You are on page 1of 10

Animal Behaviour 77 (2009) 785794

Contents lists available at ScienceDirect

Animal Behaviour
journal homepage: www.elsevier.com/locate/yanbe

Essay

Niko Tinbergen and the red patch on the herring gulls beak
Carel ten Cate*
Behavioural Biology, Leiden University

a r t i c l e i n f o
One of the classic studies in animal behaviour is that by Niko Tinbergen on the stimuli that elicit begging
Article history: behaviour in young herring gull, Larus argentatus, chicks. Tinbergen examined which features of the beak
Received 3 October 2008 induced chicks to peck by using various cardboard models of herring gull heads in different shapes and
Initial acceptance 10 November 2008 colours. Leiden University, home university to Tinbergen at the time, still has a summarizing overview of
Final acceptance 15 December 2008 the data of the initial experiment of the study. In that experiment, models with a yellow beak with
Published online 28 February 2009 patches in various colours and different positions were presented to young chicks. In this essay, I relate
MS. number: 08-00639 those earliest data to the various publications in which Tinbergen discussed them over the years.
Subsequent publications became more and more detached from the original data, resulting in sizeable
Keywords: discrepancies between the original study and later descriptions of when and how the experiment was
begging response
done and its outcome. I rst sketch the scientic context of the herring gull study. Next I present the
chick
original data and document the subsequent changes. I discuss what might have led Tinbergen to modify
colour preference
herring gull his account of the study over the years, relating it both to its historical context and to the issue of
instinct expectancy biases.
Larus argentatus 2009 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
observer bias
releasing mechanism
Tinbergen
visual perception

No textbook on animal behaviour would be complete without The herring gull study, carried out on the Dutch island of Tersch-
mentioning the important contributions of Niko Tinbergen, one of elling, was not the rst one in which Tinbergen used the technique
the founding fathers of our discipline. Tinbergen received the of presenting models (e.g. ter Pelkwijk & Tinbergen 1937; Tinber-
ultimate recognition for his work when the 1973 Nobel prize in gen & Kuenen 1939), but it was certainly the most extensive.
Physiology or Medicine was awarded jointly to him, Konrad Lorenz Carried out over half a century ago, the study has been, and still is,
and Karl von Frisch for their discoveries concerning organization mentioned in many textbooks on animal behaviour (e.g. Wilson
and elicitation of individual and social behaviour patterns (https:// 1975; Gould 1982; Goodenough et al. 1993; Drickamer et al. 1996;
nobelprize.org/nobel_prizes/medicine/laureates/1973/press/html). Manning & Dawkins 1998; McFarland 1999; Barnard 2004; Alcock
Current day studies of animal behaviour owe a lot to Tinbergens 2005; Scott 2005) and on related disciplines (e.g. Pinel 2000;
pioneering work and several of his studies serve as classic textbook Zupanc 2004), and hence has become a true classic. It also featured
examples up to the present day. This paper is about one of those prominently in a television programme (Moments of Genius, Pro-
studies: the stimuli that make young herring gull, Larus argentatus, gramme 17: An Instinct of Behaviour, produced by the BBC in 2000)
chicks peck at their parents beak to obtain food. devoted to Tinbergen as a Nobel Prize recipient. Apart from its
One of Tinbergens strengths was his ability to derive interesting pioneering experiments and its contribution to the emerging
research questions from observing wild animals and then trans- ethological theory, another likely reason why it has become such
lating these questions into elegant and simple experiments. This a classic may be that Tinbergen himself published frequently about
was often done by using articial models (dummies) to obtain it, thus making it widely known inside as well as outside the animal
responses from free-living animals, thus introducing a method that behaviour research community.
has become an indispensable tool for studies on animal behaviour. Leiden University, where Tinbergen worked until his move to
Oxford in 1949, kept some remnants of Tinbergens work from that
period including some material underlying the rst publications
* Correspondence: C. ten Cate, Institute of Biology, Leiden University, P.O. Box
about the herring gull study: a sheet with the summary of the data
9516, 2300 RA Leiden, Netherlands. from the rst eld season (1947, see Fig. 1a; for a larger image, see
E-mail address: c.j.ten.cate@biology.leidenuniv.nl (C. ten Cate). Supplementary Material), the handwritten and typewritten drafts

0003-3472/$38.00 2009 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.anbehav.2008.12.021
786 C. ten Cate / Animal Behaviour 77 (2009) 785794

