The Elephant Ethogram: A Library of African Elephant Behavior (TEE) is an ElephantVoices initiative to document the complex and diverse repertoire of African savanna elephant (Loxodonta africana) behavior and communication. In a unique, user-friendly and fully searchable online database, TEE documents the rich postural, gestural, tactile, chemical and acoustic communication and behavior of savanna elephants, including rare, novel and idiosyncratic behavior, as well as that expressed in response to people. The Elephant Ethogram is based on our decades of behavioral studies and our photographic, acoustic and videographic collections from Amboseli National Park and Maasai Mara ecosystem, Kenya, and from Gorongosa National Park, Mozambique, as well as on raw video footage from documentaries filmed in the Maasai Mara and Gorongosa, and on footage of unusual behavior filmed by our colleagues and the general public. TEE is built on ElephantVoices previous online behavior and calls databases (and see also Poole & Granli 2011 and Poole 2011), to which we have added hundreds of new behaviors, included video, and improved the structure and functionality of the database. We intend TEE to be a repository for scientific study and comparison, and to inspire broader interest in elephant behavior, conservation and welfare. We, therefore, include introductory and explanatory information suitable for both audiences.  We define 322 Behaviors, 103 Behavioral Constellations and 23 Contexts. The Elephant Ethogram is a living database and we will continue to supplement it over the coming years with currently undocumented behavior and additional or unusual video clips or photographs. We welcome comments, behaviors for inclusion, or submissions of interesting video from our colleagues or members of the public. Please contact us via this email-address. The first version of TEE was made available on www.elephantvoices.org in late April 2021. 


African savanna elephants are among the most socially complex non-human species (McComb et al 2000, Poole & Moss 2008) on our planet, as well as one of the more heavily exploited. As scientists document their extraordinary behavior, elephants are increasingly impacted by humans and their lives are changing. The Elephant Ethogram is a multimedia catalogue, or library, that aims to describe the characteristics and, where possible, the functions of the behaviors that have been recorded for African savanna elephants (Loxodonta africana). We intend TEE to be a repository for scientific study and comparison, and to inspire broader interest in elephant behavior, conservation and welfare. It replaces ElephantVoices' online Elephant Gestures and Elephant Calls Databases developed originally in 2003 and revamped in 2011 (and see Poole & Granli 2011 and Poole 2011), that were based on our decades of study in Amboseli, Kenya and the work of other scientists (e.g., Douglas-Hamilton 1972, Kahl & Armstrong 2000, 2002, Moss 1983). Unique in their detail and scope, the original databases constituted the only accessible online documentation of the repertoire of African elephant behavior. TEE combines, restructures and dramatically expands on those databases and includes many more behaviors, annotated video clips, higher-resolution images, additional audio files and examples from additional populations. We aim to create an exhaustive list of the behavior of this species.

Further to our behavior studies of elephants in Amboseli between 1975 and 2009, we carried out elephant behavior and conservation projects in Maasai Mara, Kenya and Gorongosa, Mozambique between 2011 and 2019, during which we collected images and video of the behavior of known individuals. Furthermore, in a groundbreaking collaboration with copyright owners Off the Fence (The Netherlands), Gorongosa Media Project (US/Mozambique) and Bob Poole Films, hundreds of hours of raw footage of elephant behavior, originally shot for documentaries in Gorongosa and the Mara, were granted to ElephantVoices for use in science and education. Since we worked in collaboration with the filmmakers, this footage is also of known individuals. From these two sources, as well as our own footage from Amboseli, Mara and Gorongosa, we are now able to document behaviors with video and to report behaviors not described in our original databases. 

From the hundreds of studies published, we know that elephants can show great variation and flexibility in behavior. Some of these studies have included partial ethograms to describe terms relevant to the aims of that particular study (e.g., Goldenberg & Wittemyer 2018, Moss 1983, Poole 1987, O'Connell-Rodwell et al 2011), but none have attempted to document an exhaustive ethogram (but see Poole & Granli 2011). The construction of exclusive ethograms to describe a species' behaviors or activity patterns is commonly used in behavioral studies, where the ethogram focuses on the behaviors of interest. It is more unusual to find ethograms that attempt to produce an exhaustive catalogue of all known behaviors of a given species. One example is the work of Nishida et al (1999) on the chimpanzees of Mahali, Tanzania, which aims to provide an exhaustive list of behaviors and includes 316 anatomical terms, 81 complex anatomical terms, 37 simple functional terms and 81 complex functional terms for a total of 515 behaviors. This body of work also notes the universality of these behaviors, recording whether they were idiosyncratic, limited to a small group, to one population or were, to a greater or lesser extent, cross-cultural. Another example is the work of Bolgan et al (2014) on the Arctic charr, Salvelinus alpinus, a fish species

A draw-back of these studies is that they are non-searchable, written documents without photographic or videographic documentation. Very few studies have attempted to produce an exhaustive, searchable online ethogram of a species with video examples of behavior (but see Mouse Ethogram: An Ethogram for the Laboratory Mouse developed in the Stanford Medical School in the Lab of Dr. Joseph Garner). The Elephant Ethogram, describing 322 Behaviors, 103 Behavioral Constellations, with the occurrence of each noted by age, gender, active body part, context and likely mode of communication, and containing a searchable illustration, is a unique attempt to do so.

The Elephant Ethogram is a living database and we will continue to supplement it over the coming years with currently undocumented behaviors and with additional or unusual video clips or photographs. Creating and populating TEE has been a multi-year endeavor and we acknowledge its inevitable shortcomings and mistakes. We invite people to inform us of any corrections to be made or references to be included. At the time of its publication, TEE documents only African savanna elephant behavior and includes media files primarily from three populations, Amboseli, Maasai Mara and Gorongosa, although there are a few from Tsavo, Kenya, and Kruger, South Africa. We hope to expand TEE to include files from other populations as well as references to African forest elephants (Loxodonta cyclotis). We appeal to anyone with documentation of interesting or unusual behavior to consider sharing it with us for inclusion in TEE. All such contributions will be credited. We welcome submissions from our colleagues or members of the public of descriptions, photographs, recordings or video of interesting or unusual behavior and invite you to do so via this email-address

The first version of TEE was made available on www.elephantvoices.org in April 2021, which includes an Introduction, a User Guide, the Ethogram Table, the Search Portal, The Science and Acknowledgements. This Science section includes an Abstract, Introduction, Methods, a basic set of Results and Discussion of them, and References. Once TEE has been fully populated we will expand the Results and Discussion with conclusions that can be drawn from the data. A scientific publication about The Elephant Ethogram was published in Pachyderm 30 October 2021.


