Content uploaded by Johan Van Rhijn
Author content
All content in this area was uploaded by Johan Van Rhijn on Mar 18, 2018
Content may be subject to copyright.
The Patterning of Preening and Other Comfort Behaviour in a Herring Gull
Author(s): Johan G. van Rhijn
Reviewed work(s):
Source:
Behaviour,
Vol. 63, No. 1/2 (1977), pp. 71-109
Published by: BRILL
Stable URL: http://www.jstor.org/stable/4533848 .
Accessed: 22/02/2013 07:10
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
.
BRILL is collaborating with JSTOR to digitize, preserve and extend access to Behaviour.
http://www.jstor.org
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
Behaviour, LXIII, I-2
THE PATTERNING OF PREENING AND OTHER
COMFORT BEHAVIOUR IN A HERRING GULL
by
JOHAN G. van RHIJN 1)
(Department
of Zoology, University of Groningen,
The Netherlands)
(With io Figures)
(Acc. I4-111-I977)
INTRODUCTION
A very important condition for the analysis of the internal mechanisms
underlying a certain type of behaviour, is a thorough quantitative descrip-
tion of the patterning of this behaviour. This prerequisite has been recognized
by a large number of ethologists and psychologists. Hence, much attention
has been payed to quantitative descriptions, and to the methods suitable for
this purpose.
Although at present we have a large number of methods at our disposal
for describing the patterning of behaviour, only a few of these appear to
be very popular. For instance, transition analysis is a very common tech-
nique. Its different variants have been used by a vast number of authors
(e.g.: WIEPKEMA, I96I; NELSON, I964; FENTRESS, I972; SLATER &
OLLASON,
1972; VEENING,
I975). Many improvements and extensions of
this method have been applied: factor analysis (WIEPKEMA,
I96I), interval
analysis (NELSON, I964), information analysis (FENTRESS,
I972), and
correction for impossible transitions (SLATER
& OLLASON,
I972).
We may question why these popular techniques are used so often. Are
these methods very efficient? Is it difficult to obtain more information by
the use of other methods? It is obvious that the answers to these questions
are closely associated with the purpose of the investigation. For instance, if
we wish to decide whether a given type of patterning (e.g. a Markov chain)
can be found in our behaviour records, it is usually sufficient to select only
i) I wish to express my gratitude to Professor G. P. BAERENDS,
Miss M. K. CARL.-
STEAD, Professor J. J. A. VAN IERSEL,
Dr J. REDDINGIUS,
Mrs A. C. A. SEVENSTER-BOL,
Mr W. A. SMIT,
Dr J. VEEN,
Mrs N. & Mr R. VODEGEL,
and Dr P. R. WIEPKEMA
for their
critical comments on drafts of this paper. Their help contributed greatly to the im-
provement of the final manuscript. I also wish to thank Mr D. VISSER for preparing
the figures, Mrs H. LOCHORN-HULSEBOS
for typing the manuscript, and Mrs G. BAEYENS
for correcting the French summary.
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
72 JOHAN G. VAN RHIJN
one method (in the case of a Markov chain: transition analysis); if we
wish to reveal as much as possible about the mechanisms underlying the
behaviour, however, we certainly need a larger number of analytical methods.
The importance of this latter multiple-method approach has been shown by
DELIUS (I969), HEILIGENBERG (1973) and HALLIDAY (I975). In the
present study I shall therefore consider the results of several methods in
order to determine the properties of the internal mechanisms underlying
preening and other comfort movements in a Herring Gull.
Preening occurs in all species of birds. It is associated with bathing,
shaking and scratching. For reasons of simplicity the term 'body-care' will
be used for the combination of preening and the other comfort movements
just mentioned, whereas the term preening itself will be restricted to all
movements with bill and/or head towards all parts of the plumage including
the oil gland. Body-care has been investigated in a large number of bird-
species, for instance, Buntings (ANDREW, I956a, b), Terns (VAN IERSEL &
BOL, 1958), Chaffinches (ROWELL, I96I), Anatidae (MCKINNEY, I965),
Skylarks (DELIUS, I969) and Penguins (AINLEY, I974). Part of these
studies mainly concern the 'irrelevant' occurrences of preening (e.g.:
ANDREW, 1956a, b; VAN IERSEL & BOL, I958; ROWELL, I961), while
among the other studies MCKINNEY (1965) and AINLEY (1974) provided
comparative descriptions of body-care, and DELIUS (1969) investigated the
underlying causal mechanism. Although my investigations were set up as
an extension of the work by BAERENDS (1970) on preening as an inter-
ruptive ('irrelevant') activity during incubation, the present paper only
deals with 'normal' body-care.
Unfortunately I could not isolate the gulls, because (at least when they
are accustomed to living in a group) social isolation induced abnormal
behaviour with a minimum of body-care. Although body-care is not neces-
sarily dependent on the presence of conspecifics or other animals, we have
to take account of some influences of conspecifics. For instance, there are
indications that bathing can be induced by perception of another bathing
gull. Besides, even the patterning of preening may be influenced by con-
specifics: rather long sequences of different movements of a given gull
sometimes appeared to be echoed (with very short delay) by another gull.
To avoid a possible effect of such environmental factors, the quantitative
part of this paper deals with the behaviour of the dominant gull of the
group only. This bird was seldom disturbed by other individuals. It nor-
mally took the initiative in performing bathing and preening. Evidence for
copying the behaviour of other gulls could not be found for this dominant
bird.
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BEHAVIOUR IN A IERRING GULL 73
METHODS
Observational procedures.
The observations which are relevant for this paper were made on a group of four
gulls during winter and spring 1974. This group was housed in an outdoor-cage (5 X 3
meters). Food (only dry pellets) and water in a small dish (unsuitable for bathing)
were always available. The cage contained one bath (70 X 50 centimeters) with a
cover, which could be opened and closed from an observation room. Observations
were
started by the opening of the bath after bathing had been prevented for 3 hours. The
behaviour of one gull was then (during 40 and later 20 minutes) watched in detail aii
verbally recorded on magnetic tape.
Two different series will be discussed. The first series of observations (32 X 40
minutes) included all four individuals (eight observations per gull). Every bird was
watched at four different days in the morning (II a.m.) and in the afternoon (2 p.m.).
This first series was used for a qualitative description. The second series (15 X 20
minutes) only referred to the dominant gull. All these observations started at noon.
This latter series was used for a quantitative analysis.
Classification of behaviour.
In this paper I shall use a number of terms for labelling behaviour.
I. body-care sequence: the recording (40 and later 20 minutes) of body-care behaviour
occurring during an observation.
2. group: broadest category of behaviour. I distinguished only two groups of body-
care behaviour: (a) preening (all movements with bill and/or head towards all
parts of the plumage, including the oil gland), and (b) all other comfort move-
ments associated with preening (containing the bathing-, shaking-, and scratching
movements).
3. element: narrowest category of behaviour. Each group contained a number of ele-
ments. Each element was defined according to the orientation and/or form of the
behaviour.
4. event: smallest unit of behaviour used in the recordings, namely one single occur-
rence of an element. It is either a single series of muscular contractions, or (if
intervals are smaller than 0.5 second) a repetition of the same series of muscular
contractions. The criterion of 0.5 second was estimated while observing.
The terms 'main component', 'bout' and 'string' will be defined later on. All other
terms (e.g. 'movement' and 'combination')
must be taken literally.
An extensive classification of the first group of elements (preening) must be based
upon two properties of a movement: (a) its orientation, and (b) its form ('modal
action pattern', cf.: BARLOW,
I968). The orientations were defined by means of the
locality of the plumage to be preened (in some cases the position of the neck was
taken to make an additional
distinction). A survey of the localities distinguished (with
abbreviations) is given below. It is mainly in conformity with Van lersel & Bol
(1958) and needs no further explanation.
i. throat (th), 2. breast (br), 3. belly (bl), 4. shoulder (sh), 5. wingbow (ww), 6. flank
with neck high (via back) (fh), 7. flank with neck low (via belly) (fl), 8. outside
of the wing (ow), 9. inside of the wing with neck high (via back) (hw), Io. inside
of the wing with neck low (via belly) (Iw), ii. pinions (pn), I2. back (bk), I3. tail (tl),
14. oil gland (og).
The different forms of the movements with bill and/or iead are explained briefly.
i. nibbling (nb): high frequency and low amplitude of opening and closing of the
bill at the same locality.
2. snapping (sp): low frequency and high amplitude of opening and closing of the
bill at the same locality.
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
74 JOHAN G. VAN RHIJN
3. bitting (bt): intermediate
frequency and low amplitude
of opening and closing of tht
bill at the same locality.
4. combing (cb): open bill moves through the feathers.
5. stroking (sk): feathers through closed moving bill.
6. rubbing (rb): moving head over the feathers.
7. rolling (ro): head and bill move along their axis at the same locality.
The second group of elements, containing bathing, shaking and scratching, are
described as follows.
i. drinking (dr): open bill in the water, then closing of the bill and stretching of
the neck upwards.
2. snapping-water (sw): open bill in the water, immediately followed by closed bill
out of the water (no stretching of the neck).
3. bill-washing (bw): sideways shaking of closed bill through the water.
4. stamping (st): stamping
with both feet on the bottom of the bath, often in alterna-
tion which snapping-water.
5. head-dipping (hd): dipping
of the bill and head in the water, followed by stretching
of the neck upwards with bill downwards, combined with moderate tail-shaking
and wing-tilting.
6. plunging (pl): turning around the body axis in the water, also combined with
moderate tail-shaking and wing-tilting.
7. wing-flapping (wf): from a squatted position flapping of both wings through
the water.
8. wing-beating (wb): from a standing position beating of both wings through the
air.
9. wing-tilting (wt): tilting with both wings.
1o. body-shaking (s): shaking of the whole plumage.
ii. head-shaking (hs): sideways shaking of the head (bill is horizontal).
12. tail-shaking (ts): sideways shaking of the tail.
13. head-scratching (hc): scratching with one foot of the head.
14. bill-scratching (bc): scratching with one foot of the bill.
The moment of the first contact of bill or head with the locality of the plumage
to be preened was taken as the start of a preening event; the moment of losing
contact was taken as the end. Most preening events only last a few tenths of a second;
they are usually preceded and followed by short periods without contact between
plumage and bill or head.
One complete movement of drinking, snapping-water,
head-dipping
or plunging was
taken as one event of the element involved. The other elements (bw, st, wf, wb, wt,
s, hs, ts, hc and bc) mostly occur as very rapid repetitions of the same movement.