of the rst more extensive publication (in Dutch) of the experi- important role in his and Lorenzs early theorizing and in the
ments (Tinbergen 1949), some notes with tabulated data under- emergence of ethology as a separate discipline. The herring gull
lying some gures of the 1949 publication, and a few of the models study was a key study in this phase of conceptual development.
used in that study. I became intrigued when I noticed that the data Although initially presented as a mainly curiosity-driven study
sheet and the 1949 publication did not quite seem to match, and (Tinbergen 1948a, b), Tinbergen later on wrote that the general aim
when I consulted my copy of the Study of Instinct (Tinbergen 1951) of the herring gull study was to examine the begging behaviour as
it showed yet another picture of the experiments, one that no an example of, and a way to get insight into, the release of
longer seemed congruent with the original data. This essay instinctive behaviour (the pecking of the chicks to the parents tip
describes how Tinbergen used the data of the rst eld season in of the beak) by sign stimuli that activate a releasing mechanism
subsequent publications. It provides an illustration of his devel- (Tinbergen & Perdeck 1950).
oping insights, but also shows that his presentation of the data Normally, young chicks direct their pecking mostly towards
became more and more detached from the actual experiment in the red patch on the lower mandible of the yellow beak of their
which they had been collected. parents. By the time Tinbergen developed an interest in this
All textbooks discussing the experiments refer to later accounts behaviour, earlier studies by Heinroth & Heinroth (1928) and
of the study that include several more important modications of Goethe (1937) had shown that newly hatched chicks already have
the original data, some of which are only apparent when comparing a tendency to peck at red objects. To analyse the relevant stimulus
them with the earliest publications and the original data. The characteristics for the release of this pecking behaviour more
earliest published accounts are in Dutch and German and are not systematically, Tinbergen and his collaborators (mostly under-
only difcult to obtain for those interested, but also deviate in some graduate students from Leiden University) took young gull chicks
details from the original data. Therefore, it is of scientic and out of the nest and presented to them various two-dimensional
historical interest to present the original data and trace the various painted cardboard models and some three-dimensional models
changes. As discrepancies between original data and published that were gently moved in front of the chicks head. The number
accounts can easily give rise to the suspicion of intentional fraud I of pecks by chicks over a period of about 30 s was used to assess
want to emphasize upfront that I have no reason to believe that this how strongly a particular model released the pecking. Over
is such a case. On the contrary, it is because Tinbergen himself has a period of several years, models of different shapes and colours
been explicit about some of the more important changes that it is were used to examine the role of stimulus characteristics such as
possible to reconstruct the chain of events resulting in the the presence, colour and position of the patch, the colour of the
discrepancies. beak and head, and the shape of the head. Among other results,
Below I rst provide some background, followed by a general the various experiments indicated preferences for certain patch
description of the study. Next I give a chronological account of the and bill colours and shapes. Tinbergen was also able to demon-
various publications in which the data have been presented over strate that combining various attractive features could lead to
the years, describing the changes that have accompanied them. I a so-called supernormal stimulus, a thin red rod with three
will end with trying to answer the question of what might have sharply edged white bands at its tip, which induced more pecks
brought Tinbergen to make the various changes. Several recent than a naturalistic stimulus.
books have provided a comprehensive account of Tinbergen as Most textbooks mentioning the herring gull study refer to the
a scientist and as a person (Roell 1996, 2000; Kruuk 2003; Bur- paper by Tinbergen & Perdeck (1950). That paper incorporated and
khardt 2005). They pay attention to the context for Tinbergens extended the data presented in the earlier mentioned 1949 paper
studies and therefore make it possible to put his modications in (Tinbergen 1949). However, the 1947 data were mentioned in two
a historical perspective as well as relating them to current notions publications before 1949, along with studies on some other subjects
of desirable scientic conduct. (Tinbergen 1948a, b). After 1950, the 1947 data were presented in
The Study of Instinct (Tinbergen 1951). Then, an extensive account of
BACKGROUND OF THE HERRING GULL STUDY the Tinbergen & Perdeck studies and their outcome was given in The
Herring Gulls World (Tinbergen 1953). In 1954 the ndings were
One of the core scientic disputes in the study of animal presented at a meeting of the Josiah Macy Foundation on group
behaviour during the rst half of the 20th century concerned the processes (Tinbergen 1955), a meeting of historical importance in
role of animal instincts: the tendency to show adequate behaviour which leading animal behaviour researchers from both sides of the
in a particular situation, seemingly without any relevant experience Atlantic met (see also Kruuk 2003; Burkhardt 2005). It is this series
or learning being involved. For many comparative psychologists at of publications that I discuss in this essay.
the time, instinctive behaviour was outside their interests: spon- Over the years the herring gull study has received not only
taneously shown adequate behaviour in nave animals was ascribed praise, but also criticism on both the methodology and its conclu-
to the presence of an instinct or drive for that behaviour and that sions. In his discussion of the study, Kruuk (2003) mentioned some
was it. Their focus was on intelligent behaviour involving learning. methodological issues, such as the nonstandardized way in which
Instinctive behaviour was only of interest as an evolutionary the models were moved in front of the chicks and the lack of
precursor of such intelligent behaviour. In contrast, to the early statistical analysis. Later, more systematic studies on the pecking
ethologists such as Lorenz and Tinbergen, innate instincts were response by, in particular, Hailman (1967, 1969) but also others (e.g.
seen as a major factor shaping the behaviour of animals and Alessandro et al. 1989; Eypasch 1989) replicated several of the
considered a subject to be analysed experimentally (see Roell 2000 ndings, but also conrmed the validity of several of the criticisms
and Burkhardt 2005 for historical overviews of the different views). and led to different conclusions on the stimuli releasing the
This led to concepts such as the innate releasing mechanism (IRM), begging behaviour (see Discussion). Such a fate, however, is
a perceptual mechanism sensitive to specic sign stimuli or certainly not specic to Tinbergens studies, and illustrate how the
releasers, that is, stimuli that evolved to produce (release) specic discipline matured. They do not take away the value of Tinbergens
adaptive behavioural responses within a particular context without pioneering experiments as a source of inspiration and a prime
learning being involved. While Tinbergen himself started to think example of original eld experiments.
in less rigid ways about the innateness of behaviour in later years This essay does not focus on how the experiments were done,
(e.g. Tinbergen 1955, 1963), instincts and their analysis played an but on how Tinbergen treated the data once collected. I will not
(f)
(b) (d) KLEUR VLEK
100

RODE VLEK 100


ROOD
rot
105
105
ZWARTE VLEK
ZWART
schwarz
85 85
BLAUWE VLEK BLAUW

blau 71
70 WIT
WITTE VLEK

30
(a) weiss 938 REACTIES
30
TOTAAL 1095 REACTIES GEEN VLEK

RODE VLEK - ZWARTE VLEK

C. ten Cate / Animal Behaviour 77 (2009) 785794


24
100
100
ROOD VOORHOOFD ROOD

105 86
ZWART

1276 REACTIES
(e)
85 (g) PLAATS

100

71 24
809 REACTIES

30

74
13
24 (c) 73
70 70

13

1947 1948a 1948b 1949

Figure 1. (continued)