The elephant populations

The Elephant Ethogram documents behavior observed in three primary elephant populations: Amboseli and Maasai Mara, Kenya and Gorongosa, Mozambique. We include a short contextual description of each population touching on location, habitat type, historic level of security and other factors that might affect the behavior we document in The Elephant Ethogram.


Amboseli National Park is located in Kajiado County in southern Kenya, at the northern base of Kilimanjaro. The 390 kmnational park lies within the Amboseli ecosystem, a roughly 8,000 km2 area that straddles the Kenya-Tanzania border (Western 1973, Lindsay 1982). Apart from a series of spring-fed swamps, there are no perennial rivers, and scarcity of permanent water is a salient feature of the ecosystem. Rain falls in two seasons, the short rains in November-December and the long rains in March-May, accumulating an average of 340 mm per year. The ecosystem's dominant vegetation includes open-bushed grasslands, Acacia-dominated bushland and, inside the park, swamp, swamp-edged grasslands and patches of Acacia woodlands (Pratt et al 1966). The Amboseli elephants are primarily grazers and may gather in large aggregations of several hundred elephants during seasons of abundance. 

In the late 1960s Western (personal communication) estimated the number of elephants in the ecosystem to be at least 1,200 individuals. This coincided with the beginning of a period of heavy poaching in Kenya, which in the Amboseli region lasted through the early 1970s. In 1973, a year into the Amboseli Trust for Elephants now five-decade-study (see Moss et al 2011), the population was estimated to be about 600 individuals. By 1976, when 90 percent of the population was identified, there were 512 known individuals (Croze & Lindsay 2011). By that time most of the large adult males initially registered had disappeared, presumably killed by poachers. Since that early period the population has gradually increased, tempered by drought years associated with higher mortality and lower rates of conception. In 2020 the population numbered some 1,700 individuals. While poaching for ivory has remained relatively low compared to other elephant populations in the country, conflict between elephants and people outside the park has been a continuing challenge (Browne-Nuñez 2011; Sayialel & Moss 2011). Between 2000 and 2007 a mean of 8.3 elephants were speared each year of which 5.2 died (Sayialel & Moss 2011). 

While the Amboseli elephants are very tolerant of people in vehicles, they react defensively to the sight, sound, voices or scent of Maasai and/or their livestock (Bates et al 2007, Kangwana 2011, McComb et al 2014). Interactions between elephants and people out-of-vehicles are primarily limited to outside the national park except in several specific locations where permission has been granted for watering of livestock.

Maasai Mara

The Mara ecosystem, located in Narok County in southern Kenya, is composed of open grasslands, dense bush, open wooded savanna and riverine forest. The ecosystem includes the Maasai Mara National Reserve (MMNR), surrounding conservancies, group ranches and individually held land parcels. Significant movement of elephants south across the Kenya-Tanzania border to Serengeti and east to the canopied forest of the Naimina Enkiyio makes it difficult to define the “Mara elephant population.” The Kenyan component of elephant range encompasses some 8,000 km2, of which about 1,500 km2 lies within the MMNR.

The first aerial survey in 1961 counted 1,157 elephants in the entire Mara-Serengeti ecosystem, of which 455 elephants were in the Mara (Talbot and Stewart 1964). Elephant numbers increased in the 1960s, but poaching in the 1970s and 1980s affected elephant numbers and distribution across the Mara-Serengeti ecosystem. In the two decades that followed the ivory trade restrictions that came into effect in 1990, elephant numbers in the MMNR and surrounding dispersal areas varied between 1,000 and 3,071 individuals (see Kiambi, 2012, Poole et al 2016). In 2011 poaching began again and between 2011-2016 some 400 elephant deaths were recorded. The proportion illegally killed (PIKE) ranged between 58-83% with some areas reporting a PIKE of 90% (Poole et al 2016). The June 2014 aerial count north of the border revealed a marked decline of elephants to 1,448 (Mduma et al. 2014).

Human and livestock population increases and changing land use and settlement patterns form the context of the Mara elephant picture. Many of the Mara elephants had wounds and abscesses known to have been inflicted by spears and arrows (ElephantVoices' Mara Elephant Who’s Who [LINK perhaps] and photographic collection). We noted that human activity and livestock grazing influenced the location of elephants, their behavior (frequency of vigilance and other defensive behaviors), patterns of movement and tendency to aggregate. The Mara elephants adopted strategies to augment their safety, abandoning highly insecure areas and seeking refuge in protected “safe-havens”; hiding in dense habitat during the day; moving rapidly through settlement or open areas at night or; using the cover of dense river valleys or luggas to move from one safe place to another (Poole et al 2016).


Gorongosa National Park (GNP) is located in Sofala Province in central Mozambique. The park encompasses 3,770 km2 and is characterized by dense Acacia–Combretum savanna woodlands and floodplain grasslands (Stalmans & Beilfuss 2008). The mean annual rainfall is 700–900 mm, with a rainy season from November to March and a long dry season from April to October (Stalmans & Beilfuss 2008). Lake Urema, located in the center of the park, provides a permanent source of water throughout the year. There are several rivers in the park, including the Urema and the Pungue, the latter forming the southern border of the park. During the wet season the floodplains around the lake are inundated.

During Mozambique's civil war (1977–1992), wildlife populations, including elephants, were killed for meat and ivory and declined dramatically (Daskin et al 2016; Vines 1991). Prior to the war, Gorongosa was estimated to have about 2,200 elephants (Tinley 1977), but the population was reduced to fewer than an estimated 200 individuals by 1994 (Poole & Granli 2017).

Since 2007 renewed investment in wildlife management has enabled elephant populations to begin to recover (Stalmans & Peel 2016). The current elephant population in GNP is estimated to be between 567 (aerial total count, Stalmans & Pee, 2016) and 825 individuals (based on individual registration of adults and a mean ratio of 2.4:1 of immatures to adult females; Poole & Granli 2018).

