Flexible life cycle duration is widespread in insects 12, also among aquatic species 3. Variation may concern an important portion of normal life cycle length. It arises mainly from differences in the duration of the larval stage in which most of the life cycle is spent 4. The main causes for such differences are different temperatures, day length or food in the various habitats of the population (see, for example, the review by 5).

Differences in life cycle length between individuals of a population at the same site have also repeatedly been reported for aquatic insects 678910, but the causes are less easily explained. In the case of extended hatching from eggs, early hatching individuals are often assumed to have an advantage in development resulting from the combined effects of temperature and longer growth period 111213. Such explanations focus mainly on time as factor. However, what causes a population to split into several cohorts growing and developing at different speed? There are no satisfactory explanations so far. Often, cohort splitting occurs when larval density at the respective site is high (see below). Intraspecific competition may be important in this context as shown, for example, in tadpoles 1415, water striders 16 and also stoneflies 17. In the mentioned cases, some individuals grew better than the rest, at the expense of the remaining members of the cohort. Possible causes might be competition for space and access to resources (interference), or some direct competition for food (exploitative competition; after 18).

To investigate possible causes of these frequently observed phenomena we used the stonefly, Nemurella pictetii Klapálek, 1900, as a model species to check for intraspecific competition as a possible cause of cohort splitting in experiments. Our experiments extended previous studies by testing competition under standardised laboratory conditions, for long periods of time. N. pictetii is particularly suitable for such study, for several reasons. In appropriate habitats like springs, seeps, and slow flowing spring runs N. pictetii may attain very high larval densities (see Table 1). The species is remarkable for a very flexible semi- to multivoltine life cycle (see references in 19). It also exhibits cohort splitting, even between siblings from a single egg mass 20. Some larvae hatching from eggs in spring attain the last instar within three months, emerging as adults and ovipositing in late summer of the same year. Other siblings from the same egg mass require a full year to complete development and emerge the next spring, together with offspring of their fast developing siblings 2021. Multimodal adult emergence described in other studies (for example, 222324) probably also reflects similar cohort splitting. At northern latitudes, N. pictetii may exhibit semivoltine life cycles 25.

The present study attempts to cast light on these problems and is based on the following hypotheses: (1) intraspecific competition does influence growth and development of N. pictetii; (2) the type of competition involved is interference competition. To test these hypotheses, we first compared the immediate effect of different larval densities on speed of larval growth and development (1). In a second experiment we supplied excess food and indirectly estimated the amount of ingested food by determining the amount of faeces produced per unit time, under different larval densities (2). In doing so we assumed that speed of development is positively related to the amount of food ingested per unit time.

The results of the speed of growth and maturation experiment confirm our first hypothesis that intraspecific competition influences growth and development of N. pictetii. Both parameters were distinctly affected by larval density. High densities induced significantly slower growth and retarded development than observed at low densities. There were distinct (P < 0.001) differences in HCW as well as in WPD. However, intraspecific competition was not strong enough to distinctly influence survival rate (P = 0.350), which was over 90% at both factor levels.

During preliminary tests we had noticed a striking aggressiveness of larvae during encounters, at least at high densities. Larvae stood head to head, threatening with open mandibles. Actually, antennae but also cerci were often bitten off. Larvae behaved in ways similar to Baetisr hodani (Ephemeroptera) at high larval densities 26. This author described larvae frequently pushed against others laterally, and conducted abrupt beats with the cerci against the competitor. Each larva may have a critical distance at which no others are tolerated, as described for other aquatic insect larvae with territorial behaviour (Plecoptera: 27, Odonata: 28). For example, large nymphs of Coenagrion resolutum (Odonata) excluded smaller nymphs from areas with high food resources 28. The favourable food supply might accelerate development of such larvae over others, and eventually lead to cohort splitting. However, analysis of HCWR suggests no similar mechanisms act in the present case. HCWR was only little wider at high than at low larval densities (P = 0.414). At the end of the experiment, there were no individuals in a unit that were strikingly larger than the rest; drastic suppression of some of the larvae by a few others appears improbable. The fact that HCWR of larvae was similar at both factor levels while HCW and WPD differed distinctly suggests two other causes.

First, N. pictetii larvae are very active and move around frequently 192930. At high larval density, disturbance during frequent encounters of larvae may retard development which would be interference competition. Stress during frequent encounters is supposed to reduce food ingestion and further increase locomotion. Hence, metabolic cost would increase at the expense of resources for growth. Our second consideration concerns exploitative competition. Shortage of high quality food (for example, densely microbially colonised leaf patches) may have caused retarded development at high larval density, affecting all individuals in a similar way.

The food ingestion experiment elucidated the kind of intraspecific competition more precisely. The distinctly negative correlation (P < 0.001) between larval density and amount of faeces produced by a larva does not suggest exploitative competition, for several reasons. On the one hand, biofilm is a high quality resource 3132 that N. pictetii prefers 19. On the other hand, enough biofilm remained during our short term experiments to exclude food shortage. Larvae used in the experiment were of comparable size and similarly far developed. Also, the biofilm used offered uniform food quality across the entire feeding plot. Therefore, interference (competition for space) seems to have been decisive, which supports our second hypothesis. Frequent encounters seem to have hindered ingestion of food. A recent study 30 found comparable evidence. Among other, the authors studied mass loss of conditioned alder leaves in relation to larval density of N. pictetii and also found a negative correlation. Like ourselves, they attributed their results to the high activity of the species and also regarded interference as cause of the correlation. In contrast, studies of intraspecific competition among slowly moving grazers like Trichoptera 3334 or Gastropoda 35 regarded exploitative competition as more important factor than interference.

Although intraspecific competition importantly influences the life cycle length of N. pictetii it can hardly by itself cause the cohort splitting observed both in the field and in the laboratory. Egg development of N. pictetii is direct and synchronous, egg rests that are important in other stoneflies 363738 do not occur in the present species (20, own investigations). Instead, several factors may act in combination. Possibly, intraspecific competition resulting from high larval densities drives part of the population into suboptimal microhabitats with reduced availability of high quality food, i.e., benthic algae 19 where detritus or the like must instead be consumed. MACAN'S 39 reasoning to explain why the development of Pyrrhosoma nymphula (Odonata) takes 2 years in some microhabitats but 3 years in others is similar (compare 4041).

In conclusion, intraspecific competition may importantly influence length of larval life and consequently the entire life cycle of the present species. Given the high larval densities of over 29000 specimens per square meter (42; see Table 1) observed in some habitats, interference competition must decisively influence the life cycle.

In our experiments, we observed interference competition, but no animal was able to profit at the expense of others. Therefore, cohort splitting observed in the field 2021 cannot be explained by direct competition, provided the animals inhabit restricted areas with similar food quality. However, in more heterogeneous habitats displacement of some specimens to less favourable sites with lower food quality (e.g., detritus) can be conceived, especially in view of the aggressive behaviour observed in the present species. Size differences caused by different food qualities 19 may thus be enhanced through intraspecific competition by increasing territorial behaviour, as shown in water striders 1643. Further study is needed for a precise analysis of the situation.

AF: amount of faeces; HCW: head capsule width across the eyes; HCWR: range of head capsule width of larvae in single experimental units; SR: survival rate; WPD: wing pad development.

The study was jointly planned by RL and PZ. RL performed the experiments and the statistical analyses. RL and PZ together wrote the present manuscript.