787
788
(h) (k) (n)
100 100
Red
Red Red
1

86 105 Black 86

Black 2 Black

71 85 Blue 71
Blue
3 Blue

59 71 White
59
White
4 White

25 30
25

C. ten Cate / Animal Behaviour 77 (2009) 785794


5
938 Reactions 938 Reactions

(i) (l)
A
100

B
24

809 Reactions

(j) (o)
A 2
2

(m) 74
12 12

47 47

B
75 75
70

13
13
70 70
1950 1951 1953
C. ten Cate / Animal Behaviour 77 (2009) 785794 789

deal with all the experiments on the begging response of herring


Figure 1. Various versions in which (a) the original data have been presented in (bo) successive publications. Each column presents gures from the same publication, indicated by reference to its year of publication by Tinbergen at

Verlag; 1949: Bijdragen tot de Dierkunde (copyrights 2008: Contributions to Zoology. All rights reserved. ctz@naturalis.nl); 1950: Behaviour, Koninklijke Brill Publishers NV; 1951: Study of Instinct, HarperCollins (Oxford University Press);
the bottom of the column. Arrows connect gures presenting the same data. For all gures presenting the results of different colours of the patch, the sequence of colours is always the same from top to bottom: red, black, blue, white,
no patch. Several gures give the relative number of responses. These are (again from top to bottom): (a) 100, 105, 85, 71, 30, 24; (b) 100, 105, 85, 70, 30, 24; (f) 100, 105, 85, 71, 30, 100, 86; (g) 100, 24; (h) 86/105, 71/85, 59/71, 25/30;
(i) 110, 24; (n) 100, 86, 71, 59, 25. See also Appendix with notes for variation in numbers between the gures. Figures are reproduced with permission of the various publishers: 1948a: De Levende Natuur; 1948b: Experientia, Birkhauser
gull chicks, but only those of the initial study, as presented in the
1947 data sheet. In doing so, I put just one tiny part of the herring
gull study under the microscope, to reveal details one would not
otherwise see. The danger of such an approach is that it will
inevitably distort the full picture of this study: I select this part of
the study not only because we have the original data, but also
because it shows the most striking modications over the years. I
want to emphasize that what follows should therefore not be
considered representative of the whole study or of everything
Tinbergen did.

CHRONOLOGICAL OVERVIEW OF TINBERGENS PUBLICATIONS


ON THE 1947 DATA

I discuss the different publications exclusively in relation to the


1947 data as presented in Fig. 1, in which also the relevant gures of
the various subsequent publications are reproduced.

1947 Data

The data sheet (Fig. 1a) presents the results under the heading
Pikreacties Zilvermeeuw (Pecking responses herring gull;
Terschelling 1947). It also provides the total number of responses
(1095). The score for the responses released by each of the
models is given in proportion to those released by the normal,
red-patched, model for which the score is normalized to 100. For
the model with the patch on the forehead, the number of
responses to different parts of the beak is given, adding up to
a total of 157 pecks. There are no notes left of the data as
collected in the eld.

De Levende Natuur

The paper in De Levende Natuur (in Dutch; Tinbergen 1948a;


Fig. 1b, c) is a very readable and stimulating account in a journal for
naturalists of various studies by Tinbergen and coworkers during
a 2-week animal behaviour course in the dunes of Terschelling. The
herring gull study was one of the studies. The paper provides the
background for the 1947 data sheet. The study is introduced by
noting that the conspicuous markings on skin, feathers and fur of
animals often serve as social signals releasing the instinctive
responses in conspecics. Tinbergen mentioned that an earlier
study by Goethe (1937) implied that young chicks pecked at various
red objects, which raised his interest in the pecking response. To
examine this behaviour in more detail, cardboard models were
made with patches in different colours, as well as one model
without a patch and a model with a patch on the forehead, thus
matching the six models shown on the data sheet. The models with
the differently coloured patches were offered to examine whether
1953: The Herring Gulls World, Oxford University Press.

any patch would be attractive and the model with the red patch on
the forehead to examine whether the chicks would peck at the
patch or at the tip of the beak. It is reported that chicks were
collected from the colony and tested in a hide. A chick was pre-
sented with two models sequentially, one of which was always the
standard model (red patch on the beak), for 30 s each. The same
models were presented up to 20 times after each other. To stimulate
the chicks to beg, the experimenters gave an imitation of the
parental mew call and moved the models slightly. Chicks stimu-
lated in this way could give up to 1 peck/s during the 30 s period.
Collecting the data presented on the pecking behaviour took no
more than 3 days, according to the paper. It is also mentioned that
the models were painted with watercolours and that the colouring
of some of the models got wet and spread out. The numbers pre-
sented in Fig. 1b are the same as on the 1947 data sheet (with one
790 C. ten Cate / Animal Behaviour 77 (2009) 785794