More than a quarter of a century has passed since the end of the war, but Gorongosa's elephants retain both physical and psychological scars. Tusklessness among females is a naturally occurring and inherited trait in African elephants, with the frequency of occurrence typically ranging from 2-4% in populations subjected to low rates of ivory poaching. In Gorongosa, the selective killing of tusked individuals during the war has resulted in an unnaturally high proportion of tuskless female survivors and their offspring. Approximately half of Gorongosa’s adult females are tuskless (Poole & Granli 2013). The history of killing and antagonism has instilled this population with a fear of humans and vehicles. Based on our observations, family groups in particular exhibit a high rate of aggressive behavior towards people, despite very low levels of elephant poaching at present (Gaynor et al 2018, Poole & Granli 2017, Poole & Granli 2018).


Terms used in the Combined Dropdown Search


A Behavior is a unique movement or action in response to a particular situation or stimulus. When the word Behavior(s) is capitalized we are referring to Behaviors that are defined in TEE, whereas the term without capitalization refers to generalized behavior of elephants. The structure of The Elephant Ethogram is built upon uniquely observable Behaviors (e.g. Chin-In, Ear-Wave, Trunk-Twisting). Each of 322 Behaviors is documented by a detailed written description, noting: its form; its function, if apparent; the ages and genders of those observed to engage in the behavior; the Behavioral Context(s) or "Context(s)" in which it occurs and video examples. Where relevant, we include audio examples. When the Behavior occurs in more than one Context, we note the age and sex of participants by Context. We aim to include video examples of the Behavior as it is expressed in the different Contexts.

Behavioral Constellation

A Behavioral Constellation is a suite of Behaviors that typically occur together (e.g a Musth-Walk typically includes the Behaviors: Ears-Stiff, Head-High, Chin-In, Purposeful-Walk, Musth-Temporal-Gland-Secretion, Urine-Dribbling). We define 103 Behavioral Constellations. Each Constellation is documented by a detailed written description (including its form, function, the ages and genders of those engaging in the behavior), the "Context(s)" in which it occurs, the Behaviors that are likely to be observed, and video examples. Where it is relevant we include audio examples. Similar to the pages for Behaviors, we describe the Contexts in which the Constellation occurs and, when it occurs across more than one Context, we note the age and sex categories of participants by Context. Likewise, we aim to include video examples of the Constellation as it is expressed in the different Contexts.


Behavioral Context, or "Context," refers to the particular setting, situation or circumstances in which a Behavior or Behavioral Constellation occurs. We define 23 Contexts. Most Behaviors and many Constellations occur in more than one Context, and many appear to have different functions or to signal different meanings in different Contexts. For example, Trunk-to-Genitals may be a form of assessing sexual or physical state in an Advertisement & Attraction Context, a way to solicit suckling in a Calf Nourishment & Weaning Context, a gesture of reassurance in a Calf Reassurance & Protection Context, or one of conciliation in an Aggressive Context. Ear-Folding may be a threat in an Aggressive or Mobbing & Attacking Context or be part of a suite of behaviors associated with high arousal in Advertisement & AttractionAffiliative, Birth, Courtship and Coalition Building Contexts.

We define the Contexts in the Ethogram Table and here:

Age classes

Following established Amboseli conventions, we categorize elephant ages in two ways: by absolute age, where known (years and accuracy code), and by age class.

Age classes are:

We use the terms:

Modes of communication

Sounds of Elephants

Other terms

Data Collection

Elephant sightings

Over the course of 45 years we studied elephants in Amboseli (1975-1989; 1998-2009; 2020), Tsavo (1998), Maasai Mara (2011-2015) and Gorongosa (2011-2019). We collected sightings data on each group that we observed (see Definitions for explanation of "sighting"). In the early years in Amboseli (1975-1989) we carried out focal and scan sampling (Altman 1974) and from 1985-2009 and in 2020, when were were recording sound, we collected additional data (see Audio recording below). We collected ad libitum field notes to document interesting behavior. 

The elephants of Amboseli have been monitored by the Amboseli Trust for Elephants (ATE) since 1972; all are individually recognized and most are of known age. Data on the life history of each elephant are held by ATE in a demography database and every observation, or sighting, of each individual is held in a sightings database, both in Microsoft Access. More information can be found in Moss et al 2011.

In Maasai Mara and Gorongosa we built four online databases to house similar information for the two populations: the Mara Elephant Who's Who and Whereabouts Databases and the Gorongosa Who's Who and Whereabouts Databases. Both databases are live, although the Gorongosa database is password protected. We photographed each individual in each group we sighted and registered new individuals in the respective Who's Who Database and entered the group sighting in the Whereabouts Database. Following protocols established in Amboseli, we registered individual elephants with an ID number, noted his or her family, putative mother, estimated birthdate (each with a code for accuracy), together with 6 photographs, and we coded in information on 8 physiognomic characteristics. We collected sightings data via a cellphone application and included: observer, date, time, location, group type, number of elephants, individual identities, individuals in musth or estrus, information on wounded elephants and field notes.


In a collaboration with copyright owners, Off the Fence (The Netherlands), Carr Foundation/Gorongosa Media Project (US/Mozambique) and Bob Poole Films, hundreds of hours of raw footage of elephant behavior from the Mara and Gorongosa, originally shot for documentaries, was granted to ElephantVoices for use in science and education. 2015 footage from the Mara focuses on the life of an infant elephant (from a day old to about three months). Gorongosa footage includes all of the elephant sequences captured in 2012 and 2013 for a five-part film on Gorongosa National Park, in which elephants featured prominently, as well as additional footage captured in 2014 and 2015 by the Gorongosa Media Project. In both cases we worked closely with the filmmakers and, therefore, have relevant individual identifications and sightings data for most of this third party footage. Together with our own footage, captured primarily between 2017 and 2019 in the Mara and Gorongosa and from 2020 in Amboseli, these media put us in a unique position to document the behavioral repertoire of African savanna elephants, as well as to record how environmental and historical factors, such as variation in habitat and human interference, may affect its expression. 

