Sounds like a fight: listeners can infer behavioural contexts from spontaneous nonverbal vocalisations

Figures & data

Figure 1. Acoustic characteristics of nonverbal vocalisations produced in 10 behavioural contexts. Acoustic measures are broken down by speaker sex. Min. = minimum, Max. = maximum, SCoG = Spectral center of gravity, dB = decibel, Hz = Hertz, sec = seconds.

Table 1. Acoustic characteristics of vocalisations as means per behavioural context with standard deviations in brackets.

Behavioural context	Duration	Intensity	Pitch Mean	Pitch Sd	Pitch Min	Pitch Max	SCoG Mean	SCoG Sd
Being attacked by another person	1.39 (0.63)	78.76 (3.49)	314.63 (65.34)	84.77 (39.34)	157.98 (83.80)	443.94 (68.39)	1106.73 (382.41)	765.08 (367.54)
Being refused access to food	2.17 (2.20)	77.80 (3.83)	311.51 (68.47)	69.67 (33.94)	150.54 (85.39)	426.63 (106.93)	1756.13 (1059.63)	1482.46 (761.99)
Being separated from mother	2.27 (0.76)	77.48 (2.40)	355.23 (49.60)	78.88 (20.52)	193.05 (62.61)	493.43 (55.30)	1820.51 (521.60)	1223.73 (391.97)
Being tickled	2.20 (1.33)	79.88 (3.77)	349.84 (45.73)	77.97 (32.21)	183.79 (84.72)	473.72 (36.26)	1604.90 (526.47)	861.51 (378.72)
Discovering a large food source	1.34 (1.06)	79.54 (4.80)	265.97 (102.60)	53.27 (33.38)	152.93 (98.88)	342.36 (128.17)	1240.08 (1426.76)	1157.13 (1504.70)
Discovering something threatening	1.17 (1.32)	82.61 (2.71)	344.75 (88.11)	62.63 (44.51)	235.37 (110.67)	453.30 (61.40)	1445.75 (322.43)	758.78 (355.17)
Eating high value (strongly preferred) food	1.04 (0.33)	78.54 (2.60)	193.83 (57.11)	58.99 (33.23)	110.19 (38.91)	306.49 (116.38)	494.99 (188.99)	619.05 (285.86)
Eating dispreferred food	0.92 (0.17)	79.70 (1.52)	189.62 (78.79)	37.86 (36.04)	131.07 (60.07)	253.61 (122.60)	425.78 (130.55)	611.02 (294.58)
Having sex	3.89 (4.68)	75.68 (4.75)	242.85 (90.21)	35.48 (33.77)	175.64 (89.29)	305.21 (94.39)	1302.06 (1464.05)	1356.65 (1279.13)
Threatening an aggressive person or people	1.24 (0.86)	78.18 (3.41)	298.90 (82.72)	64.76 (25.98)	175.34 (96.34)	391.70 (81.41)	1129.02 (684.97)	938.86 (627.98)

Note: Min. = minimum, Max. = maximum, SCoG = Spectral center of gravity.

Table 2. D-prime scores indicating participants’ performance in matching vocalisations to behavioural contexts, tested against chance level.

Behavioural context	Mean (Sd)	Mdn	n	V	Rank-Biserial Correlation	95% CI for Rank-Biserial Correlation
						Lower	Upper
Being attacked by another person	1.01 (0.83)	1.03	317	48455.00*	0.985	0.980	0.988
Being refused access to food	0.58 (0.52)	0.59	300	38971.50*	0.899	0.870	0.922
Being separated from mother	2.53 (0.83)	2.76	305	46324.50*	0.998	0.998	0.999
Being tickled	2.55 (0.85)	2.76	294	42725.00*	0.998	0.997	0.998
Discovering a large food source	0.54 (0.70)	0.51	311	47242.00*	0.741	0.676	0.794
Discovering something threatening	1.66 (0.63)	1.81	310	47242.00*	0.998	0.998	0.999
Eating preferred food	1.90 (0.87)	2.05	304	45904.00*	0.993	0.991	0.995
Eating dispreferred food	0.10 (1.05)	0.00	315	24344.50	0.043	−0.086	0.171
Having sex	2.11 (0.69)	2.12	305	46637.00*	0.999	0.998	0.999
Threatening an aggressive person or people	1.09 (0.79)	1.05	324	50209.50*	0.955	0.943	0.965

Note: * significance level, p < 0.001 (Wilcoxon non-parametric test). n indicates the number of participants who listened to a particular behavioural context as target category, V is sum of ranks assigned to the differences with positive signs, effect size is given by the matched rank biserial correlation.