Here it was impossible to count the number of separate movements. For this reason
an event of each of these elements was defined as a sequence of repetitions of one
and the same movement with intervals of less than half a second. Thus, for instance,
Io sideways shaking movements with the bill through the water, with intervals of
the following durations: 3 X 4", I X I" and 5 X /4", were considered
as 2 subsequent
events of bill-washing.
The verbal commentary of the behaviour contained the starts of all events of the
behaviour elements mentioned in this section. The time-lag between the occurence of
this start and its registration varied between o and I second. For most events it was
impossible to registrate their ends too. Other behaviour was occasionally recorded; it
was ignored in the analysis.
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BEIIAVIOUR IN A HERRING GULL 75
RESULTS
QUALITATIVE DESCRIPTION OF A BODY-CARE SEQUENCE
The first series of observations was not very suitable for a detailed
analysis. The main kind of information these observations gave, concerns
differences and similarities in body-care sequences between the individuals.
The large differences which were found, were strongly associated with the
dluminance
relations within the group. These relations appeared to be very
stable and came out frequently during eating and bathing. One gull was
lominant over all others; normally this bird was the first to enter the bath,
and often all other gulls had to wait before the dominant one had finished.
Consequently all observations on this gull contained much bathing and
preening. On the other hand, the lowest ranking gull started to bathe very
late, and even then, this animal was often chased away from the bath. As
a matter of fact, body-care of the lower ranking gulls was often interrupted
by intolerant behaviour of more dominant gulls, which had finished their
body-care.
In order to find similarities between the individuals, the properties of
body-care sequences of the dominant gull were investigated, and afterwards
I tried to (liscover the same properties in the records of the other gulls. The
most elaborate body-care sequence in all individuals starts with bathing (bill-
washing, head-dipping, and wing-flapping) with some scratching move-
nents, a few Ireening movements (mainly rubbing breast), and a few
shaking movements (mainly wing-beating). Bathing is immediately followed
by a series of shaking movements (wing-beating, tail-shaking, head-shaking,
and sometimes body-shaking), which may be interspersed with some rapid
preening movements. Then, within a few minutes after bathing, the bird
starts to oil. Oiling lasts about two minutes, from the first to the last move-
ment towar(s the oil gland, and it contains about eight oiling events. The
first ones are biting movements, the later ones mainly rolling movements.
Oiling with neck left and neck right occurs normally in a rigid alternation.
The first minute of oiling is characterized by a complete absence of shaking
movements. In the second minute wing-tilting and body-shaking reappear,
and after a much longer period head-shaking and wing-beating. The inter-
vals between successive oiling events contain some preening movements,
and may contain shaking movements. The nature of these preening move-
ments is strongly associated with the time elapsed since the first oiling event.
During the first intervals, preening consists of snapping toward breast,
throat, belly, flanks and tail; then during later intervals, it switches towards
stroking of the pinions, insides of the wings and wingbows, while finally
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
76 JOHAN G. VAN RHIJN
snapping towards the outside of the wings is added after some time. At
the time of the last oiling movement rubbing on the outside of the wings
is added, while all stroking movements gradually disappear and are replaced
by snapping towards breast, throat, belly, flanks and tail again. After the
last oiling event preening may be continued for several minutes. During
the first minutes of this phase snapping, rubbing and stroking are the most
common bill and/or head movements; then these movements gradually cease
and nibbling appears, sometimes via a stage with some combing movements.
After body-care the bird mostly becomes inactive.
This description strongly suggests that in a body-care sequence there
are a number of sharp switches in the behaviour, which occur in a fixed
order. Firstly, the initial entering of the bath enables the bird to perform
a number of elements which are strongly associated with the contact with
water. Secondly, by the last leaving of the bath, the opportunity to perform
these elements is terminated. Thirdly and fourthly, the first and the last
oiling event demarcate the short period of oiling. In the following, the time
phase between the first entering of the bath and the last leaving will be
indicated as "bathing" (periods out of the bath during 'bathing', if they
occur, normally last only a few seconds, and the behaviour shown during
these periods mainly consists of elements which are also shown during the
stay in the water); between the last leaving and the first oiling event as
'shaking'; between the first and the last oiling event as 'oiling'; and finally
after the last oiling event as 'preening'. After the last preening movement
eventually a phase of 'sitting down' can be (listinguished. This nomen-
clature is based on the predominant type of behaviour in each of these
phases or main components. A main component is defined as a part of a
body-care sequence, whose start and end is characterized by a sharp switch
in the behaviour, i.e. the first or last appearance of one or more elements
in a body-care sequence.
The body-care sequences always contained bathing, shaking and preening
movements. Oiling was not seen in all observations. Because all observations
in this first series were precelded
by a few hours with the bath closed, it
is not surprising that they were all characterized by the occurrence of
bathing. The occurrence of shaking and preening movements, however.
cannot be ascribed to prevention of these movements during the deprivation
period. Other observations revealed that shaking and preening may occur
without preceding bathing. Since the frequencies of both types of move-
ments are high after all bathing sequences (also bathing without depriva-
tion), it is likely that bathing and/or its immediate effect (getting wet)
promotes the occurrence of shaking and preening. Oiling has never been
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BEHAVIOUR IN A HERRING GULL 77
observed without preceding bathing. Bathing or its immediate effect seems
to be a prerequisite for oiling. Sequences with bathing and oiling usually
contain more shaking and preening movements than sequences with bathing
and no oiling, even when the frequencies of the bathing elements are similar.
Thus, additionally oiling seems to facilitate the occurrence of shaking and
preening.
QUANTITATIVE APPROACH
The second series of observations were used for a detailed analysis of
the patterning of body-care movements in the dominant gull. There were
two reasons for the use of only one bird: (I) to restrict the number of
observations, and (2) to avoid variability (lue to individual differences. The
justification for the choice of the dominant gull has been given in the
previous section. A disadvantage of the use of only one bird is that generali-
zations to the species cannot easily be made.
I shall presume that the main components of a body-care sequence occur
in a fixed order: bathing, shaking, oiling and preening. In the second series
only three out of the fifteen sequences (lid not contain all four main com-
ponents (they lacked oiling). The twelve other observations (containing
bathing, shaking, oiling and preening in this order) will be used to analyse
quantitatively (I) the frequency distribution of the behaviour elements
over the main components of a body-care sequence, and (2) the frequency
distribution of the elements within these main components. The latter part
of the analysis will not be applied to shaking (between bathing and
oiling) because its duration is mostly too short.
Distribution of the elements over a body-care
sequence.
To study the frequency distributions of the elements, I first tried to
employ two classical methods, namely (i) dividing the sequence after
bathing into periods of equal durations (cf.: VAN IERSEL, 1953; SEVENSTER,
1961), and (2) dividing the sequence after bathing into parts of equal
numbers of movements (cf.: VAN IERSEL & BOL, 1958). Both methods
yielded a very bad agreement between the frequencies of corresponding
elements in corresponding periods or parts. For this reason I designed
another method. Each of the twelve body-care sequences was divided into
thirteen periods. The first period concerned the behaviour before the first
entering of the bath; the second up to and including the fourth concerned
bathing, subdivided into three parts of equal durations; the fifth concerned
shaking; the sixth up to and including the eighth concerned oiling, also sub-
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
78 JOHAN G. VAN RHIJN
divided into three parts of equal durations; and the ninth up to and
including the thirteenth period concerned preening, subdivided (as far as
time was not a limiting factor) into five parts of the following durations:
I, I2, 2, 3 and about four minutes (up to the end of the observation).
These increasing durations of the consecutive periods were chosen because
the frequency of preening after oiling tended to decrease with time. I cal-
culated the frequency of a given element in a given period on the basis
of the total number of events of that element in that period of all twelve
observations, and the sum of the durations of that period. For most elements
this frequency was not very different from the frequencies calculated for
the separate observations.
150-
100 __/ \ a bathing
150-
150-
100-/ b preening
50- _
100-
50-1 c shaking off
0-
bathing oiling
I , , I *
6 5 10 15 20
Fig. I. The distribution of three combinations of behaviour elements over an observation.
The abscissa is a time scale (0-20 minutes). The ordinate indicates the mean number
of events per 5 minutes (the means are based on 12 observations with oiling). The time
phase spent on bathing and oiling is indicated along the abscissa. For a further
explanation see text.
A survey of the course of the frequencies of different combinations of
elements is given in Fig. I. The first combination (a: bathing) refers to
all non-preening activities which occur only during the contact with water
(dr, sw, bw, st, hd, pl and wf); the second combination (b: preening)
refers to all preening activities (oiling included); and the third combination
(c: shaking off) refers to all other non-preening elements (wb, wt, s, hs,
ts, hc and be). For reasons of simplicity the thirteen periods have been
considered as fixed points of time in this scale. These points are the averages
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BEHAVIOUR IN A HERRING GULL 79
10 to/j V ---th
20
51 , g sh
OJ ------- ^^J.-- -------'- ---
20
10- ww
40- A
10- J \
30
20-
10
0 l . . .
*:0~1 7\^ -^ - ~ow
20-
10
30-
201
10
301
20
10-
Fig. 2. The distribution of the total number of movements towards the different
preening localities over an observation. For a further explanation see Fig. i and text.
of the centres of the corresponding periods in the twelve observations
with oiling. The periods spent on bathing and oiling are also indicated. The
mean number of movements per 5 minutes is plotted along the ordinates.
Figs 2, 3 and 4 (constructed in the same way as Fig. I) show to what
extent the separate elements were performed during the consecutive time
periods. Some elements are left out of consideration because of their low
frequencies (bl, fh, lw, bk, cb, dr, st and pl); two elements were taken
together (he and be) because of their similar form (only the orientation
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
80 JOHAN G. VAN RHIJN
201
10
40
30-
\sp
20
80-
70
60-
50
40
30-
20
10
50-
40
30
20
10-
10
Fig. 3. The distribution of the different bill and/or
head
movements over
an observation.
For a furtlher explanation sec Fig. I and text.
differs) and distribution over time. The different preening localities are
shown in Fig. 2 (except: bl, fh, Iw and bk), the bill and/or head movements
in Fig. 3 (except: ob), and the non-preening movements (except: dr, st
and pi) in Fig. 4 (note the deviating scale on the ordinate of head-dipping).
It is obvious that for most elements there are frequency differences
between the main components of a body-care sequence. It is also clear
that the frequencies of the elements tend to change within the main com-
ponents bathing, oiling and preening. For instance, during bathing the
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BEHAVIOUR IN A HERRING GULL
~,,J/'-y \ ~ wb
10 S
20-
10-
0 hs
30-
20-
10-
101
ts
0
n . .-- - -_
30-
20-
10- SW
0
30-
20-
10- bw
0 . . .