The third-party footage from the Mara and Gorongosa was filmed with a variety of professional cameras. Our own footage was filmed with a Canon 80D with a 70-200 zoom lens, DJI Mavic Pro drone and, in 2020, two iPhone 11 Pro Max. We recorded audio using external microphones attached to the cameras (Shure MV88 microphones with the iPhones and a Røde Microphone with the Canon). To accurately record very low frequency elephant calls we simultaneously recorded with Earthworks QTC1 microphone connected to a Nagra Ares BB with the high-pass filter removed (see the section on audio).  


We recorded elephant vocalizations in Amboseli between 1984 and 1990, from 1998 through 2003, intermittently between 2004 and 2009 and for two months in 2020. In 1998 we also recorded elephants in Laikipia, Maasai Mara and Tsavo, Kenya and in 2017 and 2019 in Gorongosa, Mozambique. Prior to 2004, we recorded on an HHB PDR 1000 DAT using an Earthworks QTC1 omni-directional microphone (frequency response: 4Hz-40kHz + 1 dB). After 2003 we recorded elephants using a Nagra Ares BB with the high-pass filter removed by the factory and the same Earthworks QTC1 microphone. We have 156 GB of audio recorded at 48 kHz.

In addition to sightings data (see above) we collected a separate set of data when recording sound. These included: field site and specific location, date, tape and track number, track start and end times, channel settings, group size and type, and individuals present. When a vocalization was heard we noted the elapsed time as well as information on the presumed call type, caller, and distance to the source. We use the term call type to refer to the broad categories of calls (e.g. Rumble, Trumpet, Roar, Snort) and the term context-type to refer to different calls within a type (e.g. Begging-Rumble, Musth-Rumble).

Caller and context-type were recorded with level of sureness (A: certain - sound came from the direction of the elephant and it behaved in a way associated with calling - lifted head, opened mouth, flapped ears, etc.); B: fairly confident, C: educated guess; D: no idea). A call context-type assigned sureness “A” required that both the behavioral context and sound quality matched the call designation. Finally, any contextual or other comments about the situation or the behavior of the calling animal were noted.

After January 1999 we focused recording sessions on members of Amboseli's EB family (N=27 at the time). Once we located the family we parked near to a subgroup that provided good visibility (typically 5-20 meters to the nearest elephant). As we began recording we noted the closest adult female and her two nearest adult female neighbors. Once the closest elephant moved more than about 25 meters away we moved the vehicleand started a new track. Although many elephant calls travel long distances, the best quality calls and field data are typically from the closest individuals; accordingly, we varied who we parked closest to.

In Amboseli in 2020, audio recordings with the Nagra Ares BB were made simultaneously with video recordings to ensure that low frequency elephant calls were accurately captured.   


Over the years we used Canon 5D, 6D, 7D and 80D cameras. In the Mara and in Amboseli we used a 70-200 mm lens, while in Gorongosa, due to the need to work at a greater distance from unhabituated elephants, we also relied on a 200-400 mm with a 1.4 extender. In our multi-year studies of elephants, we used photography to document both individual identification and behavior. We have large photographic collections of individually known elephants and their behavior from Amboseli (N=~43,000), Gorongosa (N=~142,000) and Maasai Mara (N=~56,000). For each population, photographs (and video) can be linked to known elephants in a database registry of individuals that holds life history information, and via date and time to a sightings database. 

Naming Conventions

Behaviors and Behavioral Constellations

The name of each Behavior or Constellation is capitalized, and hyphenated if it contains more than one word (e.g. Ears-Stiff, Musth-Rumble, Lets-Go-Rumble). We have used simple anatomical terms (e.g. J-Trunk, Periscope-Trunk, Ear-Wave, Tail-Swat), complex anatomical terms (Coprophagia, Clamber), simple functional terms (e.g. Chase, Charge,) as well as complex functional terms (e.g. Guiding, Helping, Conciliation, Waiting) (e.g., Nishida et al 1999). There is a fair amount of inconsistency in the grammatical structure of the names, with some being verbs and some nouns. For instance, we refer to Ear-Folding and Ear-Lifting, but an Ear-Wave and Ears-Stiff. Where Behaviors and Constellations are mentioned in the descriptions and in the video annotations we adjust the names to make the sentence grammatically correct, thus, "Pascal Ear-Waves at Paolo," "Angelina Waits for her twins, Looking-Back to see whether they are coming," or "Little Male Ear-Folded at Tim." 

If we were able to detect consistent differences in the appearance or form of similar Behaviors we assigned them different names (e.g. Circus-Pose, Periscope-Trunk, Trunk-Sweep, Reaching-High). Conversely, we have given the same name to all behaviors that have the same appearance. Thus, for example, Periscope-Sniff (Poole & Granli, 2011) and Distant-Frontal-Attitude (Kühme, 1972) are now combined and referred to as Periscope-Trunk. Our aim has been for all Behaviors of the same form, to have the same name, even though they may have different functions or carry different meanings in different Contexts. For example, depending on the Context, Trunk-to-Mouth may be a gesture of greeting or of reassurance, a signal to initiate a coalition or to signal victory, to solicit food, as well as, possibly, to assess physical state. Periscope-Trunk may be used to sniff a scent carried on the wind, to signal "heads-up" to others, to solicit suckling, and, in anticipatory situations, possibly to signal preparedness. 

An exception to this rule is in the context, Death. There is growing interest in the study of death and dying across species and elephants stand out in the animal kingdom for the variety and duration of behavior that they display toward the dead of their own species. We have, therefore, elected to give specific names to Behaviors directed toward dead elephants. So, for example, Flehmen-Carcass, Back-Toward-Carcass, Pull-Carcass, etc. These Behaviors may be seen on the Ethogram Table under the Context Death.

Where possible, we have used previously published Behavior names. There are, however, instances where we have had to use new terms (e.g., the example above where Periscope-Sniff and Distant-Frontal-Attitude have both been renamed Periscope-Trunk). In these cases we have included the original names along with the references. In some cases we were not able to use previously published names because, without video, it was not possible to understand what the author was describing.   