small exception; see note 1 in the Appendix). Tinbergen mentioned In 1948 they discovered that repeatedly presenting chicks with the
that the results (Fig. 1b) were surprising, with the black and blue same model led to habituation of the chicks to that model. He
patches giving rise to approximately equal numbers of pecks as the inferred that this led to a reduction in the number of pecks to the
red patch, with the white patch receiving fewer pecks. He wrote red model. As a consequence, its attractiveness had been under-
that more experiments are needed to decide whether black does estimated and hence the relative preference of other patch colours
indeed produce more responses than red. The model without was overestimated. For this reason, a separate experiment was
a patch got very few responses, but these were, according to Tin- done in which a model with a red patch was tested against one with
bergen, directed at the tip of the beak. A separate gure (Fig. 1c) a black patch. The outcome, shown at the bottom of Fig. 1f, was that
illustrates the distribution of pecks over various locations at the the red-patched model now indeed received more pecks. Based on
head for the model with the red patch on the forehead (again with the nding that the black-patched model received 86% of the
a small deviation in numbers; see note 2 in the Appendix). Tin- number of pecks to the red model, Tinbergen explained he used this
bergen concluded that the chicks do not peck at something red, value to adjust the values for the models with the all other patch
but at a dark patch at the tip of the beak, and also that the position colours by multiplying them by 0.82 (86/105), arriving at the values
of the patch is important. indicated by the vertical white bars in Fig. 1f. The adjusted data
were now discussed as being the real ones, leading to the conclu-
Experientia sion that they showed two effects: the degree of contrast between
patch and beak colour was a relevant factor, but in addition the
Tinbergens (1948b) paper is Experientia (in German; Fig. 1d, e) is chicks had a preference for the colour red. These two factors were
based on several talks given at the University of Zurich in January next tested separately, and conrmed, in several experiments in
1948, again presenting the herring gull study together with several 1948, also described in the paper.
others. The data are discussed in the context of the concept of sign At the end of the paper Tinbergen presented the data for the
stimuli that release behaviour that is otherwise prevented from model with the red patch on the forehead (Fig. 1g). He noted the
being expressed. Tinbergen explained that animals respond to just outcome of the experiment as a typical demonstration of
a few stimuli out of all those perceived. Referring to Fig. 1d, he a congurational stimulus situation, indicating that the responses
mentioned that the results of the experiments show that young of the chicks are based on complex processes within the senses and
chicks do not respond to the colour of the patch, but to any patch the nervous system. He mentioned that this experiment was also
contrasting with the yellow beak, with a black patch giving an even done during 1947, but presented it as a direct comparison of this
stronger response than a red one. Three pages later, he discussed model with the standard one, which is different from the descrip-
Fig. 1e as an illustration that it is not just the presence of particular tion in 1948 where he wrote that all models were part of the same
elements that release the begging behaviour, but that their mutual set. Being part of the same data set, here too the results will have
arrangement is also of importance. He provided this as an example suffered from the habituation effect, leading to an overestimation
that sign stimuli can be complex, consisting of a congurational of pecks to the model with the patch on the forehead. However, in
(Gestalt) relationship between different elements. He did not this case the outcome was not corrected (see note 4 in the
mention how the pecks were distributed over the model. In this Appendix).
gure, Tinbergen presented the data for the pecking at the model
with the patch on the forehead with those for pecking at the Behaviour
standard model, not mentioning that these data originated from
the same experimental series as the data on patch colour Tinbergen & Perdecks (1950) paper (Fig. 1h, i, j) is the one to
preference. which most textbooks refer. This will not only be because it is the
most extensive paper, but also because it is the rst one in English.
Bijdrage tot de Dierkunde Essentially it presents the same data as the earlier papers, but adds
experiments done by Perdeck during one additional eld season
Tinbergens (1949) paper (in Dutch; Fig. 1f, g) is about the (1949). Again, the introduction mentions the work by Goethe. Then
herring gull study only, adding many new and interesting experi- it introduces the concepts of sign stimulus and the IRM (innate
ments to the 1947 data, for instance using differently coloured releasing mechanism), with reference to Lorenz (1935). The herring
beaks, and models with a grey beak with patches in different gull study is presented as an experimental study of these concepts,
shades of grey. The paper begins with a description of the begging using the pecking at the red patch as a model system. The
behaviour of herring gull chicks, again mentioning that Goethe description of the methodology explains that over the years the
(1937) did some preliminary studies suggesting that nave chicks methods became more sophisticated, outlining the issue of habit-
had a preference for pecking at red objects. Tinbergen next uation and how that was avoided in later years by a balanced
mentioned that his own study was initiated with the aim of doing design. Again the rst data presented are the 1947 ones (see note 5
a systematic experimental study on external stimuli that trigger in the Appendix). To our astonishment, black got more responses
innate behaviour patterns. He stated that the studies were done than red, Tinbergen wrote (page 7). He explained that afterwards
during the eld seasons of 1947 and 1948. The rst data presented they began to suspect that this was due to negative conditioning
in the paper are the results obtained in 1947, collected over 3 days. and that this led to a series of tests with the red-patched beak
Tinbergen began by explaining that the model with the red patch against the black-patched one. As in the previous paper, the original
received substantially more responses than a model without data and the corrected ones are presented in one gure (Fig. 1h; see
a patch, but also that patches in other colours received relatively note 6 in the Appendix).
many responses, referring to Fig. 1f (see note 3 in the Appendix). He The paper then presents several other experiments before the
mentioned that the model with a black patch received even more position of the red patch is discussed and presented in Figure 13B of
pecks than the red-patched model. He then stated that during the the paper (reproduced as Fig. 1j). It is mentioned (page 18) that
1948 season they discovered they had made a methodological some models were made in which the red patch, while of the usual
error: not all models had been presented equally often. The model size, was situated at abnormal places. The text notes that usually all
with the red patch, serving as a standard against which the other pecks are directed to the red patch when it is in the normal posi-
models were tested, was thus presented much more frequently. tion, but that the results show that the patch loses impact the
C. ten Cate / Animal Behaviour 77 (2009) 785794 791

