Strategies for improving the quality and accessibility of captive animal ethograms

Introduction

Accurate behaviour identification and monitoring in captive animals is essential. Behaviour provides an insight into an animal’s emotional state and their welfare, for example abnormal behaviours are often used as indicators of negative welfare, and changes in species-typical behaviours such as sleep, play, or aggression can also be used in welfare assessment (Manteca et al., 2016). Animal owners and/or caretakers are responsible for the welfare of the animals in their care, and to optimise welfare they may employ assessment techniques such as behavioural monitoring (Dawkins, 2004). In research, poor welfare can affect study outcomes. Studies using animals experiencing, or having previously experienced, negative welfare lack validity and generalisability (Loss et al., 2021; Poole, 1997; Prescott and Lidster, 2017). Welfare monitoring is therefore necessary for both scientific and ethical reasons.

In biomedical research, behaviour can be used to assess the effectiveness of an intervention, for example as a non-invasive measure of pain (Descovich et al., 2019) or depressive-like symptoms (Camus et al., 2014). Similarly, behavioural changes can be used to evaluate the success of environment modifications or enrichment items (Young, 2003). To quantify behaviour accurately, reliably, and systematically, having a well-defined ethogram is crucial. Ethograms originated in the field of ethology in the 1930s (Makkink, 1936), becoming widely used in the 1950s (Eibl-Eibesfeldt and Kramer, 1958), and were defined as a catalogue of a given species’ innate behaviour patterns. Modern ethograms are required to 1) include a comprehensive list of behaviours, capturing a species’ entire behavioural repertoire, and 2) define those behaviours in ways that are mutually exclusive (Renner, 2022). A well-designed and well-defined ethogram will reduce inter- and intra-observer, as well as inter-study, variation, and improve reproducibility (Neadle et al., 2025).

Our study species is Rhesus macaques (Macaca mulatta), which are a species prevalent in zoos and are one of the most used non-human primates in biomedical and neuroscience research (Cauvin et al., 2015). Despite their ubiquity, there is currently no published, comprehensive, ethogram available for this species. The initial aim of this project was therefore to construct a captive Rhesus macaque ethogram, to be used in future research and monitoring. The process was more challenging than envisaged, however, and revealed significant issues with the ethogram landscape. Discussions with researchers working with other species highlighted that these problems were not restricted to macaques. There are a number of widespread problems that are accepted as standard. Firstly, we did not want our ethogram to succumb to such pitfalls. Secondly, it was apparent there is a substantial need to vocalise and discuss these issues if the landscape is to be improved. This paper aims to begin this conversation by 1) summarising the key issues with the current ethogram landscape, and 2) discussing approaches for improvement. These discussions are framed around our experiences constructing a comprehensive Rhesus macaque ethogram. Note that this paper is intended to be a discussion of the overall landscape, not a name-and-shame exercise. Specific articles will therefore not be cited as examples.

Issues with the ethogram landscape

We identified seven key issues with current ethogram approaches, broadly fitting into two categories: 1) problems with definitions and utilisation, 2) lack of accessibility and sharing. These are summarised in Table 1 and discussed further below.

Table 1. Summary of the key issues with the current ethogram landscape.

This project began due to a lack of an available, detailed, and comprehensive ethogram for Rhesus macaques. The available comprehensive ethograms typically group behaviours into broad categories, such as ‘abnormal behaviour’ or ‘social behaviour’. This is not necessarily an issue, but problems arise when accompanying definitions are vague, making it challenging to categorise behaviours without prior knowledge. From experience, observers often discuss how to designate more ambiguous behaviours, yet these discussions are rarely reported in either ethograms or accompanying publications, raising questions about repeatability. Such uncertainty also suggests a substandard ethogram which is unlikely to produce valid and reproducible results (Neadle et al., 2025).

