1g10, while no apparent horizontal lines are presented. This indicates that the tutor song selections were consistent and that all 20 tutor songs resulted in similar comparisons with the juvenile songs. The juvenile song selections were less consistent. Some juvenile song selections resulted in systematically higher or lower values compared with the other selections. Such sampling variability is expected because juvenile songs vary greatly and have structures that are hard to categorize.

The three inconsistent selections mentioned above were retained in the final results because none of them resulted in values that were far from the final median values (Table 2. 1).

Table 2. 1: Song similarity score values for the tutored birds in the SHORT and LONG groups. The bird and group identities were coded (2nd column) to ensure unbiased (group-blind) analyses. The song similarity score of each bird were determined by averaging the 400 comparisons, and the median ± MAD values are shown in the 4th column. The group average scores were determined by averaging across all of the birds within the groups, and the group means ± SE are shown in the last column.

61

The song similarity score for each bird ranged from 18 to 28 in the SHORT group, and from 18 to 40 in the LONG group. In the SHORT group the juvenile birds received only 1 day of tutoring, and their song similarity scores were generally very low. In the LONG group, the juvenile birds received 24 days of tutoring, and their song similarity scores on average were 4.25 point higher than those in the SHORT group. However, two birds, k8g8 and k20g10, had lower scores than the group mean value of the SHORT group.

Figure 2. 19: The song similarity scores (for data, see Table 2. 1). The boxes represent the similarity scores of individual birds, and the error bars indicate the MAD. The colored bars represent the group means, with group means ± SE are written in them. The colors in the boxes and bars indicate their group (yellow: SHORT; blue: LONG).

Table 2. 2: The results of the Wilcoxon rank-sum test of the song similarity score differences in the SHORT and LONG experimental groups. The difference between the two tutored groups was not significant.

62

The group means ± SE of the song similarity scores were 23.00 ± 2.61 in the SHORT group, and 27.25 ± 5.12 in the LONG group. No significant difference was found between the tutored groups (Table 2. 2). I expected the birds in the LONG group to score significantly higher than the birds in the SHORT group, but the result showed no significant differences. This unexpected result is discussed together with the song similarity comparison result obtained from Experiment II, in section 3.3.1.