ElephantVoices' original online elephant behavior and calls databases (2003) described eight overall General Contexts (GCs) with these divided into one or more Narrower Contexts (NCs). These are referred to in Poole & Granli 2011 and Poole 2011. In building The Elephant Ethogram, we decided that the GCs were too general and some of the NCs unnecessarily specific. We dropped the two tiers and used a revised list of 23 Contexts, also in an effort to make The Elephant Ethogram more user-friendly. The 23 Contexts are defines in the Ethogram Table and in the Definitions, above. Figure 1 lists our previous GCs (in back) and, using color coding compares our previous NCs to our revised list of 23 Contexts. Many Contexts are the same (or have a slight name adjustment) as the previous GCs or NCs, while we have merged and split others and added three new Contexts to cover Birth, Maintenance and Foraging and Comfort Technique.

After studying the dramatic behavior of the Gorongosa elephants toward humans, we deliberately designated five Contexts to cover elephants' complex anti-predator response. These are Vigilance, Avoidance, Coalition Building, Conflict & Confrontation, and Attacking & Mobbing. Similar behavior directed toward other elephants, respectively, fall under: Attentive, Submissive, Affiliative, and the last two under Aggressive. 

Defining the number and type of Behavioral Contexts partly comes down to how much one wants to lump or split, highlight or ignore the specific circumstances in which behaviors occur. The more one splits, the more the risk of overlap between similar Contexts. The more one lumps, the less is offered about the circumstances of the behavior. We believe the resulting 23 Behavioral Contexts listed and defined in the Ethogram Table adequately cover the circumstances in which African savanna elephant behavior occurs. Since The Elephant Ethogram is driven by individual Behaviors, not by Context, and each video is annotated in detail, the Contexts can be seen as a guide to the general circumstances in which the behavior occurred.

Figure 1. The General and Narrower Contexts defined in the previous databases and in Poole, 2001 and Poole & Granli, 2011, compared to the Contexts defined in The Elephant Ethogram. 

Screenshot 2021-04-06 at 15.32.20.png

Coding Behaviors and Constellations

For each Behavior or Constellation defined in the database we coded in the following searchable information: 1) the Context(s) in which the behavior has been observed to occur (selected from the 23 Contexts listed on the Ethogram Table); 2) the age and sex classes of elephants that have been observed to (or assumed to) engage in the behavior (adult female, estrous female, adult male, musth male, juvenile or adolescent female, juvenile or adolescent male, calf, infant). Note that we have coded musth male and estrous female for Behaviors and Constellations that are typical of these sexually active individuals; 3) the part of the body actively used to engage in the behavior (entire body, ears, eyes, forefeet, genitals, head, hind-feet, mouth, tail, temporal glands, trunk, tusks); 4) the likely primary mode(s) of communication (acoustic-vocal, acoustic non-vocal, chemical, seismic, tactile, visual).  

Most of the Behaviors and Constellations are signals, while some are just things that elephants do (Drinking, Fell-Tree, Dig-Minerals). All are associated with sounds, scents, postures, actions, movements, touches or vibrations that can convey information to other elephants. Therefore, for each Behavior or Constellation, we have indicated what modes of communication the actor/signaler(s) might be using to communicate with other elephants and/or that other elephants might be using to gain information about the actor. In some cases (e.g. acoustic-vocal for vocalizations) this is obvious, whereas in other cases (e.g., various foraging techniques) this may be conjecture.

We are aware that we have been biased toward coding for visual, tactile and auditory communication as that is what we can see and hear, and that we have under-coded for seismic and chemical communication, because we are unable to detect much of what elephants feel and smell. We included "seismic" for all powerful Rumble calls and for all fast or purposeful movement. It is possible that elephants pick up vibrations of all calls and all locomotion within a certain radius. We coded for "chemical" for all Behaviors that obviously involved elephants using their sense of smell. It is very likely that most foraging techniques, for example, are also associated with the scents of particular plant species, or parts of those species, and that these scents change from wet to dry season. Yet, we coded "visual" not "chemical" for the various foraging techniques, undoubtedly over-estimating the role of visual communication and under-estimating the role of chemical.  

The Combined Dropdown Search can be used to select any combination of these variables to find behaviors of specific interest. Behaviors or Constellations that have an audible component, are coded as either Calls (Acoustic-Vocal) or other sounds (Acoustic-Non-Vocal) and can also be selected under Sounds of Elephants. 

Mining of Media

We "mined" or "trawled through" 18TB of raw video footage, 241K photographs and 156GB of audio recordings and extracted clear examples of any behavior or behavioral patterns that we recognized from many years of observing elephants. In the process we also discovered behavior that we had not previously noted, as well as novel and idiosyncratic . Below we describe the mining and extraction process we used for videographic, photographic and audiographic material.


Video is stored on ElephantVoices server and hard drives. We searched through and mined 18TB of footage from three elephant populations, Amboseli, Gorongosa and Maasai Mara. The raw footage included 8TB of third-party footage from Maasai Mara, Kenya; 5TB of third party footage from Gorongosa National Park, Mozambique; as well as our own footage, which included 200GB from Maasai Mara, 1TB from Gorongosa and 3.6TB from Amboseli.

The third-party footage was imported into Adobe Premiere Pro projects, which we named: Mara, Gorongosa 2011, Gorongosa 2012, Gorongosa 2013, Gorongosa 2014 and Gorongosa 2015. Our own material was imported into Adobe Premiere Pro projects: Amboseli 16, Amboseli 19, Amboseli 20, Mara, Gorongosa 16, Gorongosa 17, Gorongosa 19. Within each Adobe Premiere Pro project, the footage was organized in folders by date, by group and by camera, and placed in an overarching folder called Footage. From each day, camera, or in the case of our own footage, each individual elephant group, sequences were created. These sequences could be linked back to a particular elephant group sighting and individual ID data and field notes that had been collected on the elephants and stored in separate online databases.

In a separate Behavior folder additional sequences were prepared for each previously known and described Behavior or Constellation (see Poole & Granli 2011; Poole 2011). To this original set we added new Behaviors as we observed them in the footage. We scrolled through each sequence frame by frame, looking for clear examples of Behaviors or Behavioral Constellations. Using markers we carefully described the individuals and behaviors involved, noting the elephants' ages, sexes and individual identities, where known. Each time we detected an easily visible example of a particular behavior, we copied the selection onto the relevant Behavior/Constellation sequence. Since many of the Behaviors filmed involved a complex series of interactions between more than one elephant and multiple Behaviors, we often copied the same selection, and its marker description, onto several different sequences. The process of going through the material took several years to complete. From each Behavior and Constellation sequence we prepared multiple annotated clips as described under Editing and Annotating Clips below.


