* Danio rerio* and the sunbleak *Leucaspius delineatus,* bothof whichwere obtained from laboratory stocks.

The strain of *Danio rerio* that was used, was provided by Dr. Oberemm, Institute of Freshwater Ecology and Inland Fisheries (Berlin, Germany). The breeding groups consisted of about 12 wild-type females and 15 males of *Danio rerio *each,thatwere derived labelled as a Singapore import, from a wholesale trader. The animals were kept under standard conditions according to Westerfield (1993) and they were 160-171 days of age at the start of each experiment.

The strain of* Leucaspius delineatus *that was used, was provided by Dr. Jähninchen, Institute of Freshwater Ecology and Inland Fisheries (Berlin, Germany) and was laboratory reared offspring of a population from lake Malchower See (Germany). The animals were acclimated to a recirculation system at a temperature of 20°C and kept under standard conditions (adapted from Westerfield, 1993). They were 170- 185 days of age at the start of each experiment.

*Danio rerio* one school consisted of three females and four males. For *Leucaspius delineatus* the sex ratio could not be ascertained *in vivo.* Mean total length and mean body mass of both species are shown in Table 3.

After 3 weeks of acclimatisation to the test conditions (see 4.2), the behaviour of all fish groups was recorded under standard test conditions. Thereafter, during exposure four groups of both species were exposed to the test substances MC-LR or PCB 28 (see 4.3). New groups of animals were used for every experiment.

**
Danio rerio
**

**
Leucaspius delineatus
**

**TL [cm]**

**BM [g]**

**TL [cm]**

**BM [g]**

**MC-LR**

3.45 ± 0.26

0.25 ± 0.02

4.55 ± 0.22

0.30 ± 0.02

**PCB 28**

3.47 ± 0.25

0.26 ± 0.03

4.57 ± 0.21

0.31 ± 0.03

^{®} flakes twice a day at a ratio of 3% body mass per day, three and seven hours after light-on.

Temperature °C 26 ± 0.5 (1) 20 ± 0.5 (2) Conductivity µS cm 720 ± 10 pH- value 7.5-7.7 Total hardness °dH 16.5 ± 0.5 Total organic carbon (TOC) mg l 4.1 ± 0.4 Oxygen mg l 7.7 ± 0.2 (1) 8.2 ± 0.3 (2) Ammonium mg l <0.5 Iron mg l <0.03 Nitrate mg l-1 5 ± 0.2 Nitrite mg l <0.03**Parameter**
**Unit**
**Measured value**
^{-1}
^{-1}
^{-1}
^{-1}
^{-1}
^{-1}
^{-1}

For the experiments aerated drinking water (“Berlinwasser Holding Friedrichshagen”) was used. The physico-chemical composition of the water is listed in Table 4. Basic water quality parameters of pH, oxygen and ammonium were measured both in storage tanks and aquaria once a day. The physico-chemical parameters of the used aquarium water were constant over the exposure period and within the normal physiological ranges for fish (Schreckenbach et al., 1987, 2001; Schäperclaus, 1990).

The test substance MC-LR was purchased from Calbiochem-Novabiochem Corp. (La Jolla, CA, USA). Purified MC-LR (10 mg) was dissolved in 1 ml of methanol and diluted in 100 ml of distilled water. To keep experimental conditions as consistent as possible, fish exposed to 0.003 % methanol in water only served as controls. A comparison of behaviour during the pre-exposure standard test conditions with behaviour of the controls during exposure revealed that the single exposition to 0.003% methanol in water had no effect on behavioural parameters.

During the exposure period both fish species were exposed to four different concentrations of MC-LR (nominal concentrations): 0.5, 5 and 15 µg l^{-1} for a period of 17 days each and 50 µg l^{-1} for a period of 6 days. The storage tank water (10 l) was renewed daily and MC-LR was added at nominal concentrations.

During the exposure period four groups of fish were exposed to two different (nominal) concentrations of PCB 28 (duplicates): 100 and 150 µg l^{-1} for a period of 8 days. The storage tank water was renewed daily and PCB 28 was added at nominal concentrations. Two groups served as controls, the first group without any exposure to solvents and the second group with 0.003 % ethanol. There were no significant differences between both control groups, what revealed that the single exposition to 0.003 % ethanol in water had no effect on behavioural parameters.

