This article was downloaded by: [University of Alberta]

On: 31 December 2014, At: 03:27

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Human and Ecological Risk Assessment:

An International Journal
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/bher20

Measuring Complex Behavior Patterns in

FishEffects of Endocrine Disruptors on
the Guppy Reproductive Behavior
a
Erik Baatrup
a
Institute of Biological Sciences, Zoophysiology , University of
Aarhus , Aarhus, Denmark
Published online: 04 Feb 2009.

To cite this article: Erik Baatrup (2009) Measuring Complex Behavior Patterns in FishEffects of
Endocrine Disruptors on the Guppy Reproductive Behavior, Human and Ecological Risk Assessment: An
International Journal, 15:1, 53-62, DOI: 10.1080/10807030802615097

To link to this article: http://dx.doi.org/10.1080/10807030802615097

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
Content) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
 Human and Ecological Risk Assessment, 15: 5362, 2009
Copyright
C Taylor & Francis Group, LLC
ISSN: 1080-7039 print / 1549-7680 online
DOI: 10.1080/10807030802615097

Measuring Complex Behavior Patterns

in FishEffects of Endocrine Disruptors
on the Guppy Reproductive Behavior

Erik Baatrup
Institute of Biological Sciences, Zoophysiology, University of Aarhus,
Downloaded by [University of Alberta] at 03:27 31 December 2014

Aarhus, Denmark

ABSTRACT
An obstacle in behavioral research is in obtaining unbiased measurements and
presenting them in numerical terms, independent of human interpretation and
observational endurance. Simple locomotory behavior can be relatively easily quan-
tified with existing computerized vision systems, whereas it is much more difficult to
obtain unbiased measurements of complex behavior patterns. We have developed
a computerized vision system (DISPLAY), which identifies and quantifies complex
behavior patterns and interactions in fish. In short, the fish scenery is viewed from
above by a camera. At a frame-rate of 12 s1 , the fish silhouettes are stored in a
frame file. Subsequently, the position and orientation of each fish relative to the
other(s), the distance between them, and their body curvature are determined in
each frame. Further, frame-to-frame comparisons enable calculation of speed and
direction of fish movements. With these measurements available, composite be-
havior patterns can now be entered graphically by the user as distinct classifiers.
The system has been used for studying the effects of endocrine disruptors on the
reproductive behavior of the male guppy (Poecilia reticulata). Both estrogenic and an-
tiandrogenic substances significantly suppress the number and duration of posturing
behavior and sigmoid displays, two distinct elements in the courtship behavior of the
male guppy.
Key Words: quantification of fish behavior, computerized vision system, automatic
measurements, guppy courtship behavior.

INTRODUCTION
Potentially, behavioral changes are excellent biomarkers of internal and external
stress in animals, given that behavior is the physical manifestation of the animals
internal neuronal, metabolic, and endocrine processes, and at the same time the

Address correspondence to Erik Baatrup, Institute of Biological Sciences, Zoophysiology,

University of Aarhus, Building 1131, DK-8000 Aarhus, Denmark. E-mail: erik.baatrup@
biology.au.dk

53
 E. Baatrup

integrated physiological response to its environment (Clotfelter et al. 2004). Behavior

is regulated by the nervous and endocrine systems and even subtle disturbances in
either of these two systems may translate into serious behavioral aberration.
An absolute prerequisite for the applicability of behavioral studies in ecotoxicol-
ogy is that it can be quantified. Traditionally, behavioral studies involve a human
observer, who interprets the scenario and scores the number of events in question.
The advantage of using this approach is the superior ability of the human brain to
survey and decode even complex behavior patterns, extracting precisely the type
of behavior of interest. In contrast, an obvious disadvantage of human observation
in behavior studies is the difficulty of presenting behavior in unbiased and precise
numerical terms. Firstly, the variation in human interpretation and identification
of a specific behavior compromises result consistency and reproducibility. Secondly,
the precision and endurance of the human observer is severely limited. Anyone
Downloaded by [University of Alberta] at 03:27 31 December 2014

