Rearing Systems and Experimental Units
The flowing water system used in the experiment on speed of growth and maturation consisted of a reservoir (25 l) from which an aquarium filter pump (Eheim, Typ 1026; 11.8 l min-1, Deizisau, Germany) pumped water into plastic tubes (∅ = 15 mm) with boreholes at regular distances. Each resulting water jet entered an experimental unit (poly-propylene, base area = 4.5 × 4.5 cm, volume = 70 ml) through a screen window (mesh size = 80 μm) in the lid. Water left via similar screen windows in the sides of the unit, flowing back to the reservoir via a trough. For details and illustrations, see  and .
A different flowing water system circulating a larger volume of water was used in the food ingestion experiment (for details see ). Each experimental unit consisted of two conical plastic beakers (poly-propylene, basal ∅ = 5.5 cm, volume = 170 ml) of which one had a gauze bottom (mesh size = 1500 μm) and was placed into the other beaker.
We used water from the Breitenbach in both experiments; for notes on its chemical properties see . In the long lasting speed of growth and maturation experiment about half the water was replaced once a week. Random water samples revealed no change of water quality (conductivity, ammonium, nitrate and phosphate) during the experiments. Water temperature was 12°C, daylength was 16 h light:8 h dark in both experiments. Water temperature was recorded with a datalogger (1250 series, Grant Instruments [Cambridge] Ltd., Cambridgeshire, UK) at 10-min intervals.
In the speed of growth and maturation experiment we used chips (area ≈ 11 cm2) from soft alder (Alnus glutinosa) shade leaves that had fallen in autumn. Air dried leaves were soaked in water for half a day, the chips punched with a metal tube and conditioned in aerated Breitenbach water for 2 weeks in the dark at 12 to 14°C before they were used as food. One leaf chip was randomly assigned to each experimental unit. Food was replaced twice a week.
In the food ingestion experiment we used unglazed clay tiles (5 × 5 × 0.5 cm; for details see ) after they had been exposed in trays in the Breitenbach for establishment of a natural biofilm . This type of food resource has several advantages. It is the preferred food of N. pictetii , biofilm quantity and quality tend to be uniform across the entire area of a tile , and faeces fell directly through the gauze bottom into the lower collecting beaker when tiles were placed obliquely into the experimental units, with the biofilm on the lower side.
Experimental procedure and variables measured
The speed of growth and maturation experiment lasted for 44 days. Two factor levels (low and high larval density) were tested. Experiments started with 5 and 20 larvae, respectively, per experimental unit corresponding to 2500 and 9900, respectively, individuals/m2 stream bottom. We used 10 and 9 experimental units, respectively, for experiments at low and high density, respectively. The increase in head capsule width across the eyes (HCW) during the experiment was recorded to document possible differences in growth and development between the treatments. HCW was measured four times: before the experiment, and then every second week using a dissecting microscope (WILD M5, Wild, Heerbrugg, Switzerland) combined with a digitizing tablet (Numonics 2200, Numonics, Montgomeryville, Pennsylvania, U.S.A). At the end of the experiment, we additionally recorded the range of HCWs in each experimental unit (HCWR), the number of individuals with wing pads as a measure of degree of maturation (WPD, wing pad development), and the survival rate (SR, percentage of larvae surviving since the start of the experiment). SR was calculated separately for each experimental unit, we here present the mean across all units. Since individual larvae could not be marked and identified, HCW and WPD were aggregated at the level of experimental units by calculating means per unit.
Two factor levels (low and high larval density, 1 and 5 larvae per unit, respectively, corresponding to 400 and 2100 individuals/m2 stream bottom) were also tested in the food ingestion experiment. Lower numbers of larvae than in the previous experiment were used because the present larvae were much larger (mean HCW = 1.21 mm, SD = 0.09). We assessed the amount of ingested food indirectly by recording the amount of faeces produced (AF), on the assumption that larvae frequently disturbed by competitors would feed and defecate less. However, it was impossible to document the degree of disturbance precisely by counting the number of aggressive encounters between larvae during the experiment. The experiment lasted for 66 hours. Gut passage times in N. pictetii are only one to two hours , sufficiently many faecal pellets  were therefore produced during the experiment. N. pictetii grazes very efficiently on the biofilm, all detached pieces of biofilm are actually also eaten (personal observations). Therefore, only faeces dropped into the lower beaker of each experimental unit. A manual vacuum pump (MityvacII, Nalgene, Rochester, U.S.A.) was used to transfer the accumulated faeces to glass fibre microfilters (GF/C, ∅ = 25 mm, Whatman plc, Brentford, England). The filters were placed into small aluminium cups and dried at 105°C for 72 hrs (drying oven T6060, Heraeus, Hanau, Germany). Samples were allowed to cool in an exsiccator for three hours and were then weighed with an ultra microscale (4504 MP8, Sartorius, Göttingen, Germany). Before use, cups and filters had been in a furnace at 500°C for 2 hrs and were pre-weighed as described above. The amount of faeces produced during the experiment was the difference between the two readings.
Sexual dimorphism of N. pictetii is pronounced towards the end of the larval period [21, 53]. Therefore, sex was included in the experimental design as an additional factor by running replicates separately for each sex. This as well as the use of only the last three larval instars was also intended to exclude that some individuals were much larger than the others, and might become dominant and possibly territorial. Shortly before molts, stonefly larvae do not feed (see references in ). To exclude any possible error caused by this we analysed only replicates in which no molts had occurred during the experiment. Seven experimental units with males and 14 with females were analysed at low density, and 5 and 9, respectively, at high density. Except for sex in the second experiment, larvae were randomly assigned to experimental units.