150-
100-
50-
l
\0...... w
° I Xwf
Fig. 4. The distribution
of the different non-preening
movements over an observation.
For a further explanation see Fig. i and text.
frequencies of snapping-water (sw) and bill-washing (bw) tend to decrease,
while the frequencies of head-dipping (hd), wing-flapping (wf) and wing-
beating (wb) gradually increase. It is striking that some of the elements (br,
sh, fl, ow, tl, nb, sp, rb, hs, ts and hc + bc) seem to be bimodally distributed
around oiling (mostly with dissimilar peaks), while other elements (ww,
hw, pn, og, bt, sk, ro, wt and s) seem to be unimodal with a peak in the
activity during oiling.
Irrespective of some sharp switches in the behaviour after entering the
bath, leaving the bath, and after the first and the last oiling event, the
frequencies of most elements appear to change gradually. We cannot
conclude, however, that gradual changes really occur in the separate ob-
servations, because Figs i, 2, 3 and 4 only give the mean trends. A gradual
change in the mean trend may be found in spite of sharp frequency switches
in each of the observations, if these should occur in different time periods.
In order to gain more insight into the nature of these frequency changes,
we shall first consider how interval durations between successive events
of each of the elements are distributed. All fifteen observations will be
used for this analysis.
Bout organisation of the elements.
The log survivor function offers a useful method to analyse the distribu-
tion of interval durations between successive events of the same behaviour
6
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
82 JOHAN G. VAN RHIJN
element. Its significance has been stressed by NELSON (I964), DELIUS
(I969), SLATER (i974a, b), METZ (I975) and DIENSKE & METZ (I977).
It technically implies that the interval duration (t) is plotted (along the
abscissa) against the logarithm of the number of intervals longer than t
(along the ordinate). If the different events of an element occur randomly
over time, i.e. the probability of occurrence of that element is constant over
time (and thus independent of the period of time elapsed since the last
occurrence of that element), the log survivor function yields a descending
straight line. If, however, the different events of an element tend to cluster,
i.e. the probability of occurrence of that element is negatively associated
with the duration of time elapsed since the last occurrence of that element,
the log survivor function yields a descending curve with a diminishing slope.
The distribution of interval durations can be tackled in at least two equi-
valent ways: (I) by evaluating the slope of the log survivor function, and
(2) by calculating the probabilities of occurrence per time unit At after
an interval t since the last occurrence of the same element. I have chosen
for a direct plot of probabilities. In this study probabilities (pAt) have
been estimated over time units of 3 seconds (= At); they are based on the
number of intervals >t (nt) and on the number of intervals >t+At
(nt+At), and can be calculated with the following formula:
PAt= I- (nt+At nt).
It is obvious that the precision of the probability estimations is associated
with the size of the difference between nt and nt +At (both integer numbers).
In most cases this difference decreases with an increase of t. In order to
make rather reliable estimations, PAt after intervals with t> 9 seconds was
based on means over two or more (a) time units (At), and was calculated
with formula:
PAt I - (nt +aAt / nt) /a
The power (I/a) refers to the reciprocal value of the number of time units
(a) involved.
In Figs 5, 6 and 7 the amount of time elapsed since the last occurrence
of each element is plotted along the abscissa (0-90 seconds), while the
estimated probability of occurrence (per time unit of 3 seconds) of that
element is plotted along the ordinate. The preening localities are represented
in Fig. 5, the bill and/or head movements in Fig. 6 and the other elements
in Fig. 7. We shall first consider the solid lines in these graphs, which for
each graph are based on the distribution of all intervals between successive
events of the element concerned. It is obvious that for almost all elements
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BEHAVIOUR IN A IERRING GULL
.3-
.2-
.1-
ow
! . [n: 310(193)
X
hw
n: 227 (130)
ww
.1- n
n:158(115)
n oo
.2-
.1 -
0 0
I /~\> n
:114(
A
1 P_-o -- - -
.2-
.1 -
tl
n: 86 (61)
0 QQrJ
.3-
.2 - og
.1- n: 100(100)
0 . i
-1 T -
fl
74)
I I I
30 60 90 0 30 60 90
Fig. 5. Probabilities per 3 seconds (ordinate) of being followed by the same preening
locality after an interval t (abscissa) without movements towards that locality. Solid
lines are based on all intervals between events of the same element (= n); broken
lines are based on intervals with punctuating events of other elements (total number
is given in brackets).
the highest probability occurs shortly after the last occurrence of that ele-
ment. After long intervals (longer than i minute) the probabilities are low
in all graphs. We must therefore conclude that events of each of the ele-
.4.
th
n:87(34)
.2-
.1 -
br
ri: 328 (186)
I 0 a
-t
I
I
-,-
83
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
84 JOHAN G. VAN RHIJN
.3-
.2- nb
2 d \\. n=:563(221)
.1- .
O
d- .4-
.3- .3-
.2- .2-
sp sk
J/9- <n: 299(19) .1 n 423(230)
0- 3- 69 0-0--
.4- .2 rb
.3-
i
n 291 (203)
Fig. 6. Probabilities of being followed by the same bill and/or head movement. For
a further explanation see Fig.
ments tend to occur in some kind of bout. After very short intervals,
however, the probability does not rin most elements.
This phenomenon can be attributed, for the majority of the elements, to
/ ^^ ' ,/^^\ n:28(28)
0 30 60 90 0 30 60 90
Fi. 6.
Probabilities
of
an evenintervag
follos wed
bye me billetween starts of event. For
and to
the tendency of ther explanation
to pauseor a moment after each event,
which may be interpreted as a short suppression of body-care behaviour.
For a few elements, particularly oil gland and body-shaking, the intervals,
however,
the probability docurs mureach
later its maximum in most elements.
This phenomenon can be attributed, for the majority of the elements, to
othe duration
of an event
(intervals
following pause. Ibetween starts
ofthat
events
and
to
the elements and/or the
ird to pause for a moment after each event,
which may be interpreted as a short suppression of body-care behaviour.
For a few elements, particularly oil gland and body-shaking, the increase
of the probability occurs much later than should be expected on the basis
of the duration of an event plus following pause. It is assumed that events
of these elements and/or their effects suppress the occurrence of the same
element for a longer period that the occurrence of other body-care elements.
The observed bout organisation of the elements may be due to a tendency
of the bird to produce sequences of events of one and the same element.
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BEHAVIOUR IN A HERRING GULL 85
.3- .2
. X
n=808n:191(90)
~~~.6-1~ ~ ~ ~ ~~~~~ O'
_____--- -_______ _--
_ _ wb
,51~ n-=
(:149(133)
.3- \0-
wt
.2- bw .1 n97(94)
1I~~~~~
~~\ l~~ _ 1-)~hs
n
-80 (80)
*5- \ .11 n=164(153)
.4- I
exl
\ se Fig.t ts
.3"I *1 n= 359 (349)
.2 d 0
hd
I
1 c n:1077(225) hc+ bc
inf t *n n:62(61)
0ths
i' tv" in 1 -e o
0 30 60 90 0 30 60 90
Fig. 7. Probabilities of being followed by the same non-preening
movement. For a further
explanation see Fig. 5.
For each element this suggestion can be tested by an analysis of the distribu-
tion of the durations of a part of the intervals previously studied, namely
these intervals in which other elements occur. The probability estimations
base{d on these distributions are given as broken lines (as far as they do not
coincide with the solid lines) in Figs 5, 6 and 7. It is clear that for most
elements a relatively high probability is maintained after rather short inter-
vals. Hence, for these elements the observed bout organisation cannot only
be due to the occurrence of sequences of events of one and the same ele-
ment. For the preening localities throat (th) and shoulder (sh), however,
the broken lines appear to be more or less horizontal. For these elements
the bouts seem to be mainly homogeneous.
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
86 JOHAN G. VAN RHIJN
After very short intervals the probabilities indicated with broken lines
are much lower than those indicated with solid lines. This is due to the fact
that in the former case successive events of the same element were
always separated by at least one (time consuming) event of another ele-
ment. The elements, whose events are never or almost never produced in
homogeneous sequences, are characterized by a complete or nearly complete
coincidence of the broken with the solid lines: oil gland (og), body-shaking
(s), wing-flapping (wf), wing-beating (wb), wing-tilting (wt), head-
shaking (hs), tail-shaking (ts) and scratching (he + be).
The results of this section justify the following conclusions: (i) for
each element the events are non-randomly distributed over time; they tend
to occur in some kind of a 'bout'; and (2) for most elements this bout-
structure is not due only (for some elements not due at all) to the occur-
rence of homogeneous sequences. Thus a bout of events of a given element
may contain a number of events of other elements; in other words, if an
element is followed by another element, the former element may reappear
in the same body-care sequence. Thus, sharp switches (as between the main
components) from one element towards another do not occur.
We may still question, however, whether sharp switches from one com-
bination of different elements to another such combination occur within the
main components of a body-care sequence. To solve this question we have
to analyse the composition of the bouts of each of the elements. A bout
will be defined operationally for this purpose as any series of events of
the same element (irrespective of being interspersed by other elements) in
which the intervals between those events do not exceed 12 seconds. The
criterion of 12 seconds roughly corresponds to the maximum probability
of being followed by an event of the same element (Figs 5, 6 and 7).
Composition of the bouts.
The analysis has been restricted to eighteen elements, each of which con-
tained at least 50 occurrences of other activities in its bouts altogether.
