Antm. Behav.
, 1969, 17, 641-651
CLEANING BEHAu IN DROSOPHILA MELANOGASTER
BY ANDREW L. SZEBENYI, S.J.
Biology Department, Le Moyne College, Syracuse, New York 13214
The objective of this paper is to give a descriptive
account of a behaviour pattern complex in
Drosophila melanogaster, generally known as
grooming or cleaning behaviour, and to present
methods which make the descriptive and com-
parative study of this behaviour complex
possible.
Very little is known about grooming or
cleaning behaviour in Diptera, Apart from the
fact that cleaning movements do occur, most
other aspects of this behaviour, such as the
nature of the eliciting stimulus, spontaneity
of movements, stereotypy of patterns, the im-
portance of sensory feedback during perform-
ance, and to a large extent even the biological
significance of the behaviour remained hitherto
obscure.
Recently, Connolly (1968) described social
facilitation of preening behaviour in D. melano-
gaster and showed that facilitation is probably
induced through vision, and it is unaffected by
physical contact between flies, or by the amount
of dirt on the body. Another published work on
cleaning behaviour in Diptera is that of Heinz
(1949) describing the cleaning movements in
terms of a number of components, and investi-
gating some of the conditions which may be
responsible for releasing or strengthening clean-
ing activity. In the same article, Heinz also
reported his observations on cleaning behaviour
in twenty-eight Drosophila mutants, and found
the movements to be identical with those of the
wild type. It should be noted, however, that the
system of components used by Heinz was rather
limited and did not extend to more precise
quantitative aspects of the behaviour which
may account for this statement.
Apart from these two studies, there are only
a few sporadic references in the literature to
cleaning behaviour in Diptera. Lorenz (1956)
remarked in a conference on instinct: 'in
Drosophila wingless mutants are still in full
possession of the normal movements of cleaning
wings' (p. 55). Tinbergen (1965) referred to this
observation as a possible example of spontaneous
behaviour, cautioning however, that the appar-
ent absence of external stimuli is n o t in itself
a positive proof of spontaneity. He also main-
tained th.at 'flies will always preen their wings
641
as soon as dust particles stick to them' (p. 85).
From the author's own observations the
following general statements can be made.
Cleaning behaviour is performed with the legs
by two different basic movements: the sweeping
movements of the legs over the body surface,
and the rubbing motion of the legs along the
tarsal joints. It has been observed in D. melano-
gaster on several occasions that removal of
macroscopic particles was actually achieved by
these movements. A small piece of yeast particle
adhering to the body surface has been seen to
be picked up by the sweeping movements, while
the rubbing that followed caused the particle
to travel along the tarsal joints and to drop off
at the end. The direction of the movement of the
particle seemed to be determined by the direction
of bristles on the legs where they all point away
from the body.
It is not at all certain, however, that removal
of larger particles is the only result of this
behaviour. It is quite possible that microscopic
particles adhering to sensory receptors, Or the
need for even distribution of wax-coating on the
body surface play some role in eliciting the
movements. Further investigation is needed to
determine the causes and the effects of this
behaviour.
The behaviour presents a rich source of
material for study, and many aspects of it may
become important not only as a mere behaviour
study but also as a new source for behaviour-
genetic analysis. To investigate these aspects
and possibilities of this widespread and probably
highly adaptive behaviour complex, a survey
has been initiated. What was sought, first of all,
was some means of obtaining records which
would make it possible to analyse and compare
the cleaning performance in different individuals
and populations in appropriate qualitative and
quantitative terms.
The progressive phases of these preliminary
investigations were the following: (1) analysis of
cleaning behaviour into recurrent components
and the construction of ethograms based on
the performance of D. melanogaster wild type
flies; (2) discovery of the structural organization
of the behaviour based on the ethograms and
on continued direct observations. (Thes~ two
642 ANIMAL BEHAVIOUR,
phases of the investigation furnished the
material for the present paper.) (3) Meanwhile a
dissection of the behaviour into relevant quan-
titative aspects has been undertaken, and a
method for quick and accurate determination of
cleaning behaviour in terms of quantitative
parameters of populations has been developed.
This third phase of the investigation is now being
tested, and the results will be brought to light
in the near future.
All these phases are primarily descriptive
but they lay the foundation for meaningful
experimental and comparative study of the
behaviour.
The present account is based on studies of
D. melanogaster wild-type flies which were
obtained at Syracuse University from a popu-
lation cage in Dr Marvin Druger's laboratory.
The cage contains mixed strains derived from
single females caught at various localities in
Syracuse in October 1964. The cage was started
with 15 males and 15 females of each of 15
strains in 23 December 1964. The flies used for
observation and record in this account were
maintained in culture bottles since November
1967. They were raised on agar molasses media
at a temperature of 23~ and at 62 per cent
humidity in a constant temperature and humid-
ity room. Observations were carried out in the
same room. For observation the flies were either
transferred into clean glass vials or were trans-
ferred into food vials, and a 5-min 'settling down'
period was allowed before each observation or
recording.
Ethograms: The Descriptive Method
Looking at unetherized flies, one is immediately
struck by the high frequency of occurrence, and
by the complexity and great variety of move-
ments of cleaning behaviour. In most instances,
the movements are rapid, and they follow each
other in quick succession which renders detailed
observation an exacting task, and also limits the
precision of direct visual observation. The