who has manually scored behavioral events will know that 15 minutes is a long time.
Thirdly, and most important, the human brain has a poor capacity for the storage
of numbers and mathematical data analysis. For instance, if watching a small group
of animals in a test arena, it is impossible to deduce and hold precise quantitative
information on their locomotor activity or their pattern of mutual interactions.
Computerized vision systems, on the other hand, are unsurpassed for such data
acquisition. They can record the movements of multiple animals over prolonged
periods and calculate simple behaviors such as walked distance, maximum and av-
erage velocities, distribution of active and resting periods, turning behavior, and
visits to specific locations (Areas Of Interest). A number of University research sys-
tems (e .g., our own MOTIO system, Institute of Biological Sciences, University of
Aarhus, Denmark) and commercial systems (e .g., Ethovision, Noldus, Wargenin-
gen, The Netherlands) have been developed. Most of these vision systems, however,
are confined to quantifying behaviors that relate to the position or movement of
the animal(s), since each animal is represented only by the x,y,(z)-coordinate of its
centroid (geometric center of gravity) in the data handling (Baatrup and Bayley
1993; Jensen et al. 1997). This simple data representation of the animal(s) excludes
the determination of complex behavior patterns that involve the orientation of the
animal in space, its form, and its interactions with other animals. To approach the
need for such a vision system we developed DISPLAY, which quantifies complex be-
havior patterns in fish. The applicability of the DISPLAY vision system in behavioral
and ecotoxicological research has been documented in a number of studies (Bayley
et al. 1999; Toft and Baatrup 2001; Baatrup and Junge 2001; Bayley et al. 2002; Toft
et al. 2003; Toft and Baatrup 2003; Toft et al. 2004). This communication presents in
detail the functionality of the DISPLAY system.

QUANTIFYING GUPPY COURTSHIP BEHAVIOR

The motivation for creating the DISPLAY system was a series of studies with the
aim of quantifying the effects of endocrine disruptors on guppy (Poecilia reticulata)
sexual characteristics. Among the various hormone controlled features of this fish, its
sexual behavior is an obvious endpoint of endocrine disturbance (Baatrup and Junge
2001; Toft and Baatrup 2003). Sexually mature guppies perform courtship behavior

54 Hum. Ecol. Risk Assess. Vol. 15, No. 1, 2009

 Measuring Complex Behavior Patterns in Fish
Downloaded by [University of Alberta] at 03:27 31 December 2014

Figure 1. Posturing behavior and the sigmoid display are two key elements in the
guppys courtship behavior. In the posturing behavior (A) the small male
places himself in front of the larger female facing her. From this position
he performs the sigmoid display where his body assumes a C or S shape
while he swims sideways along the length of the female (B and C) to get
into position for copulation, or he moves away enticing her to follow.
In some populations of guppies, this behavior is performed on average
once per minute.

(Houde 1997) almost continuously during the light hours. Briefly, the male places
himself in front of the female and stays within her field of view (posturing behavior).
From this position he performs the sexual display toward the female known as sigmoid
display (Figure 1), where his body assumes the shape of an S or C, and vibrates
while he swims sideways displaying his sexually attractive orange-yellow coloration.
He either moves along the length of the female to come into position for copulation,
or he moves away from the female, remaining in her field of view to entice her to
follow. These behavioral maneuvers are so stereotyped and performed so frequently,
even in a laboratory setting, that the male guppys courtship behavior lends itself
to automatic quantification. In the following description of the DISPLAY system,
the guppy posturing behavior and sigmoid display will serve as examples of quantifying
complex behavior patterns in fish.

Experimental Setup and Hardware

Each male guppy was paired with a 4-month-old female in a sand-blown 20
15 cm aquarium (to avoid fish mirror images) containing 1.8 l of 25 C water (water
depth, 10 cm), placed on a sheet of glass 50 cm above diffusely lit white paper. The
entire setup was enclosed in a metal frame covered with a blackout curtain in order
to exclude visual disturbances. When viewed from above, this arrangement resulted
in clear silhouettes of the two fish, where the male was easily distinguishable from
the much larger female (Figure 1). A CV-M10 progressive scan (non-interlaced)
CCD camera (JAI, Inc., Copenhagen, Denmark) mounted 50 cm above the aquar-
ium displayed an image of the scenario on a monitor. Simultaneously, the analogue

Hum. Ecol. Risk Assess. Vol. 15, No. 1, 2009 55

 E. Baatrup

video signal from the camera was digitized by a DT3155 frame grabber (Data Trans-
lation, Inc., Marlboro, MA, USA) into a 768 576-pixel digital image, giving a 0.25-
mm spatial resolution of the visual field. The frame grabber was interfaced with a
300-MHZ Pentium II personal computer.

