Behavior Ethogram (San Diego Zoo) and Protocol

Giant Panda Behavior Research Methods
Dr. Ron Swaisgood, C.R.E.S

In this section an overview of the behavior research methods used by San Diego Zoo research staff is presented. The version here is the one used in Wolong, which differs only slightly from the one used in San Diego. We have found that each institution may have slight variations (for example, enclosure design, access to conspecifics) that requires minor modifications to the methodology to accommodate each unique situation. The version offered here is also a somewhat abbreviated version of our working methodology. In the longer version, we provide details of the minutiae and nuances regarding what constitutes a particular behavior, and supplemental rules that guide the observer through many decisions that need to be made in specific—and often rare—situations. Still, the version presented here is quite long and comprehensive, for it meets the needs of more than a dozen different topics subject to study by our research team. We recommend that anyone wishing to use this methodology should pick and choose behaviors and protocols tailored to their own specific research needs.

The observer is provided with a detailed ethogram (a list of behaviors with precise operational definitions), a recording protocol, a behavior data collection sheet, and a data summary sheet.

Behavior data collection sheet.

A place to record standard “header” information is found at the top of the data collection sheet. Here we record the subject’s identity, the date and time, the observer’s name, the enclosure, the temperature and cloud cover, the identity of neighboring pandas, and the number of visitors present. In addition to this standard information, we record an estimate of the quantity and quality of bamboo available during the session. Bamboo availability is rated by the following criteria on a scale from 0 to 3 at intervals of 0.5.

Bamboo quality index:

0. No bamboo present or available bamboo is in very poor condition and most of preferred portions are already eaten. Often recorded for leftovers present in the morning prior to the day’s first feeding.

1. Panda is given a small amount of fresh bamboo (e.g., less than 25% of a “typical” feed or there is little bamboo remaining from a previous feeding and most is old and of poor quality, yet some “desirable” portions remain).

2. Ample relatively fresh bamboo remaining, although some may have been eaten earlier.

3. Ample fresh bamboo present, placed in the enclosure just prior to or during the observation period.

Note: if the status of bamboo changes during the observation session, calculate an average bamboo quality index for the session based on the length of time bamboo of a particular rating was present. For example, if in a 45-min session, a “0” is scored for the first 30 min and a “3” is scored for the last 15 min, the average score is: (0+0+3)/3= 1.

Below the header are 30 rows for collection of behavioral data. Each row represents one minute. The first column is for recording all occurrences of behaviors occurring during the one-minute interval. The second column is for recording of vocalizations using the “one-zero” method. The next column is for recording the “point sample” behavior at the end of each one-minute interval. In the last three columns the panda’s location, social distance to other pandas, and “interaction potential” with other pandas are recorded. For definitions see below.

Behavior Data Recording Protocol

Terminology for behavioral categories.

• STATES (S) are generally behaviors that are long in duration, such as resting, feeding, or locomoting. We only record STATES in all-occurences if the behavior lasts more than 5 seconds (see below).

• EVENTS (E) are generally behaviors that are of relatively short duration, such as somersault, head-toss, and approach.

• STATE-LIKE EVENTS (SLE). Not all behaviors fall neatly into these two categories. Some EVENTS can occur for more than five seconds, making them "STATE-LIKE". For example, a somersault only takes approx. 1-2 seconds, so it would qualify as an EVENT. However, if a panda somersaults down a hill, it could take 7-8 seconds. Therefore, the EVENT "somersault" now becomes a STATE of somersaulting which is longer in duration. These "STATE-LIKE EVENTS" do not have to occur for five seconds before they are recorded because we treat them as EVENTS; however, in all other respects, they are treated like STATES.

We use three fundamental behavior recording methods:

All-occurrence sampling. This method is used to record each instance of a behavior as it occurs, and provides an accurate record of the frequency of the behavior. A behavior may be recorded several times within a one-minute interval. Each EVENT is recorded every time it is observed, regardless of how long the event lasts or how much time passes between each occurrence of the same event. STATES must last for > 5 seconds before it is recorded. A particular state must be interrupted for more than 5 seconds before it is recorded a second time. For example, if the panda locomotes for 8 seconds, rests for 3 seconds, then continues to locomote, only a single bout of locomotion is recorded and rest is not recorded at all. SLE’s need not last 5 seconds, but they must be interrupted for > 5 seconds before another bout of the same behavior is recorded.