Table i shows the relative frequency of occurrence of various other ele-
ments (expressed as percentages of the total number) in bouts of these
eighteen elements. This table shows that the frequency of occurrence of a
punctuating element during bouts is strongly associated with the kind of
bout. For instance, during bouts of wing-flapping (wf) 89% of the activities
consist of head-dipping (hd); during bouts of most preening elements, how-
ever, head-dipping is completely absent. The frequencies in this table give
information about overlap between different kinds of bouts. If sharp
switches between one combination of elements (A and B) and another
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BEHAVIOUR IN A HERRING GULL 87
TABLE i
Percentages of occurrence of the elements indicated on the top row
during bouts of the elements indicated in the left column
br ww ow hw pn og Ilb sp sk rb sw bw hd wf wb wt hs ts n
br - 2 21 2 0 IO 2 3 I4 2 I 8 5 II II7
ww o - i6 29 I9 I6 o o0 2 7 2 7 58
ow 14 9- 8 3 4 0 o o 2 13 2 i6 I92
hw o 21 9 - 21 17 o o o o 0 4 I I3 70
pnt 0 2 5 31 - 30 o o o o o 8 0 3 92
og 19 I 3 15 34 - 33 59 o o o o o o o 3 86
nb o - 7 35 I4 o o o o o 2 19 22 117
Sp II 5 - 'I 39 0 0 0 0 2 IO 4II I77
sk 27 i6 15 - 8 o o o o o II I 17 2II
rb I 8 27 i6 I 2 6 I 2 14 2 I4 i8i
sw 3 0 0 0 0 0 0 0 3 - 42 36 0 3 8 2 64
bw 8 o o o o o I I 0 7 22 - 53 4 I 0 2 2 IIO
hd 12 o I o o o I 2 o I2 4 0 - 31 21 o 3 I2 239
wf 2 0 0 0 0 o 0 I 0 2 0 83 0 0 2 122
wb I 2 2 O 0 0 2 3 2 2 I I 57 1 - I 10 22 I88
wt 13 10 32 II 13 6 4 I8 32 33 0 0 0 0 0 - o 3 72
hs 8 7 0 o
o 0 8 2o i 6 o o o o I7 3 - 42 71
ts 5 4 I 5 I I 8 9 13 9 o o i8 o i6 2 I8 - 277
The percentages are based on the total number of punctuating events (last column: n).
A preening movement (bill and/or head movement towards a locality) was for the calculation
of percentages considered as one event. Hence, the result of summation of the percentages
in the rows of non-preening
movements may be considerably higher than Ioo, because per-
centages are given for both preening localities, and bill and/or head movements. Percentages
of bill and/or head movements during bouts of preening localities, and, respectively, of
preening localities during bouts of bill and/or head movements, are left out of consideration.
The latter kind of relations (between movement and orientation) will be discussed separately.
combination (C and D) occurred, A-C, A-D, B-C and B-D should be non-
overlapping, while A-B and C-D should show a complete overlap by defini-
tion (combination of elements). We shall consider both possibilities. For
non-overlapping between bouts of element A and bouts of element B the
following condition must be fulfilled: the frequency of element A during
bouts of element B and the frequency of element B during bouts of ele-
ment A must equal 0. Several pairs of elements fulfil this condition, for
instance breast and pinions. For a complete overlap between A and B the
following conditions must be fulfilled: (i) the frequency of A (respectively
B) must be highest during bouts of B (respectively A), (2) during bouts
of A the same other elements must occur as during bouts of B, and (3) the
frequencies of these other elements must be similar in both kinds of bouts.
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
88 JOHAN G. VAN RIIIJN
It can easily be seen that the following pairs of elements fully or nearly
agree with the first condition (I): ww-hw, ow-wt, pn-og, og-sk, sp-rb, rb-
wt, sw-bw, hd-wf and hs-ts. Among them the following pairs approximately
agree with the second condition (2): ww-hw, ow-wt, sw-bw and hd-wf. It
is difficult to show agreement with the third condition (3) from a com-
parison between the percentages in Table i. For instance, the percentage
of head-dipping (hd) during wing-flapping (wf) is 89, and the percentage
for wing-flapping during head-dipping is only 31. This implies that the sum
of the percentages of the other activities is 69 during head-dipping and
I during wing-flapping, so that to fulfil the third condition, the percentages
of each of the other elements during head-dipping must be about 6 times
the percentages during wing-flapping. For the nine pairs of elements which
agreed with the first condition, a comparison between distributions of other
elements is given in Table 2. From that table it can be seen that in only
TABLE 2
Comlparison
between bouts of different elements
other elements
occurring
significantly (x2,
d.f. = I, p <
0.05) more often during the:
pair of elements
first kind of bout second kind of bout
ww - hw
ow - wt ts pn
pn - og hw br
og - sk sp nb, rb, ts
sp - rb og hd
rb - wt - sk
sw - bw
hd - wf -
hs - ts sp hd
three pairs of elements complete overlapping of their bouts is not disproved
(ww-hw, sw-bw and hd-wf). This does not mean, however, that sharp swit-
ches occur between these combinations. Although the elements of the first
pair (ww-hw) do not overlap with the other four elements, sharp switches
are unlikely because ww and hw occur much later in a body-care sequence
than sw, bw, hd and wf (Figs 2 and 4). Sharp switches between sw-bw and
hd-wf are unlikely because sw-hd, bw-hd and bw-wf overlap (Table I).
Overlap between sw and wf could not be shown.
From the above we may conclude that for almost each element the distribu-
tion of the events is different from the distribution of any of the other
elements. Sharp switches between combinations of different elements do
not occur. I shall therefore presume that, within the main components of a
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BETAVIOUR IN A HERRING GULL 89
body-care sequence, the frequencies of most elements tend to change
gradually. In the following section I shall concentrate upon the directions
of these changes.
Directions of the frequency changes.
A rough picture of the frequency changes of the different elements during
a body-care sequence has been given before (Figs I, 2, 3 and 4). A de-
tailed plicture will be presented in this section. It concerns an analysis of
the distribution of interval durations between events of different elements,
and can be considered as an extension of the method presented in the section
regarding the bout organisation of the elements. Interval durations were
measured between the occurrence of a behaviour element and the immediately
following or preceding occurrence of the other element. The analysis was
restricted to single events and, to avoid as much as possible interspersed
events of the same element, to the first events in the case of preceding,
and the last events in the case of following, of strings. A string was defined
as a sequence of events of the same element with intervals smaller than
30 seconds, and without interspersed events of other elements from the
same group. The criterion of 30 seconds more or less corresponds with the
decline of the probability of being followed by an event of the same ele-
ment (Figs 5, 6 and 7). For reasons of economization of work I shall
restrict myself to preening localities. The distribution of interval durations
will be used again for estimations of probabilities. In this case two sets of
probabilities will be calculated: (i) the probability per time unit At of a
single event of A, or the last event of A of a string of A, being followed
by behaviour 3, after an interval t without B since the start of that event
of A, and (2) the probability per time unit At of a single event of A, or
the first event of A of a string of A, being preceded by behaviour B,
before an interval t without B since the start of that event of A. The time
unit At equals again 3 seconds.
The estimated probabilities for the different combinations of the most
common preening localities are indicated in Fig. 8. Interval durations (t)
are plotted along the abscissa (ranging from -60 to +60 seconds), and
probabilities are plotted along the ordinate (ranging from o to o.5). Fig. 8
shows that: (I) For most combinations the probabilities after short intervals
without B are different from those after long intervals. (2) For a number
of combinations the probabilities of preceding differ from those of following.
In the graphs on the descending diagonal (following and preceding of the
same locality) the probabilities of being followed are by definition identical
to the probabilities of being preceded. For the thirty other graphs in 13
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
90 JOHAN G. VAN RHIJN
br ww ow hw pn og
11n:167 n:113 t n0186 n 130 *t n 77 n100
br - - _ L
Fig. 8. Probabils pr 3 r
-a s g of
Flig. 8. Probabilities per 3 seconds (ordinate) of a single event of or a string of evenits
towards one locality being preceded (respectively followed) by the same or another
locality before (respectively after) an interval t (abscissa) without movements towards
the latter locality. Each abscissa ranges from -60 up to +60 seconds; each ordinate
from o up to 0.5. Significant differences between the probabilities after intervals
ranging fromi -I5 up to +15 seconds (short iltervals) and the probabilities after
intervals ranging from --o up to -15 and +IS up to +go seconds (long intervals)
are indicated with dots on the ends of the ordinates (x2, p0o.o5). Significant dif-
ferences between the proportion
of being followed and the proportion
of being preceded
within go seconds are indicated with arrows towards the larger proportion.
Significant
differences between the proportion of being followed and the proportion of being
preceded
within 15 seconds are indicated with asterisks on the side of the larger pro-
portion.
Thll number
of investigated
strings (single events included) is also indicated
(n).
cases is the l)roportion of being followed within go seconds significantly
different from the proportion of being preceded; in only 3 cases is the
proportion of being followed within 15 seconds significantly different
from the proportion of being preceded.
The difference between the probabilities during short intervals and those
during long ones for a given combination of elements gives an indication
for the degree of overlap between bouts of the elements involved. It is
shown in Fig. 8 that in most combinations of the six preening localities
overlap occurs to a larger or smaller extent (high probabilities during short
intervals). Evidence for overlap could not be found in the combinations
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BEHAVIOUR IN A HERRING GULL 91
br-ww, br-hw and br-pn. The asymmetry of the curves gives information
about the frequency changes. For instance, the probability of an event or
string of oil gland (og) being followed by wingbow (ww) is considerably
higher than the probability of being preceded by wingbow, and, on the other
hand, the probability of an event or string of wingbow being followed by
oil gland is considerably lower (although not significantly) than the
probability of being preceded by oil gland. This can be interpreted as an
increase of the frequency of wingbow during and/or after a bout of oil
gland. It is also shown that during and/or after br the frequency of pn
increases, during and/or after pn the frequency of ow increases, and during
and/or after og also the frequency of ow increases. It is self-evident that
there is a relation between probability estimations of element A around B
and( estimations of B around A. Nevertheless, the upper right graph and the
lower left graph (combination br-og) seem to be contradictory: in the
upper right graph is shown that the probability of an event or string of
oil gland being followed by breast is lower than the probability of being
preceded by breast, while the lower left graph shows that the probability
of an event or string of breast being followed by oil gland is lower than the
probability of being preceded by oil gland. Presuming that a body-care
sequence contains only one bout of movements towards the oil gland, both
graphs can only be understood by assuming that there are two bouts of
breast associated with a bout of oil gland. Fig. 2 supports this assumption:
one short bout of breast occurs immediately before and during the first
oiling movements, the other long bout of breast starts a few dozens of
seconds after the last oiling movement.
The data presented so far suggest that the patterning of body-care
behaviour is controlled on at least two levels: (I) the basic sequence of the
main components with (by definition) sharp switches in the frequencies
of the different elements, and (2) the programme for the sequencing and/
or timing of the smooth changes of the frequencies of the elements within
these main components (Figs 2, 3 and 4). In the following sections we
shall consider to what extent for the detailed patterning of the behaviour
the supposition of other control levels is necessary.
Transitions between elements.
The analysis of transitions between behaviour elements may give some
information about the short term processes underlying the occurrence of
behaviour. The interpretation of a transition matrix may be very difficult:
in the case of body-care behaviour the main problem is the fact that the
behaviour is non-stationary; the frequencies of the different elements are
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
92 JOHAN G. VAN RHIJN
highly dependent on time (Figs 2, 3 and 4). This implies that the observed
transition frequency from element A to element B gives both information
about similarities between the frequency (listributions of A and B over
time, and about immediate effects of A or its underlying mechanisml on
the probability of occurrence of B.
