sampling at a fixed interval, e.g. 2, 15, 30 or 60 min (i.e. counts of its members at regu-
lar intervals). In continuous sampling, data are collected by ad libobservation of indi-
viduals over a variable time period and the size of the party includes all members
considered to belong to it from its beginning to its end; the result can be called cumula-
tive party size.^28 In scan sampling all individuals seen at the sampling times are record-
ed and this results in a series of ‘snapshot’ party sizes. This method suffers from
dependency of the results obtained from scan to scan because successive scans of a party
include some or all of the same individuals. Therefore, when comparing parties from
scan samples it is necessary to allow for data dependency. The size obtained by this
method can be called scan-based party sizeand is invariably lower than cumulative
party size based on continuous sampling.
For particular purposes, i.e. for measuring association frequencies between particular
individuals, scan sampling must be used to be sure that the individuals are indeed
together. At Sonso we currently collect data on parties using 10-min scans; cumulative
data on party size are obtained from scan data by summing the individuals making up
the scans over the life of substantive parties.
As always, data are initially recorded on checksheets in the field and entered into
the computer later. All infants seen are recorded in the field and entered on computer,
but for purposes of calculating party size dependent infants, i.e. infants whose presence
in a party is determined by the presence of their mothers, are excluded.
To obtain mean feeding party size for the Sonso population an analysis was done
on 2100 feeding parties observed on 854 observation days (0.58 of all possible days)
over a 1467-day period. A mean of 2.46 parties (s.d. 1.95, range 1–15) was recorded
per observation day. Mean cumulative party size was 7.01 individuals (s.d. 4.99, range
1–29). A subset of these parties was recorded by means of 30-min scans. A total of 5214
such scans were done, for which, after allowing for data dependency, mean party size
was 5.01 (s.d. 3.57, range 1–26). Newton-Fisher (1999d) gives a comparable figure
based on a separate set of scan data, again filtering the data to remove dependency
effects: 5.703.48 (median 5.0, range 1–19).^29
Mean party size in other chimpanzee studies is in most cases lower than at Sonso,
exceptions being Taï (Boesch 1996) and Ngogo in Kibale Forest, a very large commu-
nity (over 140 individuals) that breaks up into large parties with a mean of 10.27
individuals (Mitani et al. 2002: 105). For all sites the frequency distribution is heavily
Party size 87(^28) This is the number of individuals forming a party over the period of time from the start of observations
until the end of observations of that party. Observations end following a decision to observe another party, or
the disappearance or disintegration of the party being observed. This method is similar, though perhaps not
identical, to that of Matsumoto-Oda et al.(1998) who worked at Mahale. They wrote that ‘parties are all
individuals in a behaviorally cohesive unit that co-ordinated movement during the day’. There is no single
method that can take account of all the complexities of subgroup size changes in fission–fusion societies, in
particular the problem of when one party stops and a new one begins. A party begins either when chimpanzees
arrive or when observation begins, and it ends either when all the chimpanzees leave or when observation is
terminated.
(^29) These are inevitably smaller means than means derived from cumulative sampling because cumulative
sampling includes all the chimpanzees in the party, including those that join it during the life of the party, without
reducing it for those that leave, whereas in scan sampling the party size is reduced when individuals leave.