further away it is from the tip of the beak. Later in the paper a note tested each model against the standard model. Instead, it is
adds that experiments done in 1950, for which no details are pre- stated (page 188) that we had to take care that every model got
sented, indicate that the crucial factor for getting fewer responses the same chances next describing a design in which all models
to models with the patch further away from the tip of the beak is are presented equally frequently and in a systematically varied
that the chicks respond to the lowness of the patch, rather than to sequence. This might have been so for later experiments
distance from the tip. Lowness is next mentioned as the basis for (reported in Tinbergen & Perdeck 1950), but not for the rst
the congurational effect that is demonstrated by the nding, ones. The text of the chapter closely follows Tinbergen & Perdeck
presented in Figure 26 of Tinbergen & Perdeck (Fig. 1i), that the total (1950). It also mentions that all gures are the same as in that
number of responses to the model with the patch on the forehead is paper. Indeed they are, but with the exception of Fig. 1n. This
lower than that to the standard model. In the discussion on the gure now provides the corrected data only, without mentioning
results of the various experiments presented in this paper, Tin- that they had been corrected. Tinbergen states (pp. 191192)
bergen proposed the existence of seven sign stimuli that release or that he was forced to conclude that the red patch acted both
direct the pecking response, among them the mew call, lowness, through its colour and through its contrast with the bill.
nearness, and a red patch at the tip, characterized by colour and Figure 1o is identical to Fig. 1j, but lacks the earlier note that in
by contrast. Of these seven, Tinbergen identied the mew call and reality the models had eyes and that, by accident, these were
the red patch as true releasers, in the sense as characters evolved as missing in the gure. The description of the results mentions
adaptations to a communicative function (Tinbergen & Perdeck, that the gure shows that the red patch was a directing stim-
page 38). ulus, but received fewer pecks the further away it was from the
Although not mentioned explicitly, the text suggests that the tip of the beak.
data for the models presented in Figures 13A and B (Fig. 1j) origi-
nate from one and the same experiment. However, the data in B are Group Processes
clearly the ones collected in 1947 (see note 7 in the Appendix),
while the other data most likely originate from 1949 and hence At the Macy Foundation conference Tinbergen again presented
were collected using a different methodology. The legend of the the herring gull experiments. He illustrated them with gures
gure mentions that Model B had an eye at the usual place and taken from The Study of Instinct. The book about the meeting
that this was not drawn by mistake. For Figure 26 (Fig. 1i) the text (Tinbergen 1955) reports Tinbergens presentation including all the
refers to Tinbergen (1948b). No mention is made of the fact that the discussion arising from it. About Fig. 1n, he said (page 89) that it
Figures 13B and 26, which are remarkably different in how the showed the relative number of responses to bills with patches in
models are drawn, refer to the same model and concern the same various colors, and presented as outcome: Two interesting nd-
experiment, being part of the experimental series of 1947. ings emerge: red stimulates more strongly than anything else, and
yet other colors also release quite a few responses. He did not
The Study of Instinct mention the corrections on the data. Tinbergen was questioned
about many details of the experiments and his interpretations
In this classic book (Tinbergen 1951; Fig. 1k, l, m) the herring gull (Tinbergen 1955). In particular Schneirla, a well-known and
study is mentioned in three places. First, Tinbergen used the study important comparative psychologist, wondered whether Tinbergen
to illustrate the concept of a sign stimulus. He presented three was successful in controlling the presentation of the models to
gures that all refer to Tinbergen (1949). One of these gures is young gulls. Tinbergen admitted that his method had been crude.
Fig. 1k. Tinbergen now presented the corrected data only, without Nevertheless, he replied in the afrmative to a question about
mentioning the original outcome or the correction. These results, whether the chicks were raised from the egg and had not seen an
Tinbergen stated, demonstrate that both the contrast between adult gull. Again, this may have been so in some later experiments,
patch and beak and the colour red contribute to the releasing value but it was not true for the earliest experiments, as also noted
of the red patch on the parents beak. by Kruuk (2003). The method descriptions in the early papers
At page 77 in the book Tinbergen presented the comparison of (Tinbergen 1948a, b; Tinbergen & Perdeck 1950) all mention that
the model with the patch on the forehead with the standard chicks were collected from nests. As parents brood newly hatched
model (in this case referring to Tinbergen 1948a), using it to chicks, most, if not all, chicks used in the rst experiment saw
demonstrate the concept of congurational stimuli (Fig. 1l). adult gulls. Also, although Tinbergen mentions (page 5) that they
Finally, in a discussion on the difference between releasing and succeeded in acquiring a fair proportion of inexperienced chicks
directing stimuli he presented Fig. 1m (page 83) as a demonstra- with respect to feeding (Tinbergen & Perdeck 1950), some chicks in
tion that the red patch has both a releasing and a directing effect. the rst experiments had already been fed by parents. This can be
However, the 13 pecks given between the tip of the beak and the inferred from the very rst paper (Tinbergen 1948a) which ascribes
forehead patch are left out and not mentioned in the text (see note particular behavioural responses of some chicks to having experi-
8 in the Appendix). Again all gures are presented separate from enced parental feeding.
each other.
DISCUSSION
The Herring Gulls World
It is remarkable to see how some exploratory experiments have
In this popular account of the herring gull and its behaviour gained such a prominent role in Tinbergens publications. Perhaps
(Tinbergen 1953; Fig. 1n, o), chapter 22 is devoted to the studies even more remarkable is to see how subsequent publications get
on the begging behaviour of the chicks. As in earlier papers, more and more detached from the original data. While the changes
Tinbergen began by referring to Goethes study. As in Tinbergen from each paper to the next are limited, and to some extent
& Perdeck (1950), he stated that the rst experiments were done understandable, they accumulate to a sizeable discrepancy
during 2 weeks on Terschelling during 1946. He noted that the between the original data and the latest accounts. If one reads
chicks were presented with the models using a design in which about the experiments as presented in The Herring Gulls World,
different chicks received different models as the rst one of without knowledge of the intermediate publications, one would
a series. No mention is made of the fact that the original series believe that the study was done during 1946 instead of 1947; that
792 C. ten Cate / Animal Behaviour 77 (2009) 785794