Despite these difficulties we shall first consider to what extent the
observed transition frequencies deviate from the frequencies which should
be obtained when events of the different elements are randomly distributed
over time. We shall fix our attention upon transitions between different
elements. Thus, the descending diagonal in the matrix will be left open.
Tests for randomness on such matrices are rather complicated. Useful pro-
cedures for these tests have been developed by GOODMAN
(I968) and
SNIJDERS
(I973, I975). To calculate expected frequencies (based on the
assumption that events of the different elements occur in a random se-
quence) we shall use the first method, which was also applied by SLATER
&
OLLASON (I972).
Most body-care activities are separated by short pauses (undefined be-
haviour). It has been shown by NELSON
(I964) that the probability of one
element being followed by another is strongly influenced by the duration
of those pauses. It would be bad policy to ignore this finding. The limited
number of observed transitions, however, does not allow us to take account
of different classes of pause durations. For these reasons the analysis has
been restricted to transitions of which the pause between two events re-
mains within reasonable limits (arbitrarily chosen: 30 seconds).
Preening will be considered separately from the other movements occur-
ring during a body-care sequence, because (I) both groups largely differ
in their respective frequency distributions over time (Fig. I), (2) it reduces
the number of variables in the transition matrix and thus the number of
possible transitions, and (3) it increases the observed frequencies of these
remaining transitions (because elements from the other group are ignored)
and thus it makes the calculations more reliable. Preening localities will be
considered separately from preening movements, since the frequencies of
occurrence of most of the different combinations between localities and
movements are too small for a proper analysis.
To test whether the difference between the observed and the expecte(d
frequency of a particular transition has any significance, one has to keep
in mind that "the asymptotic distribution of the individual term is X2 with
one degree of freedom and a certain scale parameter" (SNIJDERS, I973, p. 8).
We may assume that this scale parameter is not very different from I (loc.
cit.). For this reason the significance of all terms with expected values > 5
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BEHAVIOUR IN A HERRING GULL 93
has been tested by means of a x2 distribution (p<o.o5, i d.f.). For the
term with expected values <5 the fitting Poisson distribution (j,/ = ex-
pected value) was applied. The results (+ or -) are indicated in Tables 3,
4 and 5. In these Tables the column totals are equal to the corresponding
row totals because of the introduction of the element "ps" (any pause of
more than 30 seconds, or start, or end of the observation).
It is obvious that a considerable part of the observed frequencies (far
more than 5%) significantly differs from expectation. Thus, it is very
unlikely that events of the different elements are randomly distributed
TABLE 3
Transitions between preening localities
following:
th br bl sh ww fh fl ow hw lw pn bk tl og ps totals
preceding: th X 33
br X+ + + 167
bl X 12
sh X + + 65
ww X + + + II3
fh X + 17
fl X + 72
ow X + - I86
hw -+ X + + I30
Iw X + 7
pn - + X + 77
bk + X + 34
tl - X 57
og + + - X ioo
ps + + X 97
grand total= 1167
TABLE 4
Transitions between bill and/or head movements
following:
nb sp bt cb sk rb ro ps totals
preceding: nb X -- + 188
sp X + + 176
bt + X + - 72
cb + X 31
sk + +
X +- 207
rb - + - X + 85
ro + + X 28
ps + - + X 97
grand total = 984
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
94 JOHAN G. VAN RHIJN
TABLE 5
Transitions between non-prcening move,ients
following:
dr sw bw st hld p wf wb wt s hs ts hc bc ps totals
preceding: dr X + + 2I
sw +X++ - - -+ 96
bw + + X + -- - -- + 105
st + X - 19
hd X + --- .- 192
pl X 6
wf + X - - - 80
wb - X - --+ - 29
wt - X + 68
s -- -+ X + 79
hs - - --X + + 146
ts ---- +X + 298
hc + X 32
be + X 28
ps + -
+ + X 150
grand total = 1449
over time. This conclusion, however, has already been drawn earlier (dif-
ferent frequency distributions). To decide whether apart from these fre-
quency changes other (short term) processes play a role, we have to com-
pare the directions of the deviations from randomness with the similarities
between the graphs presented in Figs. 2, 3 and 4. It appears that most devia-
tions can be predicted on the basis of those similarities. For instance, the
preening localities oil gland, pinions, inside of the wing with neck high and
wingbow are rather similar with regard to their distributions over an ob-
servation period (Fig. 2). Consequently, most of their mutual transition
frequencies (9 out of 12) are significantly higher (and none is lower) than
would be expected by chance (Table 3). On the other hand, the preening
localities breast an(l pinions are very dissimilar with regard to their distribu-
tions over an observation period (Fig. 2). Their mutual transition frequen-
cies are significantly lower than would be expected by chance (Table 3).
In other words, we may question whether other processes influence the
patterning of body-care behaviour.
If frequency changes over time would only play a role, the probability of
occurrence of an element immediately before the occurrence of another
element, should be equal to its probability immediately after that occurrence.
In other words, the transition matrix must be about symmetrical. Indeed,
in all three Tables we can observe a high degree of symmetry in respect
to the descending diagonal. The observed frequencies of a transition and
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BEHAVIOUR IN A IIERRING GULL, 95
its reciprocal (A preceding B, and A following B) were never found to
(leviate significantly from randomness in opposite directions (+ and--).
On the contrary, many significant deviations were accompanied by signif-
icant deviations in the reciprocal transition in the same direction: preening
localities 40% (14/35), bill and/or head movements 74% (20/27) and the
other movements 67% (48/72). The assumption of symmetry could be
falsified by means of X2 analysis of all pairs of reciprocal transitions with
a sum of both observed frequencies of at least o1. This analysis yielded
for the preening localities X2 = 55.1 (34 d.f., p = 0.02), for the !bill and/or
head movements x2 = 27.6 (I7 d.f., p = o.Io), and for the other movements
X2 = I55.0 (30 d.f., p<o.ooi). In fourteen pairs of transitions the observed
frequencies of the two reciprocals differed significantly (X2 > 5.0, I d.f.,
p < 0.05). Between the preening localities the following transitions occurred
more often than their reciprocals: br->fl, fl->og, sh->hw and ow->tl; be-
tween the bill and/or head movements: ro->sp and sp->nb; and between
the other movements: bw--hd, hd->wb, wb->ts, ts-->hd, ps->wt, wt->s,
s->ts and ts->wt. Most of these findings could not be predicted on the basis
of earlier results (distribution over a body-care sequence, and directions of
frequency changes). We may therefore conclude, that, although most of the
observed transition frequencies can be explained from the temporal
distributions of the different elements, the occurrence of part of the tran-
sitions must be ascribed to immediate effects from one element or its under-
lying mechanism(s) on the probability of occurrence of another element.
Three control levels have been discussed up to here: (I) the sequence of
the main components, (2) the frequency changes within these components,
and (3) the sequential patterning of some movements. The control within
separate movements has not been considered yet. In the following section
I shall try to throw some light upon this aspect by means of an analysis of
the frequencies of the combinations between preening localities (orienta-
tions) and bill and/or head movements (modal action patterns).
Orientation and form of a movement.
The total frequencies of all different combinations over 5 hours of ob-
servation are given in Table 6. The number of times that each combination
occurred is also expressed as a percentage (in brackets) of the total number
of times the locality involved was treated. It is obvious that the different
movements are not randomly (listributed over the different localities. For
instance: (I) biting and rolling are exclusively directed towards the oil
gland, which was never treated by other movements, (2) the outsides of
the wings are predominantly treated by rubbing, (3) the pinions (during
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
96 JOHAN G. VAN RHIJN
TABLE 6
Frequencies of all different comlbinations of preening activities
movement:
nb sp bt cb sk rb ro totals
83 4 o o o o 87
locality: th (95) (5)
92 139 o 21 o 76 o 328
br (28) (42) (6) (23)
Io 7 0 0 0 o 0 17
bl (59) (41)
86 8 o o o o o 94
sh (9I) (9)
44 13 0 o 99 2 0 158
ww (28) (8) (63) (I)
20 I 0 0 0 0 0 21
fh (95) (5)
55 38 o 12 8 I o II4
fl (48) (33) (i1) (7) (I)
31 6i 0 2 4 212 o 310
ow (IO) (20) (I) (I) (68)
81 2 0 0 144 0 0 227
hw (36) (I) (63)
7 1 0 0 0 0 0 8
lw (88) (12)
o o 0 0 I19 o o I9
pn (Ioo)
34 5 0 0 3 o o 42
bk (8i) (I2) (7)
20 20 0 0 46 o o 86
ti (23) (23) (53)
o o 72 o o o 28 Ioo
og (72) (28) o00
563 229 72 35 423 291 28 I711
totals (33) (17) (4) (2) (25) (17) (2)
these observations) are exclusively treated by stroking (in other observa-
tions nibbling of pinions is occasionally observed), and (4) throat, flanks
with neck high, shoulders, insides of the wings with neck low, and back are
predominantly treated by nibbling.
A statistical treatment of these data (e.g. x2 analysis) is not legitimate,
because it is very unlikely that the different performances of a given
movement at a given locality were independent of each other: the data in
Table 6 contain a considerable number of successive occurrences of one
and the same preening act.
It could be argued that the distribution of the different movements over
the localities as shown in Table 6, was due to their respective frequency
distributions over time (an earlier discussed control level: Figs 2 and 3).
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BEHAVIOUR IN A HERRING GULL 97
For instance the frequency distributions of breast and snapping, or pinions
and stroking are rather similar. Nevertheless, the above supposition cannot
be completely true, because (i) all movements towards the oil gland (coin-
ciding with high frequencies of snapping or stroking) are biting and rolling
movements, (2) single events of orientations towards pinions, inside of the
wings, or wingbows during bouts of snapping, are practically always com-
bined with stroking, while single events of orientations towards breast,
belly, and flanks during long bouts of stroking, are practically always com-
bined with snapping, and (3) rubbing is mostly directed towards the out-
sides of the wings, although during long bouts of rubbing a large variety
of orientations occurs. It must therefore be concluded that control of the
integration between form and orientation must be effected at the level of
separate movements. With regard to the functional significance of this
control system, the following remarks can be male: (i) biting and rolling
seem to be the most efficient movements to withdraw or to pick up oil
from the uropygial gland, (2) stroking seems to be the most efficient move-
ment to repair the coarse structure of the stiff parts of the plumage, and
(3) snapping seems to be used to repair the coarse structure of the loose
parts or soft feathers.