alternative to direct visual observation is to
make use of motion pictures. This latter method,
although excellent for analysis of movements,
is not the best to give routine permanent record
because it does not render the quantitative
characteristics of the performance immediately
comparable. This can be achieved better by
permanent written records or ethograms. For-
tunately, there is a simple and inexpensive way
to obtain such ethograms. The following is
recommended: cleaning behaviour is observed
l~n[il one b~om~s familiar with the general
17, 4
characteristics of the movements. Next, the
performance is analysed into components, and
each of these components is given a verbal
symbol. The symbols should be sufficiently
simple and short since they are to be pronounced
as fast as the movements occur. The next step is
to carry out systematic observations during
which the verbal symbols are recorded on tape.
At this point, one has to be completely familiar
with both the movements and their corres-
ponding symbols to avoid lag or confusion
during the recording. If the movements flow
into each other, it is sufficient to mark the end of
a component by the beginning of the next, and
to say 'stop' at the end of a series of movements.
No recorded symbol after the word 'stop'
means no observable movement. The next
step is to play back the tape and to write down
the symbols. This gives a written record of
patterns. For more elaborate ethograms the
duration of each component is to be measured
from the tape with the aid of a timer. The
ethogram obtained in this manner gives most
of the essential parameters of the behavionr
(patterns, duration, frequency) in a form which
is easily and directly comparable.
It is found that one can carry out visual
observation at different levels of complexity,
and that the study of some aspects of cleaning
behaviour does not require a specially detailed
record. It is, therefore, practical to work out a
number of systems of symbols of different com-
plexity. The selection of the appropriate system
will depend on the nature of the problem at
hand. For instance, the study of front versus
back cleaning preference, that is, a type of
cleaning which is performed with the first pair
of legs, while the second pair is moved forward
for cleaning, versus a type of cleaning which is
performed with the third pair of legs while the
second pair is moved back for cleaning can be
described quite adequately by a very simple
system of symbols (system I).
It should be remarked here that the choice of
symbols is entirely subjective. The symbols are
simply useful in converting fly movements into
generalizations for human understanding of the
fly's behaviour, and in making a written record
possible.
System I
Cleaning behaviour is performed with the
legs, and may refer to various body parts and
to the legs themselves. Thus, the components
can be described by giving a number to the legs
involved in cleaning (first, second; and third
 SZEBENYI: CLEANING BEHAVIOUR IN DROSOPHILA MELANOGASTER
pair), and by giving a letter symbol to the main
body parts cleaned (H = head, T = thorax, A
: a b d o m e n , and W--= wings). No more than
three legs may be involved in cleaning activity
at the same time because the fly requires at least
three legs for support. It can often be predicted
which leg is going to be used for cleaning by
observing the preparatory movements of the
supporti~g legs shifting the centre of gravity
of the body to the required position. The system
of symbols is the following.
Body cleaning movements (sweeping motions).
1H, any cleaning movement directed toward any
part of the head. The movements are performed
with the first pair of legs. 2H, same as above
but the movements are performed with the
second pair of legs (a rare component). 1T and
2T, any cleaning movement directed towards
the thorax and performed with the first or second
pair of legs respectively. 3A, any cleaning move-
anent directed toward any part of the abdomen
and performed with the third pair of legs. 3W,
wing cleaning movements performed with the
third pair of legs. 3TW, a symbol representing
the sweeping movement of one of the third pair
of legs over the dorsal side of thorax and wings.
3HTW, same as above except that the move-
ment also includes the head. 3T, the initial
phase of 3TW; the wings are, however, not
touched. The component may appear in itself
without going into the 3TW movement.
Leg cleaning movements (rubbing motions). 1,
the rubbing motion of the first pair of legs.
No other legs are involved in this component.
12, the movements include both first legs and
one member of the second pair. The legs ap-
proach each other on one side. 21, the move-
ments include one of the first and one of the
second pair of legs on one side. 2112, the above
two components flow into each other. The 21
part is often very short. 32, the movements
include both third legs and one member of the
second pair which is moved back for cleaning.
The performance is slightly to one side. The
legs cleaning usually touch the ground and form
additional support for the body. 3, rubbing
movements performed with the third pair of
legs. No other legs are involved in this com-
ponent. 1C, performed with one of the first
legs cleaning the coxal region of the other
member of the first pair. The component is
associated with the rigid extension of the leg
cleaned. While the leg cleaning movements
l~ormally mean ~h~ rubbin~ motio~ of the le~s
643
along the tarsal joints, this component is an
exception.
Experience in the use of symbols shows that
recording from direct visual observation cannot
go far beyond system I. The following systems
are more suitable for analysis of motion pictures.
System H
The description of components is extended to
finer subdivisions of body parts. The leg cleaning
components are the same as in the previous
system. For sake of simplicity, the body clean-
ing components are not all prefixed by the
number of legs involved in the activity but most
of them have their own letter symbol.
Head cleaning movements. F, all cleaning
movements directed towards the front part of
the head, antennae, front part of the eyes,
ocelli (while head is bent down), and proboscis.
The proboscis is not extended while the region
is being cleaned. I, all cleaning movements
directed towards the lateral sides of the head
and eyes. The head is not turned during these