That said, there are high quality definitions in the literature. High quality definitions, from our perspective, are those which enable a naive individual to identify a behaviour from the definition alone. Our general experience was that definitions for specific behaviours were of greater quality than those of behavioural categories. There is a trade-off between breadth and depth; papers containing higher quality definitions tend to focus on only a sub-group of behaviours, and vice versa (personal experience). Only including behaviours related to study aims is a widely accepted approach as it improves efficiency and reduces the risk of type I errors (false positives) during analysis (Finner and Roters, 2002). For our purposes, however, the result was needing to compile information from multiple ethograms, a common experience amongst ethologists. Rather than researchers continually repeating this process, it would be beneficial to have comprehensive lists of behaviours, alongside high-quality definitions, for each species. All researchers could then draw from this ‘master’ or ‘primary’ ethogram, improving efficiency.

A primary ethogram would also improve consistency. In the published literature, we found multiple definitions for the same behaviours. Determining which one was the most effective, if any, was challenging. It was often necessary to combine elements of different definitions or expand definitions using our own knowledge. Doing so, however, exacerbated inconsistencies, which risks preventing or limiting inter-study comparisons. A primary ethogram would encourage all studies to use the same definitions, allowing data to be compared or combined effectively between studies. In captive animal research sample sizes may be limited by the number of available animals, restricting statistical power, i.e. the probability of statistically detecting a true effect where one exists (Kuhar, 2006). Combining data from multiple sources to increase sample size is a useful technique for increasing statistical power (as seen in the various ‘Many’ projects, such as ManyPrimates (Many Primates et al., 2019), and ManyDogs (Alberghina et al., 2023)), however this is prevented by inconsistent definitions.

The final problem identified with definitions was the inclusion of quantifiable behavioural components, such as ‘for at least 3s’ or ‘at least three times’. Including measurable quantities reduces ambiguity, with the presumed goal of improving reliability and reproducibility. However, they have limitations. Quantities can be challenging to assess during live observations. For example, it is difficult to accurately assess distances between individuals in large enclosures, even with training, due to differences in perspective between locations, variations in terrain, and the presence of enclosure features. Equally, assessing whether an animal has been stationary for more or less than 1s, whilst scan sampling a multi-animal enclosure, is challenging. Quantities can also be restrictive, and result in incorrect coding. For example, the stereotypic behaviour pacing’s definition may state that the behaviour should be performed ‘at least three times’. However, there are instances where a behaviour is stereotypic by definition (Mason, 2006), yet may only be repeated once or twice, either because this is how the animal typically performs this behaviour, or because they are interrupted. In these cases, the ethogram definition will not capture the behaviour. We therefore advise against using the number of repetitions in definitions unless absolutely necessary. Regarding distances and times, caution is advised. They can be useful; however parameters should be based on the species’ behavioural ecology, which should be referenced either within or alongside the ethogram. Using arbitrary parameters is somewhat meaningless and can result in a lack of consistency.

High quality, comprehensive, ethograms are not common and their rarity is exacerbated when they are not openly available. Some ethograms are retained in-house; they are not shared with other institutions, made public, or published in scientific journals. This privacy may be driven by a fear of judgement by others, a lack of funding, time, or skills for publishing, or a lack of awareness regarding open science principles (Gagliardi et al., 2015; Gownaris et al., 2022). Where ethograms are published in academic journals they may remain behind a paywall due to high open access article publishing charges and/or a lack of funding. Open access publishing is recommended wherever possible, however it is not always feasible.

Publishing an ethogram in an academic journal typically requires for that ethogram to have been used in a particular study, on which authors are reporting. Methods papers are notably more challenging to publish, and there is usually a requirement for them to highlight something new or novel in their approach (personal experience). An ethogram does not necessarily fit this brief, creating a barrier to publication for small institutions (e.g. zoos, sanctuaries, small charities) who develop ethograms and collect behavioural data but do not have the capacity to produce academic research articles. This inability to publish prevents animals in other ex-situ facilities benefiting from study outcomes and may result in the needless repetition of research. Having a place where such ethograms can be stored and shared, such as on online depository, website, or with a society, would be beneficial. More broadly, as a community a concerted effort should be made to make ethograms public, in order to improve the outcomes for animals.