DISCUSSION
VAN IERSEL & BOL (I958) found that the frequency of preening move-
ments in terns (Sterna spp) increased very rapidly after bathing. After
reaching a maximum this frequency dropped strongly again, but could be
maintained at a moderate (fluctuating) level for a rather long period. These
authors also found that the type of movement performed (type of move-
ment as such or defined by the locality at which it is directed) was asso-
ciated with the frequency of preening movements in general. To explain
these findings they postulated a variable (preening drive), whose value
controlled both the frequency of preening movements and the type of move-
ment. Consequently, for the different types of movements, different
threshold values of the preening drive were assumed. High thresholds were
supposed for rubbing, pinions and inside of the wing; low thresholds for
head-shaking, breast and shoulder.
The hypothesis of a rapid increase (immediately after bathing) of the
value of the postulated preening drive is supported by my finding that
preening can be terminated immediately after its start by a slight disturbance,
while slight disturbances administered somewhat later are ineffective. Re-
cently several authors discussed the idea of a rapid increase of the tendency
to perform a given kind of movement caused by positive feedbacks of the
7
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
98 JOHAN G. VAN RHIJN
first occurrences of that kind of movement. WIEPKEMA (1971), for instance,
affords evidence for an increase of hunger in mice as a result of the first
eating movements. It is quite possible that the level of a motivational
factor(s) underlying preening also increases after the first preening move-
ments.
Threshold models have been discussed extensively by DAWKINS
(I969a, b) and DAWKINS & DAWKINS (I974). In his first model on choice
behaviour DAWKINS
(I969a) postulated that (I) none of the alternative
stimuli were chosen if a hypothetical variable remained under all threshold
values, (2) a Ioo%0
choice for one stimulus was made if one threshold value
was crossed, and (3) different stimuli were chosen with equal probabilities
if all their respective threshold values were crossed. A choice between alter-
native stimuli is comparable with the decision to select one out of different
behaviour patterns. We may therefore assume that if (in the latter case)
two or more thresholds are crossed, the corresponding behaviour patterns
become either equally likely (if they are mutually exclusively; DAWKINS,
I969a), or in special cases superimposed on each other (BASTOCK
& MAN-
NING, I955). The assumption of equal probabilities could not be falsified by
DAWKINS (I969a) for different kinds of choices. VAN IERSEL & BOL
(1958) do not give precise information about the relation between the level
of the "preening drive" on the one hand, and the probabilities of the dif-
ferent (mutually exclusive) behaviour elements on the other. My data, how-
ever, suggest that during bouts of high thresholds movements (e.g. pinions)
low threshold movements (e.g. breast) are completely or almost completely
absent (Table I). Thus, it appears that the assumption of equal probabilities
if more thresholds are crossed, is not applicable to preening.
From a comparison between my Figs I-4 and VAN IERSEL & BOL's Figs
3 and 4, a large number of similarities became obvious. Three salient points
will be mentioned: (I) the total frequency of preening movements (my
Fig. ib and their Fig. 4) turned out to be similarly distributed over time,
(2) the elements which were unimodally distributed in the herring gull (ww,
hw, pn and s) were rather unimodally distributed in terns (particularly
hw and pn), and (3) the elements which were bimodally distributed in the
herring gull (br, sh, ow, tl, rb and hs) were rather bimodally distributed in
terns. An obvious exception is rubbing, which may be due to the fact that
VAN IERSEL & BOL did not study the bathing period. In the herring gull
the first peak of rubbing occurs during and immediately after bathing
(Fig. 3).
From my data oiling turned out to be very important in the timing of
most other movements (see also: AINLEY, I974). It is a pity, therefore,
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BEHAVIOUR IN A HERRING GULL 99
that VAN IERSEL & BOL did (could?) not make the distinction between
back and oil gland. Assuming that VAN IERSEL & BOL'S model is correct,
oiling must be considered as one of the highest or the highest threshold
movement, because most unimodal distributions coincide with oiling, and
the elements which are bimodally distributed are grouped around oiling.
Support for the statement that oiling is the highest threshold movement,
is given by the finding that oiling is always associated with all high thresh-
old movements, while the occurrence of these high threshold movements
is not always associated with oiling. The statement can also be supported
by the finding that intervals between separate oiling acts are always smaller
than one minute or longer than two hours (at most one oiling bout per
2 hours), while the intervals between events of each of the other high
threshold movements may adopt all kinds of values (more bouts).
In the case of unimodal and bimodal distributions of the elements, VAN
IERSEL & BOL'S model predicts that the sequential patterning of the peaks
in the frequencies of the (lifferent elements before the highest value of
the preening drive is the reverse of the order after that value. The peaks
are thus arranged like mirror images around the highest value. This pre-
diction can be tested by means of the results presented in Figs 2 and 3. The
order of the peaks of the most common preening localities (Fig. 2), for
instance, may be summarized as follows: 5th period-ow; 6th period-br, og;
7th period-ww, hw, pn; 8th period-br, ow. Considering the rates of rising
before and/or dropping after these peaks, we may write the order as:
ow - br - og - pn - hw - ww - ow - br. This order, which is not in conformity
with the prediction formulated above, was also found in the separate body-
care sequences.
From the considerations above, it must be clear that VAN IERSEL
& BOL'S
explanatory model is not fully satisfactory. It is likely that one or more
variables (which may cross thresholds) are important, however, their exact
role in the mechanisms underlying preening is not very clear. In or(ler to
make some further comments about these mechanisms, I shall turn back
to the four control levels (liscussed earlier.
The sequence of main components (bathing, shaking, oiling and preening)
may (at least partly) be controlled by peripheral feedbacks. For instance,
the main component shaking may be triggered by the effect(s) of bathing
(e.g. getting wet). This possibility is sustained by the finding that intensive
bathing (causing intensive wetting: VAN RHIJN, 1977) is mostly followed
by intensive shaking. It is likely, however, that additionally some kind of
internally programmed sequence exists (cf.: FENTRESS,
I976). For instance,
if bathing movements are performed in the absence of water (see below),
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
I00 JOHAN G. VAN RHIJN
subsequent shaking looks as after normal bathing, although peripheral feed-
backs are unlikely to be similar. From this it may be presumed that bathing
and shaking are (to a large extent) controlled by the same mechanisms.
A similar reasoning can be applied for the sequence oiling - preening.
Preening follows oiling in a highly predictable way, while the behaviour
preceding oiling is less predictable (shaking and/or preening). This may
indicate that also oiling and preening are (to a large extent) controlled by
the same mechanisms. Although the mechanisms underlying bathing +
shaking and oiling + preening have been treated separately up to here, a
considerable amount of overlap between these mechanisms must be assumed.
For instance, it is almost impossible to artificially trigger preening, without
bathing preceding it. Even if strong preening stimuli (defined by their
effects: less bathing and more preening as in a normal body-care sequence)
are given in a situation without a bath, the preening movements are mostly
still preceded by bathing movements (on the ground) and subsequent shaking
movements. Besides, the degree of occurrence of oiling + preening after
bathing + shaking may be due to facilitative effects (e.g. peripheral feed-
backs) of bathing and shaking on the mechanisms underlying oiling and
preening. Despite the high degree of overlap in mechanism, both control
systems were separately considered, because the intensities of bathing +
shaking and of oiling + preening seemed to be rather independent of each
other.
The gradual frequency changes within the main components (second
control level) occur in a comparatively rigid order. It has been argued al-
ready that the sequential patterning of the peaks cannot only be explained
by a simple threshold model. It is likely that this order is internally pro-
grammed, but in a way other than by heights of thresholds. It appears that
during oiling not only the sequential patterning of the peaks is rigid, but
also the detailed temporal patterning of the different peaks. In view of
the work of FENTRESS (I972, 1976), this may be an indication for a high
degree of central control of oiling and associated preening. FENTRESS af-
fords evidence for grooming in mice that stereotype(l movement sequences
are mainly centrally controlled, while in flexible sequences peripheral control
may be important too. Indeed, (luring later preening (relative low fre-
quency of movements) the detailed temporal patterning of the different
peaks is very flexible. The gull may stick to the same locality for several
minutes. It strongly appears that in this phase peripheral feedbacks play
an important role. Artificially dirtied plumage localities are intensively
treated, while during the earlier oiling no effect of dirtying can be observed.
The patterning in the sequence of movements (third control level) may
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BEHAVIOUR IN A HERRING GULL IOI
have different causes. I shall try to make an inventory of the possibilities.
One possible explanation is postural facilitation (cf.: LIND, I959; DAW-
KINS & DAWKINS, I976). For instance the transition br->fl may occur more
often than the transition fl->br because most movements towards the breast
(br) are executed from medial towards lateral, and the transition ow->tl
may occur more often than the transition tl--ow because most movements
towards the outside of the wing (ow) are executed from rostral towards
caudal. A second possibility is that other (peripheral) stimuli play a role.
For instance, one might imagine that in the transition hd->wb, wing-beating
(wb) is elicited by the effect(s) of head-dipping (hd) (e.g. getting wet).
A third possibility is that the preceding act is an intention movement of
the following act. The transition wing-tilting (wt)->body-shaking (s) may
be interpreted in the latter way. Finally, the fourth possibility is that the
sequences are centrally programmed. In this stage of the analysis it is im-
possible to eliminate any of these explanations.
The integration between form and orientation of a movement (fourth
control level) may also be due to a number of mechanisms. Irrespective of
the fact that because of anatomical reasons some movements cannot be
directed to all localities (for instance, a gull is unable to rub its throat), two
possible mechanisms will be mentioned. Firstly, the form of a movement
may be controlled (peripherally) by incidentally received stimuli. For in-
stance, an orientation towards stiff parts of the plumage may elicit stroking,
while an orientation towards loose parts may elicit snapping. Secondly,
also at this level control by central programmes seems to be possible.
The use of terms like first, second, third an(l fourth control level suggest
a relation with hierachical models as introduced by KORTLANDT (I940),
BAERENDS (I94I) and TINBERGEN (I942) in animal psychology and
ethology. I do not contest this suggestion, on the contrary, I shall try to
fit in my data somewhat more with the ideas about hierarchical organisation.