movements. H, a characteristic head cleaning
component. The movements clean the dorsal
and ventral sides of the head at the same time
while the head is rotated to various degrees to
the right or to the left. P, proboscis cleaning com-
ponent. The proboscis is extended for cleaning.
All these components are performed with the
first pair of legs. 2H, a rare component (see
system I).
Thorax cleaning movements. 1V, the ventral
side of the thorax is cleaned with the first pair
of legs. The fly usually takes up a posture at an
angle to the horizontal and balances on four
legs forming three supporting points, one of the
second legs being extended parallel with one of
the third legs on the same side. 2T, any cleaning
movement directed towards the thorax and
performed with the second pair of legs. One
may distinguish between 2TD, 2TS, and 2TV
depending on whether the dorsal, lateral, or
ventral sides of the thorax are being cleaned.
3T, the dorsal side of the thorax is cleaned with
one of the third pair of legs. The component is
identical with the initial phase of TW move-
ments.
Abdomen cleaning movements. D, S, V, and G.
These symbols refer to dorsal, lateral, and ventral
sides of the abdomen, and to genital and anal
regions ((3) cleaned with the third pair of legs.
The anal plates and the reproductive organs may
or may not be ~xLen0r during cleaning.
644 ANIMAL BEHAVIOUR,
Wing cleaning movements. B, the wing base
is cleaned with one of the third pair of legs which
is inserted at the inside angle of the wing. The
movements are usually slow and protracted
and are accompanied by slight opening and
raising of the wing. K, a rather rapid move-
ment which consists of a sharp kick against the
underside of the wing. The wing is lifted by the
movement. VS, SS, DS. The symbols represent
the sweeping movements of the third pair of
legs over the wing surfaces. The ventral sweep
(VS) is often not distinguishable from move-
ments cleaning the dorsal side of the abdomen
(D). The dorsal sweep (DS) is similar to TW
but the movement does not include the thorax.
The lateral or side sweep (SS) refers to the move-
ment along the outer edges of the wings in a
parallel fashion. TW, and HTW, the same com-
ponents as 3TW and 3HTW in system I. W, the
most characteristic wing cleaning component.
The wing is taken between the two third legs
which move in a see-saw fashion along its dorsal
and ventral surfaces. Several phases can be
distinguished in this component. At an early
phase, the wing is still in a near horizontal
position, at the end phase, the wing is vertical.
All these wing cleaning components are per-
formed with the third pair of legs. Various
modifications of the W component can be
observed clearly in D. melanogaster vestigial
mutants, and in other wing mutants of this
species, such as unfolded and pupal. This is
probably the component referred to by Konrad
Lorenz (1956, p. 55).
System III
In addition to systems of symbols used in the
previous account, the problem of symmetry
may be dealt with by using subscripts R, L,
and P meaning right, left and parallel, depending
on the symmetry of the movements. For instance,
WR, DSP, and 12L would mean cleaning of the
right wing according to the W component,
sweeping movement along the dorsal surfaces
of the wings performed on both sides simultan-
eously, and the performance of the one-two
component on the left side.
Figure 1 shows the most commonly occurring
cleaning behaviour components in D. melano-
gaster wild-type flies. The drawings, obtained
by tracing single frames of motion pictures,
are schematic, and in them several structural
details were omitted to place emphasis on pos-
tures and movements. For the same reason,
the wings are shown only in outline. The addition
of v~n~tion on the wings would have. obsc_ur~d
17, 4
some of the legs. Each drawing is accompanied
by the symbol of the component it represents
according to system II.
Figure 2 is an example of ethograms of clean-
ing behaviour in the D. melanogaster wild-type
studied. The ethogram was obtained by analysis
of motion pictures, using the described method.
The movie was taken at 24 frames per sec, and
was played back at 6 frames per sec. The dur-
ation of each component was measured at the
low speed, and the true durations were calcu-
lated for the ethogram. The symbols follow
system II. The ethogram is a sample observation
of 280 sec duration of which 4.9 sec were spent
in locomotion, 140.9 sec in cleaning, and 134.2
sec in motionless state. The performance is
presented at three levels: the top level shows
locomotion, the middle level cleaning, and the
bottom level the rest periods during which no
observable movement has occurred.
The Problem of Error
In preparing an ethogram, the final written
record is based on a series of transformations.
In case of direct visual observation, the event is
first seen, and what is seen is expressed and
recorded on tape by verbal symbols. Playing
back the tape, what is heard is timed and written
down. At each step in this series of transform-
ations from the actual event to the written record
there are a number of sources of error due partly
to the observer and partly to the instruments
used by the observer, in this case the tape re-
corder and the timer. If the visual observation is
indirect, that is, if one makes use of motion
pictures, additional sources of error are intro-
duced by the movie camera and the projection
equipment. Thus, the written record is rather far
removed from the actual event it is supposed to
represent. In order to see how closely the written
record approximates the event, it is necessary
to form some ideas about the magnitude and the
direction of errors introduced by the trans-
formations.
It is customary to distinguish between system-
atic and random errors where the former refers
to the accuracy, the latter to the precision of the
results. It seems that systematic errors inherent
in the equipment are of no real significance
because the comparison of records requires
no more than relative accuracy, and because it
was found that random fluctuations in the re-
cording and timing instruments were in practice
negligible. The major cause for concern is the
error introduced by the observer at the point of
tra nfforrnatior~ of visual impressions into verbal
 9ZEBENYI: C L E A N I N G B E H A V i O U R i N DROSOPHILA MELANOGASTER 64~J

