Data from: Leopard seal song patterns have similar predictability to nursery rhymes
Data files
Apr 16, 2026 version files 132.63 KB
-
AnnotatedLeopardSealSequences.zip
9.56 KB
-
Information_Theory_Analysis_of_Leopard_Seal_Vocalisations.zip
120.81 KB
-
README.md
2.26 KB
Abstract
During their breeding season, male leopard seals (Hydrurga leptonyx) spend hours perfecting their solo performance: singing unique sequences of stereotyped calls underwater to create their ‘song’. These song bouts are made up of discrete call types common across leopard seals within a region, which begs the question – what determines the individually unique patterns of these calls? Information entropy quantifies the amount of randomness in a sequence, providing insight into the statistical patterns governing a sequence. The songs produced by 26 different Eastern Antarctic leopard seals have more predictable temporal structure than humpback whale songs and dolphin whistle sequences. The estimated information entropy of the leopard seal songs is comparable to nursery rhymes but unsurprisingly, lower than contemporary, classical, and baroque music. The greater structure of the leopard seal’s song improves the ability of distant listeners to accurately receive signals and identify singers, which is essential for this widely dispersed species. Future studies examining animal mating song would benefit from incorporating entropy analysis to question if information is conveyed through temporal structure alongside other acoustic variables.
Dataset DOI: 10.5061/dryad.tdz08kq9b
Description of the data and file structure
Recordings were made from November to January, in 1992 to 1994, and 1997 to 1998, in the Davis Sea, Eastern Antarctica (from 68'25'S, 77'10'E to 68'35'S, 77'50'E). Recordings were made on 45-minute tapes with a Sony WMD6C audio-cassette recorder, a Brüel and Kjær 8103 hydrophone, and a custom preamplifier, and this system had a frequency response from 35 to 15 000 Hz 3 dB. The hydrophone was deployed 6m below the surface of the water. The dataset contains txt files of annotated leopard seal sequences in symbolic code and MATLAB files for analysis. .txt files are named according to an individual seal ID code assigned to an individual male seal observed during the field study.
Files and variables
File: AnnotatedLeopardSealSequences.zip
Description: .txt files of symbolic code describing leopard seal vocalisation sequences.
For the purpose of information theory analysis the leopard seal vocalisation sequences were converted to a symbolic code. Five distinct call types were identified in these recordings and annotated with the following symbol: high double trills (H), medium single trills (M), low descending trills (D), low double trills (L) and a hoot with a low single trill (O). A break between calling bouts, identified by a period of silence between bouts, was marked with a ‘Z’ in the sequence.
File: Information_Theory_Analysis_of_Leopard_Seal_Vocalisations.zip
Description: MATLAB live script and function files used for information theory analysis.
Code/software
Software files run on MATLAB R2024b. ZIP file includes one live script for primary information theory analysis and one live script for bootstrap testing, with three function files for entropy estimators.
Entropy_Analysis - Main MATLAB live script for estimating Shannon entropy (in bits) for each leopard seal sequence in the sample. Uses 3 function files contained within the ZIP file.
Entropy_Bootstrap_Testing - MATLAB live script for statistical validation of leopard seal sequence entropy estimates using bootstrapping.
