7. Measuring Behaviour - choosing appropriate recording methods
Once the observer has chosen to use a sampling method, the second decision that must be made is which recording rule/s to use (how the behaviour is recorded). There are two categories of recording methods; the observer can record the whole period of observation or take samples of behaviour at intervals:
The observer records a complete account of all behaviour units of interest (Lehner, 1996). This technique aims to provide an exact and faithful record of the behaviour and should provide the most complete and accurate data (frequency, duration and sequence). It can be used to record both events and states. The main problem is that it might be impractical or unreliable as whilst the observer is recording one behaviour some other information is likely to be missed (Mills and Nankervis, 1999).
The observation session is divided into short sample intervals and the observer records the behaviour at the end of each sample interval, this time point being known as the sample point. A major practical advantage of this method is that, considering the information recorded and the observer’s work load, it allows a larger number of categories to be measured as well as a larger number of individuals. The observer watches the animals cyclically and records each animal’s behaviour at the designated sample point (Martin and Bateson, 2007). It is important to understand that time sampling methods do not give true frequencies or durations of the studied behaviour, but instead a representation of the behaviour observed. Furthermore these techniques are not suitable for recording sequences of behaviours since the observer takes into consideration only short sample intervals, and records the behaviour that the animal is performing at sample points in time. There are two principal time sampling methods:
Instantaneous recording (or point sampling): the observer only records the behaviour of the animal he or she is observing at certain time points (e.g. every 5 seconds). The time point may be indicated by a stop clock that beeps every 5 seconds. This is particularly useful when the period between time point samples is relatively short. The score obtained is expressed as the proportion of all sample points on which the behaviour pattern occurred (Martin and Bateson, 2007).
For example, the entire observation period is 75 seconds and the sample points occur every 15 seconds (i.e. a sample interval of 15 seconds is utilised). The following behaviours are recorded by the observer.
Diagram to come
The score is 2/5 for “sit”, 2/5 for “stand up” and 1/5 for “lay down”.
An observer could use this technique if interested in recording behavioural states that can unequivocally be said to be occurring or not at any instant of time (body posture, orientation, proximity, body contact, general locomotor activity). This technique is relatively easy and it can be used to obtain data from a large number of group members (Altmann, 1974). However, it can lead to a lower representation of event behaviours, particularly if the sample interval is long.
On the instant of each sample point the observer records the behaviour shown by a cat. If a 30 minute recording session is divided into 15 seconds sample intervals (the observer records the behaviour every 15 seconds) and the cat has been seated on 30 out of the 120 sample points (4 samples of 15 seconds per minute in 30 minutes are equal to 120), the score for sitting behaviour would be 30/120=0.25.
The observer is interested in studying a crib-biting horse. The recording session is 2 hours and it is divided into 1 minute sample intervals The horse has been crib biting on 60 out of 120 sample points, the score would be 60/120=0.50.
One-zero sampling: in each sample period the observer records occurrence or non-occurrence of a particular behaviour (Martin and Bateson, 2007). This is done irrespective of how often, or for how long the behaviour has occurred. The score obtained is expressed as the proportion of all sample intervals during which the behaviour occurred.
For example the observation lasts 75 seconds with a sample interval of 15 seconds. The observer is interested in sitting behaviour.
Diagram to come
The score is 3/5, because the studied behaviour has occurred in 3 intervals.
Two problems with this type of recording is that it over-estimates durations of the behaviour (as if the behaviour occurs during the whole interval, in this case 15 seconds) and it under-estimates the frequency of the behaviour (the behaviour may occur several times during the interval but is only counted as once, for example sitting occurred 3 times between 30 and 45 seconds but it only noted as once).
On the instant of each sample point (e.g. on the “beep”) the observer records whether or not the cat lay down during the preceding sample. If a 2 hour recording session is divided into 1 minute samples intervals and lying behaviour was displayed on 30 out of the 120 sample intervals, the score would be 30/120=0.25.
The observer is interested in stress signals displayed by dogs during the veterinary examination and records whether or not the dog has licked its lips during the preceding sample. The recording session is 10 minutes and it is divided into 15 second sample intervals. Lip licking was displayed on 15 out of the 40 sample intervals, the score would be 15/40=0.38.
Choosing the sampling interval: choosing an appropriate sample interval is essential. The size of sample interval depends on how many behavioural categories are being recorded and their nature. Shorter sample intervals are more accurate, but it is more difficult to reliably record many different behaviours, especially if they are complicated or occur rapidly. If the time period is long, collecting an adequate amount of data could be time-consuming. In practice the observer must strike a balance between the theoretical accuracy of measurement, which requires the shortest possible sample interval, and ease and reliability of measurement, which require an adequately long interval (Martin and Bateson, 2007).
Choosing video or live recording:
Video recording gives a visual and audio record (provided there is a microphone) of the studied behaviour. Video recording allows playback many times so the observer has the opportunity to watch and identify the behaviour more accurately over multiple viewings. Furthermore recordings can be analysed repeatedly, in different ways and by different observers (e.g. to measure inter-observer reliability); this is useful especially if the behaviour that the observer wants to study is very fast or too complex to analyze in real time. In addition, it removes the problems of observer bias and thus as long as the animal is well habituated to the video camera, there are less likely to be external influences on its behaviour.
However, there are problems associated with video recording as well and include a restricted field of view, so details and context may be lost; and the analysis of the recordings can be exceedingly time-consuming. Furthermore video recording is rarely a complete replacement for direct live observation. Behaviours are generally easier to observe and analyse live and in context, rather than by watching them later on a screen.
For live recording, particularly if the observer is making handwritten notes to record the behaviours; he or she may miss important behaviours which are displayed while they are writing. This is especially likely if the observer is studying many animals. Other potential problems with live recording include the fact that during live recording the observations could be biased by something that attracts the observer’s attention (e.g. vocalizations) and that after many hours of observations the experimenter may fatigue and thus less able to identify brief behaviours (e.g. lip licking).
It is important to remember that the position of the observer may influence the animal’s behaviour (operator effect or observer bias), for this reason it should always be stated whether the observer is in full view of the animals or hidden from them. If the observer is in view it is necessary to give some time to the animals to habituate their presence. However, even with an observer out of view (e.g. in a den), the animal may detect the observer’s smell and habituation is still therefore important. Likewise, with a camera set up where the animals can see it and no operator is present, they need to be given an opportunity to habituate to the novel object within their environment.
Task: Choose a species and scenario and observe live using sampling method of your choice whilst at the same time video recording. later, analyse the video recording utilising the same sampling methods. Compare your results and consider the advantages and disadvantages of the two methodologies.