I 12 -21 IC 52

f i ~ .

3 G I H 3T

-
%
P B IV TOP VIEW IV SIDE VIEW

f
< L.
"l iJ

K SS VS W EARLY AND LATE PHASES

Fig. I. The most commonly occurring cleaning behaviour components in D. melanogaster

wild type flies. For the definition of symbols see system II in text,
~46 ANIMAL BEHAVIOUR, 17, 4

I_ ,.oO

oo ~
12
I
I
r

II
o o

Z ~
0 ~ No

O--Z
~Zo
UW
o-0
 SZEBENYI: CLEANING BEHAViOUR iN DROSOPHILA MELANOGAST~ER d4~

symbols, and at the point of timing the duration Table I.

of components from the tape. Each of these
transformations had to be tested for the mag- Sequence of 20 Sampleaverage 2or
nitude and direction of error, and for any components of 10 measure- cr - - x 100
cumulative effect. ments of the same
First, the precision of timing was tested. A components (sec)
randomly selected portion of the tape was
1 0-9 0'06 13-3
played back ten times, and each time the dura-
tion of behaviour components was measured with 2 4"8 0.08 3-3
an electric timer which reads in sec and tenths
of sec. To aid the precision of measurements, the 3 1'8 0.06 6.7
sequence of the components was first written 4 1.6 0"07 8.8
down. In this way each component could be
anticipated. To aid accuracy, the bar switch 5 1"2 0'07 11-7
of the timer was eliminated and a sensitive 6 1"8 0"10 11'1
microswitch was installed in its place getting
rid of most of the lag associated with pressing 7 0"6 0"04 13"3
down on the bar switch. The portion of the tape
contained twenty behaviour components of a 8 7-5 0'06 1"6
total duration of approximately 118.5 sec. of 9 1"5 0"11 14'7
cleaning. The components were measured in a
cumulative fashion, that is, the timer was not 10 3.7 0'06 3"2
reset to zero after each measurement but only after 11 8"6 0"07 1-6
all the twenty components had been timed.
From these cumulative data the actual duration 12 4"5 0"15 6'7
of each component was computed by subtraction.
The ten measurements of each component were 13 2'6 0"14 10'8
then processed for sample average, standard 14 14"2 0"2 2'8
deviation, and relative percentage error expressed
2a 15 36'0 0"18 1"0
as = • 100 where N is the average of the ten
x 16 1"1 0'06 10"9
measurements of the same component. The
relative percentage error was calculated to show 17 0"5 0'05 20'0
the relationship between duration of component 18 6"5 0'08 2"5
and error of measurement. The results were
rather encouraging. The range of relative per- 19 1"7 0'09 10'6
centage error was between 1 per cent and 20 20 17"5 0"10 1"1
per cent with an average value close to 8 per
cent. The error is clearly related to duration of
component and not to any cumulative effect of events. The portion of the film contained twenty-
two behaviour components. The average relative
successive measurements. Table I gives the
percentage error was found to be close to 16 per
results of these computations.
The next step was to see the precision of cent, approximately twice as great as in the
transformation of visual observation to verbal previous case. Thus, it may be concluded that
the process of transformation from visual ob-
symbols. This was obtained in the following
manner. Cleaning behaviour was recorded servation to verbal symbols, on the average,
introduces approximately 8 per cent error. This
on film which was then replayed ten times, and
random error was found to be relative to the
each time the components were recorded on
duration of components, and was free of
tape with the aid of verbal symbols. Each in-
dependent recording was then measured for cumulative effects.
duration of components, and the measurements The Organization of Cleaning Behaviour
were subjected to the same statistical analysis as At first sight, as illustrated by the ethogram,
in the previous case. To aid the precision of cleaning behaviour is not only a complex
recording, a number of preruns were made with phenomenon but the observation of its per-
the film to become familiar with the sequence of formance gives the impression of a bewildering
648 ANIMAL BEHAVIOUR,
variety of movements apparently without any
definable system. The comparison of many
ethograms and continued direct visual obser-
vation, however, begin to reveal some consis-
tencies of performance, such as the regular
sequence of the two basic types of movements
(sweeping and rubbing) separating the behaviour
into body and leg cleaning types; the character-
istic shifts of the centre of gravity of the body
prior to a given type of cleaning involving one,
two or three legs; well-definable postures
which often accompany some sets of cleaning
movements; the relatively high frequency of
occurrence of some components over others;
the stereotyped nature of some movements and