the methodology consisted of a balanced design offering each Over the years, discussions on the issue of desirable or
model with equal frequency instead of repeated testing against responsible conduct in research have identied strong views and
a standard model; that Fig. 1n and o were based on separate expectations on the outcome of experiments as a risk factor that
experiments; that the experiment reported in Fig. 1o was done with may, intentionally or unintentionally, bias researchers in consid-
models lacking the eyes, and that Fig. 1n represented the proper ering the evidence (Anonymous 1995; Bateson 2005) and result in
responses to models with differently coloured patches on their irresponsible scientic conduct. The herring gull story seems
beak. In fact, when I gave students both the 1948a paper and a clear illustration of how an expectancy bias can colour the data
chapter 22 of The Herring Gulls World to compare, quite a few treatment and conclusions. Nevertheless, opinions of what is
believed these must have been different experiments. The data in irresponsible conduct vary between people (Montgomerie &
The Study of Instinct also deviate from the 1947 summary, and Birkhead 2005) and are likely to vary over time. Hence, views may
suggest the presence of three separate experiments. In addition, differ on the extent to which Tinbergens behaviour should be
Fig. 1c, j and m, based on the 1947 data, differ in numbers from the seen as irresponsible. In my view it should if we apply todays
original data (albeit marginally), as well as from each other (see standards. However, Tinbergen did the study in another era.
Appendix). In fact, none of the publications matches the original Todays practice that papers are subject to independent and crit-
numbers. ical refereeing, which can serve to counteract such biases, was not
We cannot be sure about the causes for the changes over the an established practice at the time, and is very unlikely to have
years, but may speculate about these. One factor might be related happened. More importantly, the postwar years were a decisive
to Tinbergens move from Leiden to Oxford during 1949. By all period for the new discipline to establish itself as a serious and
available accounts (Roell 1996, 2000; Kruuk 2003; Burkhardt competitive branch of science, particularly in relation to the
2005), it is obvious that this was a very hectic period in his life. He comparative psychologists dominating the study of animal
had to teach many students, was coping with an increasing burden behaviour in the U.S. (Burkhardt 2005). Apart from Tinbergens
of administrative duties and was disappointed about the atmo- strong expectations, one can imagine that, put in the context of
sphere in postwar Holland. Anyone makes mistakes and small the scientic controversies surrounding the birth of the discipline,
inaccuracies such as wrong numbers for N values or mixing up he was tempted to present the data as more clear-cut than they
dates might be consequences of this hectic period. They may be were. Although it cannot be a justication, one might even argue
sloppy, but these issues are details and irrelevant to the main that the discipline might not have been where it is today had he
conclusions of the herring gull study. However, the modications not done so.
to the original data, such as presenting the corrected data as the Finally, how about the conceptual insights provided by the
real ones without mentioning the corrections, presenting data as study? Overlooking the publications, it can be noted that what
though they originate from different experiments, or presenting started off as a relatively simple experiment on what stimulated
the methodology as different from what is has been, extend chicks to beg grew into a story about the analysis and complexity
beyond such details. The fact that Tinbergen in the 1949 and 1950 of sign stimuli, releasers and releasing mechanisms in general. It is
papers extensively discussed the methodological error of pre- a story that denitely contributed to the rise of ethology. However,
senting the standard stimulus more frequently than the other one might also say that the same story, rather than making these
models shows he had no intention to manipulate or manufacture concepts widely acceptable, may have contributed to their demise.
the data. Tinbergen also stated that in subsequent experiments he This is because Tinbergens inference about the presence of more
altered the test procedure to avoid the problem, and hence the and more abstract and complex congurational sign stimuli as
correction affects only a small subset of the data. Nevertheless, the nearness or lowness required the concepts, and therefore the
question remains why he changed the early data and presented proposed releasing mechanisms needed to explain them, to
these as the proper data, rather than repeating the experiment in become more and more intricate. At the same time, the experi-
a more balanced way. ments were not sufciently rigorous to support these ideas. This
In his biography of Tinbergen, Kruuk (2003) expressed worries was already questioned by Schneirla and others at the Macy
that the herring gull study may have suffered from observer conference mentioned before, and to some extent admitted by
biases, expectations of how the chicks should behave. Such biases Tinbergen. Later on, this was conrmed by extensive and more
may unintentionally have affected the way in which the experi- controlled experiments on the begging of gull chicks, in particular
menters presented the various models and hence the outcome of and most thoroughly by Hailman (1967, 1969) and Gould and
the experiments. The studies discussed above suggest that the coworkers (Gould 1982; Margolis et al. 1987; Alessandro et al.
inuence of biases did not stop here, but also affected the 1989) but also others (Evans 1979; Eypasch 1989; Zippelius 1992).
subsequent interpretation and presentation of the results. With respect to the colour preferences, studies by Hailman (1967)
The majority of the herring gull publications present the results and Eypasch (1989), while conrming that nave chicks did have
not as a test, but rather as an illustration of ethological concepts. a preference for red objects over blue ones, and for blue over
The corrected data for the rst experiment in the herring gull yellow, showed that these preferences were not so clear as Tin-
study tted better with the expectations Tinbergen expressed bergen thought at the time (see also ten Cate et al. 2009 for
early on, and the experiment to test the effect of habituation a discussion). As for other features, Hailman (1967, 1969) and
probably convinced him that the early data were anomalous. From others (Alessandro et al. 1989) demonstrated that chicks collected
this it is a small step to considering and treating the corrected data from the eld and with some experience with parents had specic
as the real ones. It may also explain why Tinbergen apparently saw preferences, also for some congurational features. However, truly
no problem in separating the data obtained with the model with nave chicks, that is, hatched in an incubator rather than collected
the patch on the forehead from the other ones and in recombining from the eld were less clear in their preferences and the results
and presenting them in different contexts. While such data use showed no need to invoke a sensitivity to complex, congura-
may reect the less rigid procedures of those early days, it also tional sign stimuli. For instance, when movements of the model
seems an example of Tinbergens tendency, observed by Kruuk were controlled for, nave chicks pecked equally often at the patch
(2003), to t the data to his ideas rather than the other way wherever it appeared on the head. Hence they appeared to
around. It raises the question whether Tinbergen went too far in respond to discrete features, rather than congurations. So, these
his adjustments of the data. studies demonstrated that it was possible to link several of the
C. ten Cate / Animal Behaviour 77 (2009) 785794 793