Recording the Scenario

Prior to recording, the interior of the aquarium was framed by a ROI (Region Of
Interest) software window, and appropriate size and gray-level ranges corresponding
to the fish silhouettes were likewise set in the software. These criteria were used for
the conversion of each 8-bit gray-scale image into a binary (1-bit) image. Thus, all
pixel assemblages fulfilling both size and gray-level criteria (fish silhouettes) inside
the ROI were assigned the value 1, while the remaining pixels in the image were
Downloaded by [University of Alberta] at 03:27 31 December 2014

given the value 0. This new binary image was stored in a frame file on disk for
subsequent analysis. During each 10-min recording, an image was captured and
processed approximately every 1/12 s, so each 10-min frame file contained about
7,200 binary frames (occupying less than 6 MB of disk space).

Basic Object Analyses

The frame files, containing the time-series of fish contours, were subsequently
analyzed by the DISPLAY program. An exhaustive description of this software is
beyond the scope of this article, but the most important steps in the characterization
and quantification of the complex courtship behavior are outlined in what follows.
First, the position and orientation of the two fish within the digital image (global
coordinate system) must be established in each frame.
The two oblong pixel assemblages, representing the fish silhouettes, are con-
verted into two small coordinate systems by determining their principal axes. These
axes can be found by computing the eigenvectors (Noble and Daniel 1988) for the
position vectors of all the pixels constituting the fish silhouette. The mean pixel
position vector is the origin of the coordinate system spanned by the eigenvectors.
The eigenvectors constitute a unitary transformation that enables transformation
of other coordinate systems into a particular coordinate system, such as the other
fishs coordinate system or the global coordinate system. Thus, all angle and distance
measures can easily be computed once the principal axes of the pixel masses are
determined. In order to associate a directed coordinate system to the pixel mass, the
numbers of object-pixels inside two circles of equal radius are counted two standard
deviations from the origin along the y -axis in both directions (Figure 2A). The end
with the largest pixel count gives the direction of the fishs head. Furthermore, the
bending of the fishs body can be computed by dividing the silhouette pixel mass
into two parts, separated by the x-axis (Figure 2B). Each of the two new pixel masses
is subsequently treated as a new object, and the angle between the two new y axes
provides a measure of the fish curvature (Figure 2B).
For each frame it is now possible to determine the position and orientation
of the fish relative to each other, the distance between them, and the curvature
of the male. Further, frame-to-frame comparisons enable calculation of speed and
direction (relative to the bodys longitudinal axis) of fish movements.

56 Hum. Ecol. Risk Assess. Vol. 15, No. 1, 2009

 Measuring Complex Behavior Patterns in Fish

Figure 2. A: The pixel assemblage, representing the fish silhouette in each cap-
tured image, can be resolved into its principal axes by computing the
eigenvectors. These axes can be transformed into an oriented coordi-
nate system by finding the head and tail ends of the fish. This is done by
Downloaded by [University of Alberta] at 03:27 31 December 2014

counting the number of fish pixels inside two circles of equal radius,
placed two standard deviations from the origin of the y-axis in both direc-
tions. The direction with the largest pixel count signifies the direction
of the fishs head. B: A measure of the curvature of the fishs body can be
found by dividing the pixel assemblage into two parts, separated by the
smallest of the principal axes (x-axis). Subsequently, the two pixel masses
are treated as separate objects of which new sets of principal axes can
be computed. The angle between the two y-axes represents a measure
of the fish curvature.

Decoding Complex Behavior Patterns

The composite courtship behavior of the guppy, including posturing behavior and
sigmoid displays, can be broken down into its basic elements, including the mutual
position/orientation of the two fish and their respective movement patterns. The
following parameters were extracted from each of the approximately 7,200 frames
in the frame file: (a) position of female relative to the male measured as the angle
(0180 ) between the males y -axis and a line between the origins of the two fishs
coordinate systems (centroids; in Figure 3A); (b) position of male relative to
the female measured as the angle (0180 ) between the females y -axis and a line
between the centroids of the two fish ( in Figure 3A); (c) distance between the
centroids of the two fish (a in Figure 3A); (d) male swimming speed defined as the
frame-to-frame displacement of his centroid divided by the time between successive
frames (v in Figure 3B); (e) male angular displacement measured as the angle
(0180 ) between the male y -axis and the position vector of his centroid in the
next frame ( in Figure 3B), where high values indicate sideways and backward
swimming; (f ) male lateral velocity calculated as the magnitude of the male velocity
component perpendicular to the y -axis in the preceding frame (u in Figure 3B),
signifying the intensity of sideways swimming; and (g) male curvature measured as
the angle between the y -axes of head region and the tail region, respectively ( in
Figure 2B).
The frame-to-frame measurements of the seven parameters were subsequently
used to identify periods with posturing behavior and sigmoid displays within the entire
10-min frame file. This was done by assigning a range of pass values (search crite-
ria) to each parameter, hereafter termed classifiers. Basically, a specific behavior