patterns; the repetitive character of components
and of subdivisions of components.
These are no more than scattered observa-
tions. Nevertheless, they indicate that cleaning
behaviour in its complexity does have organ-
ization. The behaviour is found to be built up
of simple movements and of the repetition of
these movements forming complexes which are
recognized as behaviour components, while the
components themselves are repeated forming
major structural elements of the behaviour.
One of the most fundamental features of this
behaviour is its repetitive character.
During the course of a longer period of time,
such as during the course of a day, cleaning
behaviour is being repeatedly performed. Each
cleaning bout may be regarded as a complete
unit on the behaviour level. It is not known
whether these bouts are spread out in time
evenly, having approximately the same duration
and the same length of intermission periods
between them, or whether there exists some
variation of behaviour density in time, and if so,
whether such variation is in any way correlated
with particular times of the day. The random
variation of behaviour density cannot be a
priori excluded either. In any case, at this higher
level of organization, the behaviour consists
of the repetitive performance of cleaning bouts.
There is some difficulty about the unambig-
uous definition of cleaning bouts in time. The
sources of uncertainty are partly the indefinite
end of the bout, and partly the fact that other
behaviour types, namely, walking, feeding,
mating, motionless state, and so on, often intrude
during the performance of a cleaning bout.
It is more realistic to speak of the dominance
of a given behaviour type over others at any
given period of time in the range of the bouts.
The situation is similar to a woven meshwork
i?, 4
of cotoured fabric where at any area in the range
of patterns a certain colour dominates others.
In spite of these ambiguities, there are some
well-definable features of the cleaning bout.
Its beginning is often clear cut. The fly may be
engaged in various kinds of activities when
suddenly cleaning begins with an eruption of
movements. During the cleaning performance,
other behaviour types may intrude, most
commonly walking and motionless state. If care
is taken not to disturb the flies, the motionless
state becomes more and more predominant over
cleaning, and finally cleaning disappears alto-
gether. Then cleaning may start again, or some
other behaviour type takes over. Because of the
intrusion of other behaviours each cleaning bout
may be subdivided into smaller activity units
on the behaviour level.
As it has been mentioned earlier, cleaning
behaviour is performed by two basic types of
movements, the sweeping and the rubbing
movements. Based on this distinction, each
cleaning activity unit may be broken up into
the repeated alternations of the two movement
types, that is, into a series of alternating body
and leg cleaning complexes. This subdivision of
the behaviour is probably artificial. It is sug-
gested that one set of body cleaning movements
is followed by one set of leg cleaning movements,
and they together represent the natural unit
cycle of cleaning activity. The concept of unit
cycle is consistent with the view that the primary
function of the behaviour may be the removal of
foreign material from the body surfaces, and
thus the two movement types are part of the
same natural unit of cleaning activity. That this
may be so is indicated by the ethogram in the
53, 56, 69, 73, 152, 169, 171, 188 and 225 sec
regions. The ethogram was selected to show
some of the exceptions to this interpretation.
Front cleaning activity--that is cleaning per-
formed with the first pair of l e g s - i s often pre-
ceded by an initial set of rubbing movements, as
can be seen in the ethogram in the 0 to 33,
108, 135, 148 and 187 sec regions. In addition,
the H, W, and VS components often terminate
the cleaning activity without being followed by
the rubbing movements of the legs, as in the
16, 33, 103 and 205 sec regions.
Each unit cycle may be subdivided into be-
haviour components depending on the leg or
legs involved in the cleaning activity, or in the
body part toward which the behaviour is directed.
In this paper seven leg-cleaning and twenty
body cleaning components have been described.
 SZEBENYI: CLEANING BEHAVIOUR IN DROSOPHILA MEZANOGASTER
A given behaviour component may occur several
times within the same unit cycle, and there may
be several different components in each activity
type.
The components are made of repeated sets of
movements. The leg cleaning components, for