How can we improve the landscape? A case study with Rhesus macaques

When developing our Rhesus macaque ethogram, we aimed to mitigate some of the issues described above. The approach described is not necessarily the perfect solution, however we hope that by sharing the process other ethologists are inspired to trial similar techniques.

The goal was to create a comprehensive and detailed ethogram, for use by Rhesus macaque researchers and caretakers. To achieve this, a ‘pick and play’ ethogram, containing three ‘levels’, was designed (Table 2). Level one would consist of broad behavioural groups, level two would divide these groups into behavioural sub-groups, allowing a more detailed look at certain aspects of behaviour, but remaining general enough that behaviours are relatively simple to identify during live observations. Finally, level three would contain a complete list of all fine-grained behaviours performed by captive Rhesus macaques. Importantly, all levels were to contain clear definitions of each behaviour or behavioural group.

Table 2. An extract from the three-level Rhesus macaque ethogram, depicting how each level is nested within the one above it.

Having three levels allows users to select their required level of detail, based on their aims. For example, a study examining the effect of an intervention on stereotypic behaviour might select to work at level three for stereotypic behaviours, level two for other abnormal behaviours, and level one for all other behavioural groups, thus creating an exhaustive ethogram detailed only in their focus area. In addition, having descriptions of all potential behaviours (level three), nested within their behavioural groups (level one) and sub-groups (level two), allows users to clarify which group a particular behaviour falls within.

The existing literature were used to create an initial three-level ethogram by KL. KL used their prior experience to combine and/or edit definitions where appropriate. KS and CW then 1) reviewed the definitions for each level three behaviour and provided feedback and, 2) arranged the behaviours into groups, allowing KL to assess whether behaviours had been appropriately organised into level two. There was overall agreement between KL, KS and CW in how behaviours should be arranged, and the feedback was utilised to modify and improve the ethogram prior to testing. See Supplementary Text for further details.

Preliminary testing took place during the pilot of a separate project, led by KS, examining the impact of an enrichment object rotation on activity budgets. In this paper we will focus only on the ethogram implementation aspect of the project. Firstly, the levels at which behaviours needed to be measured to meet the study aims were identified, and a study-specific ethogram generated. Following this, coders (KL and KS) sampled a group of animals together, in which any coding ambiguities were discussed and it was decided, based on the aims of the study, which behaviour would take priority when behaviours occurred simultaneously. Once coders were satisfied, pilot data collection began. Instantaneous scan sampling at 1-minute intervals was utilised to code video footage of three animal groups. Monkeys at the Centre for Macaques are continuously recorded using CCTV cameras. From this footage seven 20-minute segments of video were extracted per day, starting at 08:30, 09:30, 10:30, 11:30, 12:30, 13:30, and 14:30 (hours when staff are present). Altogether, 21 hours of video footage were coded (3 groups over 3 days, 7 segments/day). Excluding occasions where monkeys were coded as Out of Sight, inter-observer analyses indicated “substantial agreement” between coders (weighted Cohen’s Kappa = 0.72). Both observers were integral in ethogram creation and are highly experienced with macaque behaviour, yet some discrepancies did occur. Some variation was expected, however there were two more systematic differences that are worth highlighting. Firstly, KL coded ‘Out of Sight’ more regularly than KS. It was identified that our thresholds for determining the exhibition of a given behaviour differed when an animal was partly obscured. This was not something we had considered prior to coding, however will now form part of pre-data collection conversations. We recommend that all ethograms clearly define what constitutes ‘Out of Sight’ to avoid such issues. Secondly, our macaques perform a behaviour whereby they roll balls of their bedding along the floor, or against a window or solid object, with the palms of their hands. The exact purpose of this behaviour is unclear, however we believe it may be related to the rolling of food items in the wild. Certainly stone rolling behaviour, which presents similarly, has been proposed as a pseudo-foraging behaviour (Pelletier et al., 2017). Bedding rolling had not been included in the initial primary ethogram, resulting in a lack of clarity as to how it should be coded, but was subsequently added to the ethogram. This highlights the importance of investing time observing your animals prior to data collection and amending ethograms accordingly. The updated ethogram will be utilised for the enrichment study, and further inter-observer analyses conducted.