Hum. Ecol. Risk Assess. Vol. 15, No. 1, 2009 57

 E. Baatrup
Downloaded by [University of Alberta] at 03:27 31 December 2014

Figure 3. The composite courtship behavior of the male guppy is automatically

quantified using seven descriptive components. Abbreviations: a, dis-
tance between female and male; u, male lateral velocity measured as the
velocity component perpendicular to the longitudinal axis of the male;
v, male swimming velocity; , angular position of the male relative to
the females longitudinal axis; , angular position of the female relative
to the males longitudinal axis; , male angular displacement relative to
his longitudinal axis. A: The two oblong pixel assemblages representing
the male and female silhouettes from the digital image were converted
into directed coordinate systems, which enabled the calculation of posi-
tion and orientation of each fish relative to the other and the distance
between them. B: Frame-to-frame comparisons enabled calculations of
speed and direction (relative to the bodys longitudinal axis) of fish
movements.

is recognized when all classifiers are fulfilled simultaneously. Thus, before analyzing
the frame files, the appropriate combination of search criteria describing the spe-
cific behavior is set once for all. This is done by repeatedly refining and verifying
the search criteria until the softwares interpretation of the behavior in all situations
agrees with this particular behavior checked by human observer. This manipulation
is easily performed by combining a graphical user interface for entering search crite-
ria (Figure 4) with a real-time replay facility in the DISPLAY program. It is therefore
the human operator with his/her superior integrative capability that defines the
behavior for which to search. The computer will then untiringly and consistently
quantify the particular behavior resulting from this interpretation.
The measurement of the males posturing behavior (i.e ., the time he spent in front
of the female introducing the next sigmoid display) involves three search criteria.
First, the male must be in front of the female (pass values of in Figure 3A set to
0 < < 90 ). Second, the male must at least partly face the female ( criterion
in Figure 3A set to 0 < < 60 ). Note that all angles are presented without signs
because there is no distinction between the right side and left side of the fish. Finally,

58 Hum. Ecol. Risk Assess. Vol. 15, No. 1, 2009

 Measuring Complex Behavior Patterns in Fish
Downloaded by [University of Alberta] at 03:27 31 December 2014

Figure 4. The analyses screen in the DISPLAY program. In the display area,
the frame file of binary images can be replayed and analyzed. In
the captured image shown here, the male is performing a sigmoid
display toward the much larger female. In the lower screen area is
shown a pop-up window with the classifiers defining the sigmoid dis-
play. Note that the small circles are all within the rectangles lim-
iting the pairwise classifier values (defined in Figures 2 and 3),
implying that all criteria for the sigmoid display are met in this
frame. Distance: distance between the centroids of the two fish (a in
Figure 3); FAngle: position of the female relative to the male measured
as the angle (0180 ) between the males y-axis and a line between the
two fishs centroids ( in Figure 3); MAngle: position of the male relative
to the female measured as the angle (0180 ) between the females y-axis
and a line between the two fishs centroids ( in Figure 3); MShape: male
cuvature ( in Figure 2).

the two fish must be within a distance of a few centimeters (a in Figure 3A set to <
60 mm).
The guppy sigmoid display is a much more complex behavior. Firstly, the combi-
nation of involved parameters and their ranges of pass values change during the
course of the display. Accordingly, a positive identification of the entire display by
the software requires temporal adjustments of the search criteria. This is achieved
by associating a timer to each search criterion, engaging and disengaging its function