instance, are built up of a number of repeated
rubbing movements. The individual rubbing
movements are similar and they represent the
lowest level of organization of leg cleaning be-
haviour. Since the rubbing motions are mostly
along a straight line, they may be regarded as
one dimensional motor units.
The organization of the body cleaning com-
ponents is more complex because of the bi-
lateral symmetry of the body surfaces, and
because the sweeping of surfaces requires two
dimensional movements. The repetitive char-
acter of the behaviour is extended to both of
these aspects. The most characteristic wing
cleaning component (W) may serve as an ex-
ample. The wing on the right or on the left side
is taken between the third pair of legs near the
wing base. It is held at a slight angle to the
horizontal. The legs then perform one sweeping
movement from base to tip of the wing and at
the same time many small sweeping movements
over the wing surfaces between the inner and
outer edges. The motor units are thus two
dimensional. Toward the end of the component,
the wing is bent and is nearly vertical. This set of
movements is then often repeated on the same
wing several times, and then on the other wing,
or on alternate wings. Thus the repetition of
similar elements within the same component
occurs at a number of different levels of organ-
ization: bilateral repetitions of the entire
component, unilateral repetitions of sets of
movements, and finally the repetitions of two
dimensional simple movements (Fig. 3).
Not all body cleaning components have the
same complexity of organization. For instance,
the dorsal sweep (DS), another wing cleaning
component, has three aspects of repetitive
symmetry: the bilateral alternating, the bilateral
parallel and the unilateral repetitive perform-
ances. It lacks the two-dimensional motor
units, there being only repeated, smooth, one-
dimensional sweeps over the dorsal surfaces
of the wings (Fig. 3).
The side sweep (SS), another wing cleaning
component, is even simpler. It consists of a few
times repeated, uninterrupted sweeps along the
outer edges of both wings at the same time. Thus
the repetitive symmetry of the component is
649
reduced to the parallel state, :while the motor
units are one dimensional (Fig. 3).
The bilateral character of the components may
also appear as movements to the right or to the
left of the body part cleaned, as in the most
typical head cleaning component (H). Here
the head is rotated to the right or to the left
while the first pair of legs carry out two dimen-
sional sweeping movements over the dorsal
and ventral surfaces of the head (Fig. 3).
Thus it is found that cleaning behaviour in
D. melanogaster wild-type flies is not a random
aggregate of a great variety of movements, but
the behaviour is organized into a hierarchy of
repetitive complexes. The dgily cycles of the
behaviour(?) are composed of cleaning bouts
which are broken up by the intrusion of other
behaviour types into activity units. Each activity
unit is composed of sets of body and leg clean-
ing movements (unit types) which together
probably form the natural units of cleaning on
the behaviour level, the unit cycles. Each unit
type consists of sets of behaviour components.
Because of the repetitive nature of the be-
haviour, the movement complexes associated
with the components may be classified according
to three aspects of repetitive symmetry: bilateral
alternating, bilateral parallel, and unilateral
performances. Each of these complexes is built
up of repetitions of simple movements along
one or both coordinates of a Cartesian space,
forming two- or one-dimensional motor units,
the fundamental building blocks of the be-
haviour.
Summary
A descriptive account of cleaning behaviour in
D. rnelanogaster wild-type is given. The be-
haviour complex is analysed into seven leg
cleaning and twenty body cleaning components.
A method of obtaining ethograms is presented.
The method consists of a series of transform-
ations from event to written record with the aid
of verbal symbols recorded on tape. Three
systems of symbols of increasing degree of
complexity are proposed, and their usefulness
is illustrated by an actual ethogram. The problem
of error introduced by the series of transform-
ations is discussed. It is found that ethograms
may be constructed by the method with reason-
able degree of precision. The overall average
relative percentage error is found to be approx-
imately 16 per cent, one half of which is caused
by the transformation from event to verbal
record, the other half by the process of timing
the duration of components from the tape. The
650 ANIMAL BEHAVIOUR, 17, 4
A