Audio recordings of elephants are stored on ElephantVoices server. Details of elephant call acquisition and measurement prior to 2004 can be found in Poole 2011. In recent years calls are logged from field notes into a FileMaker Database. The audio files included in The Elephant Ethogram were moved directly from ElephantVoices previous Elephant Calls Database. To these we have included additional examples of poorly represented calls. 


Elephant photographs are stored on ElephantVoices server. Photographs from each population are stored in sub-folders by sighting and labeled with the date, time and number of elephants. Each folder matches the group sightings data that we collect in the field. We used Adobe Bridge and NeoFinder to view and key-word each photograph with individual elephant IDs and, where relevant, with Behavior and Constellation names. From these key worded photographs we selected illustrations for the Slideshow of Behaviors on the Search Portal page.

Editing and Annotating Video Clips

Once Behavior/Constellation sequences were populated from available footage, we began the process of editing them for use in The Elephant Ethogram. To each selected example, we added the Behavior or Constellation name on the lower left, ElephantVoices name on the lower right, and footage credits at the end. If the Behavior/Constellation was a vocalization we added a speaker icon after the Behavior/Constellation name.

To help the viewer focus on the Behavior in question, we frequently highlighted it with a circle, as it occurred, or followed its movement with a moving circle. In some cases we followed the main clip with the specific Behavior replayed in slow motion.

In the Behavioral Constellation clips we added the individual Behavior names in italics as they occurred in the clip. Many of the clips of Behavioral Constellations involved several individuals engaged in numerous Behaviors. It was not always possible to include all of the Behaviors. 

We annotated each clip with a detailed caption describing the context, the individuals involved and the specific Behaviors they engaged in. Since we studied each of these populations, we were familiar with the individual elephants in the vast majority of the video examples. We referred to known elephants by name or ID code, otherwise we noted them by age, sex and characteristics.

Populating The Elephant Ethogram


As we prepared clips for export we entered all information (Behavior or Constellation Name,  Context, Population, Filename, Caption) into a FileMaker Database. Thus, the marker name and caption on each of the prepared examples match the entry in the FileMaker Database. We exported clips in H264, 1080px, .mp4. format and stored them on ElephantVoices' server.

We uploaded prepared clips, together with their captions and the context in which the Behavior/Constellation occurred to ElephantVoices Vimeo-account and, simultaneously, linked them from there to the specific page on The Elephant Ethogram where we described the respective Behavior or Constellation. The date of upload and the Vimeo URL were entered on the same line in the FileMaker Database.


We selected a sample of elephant calls (n=~300), in .wav format, from our FileMaker Elephant Calls Database and uploaded these and related metadata (population, annotations regarding callers and behavioral context, file name) to ElephantVoices account on SoundCloud. We, consequently, linked each call from there to the page on The Elephant Ethogram where we described the respective vocalization.  


From the key worded images in our collections, we selected one illustration each for a sample (n=~120) of Behaviors and Constellations and uploaded these to The Elephant Ethogram on elephantvoices.org to create a Slideshow that appears on the Search Portal [Link]. Each photograph includes a title depicting the Behavior or Constellation it illustrates and is linked to the page where the Behavior or Constellation is described and documented with video and/or audio. 

Results and Discussion

The results presented below offer a summary of the collection. We present results based on two types of data: 1) the number of videos and 2) the characteristics of each Behavior coded into the database.

Comparing frequencies of Behaviors within or across populations was not an original objective of the TEE project and we acknowledge numerous biases in the data (see below). While the data indicate broad differences between populations that appear to be related to environmental opportunities and perceived threats as well as historical events, the results must be viewed with caution.

Biases in the Data

Biases in the data were introduced at various steps along the way. First, we began this work with some 40 years of experience observing elephants and had already produced two  savanna elephant ethograms - one on vocal repertoire and another of non-vocal behavior. We, therefore, came to this task with well-formed ideas about elephant behavior as well as with the majority of behaviors and calls already named. While we aimed to stay open to proving previous assumptions wrong, our long experience may well have introduced biases to the data. We hope that those who use TEE can help us to correct any mistakes.

The film footage used has a number of inherent biases caused by the aims of the filmakers or the scientists, or by the responses of the elephants to them. For example, the third party Mara footage focused on the life of an infant; our fieldwork in Amboseli in 2020 aimed to capture missing reproductive behavior; Gorongosa's elephants are notoriously fearful and aggressive toward people. 

Additional biases were introduced during the mining and editing processes. Joyce Poole trawled through all of the footage, and selected examples of representative behavior. Her decision to make a clip was based on three criteria. First, and foremost, was how representative it was of a particular Behavior. The second criteria was how well the photography captured the Behavior so that it could be clearly seen and understood by an inexperienced observer. The third criteria was the aesthetic quality of the photography (e.g., focus, lighting, camera jitter, unwanted background noise, etc).

While a documentary filmmaker selects only the very best footage, we know from experience that rare or interesting behavior always seems to occur just as the photographer struggles to get the vehicle into position, re-focuses the lens, adjusts the aperture, or is in the middle of having a cup of coffee. We apologize to the cinemaphotographers for sharing these less flattering moments of their work, but aesthetics was not our top priority. While, the main criteria was how well the behavior was represented, we also remained mindful of not creating so many examples of common behaviors that database population would become an impossible task. We originally aimed to have about five video examples of each Behavior, but since many Behaviors and some Constellations occur in very different Contexts, we quickly realized that five would not be sufficient to offer a good range of examples. As the number of described Behaviors increased, so too grew the task ahead. This was a massive undertaking and selecting footage often came down to a gut response. 

From the perspective of building an ethogram, it might be ideal to have obtained a similar-sized set of examples of each Behavior and Constellation. From a scientific perspective we might prefer a sample size of behaviors that represent the frequency of occurrence of each in the populations that we studied. We have neither. Although we have many examples of some very common Behaviors and none, or very few, of some very rare Behaviors, and a fair sample of the rest, we cannot say how representative these are of the frequency that they naturally occur in each population. 