Hum. Ecol. Risk Assess. Vol. 15, No. 1, 2009 59

 E. Baatrup

(Figure 4). The onset of the display is characterized by the male being within the
anterior part of the females field of view for at least 0.2 s (0 < < 90 ; time out
0.2 s) at a distance of at least 22 mm (a > 22 mm) exposing the side of his body
( > 68 ), which is locked in a distinct curvature for at least 0.5 s ( > 20 ; time out
0.5 s). After this initial phase the rules are changed allowing the male to move within
the females entire field of view (0 < < 140 ), but with his body still locked in a
curvature ( > 5 ) with the convex side continuously facing her ( > 68 ). Display
termination is registered when one or more of these criteria are no longer fulfilled.
The frame files of all treated and control fish were analyzed by the DISPLAY
software. The complete analysis of a 10-min frame file takes about 25 s with a
300 MHz Pentium II computer. The time devoted to posturing behavior, number and
duration of the sigmoid displays and the average values of all the measured parameters
are saved in a data file for subsequent statistical analysis.
Downloaded by [University of Alberta] at 03:27 31 December 2014

More Than Two Fish

In the earlier example with one male and one female, the frame-by-frame identifi-
cation of the two fish is easily determined by their difference in size. In experiments
with more than two fish (many-fish-mode in the DISPLAY software), more sophisti-
cated identity algorithms must be engaged. Obviously, it is absolutely essential that
the identity of each fish is preserved during the entire observation period. There-
fore, the paramount challenge is to keep track of each individual, especially after
optical merging of two or more fish silhouettes. In most cases this is automatically
solved by comparing speed and direction of fish movement before and after the
amalgamation. However, occasionally the software cannot with certainty assign the
right identities after a visual amalgamation. In such cases, it is necessary to manually
adjust the fish identities, which is achieved onscreen thanks to the built-in conflict
finder in DISPLAY.

ENDOCRINE DISRUPTION OF THE GUPPY COURTSHIP BEHAVIOR;

AN EXAMPLE
It is well documented that several chemicals from agricultural, industrial, and
household sources possess endocrine-disrupting properties, which potentially pose
a threat to human and wildlife reproduction (Tyler et al. 1998; Vos et al. 2000).
Most work has focused on the adverse effects of estrogenic substances (Mills and
Chichester 2005), but certain environmental contaminants possess antiandrogenic
properties, that is, disrupt the normal function of the male sex hormones (Chapin
et al., 1996). Intensively studied environmental antiandrogens include the fungicide
Vinclozolin (Gray et al. 1994) and the DDT metabolite p ,p  -DDE (Kelce et al. 1995).
Animal courtship behavior is an endpoint of endocrine disruption. The applicability
of the vision system in such ecotoxicological research is exemplified by the effects of
antiandrogens on the guppy (Poecilia reticulata) sexual behavior (experimental pro-
tocol presented in Baatrup and Junge [2001]). Groups of male guppies were exposed
through the fodder for 30 days to three antiandrogens: (1) the fungicide Vinclozolin
(1.0 and 10.0 g/mg fodder); (2) the insecticide DDT metabolite p , p  -DDE (0.1
and 1.0 g/mg fodder); and (3) the pharmaceutical antiandrogen Flutamide (1.0

60 Hum. Ecol. Risk Assess. Vol. 15, No. 1, 2009

 Measuring Complex Behavior Patterns in Fish

Figure 5. The three antiandrogens Vinclozolin (1.0 g/mg fodder), Flutamide

(1.0 g/mg fodder), and p,p  -DDE (0.1 g/mg fodder) seriously com-
promised posturing behavior and sigmoid displays in male guppies. ANOVA
Downloaded by [University of Alberta] at 03:27 31 December 2014

followed by Dunnetts test for multiple comparisons, *p < .05; **p <
.0001.

and 10.0 g/mg fodder), which was used as the positive antiandrogenic control. A
group of male guppies receiving uncontaminated fodder served as controls. Follow-
ing exposure, each male was paired with a non-receptive (gravid) female in the test
aquarium. Non-receptive females were used to preclude copulations and hence to
ensure constant female responses toward all males. The scenario was then recorded
for 10 min after which the posturing behavior and the sigmoid displays were quantified
together with the each of the seven behavioral elements. In Figure 5 the effects of
the lowest application rates of the three antiandrogens is shown. It can be noted that
the sexual instinct of the male guppy was seriously compromised by these chemicals.
The number of posturing behaviors (and time devoted to this behavior, data not shown
here) were significantly inhibited after exposure to the lowest doses of Vinclozolin
and p ,p  -DDE, and all three chemicals significantly suppressed sigmoid displays.