One-zero sampling. This method is most useful for recording behaviors that occur too frequently to record each instance (e.g., vocalizations, tongue-flicking) or for behaviors in which the onset and termination are difficult to identify (e.g., tail-up and many forms of play). If a behavior designated as a one-zero behavior occurs, the behavior is recorded once and only once during that one-minute interval. If it is observed several times during an interval, it is still recorded only once. If it is not observed during that interval, it is not recorded. Thus, it is recorded one or zero times per interval, hence the label “one-zero” sampling. Note that if a behavior occurs at the end of one interval and carries over into the next interval, it is recorded in both intervals: the behavior need not stop and start as in all-occurrence sampling. This method of sampling yields a type of frequency, but not the “true” frequency provided in all-occurrence sampling. Only behaviors designated as “1-0” in the ethogram are recorded in this manner.

Point-in-time sampling. This method is used to estimate the percent time that animals spend in various STATES or STATE-LIKE EVENTS. At the end of each one-minute interval a “beep” will sound, and the observer records the STATE or SLE that the animal is engaged in at precisely the time of the beep. Percent time is calculated by dividing the number of intervals the behavior was observed at the “point sample” by the total number of intervals during the observation session (e.g., 30/100=30%). A hierarchy for which behavior to record when more than one behavior is observed at the beep is provided. In general, only STATES and SLE’s are recorded at the point sample. However, a few EVENTS and 1-0 behaviors are also recorded at the beep, designated by “at beep” in the ethogram. These are behaviors that we wish to record frequencies using one-zero or event protocols, but wish to obtain a percent time estimate for these behaviors at the point sample as well. A good example is the stereotypic behavior “head toss.” Most pandas generally head-toss in a very event-like way, e.g,, once will turning a corner. A few pandas, however, spend a great deal of time head tossing, and may stop and head toss 10-15 times in a row. In this case, we’d like to know how many head tosses were observed (i.e., as events), but we’d also like to know how much time they spend head-tossing (point sample). Note that since true events are very short in duration (e.g., a vocalization, though recorded using 1-0 method) an animal is usually engaged in some behavioral state (e.g., locomotion) simultaneously. Thus, if an animal is locomoting and vocalizing at the beep, locomotion is recorded. Note that the behavior data sheet also has columns at the end of each interval to record other information at the point sample, such as location and social distance.

Giant Panda Ethogram, Abbreviated Version

This ethogram is the product of several years of intensive observations of giant pandas. The original version was based largely on published accounts of panda behavior available at the time from the work of Devra Kleiman and George Schaller. Initially, we used a recording system in which each behavior was identified by a numeric code. Often we found that our research needs required that these general categories be divided into several subcategories, and we instituted the practice of identifying subcategories by following the numeric code with a letter (e.g., feed on bamboo = 1a). As time passed we began to see many behaviors that had not been reported before, and we added these to the ethogram. Panda behavior can be very idiosyncratic (especially stereotypies) and as we increased the number of pandas observed in China, we added new behaviors to describe what we observed. We also greatly expanded the scope of our research, and with each new project we added numerous new behaviors to meet the needs of the research question. For example, we now have ongoing research programs on chemical communication, acoustic communication, several aspects of reproductive behavior, social behavior, pregnancy monitoring, mother-infant relationship, behavioral development, psychological well-being, environmental enrichment, cognition, color discrimination, and bamboo feeding. Eventually the numeric system grew unwieldy, and as we added new behaviors, we used a shorthand code that used 2 or more letters from the behavioral descriptor (e.g., BO = Bite Object, AP = approach). The resulting ethogram, therefore, reflects this evolutionary history.

GIANT PANDA ETHOGRAM, SAN DIEGO

Miscellaneous Individual Behaviors.