Although we are pleased with the initial output of this process, our Rhesus macaque primary ethogram will remain a working document which can be updated in response to new insights. The ethogram is published open access on the Open Science Framework (OSF) (https://osf.io/74na8/overview), and will also be published on the macaque section of the NC3Rs website. In both locations we invite feedback and request that users make us aware of any required updates. There is also an accompanying set of videos demonstrating the majority of behaviours, intended to improve ethogram usability.

Discussion

This project started with the seemingly simple task of constructing an ethogram for captive Rhesus macaques, however the process led us in an unexpected direction, highlighting issues with the ethogram landscape which call into question the validity and reproducibility of outputs. Discussions with other researchers revealed they experienced similar challenges; however we were reassured by their shared recognition of the need for change. This article will therefore finish with some recommendations for improving the landscape going forwards.

Firstly, we suggest generating primary ethograms for your study species. These could follow a ‘pick and play’ structure or something entirely different. Of greatest importance is finding a suitable method for the species. Whether creating a primary or a study specific ethogram, definitions should be high quality and unambiguous. That said, it can be difficult to reflect behavioural nuances using words alone. Accompanying videos should thus be used to support definitions where possible. For further guidance regarding creating a reliable ethogram, as well as making behavioural research more reliable more generally, we highly recommend following the detailed guidance laid out by (Neadle et al., 2025).

Once an ethogram has been produced it should be shared and, where possible, made open access. Outside of academic publishing there are many online depositories, such as the OSF, Figshare, or Dryad, where ethograms can be uploaded and made public. The presence of the ethogram will need to be signposted; present or mention it at conferences, share it with societies, and/or discuss it on social media, in blog posts or newsletters. Ideally, primary ethograms should remain working documents, with feedback from users encouraged, and reviewed and updated regularly (e.g. biannually or annually). This will allow the capture of behaviours not observed in the initial study group/s, and definitions to continually be developed and improved. We are in the privileged position to be able to share our macaque ethogram on a website frequently used by macaque keepers and researchers. Researchers should explore whether similar options are available for their species and utilise these where possible.

When using an ethogram, it is important to note all coding decisions and include these in any resulting publications, as it will not be possible to replicate the study without this information. This should include, but is not limited to, what was coded when two or more behaviours occurred simultaneously, whether behaviours were coded as state or event, and any definition modifications (these should be made to the ethogram itself).

The remaining recommendations are aimed at publishers, editors and reviewers. Journal publishers should make it a prerequisite that behavioural papers must include an ethogram containing all behaviours monitored in the study. Where an ethogram from another publication has been used, simply citing this article may replace the full ethogram, but only if 1) the cited paper is open access, and 2) the citing article explains if and how the ethogram was modified. Editors and reviewers should scrutinise ethogram quality, and provide constructive feedback to authors where definitions or ethogram structure are inadequate. Finally, more journals should permit the publication of ethograms without the need for an accompanying study, for example as methods papers, to encourage the sharing of ethograms and improve the reproducibility of behavioural science.

Conclusions

This paper aimed to spark conversation around ethogram design, use and availability, and approaches for improving the ethogram landscape. The paper explored some of the key issues with currently available (or not) ethograms, such as poor-quality definitions and inter-study inconsistencies, framed around the authors’ experiences with Rhesus macaques. To improve the reproducibility of behavioural studies, access to freely available, high-quality, ethograms is essential. One potential solution is the creation of an open-access primary ethogram for each species. Our attempts to implement this for Rhesus macaques have thus far been successful, however the ethogram will continue to be developed and updated, and user feedback encouraged, to allow continuous improvement.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fetho.2026.1869694/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Paul Rose, University of Exeter, United Kingdom

Reviewed by: Robert Kelly, University of Exeter, United Kingdom