^{®} (Spieser et al., 2000). The experimental design is shown in Figure 4. Fish were observed by infrared video cameras, one in front of each tank which were able to handle normal as well as infrared light, enabling the continuous observation even during the night. The positions of the untagged fish were recorded at a two-dimensional area, data were digitised and paths of individual fish afterwards tracked by the object recognition software. Thus it was possible to reconstruct the real movements of every fish of the school.

The screen shot (Fig. 5) shows an example of the movement tracks of one fish school. Video was filmed at a frequency of 25 frames per second, and overall there were 69 measuring cycles of 2 min per day. Every picture of the experimental chamber was in real time compared point-by-point with a background reference picture. The *x*-*y* positions of recognized objects are written to disk for every measuring interval. The raw data were pre-processed and converted into tables which contained the behavioural parameter values: the __motility__ is the swimming velocity in video-pixels per second and the __turns__ are the number of changes of the direction per second. Motility and turns characterised the swimming activity.

*Danio rerio *and* Leucaspius delineatus* was registered 23 hours per day, and data were averaged per hour. Mean motility and mean number of turns were analysed in different ways:

over the whole measured time per day (23 h d^{-1}) and over the whole exposure period

divided into light and dark phases over the whole exposure period

On these bases all the results of the exposed groups were compared with those of the controls. Statistical analysis of all results was performed in SPSS 9.0. using the ANOVA procedure, as well as the Dunnett T3 post hoc test for comparison of groups with unequal variances. Generally significant differences were accepted at *p *< 0.05. In all figures and tables a significance level of

*p *< 0.05 is indicated by one asterisk,

*p *< 0.01 is indicated by two asterisks, and

*p *< 0.005 is indicated by three asterisks.

For each exposed group and for the control a regression between motility versus number of turns was fitted. A linear model (y = a*x + b) proved to be most appropriate for the totality of evaluated relations. F-statistics were performed for testing the significance of the determination coefficients r^{2}. Student’s *t*-tests were used to compare the values for slope and intercept to 0.0. Deviations between regression parameters of control and exposition in detail were evaluated by calculating and comparing their 95% confidence intervals.

*t-*Test to compare the group values to 0.5 and furthermore the exposed group to controls using the computer program SPSS 9.0.

Oscillations are normally characterized by up and down, left and right or back and forth movements of measurable parameters (*y*-values) during the course of time (*x*-values). So they appear in a diagram as wavelike or vibration or pendulation curves.

These curves are mathematically modelled by trigonometric functions like sinus or cosinus which are adapted to the exact shape of the curve by multiplicative and additive terms for example: *y* = *A* * sin (*B* * *t* + *C*) (*t* being time, *A,B,C* being variables).

*f*(*x*) = *M* + *A* * cos (6.283/*P* * (*x* - *K*)).

The variables are explained in Figure 6.

The acrophase is defined as phase angle corresponding to the maximal value of the rhythmic parameter studied, in the present study this was the time between local midnight and maximum peak of activity. The term inside the cosinus brackets was transformed to radians by the factor 2*π (equals 6.283) so the original time units could be used inside the equation. Calculated regression parameters were considered significant if its 95% confidence interval did not include zero.

The calculated overall solutions for exposure and control were compared by an F-test (Zar, 1996). Deviation between regression parameters of control and exposition in detail was evaluated by calculating and comparing their 95% confidence intervals.

For quantification of the harmonic frequency structure of activity rhythms of *Danio rerio* and *Leucaspius delineatus *a power spectral analysis of the motility which is a Fourier transformed autocorrelation function was used. The calculations were performedby the mean of the program “Zeit” which application is described in Scheibe et al. (1999, 2002). Periodic frequencies which explain a significant proportion of the total variation of the original data series and which are furthermore harmonic to the circadian period were ascertained. Periods are called harmonic in a chronobiological context if their lengths are integer dividers of 24 h. All periods of the power spectra were tested for significance by the integrated function of the program “Zeit” (see Scheibe et al., 1999, 2002).

Degrees of Functional Coupling (DFCs) were used for comparison of the rhythmic structures (Sinz and Scheibe, 1976; Scheibe et al., 1999). DFCs express the percentage of the circadian component and harmonic ultradian components in relation to all significant rhythmic components of a spectrum based on the respective relative parts of their assigned variance. Therefore, the DFC describes the percentage of cyclic activity components which is synchronised with the circadian rhythm.

SP(total) being the sum of all significant periodogram ordinates (i.e. the variance assigned to significant periods) and

SP(harm) giving the sum of those periodogram ordinates which are significant and harmonic to the circadian period.