phenomena that Tinbergen claimed to be complex and innate to References


simpler perceptual properties of the sensory system or to inter-
pret them as the outcome of a learning process. Evans (1979) and Alcock, J. 2005. Animal Behavior: an Evolutionary Approach, 8th edn. Sunderland,
Massachusetts: Sinauer Associates.
Eypasch (1989), for instance, demonstrated that a few feeding Alessandro, D., Dollinger, J., Gordon, J. D., Mariscal, S. K. & Gould, J. L. 1989. The
trials and 510 min of exposure were sufcient to give chicks ontogeny of the pecking response of herring gull chicks. Animal Behaviour, 37,
a preference for a variety of odd, nongull-like, models. How- 372382.
Anonymous 1995. On Being a Scientist. Responsible Conduct in Research, 2nd edn.
ever, whether this, in combination with some methodological Washington, D.C.: National Academy Press
problems, should serve as a basis to dismiss the study altogether, Barnard, C. 2004. Animal Behaviour. Mechanisms, Development, Function and
as argued by Zippelius (1992), is a different matter. Rather, Evolution. Harlow: Pearson.
Bateson, P. 2005. Desirable scientic conduct. Science, 307, 645.
Tinbergens outcome might be considered to reect the prefer- Burkhardt Jr, R. W. 2005. Patterns of Behavior: Konrad Lorenz, Niko Tinbergen, and
ences of chicks with some experience with the parents, which is to the Founding of Ethology. Chicago: University of Chicago Press.
be expected with chicks obtained from the eld. Drickamer, L. C., Vessey, S. H. & Meikle, D. 1996. Animal Behavior. Mechanisms,
Ecology, Evolution. Dubuque, Iowa: W. C. Brown Publishers.
What has so far remained an open question is whether
Evans, R. M. 1979. Responsiveness of young herring gulls to stimuli from their own
Tinbergens corrected data do, as he assumed, match the outcome and other species: effects of training with food. Canadian Journal of Zoology, 57,
that would result using a balanced design. This question has been 14521457.
Eypasch, U. 1989. Das Zusammenwirken von angeborener Bewertung und
addressed in a recent experiment using replicates of Tinbergens
Erfahrung bei der Losung eines Erkennungsproblems untersucht an jungen
models (ten Cate et al. 2009). It shows that Tinbergens intuitions Silbermowen (Larus argentatus Pontopp.). Ph.D. thesis, Friedrich-Wilhelm
were largely correct. Universitat, Bonn.
To conclude, looking backwards it is important to note that the Goethe, F. 1937. Beobachtungen und Untersuchungen zur Biologie der Silbermowe
auf den Vogelinsel Memmertsand. Journal fu r Ornithologie, 85, 1119.
herring gull study was an inspiring piece of experimental work Goodenough, J., McGuire, B. & Wallace, R. 1993. Perspectives on Animal Behavior.
that fullled an important role by making clear how questions New York: J. Wiley.
about the mechanisms underlying behaviour can be tackled with Gould, J. L. 1982. Ethology. The Mechanisms and Evolution of Behavior. New York:
Norton.
ingenious experiments and that fostered theoretical progress as Hailman, J. P. 1967. The ontogeny of an instinct. Behaviour, Supplement, 15, 1159.
well as subsequent studies. At the time the study was carried out, Hailman, J. P. 1969. How an instinct is learned. Scientic American, 221, 98108.
behavioural biology still had to nd its place as a scientic disci- Heinroth, O. & Heinroth, M. 1928. Die Vo gel Mitteleuropas. Berlin: Lichterfelde.
Kruuk, H. 2003. Nikos Nature. The Life of Niko Tinbergen and His Science of Animal
pline and the appeal of the herring gull study has certainly Behaviour. Oxford: Oxford University Press.
contributed to achieving that aim. The many publications Lorenz, K. 1935. Der Kumpan in der Umwelt des Vogels. Journal fu r Ornithologie, 83,
Tinbergen devoted to the study helped spread the message to 137213. 289413.
McFarland, D. 1999. Animal Behaviour. Psychobiology, Ethology and Evolution, 3rd
a wider audience, as is reected in the many textbooks that still
edn. Harlow: Longman.
refer to the study. Tinbergen deserves the credit for this. Also, as Manning, A. & Dawkins, M. S. 1998. An Introduction to Animal Behaviour, 5th edn.
stated above, the part of the herring gull study presented here was Cambridge: Cambridge University Press.
Margolis, R. A., Mariscal, S., Gordon, J., Dollinger, J. & Gould, J. L. 1987. The
selected because of its anomalies, not because it is representative
ontogeny of the pecking response in laughing gull chicks. Animal Behaviour, 35,
of the study as a whole, let alone of everything Tinbergen did. 191202.
Nevertheless, it seems that even a great scientist may not have Montgomerie, B. & Birkhead, T. 2005. A beginners guide to scientic misconduct.
been immune to the scientic biases and prejudices that can affect ISBE Newsletter, 17, 1621.
Pinel, J. P. J. 2000. Biopsychology, 4th edn. Boston: Allyn & Bacon.
any of us. Roell, D. R. 1996. De Wereld van Instinct. Niko Tinbergen het Ontstaan van de Ethologie
in Nederland (19021950). Rotterdam: Erasmus Publishing.
Roell, D. R. 2000. The World of Instinct. Niko Tinbergen and the Birth of Ethology in the
Acknowledgments Netherlands (19021950) (Translation of Roell, D.R. 1996). Assen: Van Gorcum.
Scott, G. 2005. Essential Animal Behavior. Malden: Blackwell.
It took time for this paper to materialize. While it was not too ten Cate, C., Bruins, W. S., den Ouden, J., Egberts, T., Neevel, H., Spierings, M., van
den Brug, K. & Brokerhof, A. W. 2009. Tinbergen revisited: a replication and
difcult to get an overview of the various publications, I wondered extension of experiments on the beak colour preferences of herring gull chicks.
whether it was a story that should be published. The responses I got Animal Behaviour, 77, 795802.
upon presenting some of the material at a few small meetings and ter Pelkwijk, J. J. & Tinbergen, N. 1937. Eine Reizbiologische Analyse eniger
Verhaltungsweisen von Gasterosteus aculeatus. Zeitschrift fu r Tierpsychologie, 1,
in my teaching encouraged me to give it a try and send a draft to 193200.
various colleagues. Some expressed worries: the story might be Tinbergen, N. 1948a. Dierkundeles in het meeuwenduin. De Levende Natuur, 51,
detrimental to Tinbergens heritage or to our eld. Others, however, 4956.
Tinbergen, N. 1948b. Physiologische Instinktforschung. Experientia, 4, 121133.
convinced me that both the Tinbergen legacy and our eld do not
Tinbergen, N. 1949. De functie van de rode vlek op de snavel van de zilvermeeuw
depend on this story and found it evident that it should be (Larus a. argentatus Pontopp.). Bijdragen tot de Dierkunde, 28, 453465.
published. I took all comments to heart, and can only hope that, in Tinbergen, N. 1951. The Study of Instinct. Oxford: Oxford University Press.
Tinbergen, N. 1953. The Herring Gulls World. A Study of the Social Behaviour of Birds.
the end, I managed to present the story in a balanced way, doing
London: Collins.
justice to Tinbergen. If I did not succeed, I take the full blame for it. Tinbergen, N. 1955. Psychology and ethology as supplementary parts of a science of
For their comments and views on the issue I am most grateful to Pat behavior. In: Group Processes. Transactions of the First Conference (Ed. by
Bateson, Innes Cuthill, Rob Lachlan, Kevin Laland, Katharina Riebel, B. Schaffner), pp. 75176. Madison: Madison Printing Company.
Tinbergen, N. 1963. On aims and methods of ethology. Zeitschrift fu r Tierpsychologie,
Candy Rowe, Hans Slabbekoorn, Fritz Trillmich (who also provided 20, 410433.
access to the studies by Eypasch and Zippelius), Dirk van Delft, Jos
Tinbergen, N. & Kuenen, D. J. 1939. Uber die auslo senden und die Richtung-
van den Broek, Ed van der Meijden, Peter Slater, four anonymous gebenden Reizsituationen der Sperrbewegung von jungen Drosseln (Turdus
m. merula L. und T. e. ericetorum Turton). Zeitschrift fu r Tierpsychologie, 3,
referees and Louise Barrett. All the material left from Tinbergens 3760.
period at Leiden has been donated as a permanent loan to the Tinbergen, N. & Perdeck, A. C. 1950. On the stimulus situation releasing the
Boerhaave science museum at Leiden. begging response in the newly hatched herring gull chick (Larus argentatus
Pont.). Behaviour, 3, 139.
Wilson, E. O. 1975. Sociobiology: the New Synthesis. Cambridge, Massachusetts:
Supplementary Material Harvard University Press.
Zippelius, H.-M. 1992. Die Vermessene Theorie. Wissenschaftstheorie 35, Wissenschaft
und Philosophie. Braunschweig/Wiesbaden: Vieweg.
Supplementary material associated with this paper can be Zupanc, G. K. H. 2004. Behavioral Neurobiology. An Integrative Approach. New York:
found, in the online version, at doi:10.1016/j.anbehav.2008.12.021. Oxford University Press.
794 C. ten Cate / Animal Behaviour 77 (2009) 785794