Recently DAWKINS (I976) considered that, from a logical point of view,
hierarchical organisation is a parsimonious explanatory principle of behav-
iour. He defined a hierarchy as a set of elements with a relation "is boss
of" (p. 9). This relation is based on the direct connection between elements;
it may be interpreted as "feeds motivational impulses to", "has a causal in-
fluence on", etcetera (p. II). The relation "is superior to" indicates either
directly "is boss of", or indirectly "is boss of an element which is boss of
an element which is boss of, etcetera". The term hierarchy was restricted
to sets in which (I) none of the elements is superior to itself, and (2) one
of the elements (the "hierarch") is superior to all others. Hierarchies were
classified into linear, branching and overlapping hierarchies (p. Io). Further
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
102 JOHAN G. VAN RHIJN
one should realize that a distinction should be made between hierarchies of
connection and hierarchies of embedment (NELSON,
1973) or classification.
In the latter type of hierarchies the relation "is boss of" is equivalent with
"contains" or "includes" (DAWKINS,
1976, p. I4). In point of fact I regret
the term "is boss of", which, by associations with social hierarchies, may
call up anti-hierarchical emotions. It is very likely that a few years ago
feelings against social hierarchies in humans have led to an ideological
repudiation of hierarchies in any context. Even the use of hierarchies in a
non-social context as a scientific explanatory principle (hierarchical organi-
sation of behaviour) has been subject to vigorous attacks.
In a computer programme "is boss of" might mean "calls up as a sub-
routine (procedure)" (DAWKINS, I976, p. II). Analogous to this, my data,
as presented and discussed up to here, may be fitted into the following
hierarchical model: (I) There exists one main programme (in DAWKINS'
terminology: the hierarch) which controls the sequence of at least two, first
order sub-routines: bathing + shaking and oiling + preening. (2) Both
first order sub-routines control repetition, mixing and sequencing of second
order sub-routines: separate events of the different elements. (3) The second
order sub-routines control the combination of third order sub-routines:
separate components of the different movements. Assumption (I) is com-
parable with the first control level discussed before, assumption (2) is
related to the second and third control level, and assumption (3) is com-
parable to the fourth control level. The junction between the second and
third control level has made because this second level would add sub-
routines controlling bouts of events. Since an element is not only char-
acterized by the kind of event, but also by the kind of bout, the number
of sub-routines controlling bouts would be equal to the number of sub-
routines controlling events. The control of bout and event may thus be
realized by the same device. The addition of sub-routines controlling bouts,
does not seem to be a parsimonious extension of the model, and has there-
fore been rejected.
To conclude this paper, I shall try to connect the model, as presented
for body-care behaviour of a herring gull, with the concept of a motivational
system, which is often used in connection with the "boss" levels in hier-
archical models of behaviour (e.g.: BAERENDS,
I976). A motivational system
can be considered as a state variable, whose value can be compared with
a desired condition (norm, Sollwert). This condition may lead to negative
feedbacks (e.g.: MILLER, GALANTER & PRIBRAM, I960; BAERENDS, I970).
The first output of the system may lead (if the desired condition is not
immediately achieved) to positive feedbacks (e.g.: WIEPKEMA, I97I). The
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BEHAVIOUR IN A HERRING GULL 103
value of such a state variable is supposed to be determined by internal and
external influences. These state variables have not yet been incorporated
in my model. However, the decisions to start a body-care programme and
to determine the course of that programme must be under control of some
kind of variables. The decision to enter into the main programme, for in-
stance, may be thought to be under control of a general body-care state
variable, whose value is connected with the need to perform body-care
behaviour (difference from a desired condition: external and internal in-
fluences, of which the latter may be associated with previous external
influences, rhythms, etcetera). It is self-evident that this decision can be
carried out only, if the external circumstances are appropriate (presence of
water to bathe, etcetera).
Figure 9 illustrates which role a general body-care state variable may
play in the initiation, continuation, and cessation of body-care behaviour.
The gull is depicted as a black box. Within the black box a flow-chart is
drawn of a simple programme, enabling the bird to receive two kinds of
inputs (left side), and to produce two kinds of outputs (right side). The
sequence of instructions in this programme is indicated by solid lines, the
flow of information by broken lines. In the top left hand corner it is tested
store ------- .... BS------ -- BB
-'''- '''''--- - -l---
store ................ OS ----- OB
---BV< NO
YES
INPUT BLACK BOX OUTPUT
Fig. 9. A simple model for the initiation, continuation, and cessation of body-care
behaviour. For a further explanation see text.
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
104 JOHAN G. VAN RHIJN
whether stimuli are present necessitating body-care behaviour (rhomb BN =
Body-care Necessity). If so, the value of a corresponding state variable
(small square BV = Body-care state Variable) is adapted by means of
instruction "store". If not (or after the first "store"), it is tested whether
stimuli are present necessitating other behaviour (rhomb ON = Other
Necessity). If much stimuli are present, the value of OV (= Other state
Variable) is adapted. If they are not present (or after the second "store"),
the values of both state variables are compared (rhomb BV<OV). De-
pending on the result of this last test, one of the two systems indicated by
circles (BS = Body-care System, or OS = Other System) is activatel.
This leads to the production of an output (large square BB = Body-care
Behaviour, or OB = Other Behaviour). After this, the next cycle again
starts with testing whether stimuli are present necessitating body-care
(rhomb BN).
By producing an output the environment of the black box may be changed,
and thus give rise to a new input (e.g. positive and negative feedbacks:
broken line from BB towards BN). Positive feedbacks are important to
prevent cessation of body-care behaviour immediately after its start. If
positive feedbacks did not occur, any slight increment of OV would inter-
rupt body-care behaviour. Negative feedbacks are important to stop body-
care when its goal is achieved.
As a matter of fact, body-care behaviour (BB) consists of a large
number of different outputs. The decision by BS to select one output out
of different possibilities imight be thought to be determined by the value of
BV (dotted line from BV towards BS). In fact this explanation of output
control is analogous to VAN IERSEL & BOL'S threshold model, which has
been considered earlier as being too simple. Consequently, more state
variables are needed for a satisfactory explanation of body-care behaviour.
The decision to enter into a sub-routine may also be thought to be under
control of a specific variable.
To illustrate this idea, an extension of the previous model is given in
Fig. lo. This extended model is able to continue its output programme
after each cycle of input processing. This property has been realized (i) by
depicting the input processing (IP) as a sequence of instructions (from
"start IP" up to "continue"), which must be executed after each output ("do
IP"), and (2) by comparing the actual priority in the black box with the
previous priority (rhombs: "was OV" and "was BV"). If these priorities
are equal, the output programme is continued (continue), if they are un-
equal, the output control is transferred to the first instruction of the pro-
gramme of the system which has now the highest priority.
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BEHAVIOUR IN A HERRING GULL 105
I
do IP m __~____
oiling
+
start IP ./ preening
-< 9 9--------------
YES
store }---- LXj do iP
n--- B ---.bathing+
) '
, ON INO
9 I YES---
I cotin l-T yshaking
YES --
m
I--- -----I 1-[---
/-o
store I--- NO
^^BV< 0 /
In this model the circle BS from 1ig.
9 has been replaced by a sequence
Each sub-routine c s of a do IP
?wasOVVNO ,YE
___a
continue
Fig. Io. Extension of the previous model to show the programming of the sequence
of main components. For a further explanation see text.
In this model the circle BS from Fig. 9 has been replaced by a sequence
of two sub-routines (B = Bathing + shaking, and P = oil,ing + Preening).
Each sub-routine consists of a state variable (small square), a device to test
whether this state variable is above a critical value (rhomb), a system for
the coordination of outputs (circle), and an instruction to execute the input
processing sequence (do IP). The state variables B and P may be adapted
by the same stimuli as BV (broken lines), which does not necessarily imply
that their values are similar. To include the possibility that P and B are
influenced by external stimuli which do not affect BV, the information
flow towards P and B is not drawn through BV. At this stage I cannot
conclude whether this possibility is realized. As soon as BV rises to at least
a value equal to OV, the animal will start bathing and shaking and continue
as long as state variable B remains above its critical value (and BV remains
at least equal to OV); after this the animal proceeds with oiling and preening
as long as state variable P remains above its critical value. If P drops under
this value, and BV is still at least equal to OV, the sub-routine bathing +
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
o06 JOHAN G. VAN RHIJN
shaking is restarted; if BV becomes smaller than OV, and P is still above
its critical value, body-care behaviour is terminated and other behaviour
starts. One might imagine that there is no specific preening state variable
(P), and that the occurrence of oiling and preening after bathing + shaking
only depends on the state BV>OV. In this stage of the analysis I can not
decide which of both solutions must be chosen. Both solutions are equally
parsimonious: the first one (BV + P) because of uniform sub-routines,
the second one (only BV) because of fewer state variables.
This model implies that the value of state variable B must always be
tested before the executement of each single preening act, or short preening
sequence, interrupting other behaviour, as for instance incubation (BAE-
RENDS, I970). This seems to be a round-about procedure, and we may
consider whether easier procedures can be postulated. I shall discuss one
alternatively possibility, namely that B and P are tested against each other
(instead of against critical values), and that, depending on the result of
this test, the animal starts with the sub-routine with the highest priority.
However, to explain why bathing and preening mainly occur in this order,
if they occur in the same body-care sequence, the alternative model must
be extended and becomes less parsimonius than the original one.
It is obvious now, how the sequence of the two first order sub-routines
may be controlled by the main programme for body-care behaviour. Instead
of the circles B and P extensions can be made to illustrate how the first
or(ler sub-routines control repetition, mixing and sequencing of second
order sub-routines (separate events of the different elements). However,
Fig. io contains already some speculation, and further extensions of that
model will certainly contain a considerable amount of speculation. There-
fore, such extensions will not greatly contribute to a further clarification
of the mechanisms underlying body-care.
SUMMARY
The patterning of body-care behaviour in the Herring Gull has been studied by
means of:
(a) qualitative observations on four individuals, and
(b) application
of six quantitative analytical methods on the behaviour recordings of
the dominant gull (which was least influenced by other individuals).
The qualitative observations led to the conclusion that in a body-care sequence a
number of sharp behaviour switches occur in a fixed order; between successive
switches the following "main components"
could occur: bathing, shaking, oiling, and
preening.
The quantitative
methods with their results can be recapitulated
as follows:
I. An analysis of frequencies of the different behaviour elements suggested that
within the "main components"
these frequencies tended to change gradually over time.
2. The distributions
of the behaviour elements were further studied by means of the
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BEHAVIOUR IN A HERRING GULL 107
temporal relations between events of the same element. This analysis revealed that
each element occurred in "bouts", which could be interspersed with events of other
elements.
3. It was also studied whether different elements had similar distributions over the
observation time. This method was concentrated on the behaviour during short inter-
vals between events of the same element; its results indicated that different elements
had different distributions over time, although these distributions could be partially
overlapping.