Figure 2. Heatmap of confusion matrix (%) for behavioural context categorisation.

Figure 3. Raincloud graph for (A) d prime scores (Experiment 1b), (B) arcsine Hu scores (Experiment 2). Raincloud graphs combine split-violin plots and boxplots: split-violin plots show the data distribution, and boxplots display the range of the data, their full range from minimum to maximum, the 25–75% range (box), and the median (middle of the box). Horizontal dashed lines indicate the chance level for each experiment. In both experiments, listeners performed better than chance for all contexts, with the exception of eating dispreferred food in Experiment 1b (A). Code adapted from Allen et al. (2019).

Table 3. Hu scores indicating listeners’ performance in categorising contexts, tested against chance level.

Behavioural context	Mean (SD)	Mdn	n	V	Rank-Biserial Correlation	95% CI for Rank-Biserial Correlation
						Lower	Upper
*Being attacked by another person	0.45 (0.11)	0.45	327	59.60	3.30	3.018	3.568
*Being refused access to food	0.30 (0.13)	0.30	327	27.75	1.54	1.374	1.694
Being separated from mother	0.81 (0.17)	0.82	327	53628.00	1.00	1.000	1.000
Being tickled	1.03 (0.12)	1.03	327	53627.00	1.00	1.000	1.000
Discovering a large food source	0.49 (0.13)	0.49	327	53620.00	1.00	1.000	1.000
Discovering something threatening	0.47 (0.24)	0.49	327	52338.00	0.96	0.954	0.972
Eating preferred food	0.66 (0.09)	0.67	327	53628.00	1.00	1.000	1.000
Eating dispreferred food	0.31 (0.16)	0.30	327	51618.00	0.96	0.950	0.969
*Having sex	0.90 (0.11)	0.89	327	135.53	7.50	6.904	8.075
Threatening an aggressive person or people	0.65 (0.13)	0.66	327	53628.00	1.00	1.000	1.000

Note: * one sample t-test was used for testing against chance level given normality in the distribution of the data. For these contexts, the V column represents t values and effect size column represents Cohen’s d effect size. Wilcoxon-signed rank test was used for the other contexts in order to test the performance against chance level; effect size is given by the matched rank biserial correlation.

Figure 4. Raincloud graph for (A) d prime scores (Experiment 1b), (B) arcsine Hu scores (Experiment 2). There was no difference in d scores between positive and negative contexts in Experiment 1b (A). Listeners were better at recognizing positive contexts compared to negative contexts in Experiment 2 (B).

Table 4. Comparisons of the frequency of choosing the target category with the frequency of choosing the most confused category for each context.

Stimuli context	Response	Mstimuli	Sdstimuli	Mresponse	Sdresponse	t
Being attacked	Threatening another	9.29	2.40	4.06	2.01	24.93*
Being refused	Being separated	4.76	2.47	4.19	2.34	2.37**
Being separated	Being refused	14.04	3.86	5.10	3.65	21.87*
Being tickled	Being attacked	18.26	1.94	0.44	0.71	135.06*
Discovering food	Discovering threat	6.93	2.25	3.91	1.97	15.72*
Discovering threat	Being attached	9.35	5.15	7.49	5.15	3.32**
Eating preferred food	Eating dispreferred food	15.21	2.84	3.29	2.58	41.13*
Eating dispreferred food	Eating preferred food	4.95	3.06	13.08	3.16	−24.39*
Having sex	Eating preferred food	18.13	1.57	0.53	0.78	149.69*
Threatening another	Being attacked	11.31	2.78	3.68	2.50	28.41*

Note: * p < .001 ** p < 0.01significance level; paired t-test was used for comparing the frequency of choosing the correct category with the frequency of choosing the most confused category.

Allen, M., Poggiali, D., Whitaker, K., Marshall, T. R., & Kievit, R. A. (2019). Raincloud plots: A multi-platform tool for robust data visualization. Welcome Open Research, 4, 63. https://doi.org/10.12688/wellcomeopenres.15191.1

 PubMedGoogle Scholar

Data availability statement

The data that support the findings of Experiment 1b are openly available from https://doi.org/10.21942/uva.13560374

 Google Scholar

, and the data that support the findings of Experiment 2 are openly available from https://doi.org/10.21942/uva.18972962

 Google Scholar

.