Biomedical Engineering Reference
In-Depth Information
(i.e. was the behavior seen during the preset interval of
observation? Yes or no?) to calculate probabilities.
Bernstein (1991)
argued that probabilities are every bit as
useful a quantitative measure as is frequency or duration
and that the research investigator must choose measure-
ment techniques according to the properties of the variables
of interest.
sufficient duration so the event is unlikely to be missed as
the members of a group are scanned; and (4) be equally
observable in all possible subjects. These conditions can be
met under many circumstances, but when observers try to
use this technique to measure less visible actions or try to
increase the number of actions being scored during an all-
occurrence scan, the technique becomes equivalent to ad
libitum scoring.
Ad libitum sampling may seem to have no justification,
but it is still useful in obtaining pilot data, in searching for
unusual patterns, and sometimes for detecting the direc-
tionality of social behavior independent of frequency and
duration. The directionality of agonistic behavior in groups
with well-established dominance relationships may be all
that is needed to document dominance relationships. True
frequencies, durations, and probabilities may add little
additional data about which of two individuals is dominant.
Instantaneous Scans
Instantaneous scanning to assess involvement in a behav-
ioral state, like sleeping, requires that a “snapshot” be taken
of the subject at preselected times. When this technique is
applied, the percentage of time spent in the behavioral state
is calculated by dividing the number of snapshots in which
the subject is in that behavioral state by the total number of
snapshots. These point samples of the animal's behavioral
state can be extrapolated to determine the total amount of
time the subject spends in that state. Two caveats are in
order. First, the snapshot cannot be truly instantaneous; it
must be long enough to illustrate that the subject was, for
example, sleeping and did not just happen to blink or close
its eyes. Second, instantaneous scans must be performed at
intervals that exceed the mean duration plus two standard
deviations of most instances of the behavior of interest to
ensure independence of sampling 95% confidence. For
example, because samples taken at 1-minute intervals are
separated by far less than the average duration of a sleeping
period, they are not independent of one another. One
hundred samples of 1-minute intervals during a 2-hour nap
may suggest that the subject sleeps 100% of the time.
Instantaneous scans must be performed at intervals that
exceed the duration of most instances of the behavior of
interest to ensure independence of sampling.
Focal Animal Sampling
Focal animal sampling techniques allow the observer to
record the onset and cessation of behavioral states and
the time of occurrence of an action with a short duration
and to do so for a large number of behavioral categories.
Sequences can also be preserved using this technique;
however, data are obtained by these techniques at a cost.
Data involving animals other than the focal animal and not
directed toward the focal animal are lost. Thus, in a group
of 100 animals, 100 hours would be necessary to obtain
1 hour of data for the group. If the data represent dyadic
social interactions, each hour actually represents 1 hour on
the focal animal plus 1/99 of an hour on each of the possible
interacting animals. Recording the behavior of all animals
for 1 hour each would yield 2 hours of data for the group.
These techniques require a lot of time to collect as much
data as could be collected using instantaneous scans or all-
occurrence scans. The advantage of focal animal sampling,
of course, is that the observer gathers information on the
frequencies, durations, and sequences for many more
behavioral categories than could be studied using the other
techniques.
Ad Libitum and All-Occurrence Scans
Ad libitum scans are defined as free-form observations of
and recording all occurrences of every response without
regard for time. Since no one can watch everywhere at
once,
Altmann (1974)
criticized ad libitum data collection,
on a large battery of possible responses for multiple
subjects, as inherently inaccurate. This type of data
collection obviously tends to include more data on readily
visible individuals and more examples of prominent
attention-catching behavior, like noisy fights, than on
quieter, more subtle, and more rapid responses like a glance
or yawn. Altmann states that if the observer focuses on
a single, readily observable response, he or she can score all
instances of the behavior that occur among a group of
visible individuals. The “all-occurrence” scan requires that
the responses: (1) be readily recognized (perhaps because
the response has a “loud” activity component); (2) have
a low to moderate frequency so that it is unlikely that
multiple episodes will start simultaneously; (3) be of
Sequential Analysis
Sequence data may prove especially difficult to analyze.
One can record all the items in an ethogram during focal
animal scoring and then search for the frequencies of
combinations. These combinations are considered signifi-
cant if they occur more often than the individual frequen-
cies of the two items would predict. Such a comprehensive
analysis may prove exhausting, even if large computers are
used, when one considers an ethogram of 100 items taken
n at a time.
An alternative is to predetermine the combinations or
sequences of interest and to record each combination or
