MANNIKIN VIGILANCE BEHAVIOUR
569
Time budgets
At any session, we observedtime-budgetsof focal individuals,and designatedthe rst colour
banded individual as a focal bird, with the following subject as the next identi ed. We ended
focal samples when the bird left the feeder, or after 5 mins. We recorded the following
information for each sample: colour-bands, location (feeder number), sample duration, total
number of birds on that feeder (calculated as the average of the number at the start and at
the end of the focal session), and total number of birds at all ve feeders. Every 10 s we
noted the behaviour of the focal bird, as either pecking, vigilant for predators, or vigilant for
conspecics (other behaviors were extremely rare while foraging and are excluded).
We de ned behaviours as follows: (1) pecking: a bird with its head in a feeding hole
pecking for seed. (2) conspecic vigilance: a bird raising its head slightly from the feeding
hole, but with its head retained below horizontal and looking at its neighbours. (3) predator
vigilance:a bird liftingits head above the horizontaland cocking its head around in a clear act
of scanning. Lima & Bednekoff (1999) provide an experimental demonstration that cocking
the head in the manner described here as vigilance for predators increases predator detection
ability. We calculated(1) pecking, (2) conspeci c vigilance, and (3) predator vigilance rates
by dividing the number of occurrencesof each behaviour by the duration of the focal sample
(
in s). We made 331 focal observations over 40 days, with an average of 15 (range = 12
to 21) focal observationson each of 22 colour-bandedindividuals.
At each session, we made observations with either all holes on all feeders open (control
treatment), or with random numbers of holes on each feeder closed. We used the control
treatment (all holes open) to test whether mannikins prefer to feed at locations close to
cover, i.e. whether they perceive a predation risk while foraging. In addition, we used control
treatments to describe effectsof distance from cover and group size on feeding and vigilance.
We manipulated aggression levels by closing a random numbers of holes on each feeder.
We increased the probability of displacement by progressively reducing access to the feeders
close to cover (where probability of displacement in highest). This provided a manipulative
test of the prediction of increased vigilance for conspecics at higher probabilities of
displacement(i.e. higher level of aggression).
We selected combinations of holes to close randomly. The control treatment had eight
holes open at each of ve feeders. We performed an additional 10 treatments with various
combinations of holes closed (Table 1). We applied treatments in a random order. On any
one day, the same treatment was applied in the morning and afternoon sessions (with each
treatment being applied for at least two 2 h sessions).
Statistical analyses
Distance from cover
We tested the prediction that bronze mannikins will prefer to forage at the feeders closer to
cover using an ANOVA, with the number of birds on each feeder as the dependent variable,
and the feeder number as the class variable. We used total birds on all feeders, temperature,
and time of day as covariates. Residuals were normally distributed.
We averaged the number of birds on each feeder at each 15 min sample for each 2 h
observation session. This gave one datum per session ( N = 34 sessions). Because the same
individuals may be resampled, there is the potential for pseudoreplication(Hurlbert, 1984).
However, because each session is separated by at least 6 h, and there is a lot of movement
between each 15 min sample, we believe each datum to be independent.