[3

W --r

IL ,,~, [3 I c

[35 C

$9
A

H 1
[3
Fig. 3. Diagrammatic representation of the repetitive organization of four
cleaning behaviour components in D. melanogaster wild type flies. The
components are in sequence W, DS, SS, and H. (For the definition of these
symbols see text.) The movement complexes associated with these com-
ponents may be classified according to bilateral repetitive symmetry (A),
unilateral repetitive symmetry (B), and bilateral parallel symmetry (C).
The components are built up of repetition of simple movements performed
along the two axes of a Cartesian space (two-dimensional motor units), or
along a single line (one-dimensional motor units). The vertical arrows re.
present the direction of the main movement, while the double-headed
arrows indicate the direction of small movements superimposed upon the
main movement. The broken arrows show the direction of movement of
body parts cleaned, as in the case of the H component.
 SZEBENYI: CLEANING BEHAVIOUR IN DROSOPHILA MELANOGASTER
magnitude of this error is dependent on the
duration of components, and is free of cumula-
tive effects.
The organization of Cleaning behaviour is
discussed. The behaviour is not a random aggre-
gate of a great variety of movements, but it is
organized into a hierarchy of repetitive com-
plexes from the daily cycles down to the smallest
building blocks of the behaviour, the m o t o r
units. One of the most fundamental features of
this behaviour is its repetitive character.
Acknowledgments
The author would like to express his sincere
gratitude to D r M. Druger and to G. Smith for
their professional assistance and to the Society
651
of the Sigma Xi for a generous research grant
which made this work possible.
REFERENCES
Connolly, K. (1968). The social facilitation of preening
behaviour in Drosophila melanogaster. Anita.
Behav., 16, 385-391.
Heinz, I-I-J. (1949). Vergleichende Beobachtungen fiber
die Putzhandlungen bei Dil3teren im allgemeinen
und bei Sarcophaga carnaria L. im besonderen.
Z. TierpsychoL, 6, 330-371.
Lorenz, K. (1956). The objectivistic~theory of instinct.
In L'instinct clans le comportement des animaux
et de l'homme (ed. P. P. Grassr). Paris: Masson
et Cie.
Tinbergen, N. (1965). Animal Behavior, pp 85-86.
New York: Life Nature Library, Time Inc.
(Received 3 July 1968; revised 2 February 1969;
Ms. number: A724)