Monk parakeet and yellow-naped amazon / Parrot acoustic identity and social-scale comparison

Monk parakeet calls carry individual identity information

Published 27 July 2023 2023 · Note reviewed 2026-10-13

The authors report stronger acoustic convergence at the individual than site scale in native-range Uruguay and introduced U.S. monk parakeet populations. Their yellow-naped amazon comparator showed stronger convergence at the regional-dialect scale. This is evidence about acoustic identity structure and a proposed identity-signaling framework, not decoded call meaning or demonstrated receiver comprehension.

What we reviewed · Targeted primary article, dataset metadata, and code-provenance review; no file acquisition or reproduction

Reviewed the PLOS article’s sampling, methods, results, discussion, ethics, and linked-resource sections. Refreshed the Figshare v2 API and streamed its README.md file 41233869 for review; it was not retained. Resolved the GitHub v1.0.0 tag to commit 1f7031dc996c79914ccf5d042181f947793fb459 and read the site-scale selection R Markdown file at that revision without retaining or executing it. The code explains the 14-row selection difference. The PLOS supplement, remaining code/dependency workflow, selection CSV, and audio were not reviewed. No dataset files or audio were acquired; no code was executed.

Not independently reviewed by a bioacoustics or animal-communication specialist.

Study at a glance

Research question: At which social scale do monk-parakeet contact calls show the strongest acoustic convergence in native and introduced populations, compared with a parrot species with regional dialects?
Study scope: Monk parakeets: 229 individual-scale calls from 8 native-range and 9 introduced-range individuals, plus 1,353 site-scale calls from 63 sites (37 Uruguay, 26 U.S.). Yellow-naped amazon contact calls from Costa Rica in 1994 provide a separate comparator.
Methods: Spectrographic cross-correlation and random-forest similarity; 1,844 candidate acoustic/image features were reduced to 114 for the final model. The random forest trained on 156 calls from 8 known individuals; 73 calls from 9 other individuals supported proximity/clustering analyses.
Reported pattern: Monk-parakeet calls showed strongest convergence at the individual scale in both native and introduced ranges; the yellow-naped amazon comparison showed strongest convergence at the regional-dialect scale.
Score boundary: 97.44% (95% CI 93.57–99.30) is reported as training classification accuracy. It is not a held-out accuracy for identifying unseen callers and does not measure call meaning.
Inventory reconciliation: The README reports 1,596 monk-parakeet selection rows, with 1,582 used in the paper. At pinned GitHub v1.0.0, the site-scale script identifies and removes 14 calls from birds already represented in the individual-scale sample before comparing scales. The row-level CSV/audio mapping has not been independently recomputed.
Rights and reproducibility: Article CC BY 4.0; Figshare item-level CC BY 4.0; GitHub code MIT; associated Zenodo archive says other-open. All dataset/code files remain metadata-only; no results independently reproduced.

Reading location: PLOS Methods 2.2, 2.4–2.10; Results 3.1–3.5; Discussion 4.1–4.3; Ethics and Data Availability. Figshare v2 API and README.md file 41233869. GitHub v1.0.0 commit 1f7031dc996c79914ccf5d042181f947793fb459, IdentityEncoding_Code_05_SiteScaleDatasets.Rmd, lines 129–175; repository MIT LICENSE. Zenodo v1.0.0 API metadata.

What this does — and doesn’t — establish

The article’s supervised random forest was trained on 156 calls from 8 known monk parakeets using 114 selected features. The reported 97.44% accuracy (95% CI 93.57–99.30) is training accuracy. The remaining 73 individual-scale calls from 9 other birds were used to produce proximity matrices and inspect clustering patterns; the paper does not report that 97.44% as held-out unseen-caller accuracy. The study reports 229 individual-scale calls and 1,353 site-scale calls. The Figshare README says its monk-parakeet selection table contains 1,596 rows, of which 1,582 were used in the paper. The pinned v1.0.0 site-scale script explains the difference: 14 site-scale calls from birds also represented in the individual-scale set are removed before the scale comparison. This source-level reconciliation was not independently recomputed from the selection CSV or WAVs. The full supplement, exact file-level rights/provenance, remaining code and dependencies were not audited. Acoustic similarity cannot by itself show what a caller intends or what a receiver understands.

What we can build on

For Talk2Nature, separate known-individual identification, population/site similarity, call categories, independently observed behavior, receiver response, and meaning. This is a valuable parrot-specific methods reference and a reminder to report exactly what a validation set tests. A receiver-use claim needs an appropriately reviewed behavioral experiment; the authors point to playback work for that question.

A practical starting point

Metadata-only original summary with links; no article passages, figures, dataset files, audio, code, or weights copied.

Available starting point: PLOS displays CC BY 4.0 for the article; Figshare declares CC BY 4.0 for its item; GitHub identifies its code repository as MIT; the related Zenodo archive uses separate other-open metadata. These license statements apply to different assets and were not treated as interchangeable.

Reuse limits: Cite each source and follow its own terms. No dataset file was acquired or retained; one MIT-identified R Markdown source file was read without being retained or executed. Provider checksums are not locally verified and the archived Zenodo snapshot license is unresolved.

Our proposed experiment: A literature and metadata review only. No Talk2Nature model was trained, run, or evaluated; no animal communication or context inference was attempted.

The next reading task

If a caller-identity reproduction is selected, independently verify the 14-row exclusion against the exact selection CSV and audio filenames, review the full supplement and code/dependency environment, confirm file-level provenance and rights, and have a qualified bioacoustics reviewer assess the grouping and analysis. Keep all dataset assets metadata-only until that bounded case is justified. Do not use this dataset as a context or meaning benchmark.

Go to the sources

Individual identity information persists in learned calls of introduced parrot populations ↗
Peer-reviewed PLOS Computational Biology study comparing acoustic identity patterns in native- and introduced-range monk parakeets, with yellow-naped amazons as a regional-dialect comparator.
Smith-Vidaurre et al. identity-encoding calls and selection table (Figshare v2) ↗
Official Figshare dataset record for the paper. Its 22-file inventory totals 187,820,138 bytes under an item-level CC BY 4.0 declaration. The README describes a 1,596-row selection table and a WAV archive with matching sound.file names; 1,582 calls were used in the paper. No table or audio was retrieved.
Analysis code for identity information in introduced parrot populations ↗
Archived GitHub repository accompanying the PLOS paper. Its README describes ten R Markdown analyses and custom functions; GitHub identifies the repository license as MIT.
Zenodo v1.0.0 archive of the parrot identity analysis code ↗
Versioned Zenodo code artifact related to the archived GitHub repository. The previously reviewed Zenodo API metadata records DOI 10.5281/zenodo.8030605, version v1.0.0, and an artifact license value of other-open.
Pinned site-scale call selection code for the monk-parakeet identity study ↗
R Markdown file at GitHub v1.0.0 commit 1f7031dc996c79914ccf5d042181f947793fb459. The inspected lines explain which 14 site-scale rows are excluded from the paper's 1,582-call analysis.

Original editorial note by Talk2Nature contributors. AI-assisted, source-grounded synthesis; not a peer review or an independent replication. Linked works retain their own rights.