The Videos

Frequency of Video Examples by Context and Population

Currently (August 2021), The Elephant Ethogram holds over 2,400 video clip examples in 23 behavioral Contexts, with 813 video files from Amboseli, Kenya, 934 from the Maasai Mara, Kenya, 658 from Gorongosa, Mozambique, and three from Kruger, South Africa. While the total number of videos from the main three populations is roughly similar, examples of Behaviors and Constellations from some Contexts  are disproportionately represented from the three populations. Figure 2 shows the percentage of videos across all Contexts from Amboseli, Mara and Gorongosa. 

Amboseli has long been regarded as the best place in Africa to film musth males, estrous females and reproductive behavior, in general. Our specific aim in Amboseli was to document reproductive behavior as this was underrepresented in the footage from the two other populations. Behavior that falls in the Contexts of Advertizement & Attraction and Courtship makes up 30% of the Amboseli clips as compared to 5% of those from Gorongosa and the Mara. We knew, from long experience in all three populations, that we would be more likely to see and to capture reproductive behavior in Amboseli than in either of the other two populations. From a biological perspective it may seem strange for the prevalence of reproductive behavior to be so apparently dramatically different. Amboseli's open landscapes and grassland habitat is conducive to elephants forming large aggregations of individuals during and after the rains when the herb layer biomass is plentiful and when reproductive activity is concentrated (Poole & Moss, 1989). Large musth males and receptive females gravitate to these large aggregations (Poole & Moss, 1989) making reproductively active individuals easier to find, both for elephants and for the scientists hoping to study or film them. Our Amboseli filming took place following a period of historically high rainfall. We were able to observe large aggregations of individuals that included musth males and estrous females almost daily.

Correlated to the high capture of reproductive behavior in Amboseli is the greater percentage of video clips of agonistic behavior. In Amboseli 13% of videos fall in the Contexts of Aggressive and Submissive, reflecting the increased competition between males for mates that we observed and recorded, while only 7% and 4% of videos from the Mara and Gorongosa, respectively, were in these two categories.

The third party Mara footage focused on the first few months of a baby elephant's life, and is, not surprisingly, biased toward mother-offspring and allomothering behavior. Behavior that falls into the Contexts Calf Nourishment & Weaning and Calf Reassurance and Protection makes up 23% of the clips from the Mara, as compared to 10% of the clips from Amboseli and 4% of the clips from Gorongosa. The low representation from Gorongosa is likely due to calves being less visible in the tall and dense vegetation, and because they were often protected within tightly bunched families. In addition, the defensive nature of the elephants meant that we filmed from a greater distance from the animals than we did in either Amboseli or the Mara and, therefore, were more likely to miss behavior involving smaller members of the group.

In Gorongosa, elephants were difficult to find and to film due to the thick vegetation and to their fear of and aggression toward people. While there was no particular bias in the footage the filmmakers aimed to capture, the response of elephants to our presence means that the footage from this population is heavily biased toward defensive behavior. Behaviors and Constellations that fall in the Contexts of Vigilance, AvoidanceCoalition BuildingConflict & Confrontation and Mobbing & Attacking make up 46% of the clips from Gorongosa, as compared to 2% of the clips from Amboseli, and 6% of the clips from the Mara.

Figure 2. The percentage of videos from Amboseli (N=772), Mara (N=936) and Gorongosa (N=666) in the different Behavioral Contexts Precentage of video clips by context by population.jpg

During the period we worked in the Mara some 400 elephant deaths were recorded with the proportion of illegally killed elephants (PIKE) fluctuating between 58-83% and some areas reporting PIKE figures as high as 90% (Poole et al 2016). Our Mara Who's Who and Whereabouts Database and photographic collection documented that a very high proportion of the Mara elephants had arrow wounds. The footage from the Mara was primarily filmed on Naboisho Conservancy where Maasai herdsmen and their livestock are free to graze, and we observed that elephants were wary of people on foot or the presence of livestock. They stopped to listen to the sounds of approaching people and cattle bells, bunched together, and moved away; they often crossed open areas in a tight group. It is worth noting that Behavior included in the Context Attentive made up a higher percentage in the Mara (10%) than in either Amboseli (3%) or Gorongosa (3%). We suspect this is because we may have thought that elephants were paying attention to other elephants when in fact they were concerned about people we could not see or hear. In Gorongosa it was obvious when to score a Behavior as Vigilance versus Attentive because visible or audible vehicles (often our own) were obviously the source of their concern. Likewise, in Amboseli it was easy to differentiate because the landscape is so open and the elephants were very rarely in the vicinity of herders when we were filming. When we combine Behaviors that occurred in the Contexts Vigilance and Attentive, the number of clips in these two contexts make up only 4% of the clips from Amboseli, while they make up 12% and 13% of the clips fromthe Mara and the Gorongosa, respectively (Figure 3). In Amboseli, within the national park boundaries, encounters with people and livestock occur at predictable times and places and may permit the Amboseli elephants to avoid confrontations and put less energy into vigilance and other defensive behaviors. This situation only applies to the area within the confines of the small protected area, and outside the elephant would likely show more Vigilance

Figure 3. Percentage of video clips of Behaviors occurring in the Contexts Vigilance and Attentive


 Behavior that fell into the overall categories Play, Logistical and Affiliative were more similar across populations. Lone & Object Play and Social Play made up 14% of Amboseli clips, 12% of Mara clips and 13% of Gorongosa clips. Likewise, logistical behavior in the context Movement, Space & Leadership, was also similar across populations, making up 5% of clips from Amboseli, 6% of those from Gorongosa and 7% of those from the Mara. Affiliative behavior was also fairly similar across populations and represented by 5% of Amboseli's clips, 6% of Gorongosa's and 9% of the Mara's. 

One more comparison that is worth noting are videos documenting foraging techniques. In Amboseli (n=18) all videos show techniques related to grazing, while in Gorongosa 92% (n=52) and Mara 98%  (n=46), of the videos document browsing techniques. Based on the habitat in Amboseli National Park and our knowledge of elephant feeding habits there, we expected that the videos would document almost entirely grazing and they did. The videos from both Gorongosa and the Mara reveal that these elephants are largely browsers. One explanation for the very stark difference is that we filmed in Amboseli immediately following a period of historic rainfall and the grasslands were lush, while much of the footage from Gorongosa was from the late dry season and, from the Mara, during a year of low rainfall. Still, we know from our work in the Mara that there is heavy competition with livestock for available grass across the conservancies as well as in a 5-10 km wide strip along the northern boundary of MMNR, where cattle are taken for night grazing. 