FUTURE DEVELOPMENT
The behavior vision system presented here was developed with the aim of quanti-
fying complex behavior patterns in guppy and mosquitofish, and as such is confined
to establish the head and tail ends of torpedo-shaped fish. In order to extend the
applicability of the system, it is desirable to include other silhouette shapes such
as flatfish as well as other animal species. Also, it is planned to upgrade from the
present two-dimensional system to a three-dimensional system by adding a second
camera. This will not only allow quantification of behavior patterns where depth is
involved, but will also greatly facilitate the frame-by-frame identification of the fish.

ACKNOWLEDGMENT
This work was supported by the Danish Natural Science Research Council.

Hum. Ecol. Risk Assess. Vol. 15, No. 1, 2009 61

 E. Baatrup

REFERENCES
Baatrup E and Bayley M. 1993. Quantitative analysis of spider locomotion employing
computer-automated video tracking. Physiol Behav 54:8390
Baatrup E and Junge M. 2001. Antiandrogenic pesticides disrupt sexual characteristics in the
adult male guppy (Poecilia reticulata). Environ Health Persp 109:106370
Bayley M, Nielsen JR, and Baatrup E. 1999. Guppy sexual behaviour as an effect biomarker
of estrogen mimics. Ecotoxicol Environ Safety 43:6873
Bayley M, Junge M, and Baatrup E. 2002. Exposure of juvenile guppies to three antiandrogens
causes demasculinization and reduced sperm count in adult males. Aquat Toxicol 56:227
39
Chapin RE, Stevens JT, Hughes CL, et al. 1996. Endocrine modulation of reproduction.
Fundam Appl Toxicol 29:117
Clotfelter ED, Bell AM, and Levering KR. 2004. The role of animal behaviour in the study of
Downloaded by [University of Alberta] at 03:27 31 December 2014

endocrine-disrupting chemicals. Anim Behav 68:66576

Gray LE, Ostby J, and Kelce WR. 1994. Developmental effects of an environmental antian-
drogen: The fungicide vinclozolin alters sex differentiation of the male rat. Toxicol Appl
Pharmacol 129:4652
Houde AE. 1997. Sex, color, and mate choice in guppies. In: Krebs Jr and Clutton-Brock
T (eds), Monographs in Behaviour and Ecology, pp 2943. Princeton University Press,
Princeton, NJ, USA
Jensen CS, Garsdal L, and Baatrup E. 1997. Acetylcholinesterase inhibition and altered loco-
motor behavior in the carabid beetle Pterostichus cupreus. A linkage between biomarkers at
two levels of biological complexity. Environ Toxicol Chem 16:172732
Kelce WR, Stone CR, Laws SC, et al. 1995. Persistent DDT metabolite p,p  -DDE is a potent
androgen receptor antagonist. Nature 375:5815
Mills LJ and Chichester C. 2005. Review of evidence: Are endocrine-disrupting chemicals in
the aquatic environment impacting fish populations? Sci Tot Environ 343:134
Noble B and Daniel JW. 1988. Applied Linear Algebra. Prentice-Hall, London, UK
Toft G and Baatrup E. 2001. Sexual characteristics are altered by 4-tert octylphenol and 17-
estradiol in the adult male guppy (Poecilia reticulata). Ecotoxicol Environ Safety 48:7684
Toft G and Baatrup E 2003. Altered sexual characteristics in guppies (Poecilia reticulata) after
exposure to 17-estradiol and 4-tert-octylphenol during sexual development. Ecotoxicol
Environ Safety 56:22837
Toft G, Edwards T, Baatrup E, et al. 2003. Disturbed sexual characteristics in male mosquitofish
(Gambusia holbrooki) from a lake contaminated with endocrine disrupters. Environ Health
Persp 111:695702
Toft G, Baatrup E, and Guillette LJ Jr. 2004. Altered social behaviour and sexual characteristics
in mosquitofish (Gambusia holbrooki) living downstream of a paper mill. Aquat Toxicol
70:21322
Tyler CR, Joblin S, and Sumpter JP. 1998. Endocrine disruption in wildlife: A critical review
of the evidence. Crit Rev Toxicol 28:31961
Vos JG, Dybing E, Greim HA, et al. 2000. Health effects of endocrine-disrupting chemicals on
wildlife, with special reference to the European situation. Crit Rev Toxicol 30:1133

62 Hum. Ecol. Risk Assess. Vol. 15, No. 1, 2009