APPENDIX data sheet, the 652 pecks to the standard model are not. This
suggests that the total number of responses as mentioned in both
Note 1: Fig. 1b gives 70 as the value for the model with the white the data sheet (1095) and in the gure (938 pecks) are an adjusted
patch, instead of 71 as on the 1947 data sheet. gure, perhaps using the average number of pecks to the standard
Note 2: the number of pecks at the forehead is mentioned as 73 model, rather than the total.
in this gure, while the number in the 1947 data is 74. Note 5: surprisingly, the introduction of the paper mentions that
Note 3: the value for the model with the white patch, which was the rst experiments were done in 1946. This does not match with
reduced to 70 in Tinbergen (1948a, Fig. 1b), is now again at the the statements in the earlier papers (Tinbergen 1948a, 1949), nor is
original value of 71. there any hint in the Behaviour paper, or elsewhere, that herring
Note 4: there is a puzzling feature about these data. According to gull data were or might have been collected in 1946.
the gure, both models together got 809 responses. A hand-written Note 6: the corrected values that are given for the blue and the
note about this gure reveals that the responses consist of 652 white patch are 71 and 59, whereas a proper correction of the
pecks to the standard model and 157 pecks to the model with the original data should have led to the maximum values of 70 and 58.
patch on the forehead. However, as mentioned above, Tinbergen Note 7: the number of pecks at the red patch is now given as 75,
noted that all models were tested against the standard model, but rather than the original 74.
while the 157 pecks to the model with the patch on the forehead Note 8: the number of pecks at the forehead is now again
have been subtracted from the 1095 pecks mentioned in the 1947 provided as 74.

You might also like