4. The sequential patterning of these different frequency distributions was studied
in detail by measuring the interval durations between events of different elements.
The results of these four analytical methods could be explained by postulating a
programme for sequencing and/or timing of smooth changes in the probabilities of
occurrence of the different behaviour elements within eacli "main component".
5. By means of transition analysis it was investigated whether apart from the slow
frequency changes, rapid processes also played a role. Because of differences between
reciprocal transitions, it was concluded that those rapid processes were important too.
6. From the frequencies of the combinations
between movements and the places to
which they were directed, a certain amount of dependence between the components of
separate behaviour events could be proved.
The results of this study have been discussed in relation to Van Iersel & Bol's
study on preening in terns. It was argued that their threshold model was too simple
for a satisfactory explanation of the patterning of body-care behaviour.
The discussion
of the present results in relation to hierachical models was more successful. The
mechanism underlying body-care behaviour can be considered as a programme with a
main routine (order of main components) and several sub-routines.
The role of state
variables in this programme has been illustrated in two diagrams.
REFERENCES
AINLEY, D. G. (1974). The comfort behaviour of Adelie and other penguins. -
Behaviour 50, p. I6-5I.
ANDREW,
R. J. (I956a). Some remarks on behaviour in conflict situations, with special
reference to Emberiza spp. - Brit. J. Anim. Behav. 4, p. 41-45.
-- (56b). Normal and irrelevant toilet behaviour in Emberiza spp. - Brit. J.
Anim. Behav. 4, p. 85-9I.
BAERENDS,
G. P. (I94I). Fortpflanzungsverhalten und Orientierung der Grabwespe,
Ammophila campestris Jur. - Tijdschr. Entomol. 84, p. 68-275.
-- (970). A model of the functional organization of incubation behaviour. - In:
The Herring Gull and its egg, part I (Eds G. P. BAERENDS & R. H. DRENT). -
Behaviour Suppl. 17.
-- (976). The functional organization of behaviour. - Anlim. Behav. 24, p. 726-738.
BARLOW,
G. W. (I968). Ethological units of behaviour.
- In: Central nervous systems
and fish behaviour (Ed. D. INGLE). Univ. Chicago Press.
BASTOCK,
M. & MANNING, A. (I955). The courtship of Drosophila mclanogaster.
Behaviour 8, p. 85-III.
DAWKINS, M. & DAWKINS, R. (1974). Some descriptive and explanatory stochastic
models of decision-making. - In: Motivational control systems analysis (Ed.
D. J. MCFARLAND). Academic Press, London.
DAWKINS, R. (I96ga). A threshold model of choice behaviour. - Anim. Behav. 17,
p. 120-133.
(I969b). The attention threshold model. - Anim. Behav. 17, p. I34-I4I.
(1976). Hierarchical organisation: a candidate principle for ethology. - In:
Growing points in ethology (Eds R. A. HINDE & P. P. G. BATESON). Cambridge
University Press.
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
o18 JOHAN G. VAN RH1JN
- & DAWKINS,
M. (1976), Hierarchical organisation and postural facilitation.
Anim. Behav. 24, p. 739-765.
DELIUS,
J. D. (I969). A stochastic analysis of maintenance
behaviour of skylarks. -
Behaviour 33, p. 135-178.
DIENSKE,
H. & METZ,
H. A. J. (1977). Mother-infant
body contact in macaques: a time
interval analysis. - Biology of Behaviour 2, p. 3-37.
FENTRESS,
J. C. (1972). Development and patterning of movement sequences in inbred
mice. - In: The biology of behaviour (Ed. J. A. KIGER Jr.). Oregon State
University Press.
-- (976). Dynamic boundaries
of patterned behaviour: interaction and self-organi-
zation. - In: Growing points in ethology (Eds R. A. HINDE & P. P. G. BATESON).
Cambridge University Press.
GOOIMAN,
L. A. (I968). The analysis of crossclassified data: independence, quasi-
independence and interactions in contingency tables with or without missing
entries. - J. Amer. Statist. Ass. 63, p. I091-II3I.
HALLIDAY,
T. R. (1975). An observational and experimental study of sexual behaviour
in the smooth newt, Triturus vulgaris (Amphibia: Salamandridae).
- Anim Behav.
23, P. 291-322.
HEILIGENBERG,
W. (1973). Random processes describing the occurrence of behavioural
patterns in a Cichlid fish. - Anim. Behav. 21, p. 169-182.
IERSEL, J. J. A. VAN
(I953). An analysis of the parental behaviour of the male three-
spined Stickleback (Gasterosteus aculeatus L.). Behaviour Suppl. 3.
& BOL,
A. C. A. (I958). Preening in two tern species. A study on displacement
activities. - Behaviour 13, p. 1-88.
KORTLANDT,
A. (1940). Wechselwirkung zwischen Instinkten. - Archs neerl. Zool. 4,
p. 443-520.
LIND, H. (1959). The activation of an instinct caused by "transitional action". -
Behaviour 14, p. 123-135.
McKINNEY, F. (I965). The comfort movements of Anatidae. - Behaviour 25,
p. I21-217.
METZ, J. H. M. (I975). Time patterns of feeding and rumination in domestic cattle.-
Mededelingen Landbouwhogeschool,
Wageningen, 75 (I2), p. 1-66.
MILLER,
G. A., GALANTER,
E. & PRIBRAM,
K. H. (I96o). Plans and the structure of
behavior. - Holt, Rhinehart & Winston, New York.
NELSON, K. (I964). The temporal patterning of courtship behaviour in the glandulo-
caudine fishes (Ostariophisi, Characidae). - Behaviour 24, p. 90-I46.
- (973). Does the holistic study of behavior have a future? - In: Perspectives
in ethology I (Eds P. P. G. BATESON
& P. H. KLOPFER).
Plenium Press, New
York & London.
RHIJN, J. G. VAN (I977). Processes in feathers caused by bathing in water. - Ardea.
ROWELL,
C. H. F. (I96I). Displacement
grooming in the Chaffinch. - Anim. Behav. 9,
p. 38-63.
SEVENSTER, P. (1961). A. causal analysis of a displacement activity (fanning in
Gasterosteus aculeatus L.). - Behaviour Suppl. 9.
SLATER,
P. J. B. (1974a). Bouts and gaps in the behaviour of Zebra finches, with
special reference to preening. - Rev. Comp. Animal 8, p. 47-6I.
(1974b). The temporal pattern of feeding in Zebra finch. - Anim. Behav. 22,
P. 506-5I5.
& OLLASON,
J. C. (1972). The temporal pattern of behaviour in isolated male
Zebra finches: transition analysis. - Behaviour 42, p. 248-269.
SNIJDERS, T. (I973). A test for randomness in behaviour. - Rapport TW-I34, Mathe-
matisch Instituut, Rijksuniversiteit Groningen.
(I975). A test for randomness in behaviour. - Statistica Neerlandica 29/2.
p. 39-48.
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
PATTERNING OF COMFORT BEHAVIOUR IN A HERRING GULL I09
TINBERGEN, N. (1942). An objectivistic study of the innate behaviour of animals.
Biblioth. biotheor. I, p. 39-98.
VEENING,
J. G. (I975). Behaviour functions of the VMH in the rat: an ethological
approach.
- Thesis, University of Groningen.
WIEPKEMA,
P. R. (I96I). An ethological analysis of the reproductive behaviour of
the Bitterling (Rhodeus amarus Bloch). - Archs neerl. Zool. 14, p. IO3-I99.
(- I971). Positive feedbacks at work during feeding. - Behaviour 39, p. 266-273.
RESUME
La structure du comportement
de se toiletter du Goeland argente est examine par:
(a) des observations qualitatives de quatre individus, et
(b) une application
de six methodes d'analyse quantitatives aux registrations du com-
portement du Goeland dominant,
qui etait interrompu
le moins par les autres individus.
J'ai conclu par suite des observations qualitatives, que, dans une serie de comporte-
ment de se toiletter, l'on trouve, en une succession fixe, quelques changements
abruptes
dans la composition
du comportement.
On peut indiquer entre ces changements succes-
sifs les "composants
principaux" suivants: se haigner, se secouer, se grasser, et se
nettoyer.
On peut recapituler les methodes quantitatives et leurs resultats comme suit:
I. Les resultats d'une analyse de frequences des elements de comportement
differents
suggerent, que, dans les "composants
principaux",
les frequences des elements changent
graduellement au cours du temps.
2. L'analyse poursuivie des distributions des elements, par les relations temporelles
entre les evenements du meme element, revele que les evenements de chaque element
se passent en series, qui peuvent etre entrecoupees d'evenements des autres elements.
3. J'ai egalement examine si les distributions d'elements differents, dans une periode
d'observation,
etaient identiques. Cette methode a vise sur le comportement
dans les
intervalles courts entre des evenements du meme element. Les resultats indiquent que
les elements differents ont des distributions differentes, quoique ces distributions
peuvent etre partiellement coincidentes.
4. J'ai examine en detail la structure de la succession de ces distributions differentes
en mesurant les intervalles entre les evenements d'elements differents.
II m'est possible d'interpreter les resultats de toutes ces quatre methodes par le
postulat qu'il y a dans chaque "composant principal" une programmation pour au
moins la succession de changements graduels des probabilites
d'apparition
des elements
differents.
5. Au moyen d'une analyse de transitions, j'ai cherche si, a cote de la programmation
pour les changements graduels (et lents), des processus rapides jouent un grand role.
Parce qu'il y avait des differences entre les transitions reciproques,
j'ai conclu que ces
processus rapides sont egalement importants.
6. Par les frequences des combinations
entre les mouvements et les orientations de ces
derniers, je pouvais montrer qu'il y a une coordination entre les composants des
evenements de comportement
separes.
J'ai discute les resultats de mon investigation en relation avec les recherches de
Van lersel & Bol du comportement
de se toiletter des Sternes. J'ai argumente que
leur modele de seuils est trop simple pour une explication satisfaisante de la structure
du comportement
de se toiletter. J'ai obtenu plus de succes a adapter mes resultats
a un modele hierarchique. On peut considerer le mecanisme causal du comportement
de se toiletter comme une programmation
avec une routine principale (succession des
"composants
principaux")
et plusieurs routines sub-ordonnees.
L'importance
des variables
d'etat dans cette programmation
est indiquee en deux diagrammes.
This content downloaded on Fri, 22 Feb 2013 07:10:59 AM
All use subject to JSTOR Terms and Conditions