Figure 4. Video clips of foraging behavior in Amboseli, the Mara and Gorongosa populations.


 The Number of Video Examples per Behavior/Constellation

Once we have completed populating the database with our own material, we will present a frequency distribution of the number of video examples per Behavior/Constellation. Currently, the TEE holds over 2,300 clips have been uploaded illustrating 425 Behaviors and Constellations. The number of video examples of a Behavior ranges from 0 and 42. 

Results from the Dropdown Search

Using the Dropdown Search Portal we summarize information in The Elephant Ethogram.

The Number of Behaviors by Age/Sex Class

African elephants are known to show age and sex differences in behavior (e.g., Chiyo et al 2011a&b, Lee, 1986, Lee 1987, Lee & Moss 1999, Poole 1987, 1994, 2011, Poole & Granli 2011) as well as to express novel, idiosyncratic and socially learned behavior (e.g., Bates et al 2010, Fishlock et al 2015, Lee & Moss 1999, Poole et al 2005). An ongoing study by Bates and colleagues is examining possible cultural differences in behavior across multiple populations. While there have been hundreds of studies of African savanna elephants, and thousands of scientific papers published on wild and captive elephants, surprisingly few specifically have compared male and female behavior (but see Poole 1994) or examined developmental changes in behavior.

We will carry out a more detailed examination of sex differences and developmental acquisition of Behaviors at a later stage. Here we examine general trends. We coded in the age/sex categories of elephants that we observed or knew to engage in each Behavior. We acknowledge some educated guesses in the aquisition of some types of Behaviors by infants and calves. This is particularly true of the acquisition of various foraging tecniques. For examples, since tusks do not erupt until age two, infants were not coded for any Behaviors that employ tusks. We also acknowledge some educated guesses regarding the stage (calf/juvenile/adolescent) at which males transition from engaging in family Behaviors (e.g., participation in Greetings or other Bonding Behaviors) to those more represenative of adult males. In general we found that number of Behaviors employed increases with age in both sexes, but females engage in a greater variety of Behaviors than do males.  

Figure 5. The number of Behaviors by age/sex class.

Age Sex Class red


The Number of Calls by Age/Sex Class

We also coded in the age/sex classes of elephants employing each of the described laryngeal, trunk and combination calls and idiosyncratic sounds. While the number of calls employed by elephants increased from infant to calf, thereafter the number used by males decreased with age, while the variety used by females continued to increase with age (Figure 6).   

Figure 6. The number of laryngeal, trunk, combination calls and imitated and idiosyncratic sounds by age/sex class.

Age sex class calls.jpg


Body Parts Actively Employed in Behaviors

We also coded the primary body parts employed in each Behavior. For example, for Ear-Folding we coded the ears, for Foot-Forward, the forefeet, for Head-High, the head. Certain feeding techniques involve movement of several parts of the body in coordination. For example, for Grasp-Branch-Break-with-Tusk we coded the trunk, tusks and head. Elephants are known to use both the mouth (larynx-source) and the trunk (as a resonator) in the production of Rumbles. While some Rumbles make use of the nasal cavity more than others, all Rumble subtypes were coded as employing both mouth and trunk. For those Rumble sub-types that are associated with flapping of the ears we also coded, ears. For Behaviors that involved movement of the entire body we coded, body, but not the forefeet and hindfeet. As can be seen in Figure 7, the trunk is employed in many more Behaviors than the entire body or any other body part. In its role as a nose, a hand and a sound resonator, the trunk is use in chemical, tactile, visual and vocal communication. Furthermore, receptors at the tip of an elephant's trunk detect seismic vibrations, making it a truly versatile appendage for communication.   

Figure 7. The frequency with which different parts of the body are employed across Behaviors

Active Body Part.jpg


Modes of Communication

All Behaviors are associated with sounds, scents, postures, actions, movements, touches or vibrations that carry information to other elephants. When coding Behaviors we selected the primary mode(s) of communication we judged that the actor/signaler(s) would use to communicate with other elephants and/or that other elephants would use to gain information about the actor. In doing so we are aware that we have been biased toward coding what we can see and hear (visual, tactile and auditory communication), and that we have under-coded seismic and chemical communication, because we are unable to detect much of what elephants smell or feel through their feet or trunk tip (Figure 8). This graph is presented to stimulate feedback from our colleagues who have more experience with chemical and seismic communication, in the hope that their input can help to make The Elephant Ethogram more accurate.  

We included "seismic" for all powerful Rumbles (not Roars or Trumpets) and for all fast or purposeful movement. The behavior of elephants sometimes indicates that they are able to pick up vibrations of all calls and all locomotion within a certain radius (e.g., a matriarch begining to walk at the back of a long line of stationary elephants, causes those ahead of her to begin moving, suggesting that they detect her movement). If so our coding seriously underestimates the use of seismic detection in communication.

We coded "chemical" for all Behaviors that obviously involve elephants using their sense of smell. We also coded many of these as visual, because it is our opinion that elephants take cues from the movement and positioning of another elephant's trunk. For example, the use of Periscope-Trunk as a sign that there is something of concern in the direction another elephant is directing its trunk (see Smet & Byrne 2020).

It is very likely that many additional Behaviors are associated with scents that can be detected by elephants. For example, Behaviors involved in foraging are likely associated with the scents of particular plant species, or parts of those species, and these scents likely change as food quality improves or declinesthrough the seasons. As TEE currently stands, we coded "visual" not "chemical" for the various foraging techniques, undoubtedly over-estimating the role of visual communication and under-estimating the role of chemical communication.

While it is often claimed that elephant have rather poor vision, we know from our own observations that they can detect very subtle signals (e.g. Ear-Folding from up to 50 meters away) and that they have good peripheral vision. Each eye takes in 190o (123o uniocular and 67o binocular vision). By pressing an ear close to its body an elephant is able to see what is behind it. Furthermore, while elephants may not notice something stationary, they are quick to detect movement, particularly if it is silhouetted against the horizon.

Figure 8. Modes of communication 

Modes of Communication.jpg



You will find a full reference list for both The Elephant Ethogram and the Elephant Communication section through this link.