Song rates of dark-eyed juncos do not increase when females are fertile

Frequency of singing by birds may vary with reproductive stage in ways that reflect variation in the functions of song in intersexual and intrasexual communication. In dark-eyed juncos (Junco hyemalis) high-amplitude song is produced only by males. To investigate the function of this song, we tested whether fertility of females affected singing by their mates or by neighboring males. Using focal observations, song censuses, and radiotracking data, we determined whether song production varied between and among periods when females were fertile and non-fertile. Our findings show that males do not increase song production when their mates are fertile, nor do they increase song production when neighboring females are fertile. These results suggest that male juncos do not signal their intent to defend territories (or mates) more when females are fertile and that they do not use song to advertise to specific potential participants in extra-pair fertilizations. Received: 13 February 1997 / Accepted after revison: 2 May 1997     

Common goldeneyes adjust maternal effort in relation to prior brood success and not current brood size

Parental investment theory predicts that parental effort should be related to the reproductive value of the current brood. This depends on both the number of young and the survival prospects of each of them. Thus parents may provide more care to larger broods either because of (1) the direct effect of brood size per se on reproductive value (the “brood size” hypothesis) or because (2) past mortality, reflected in current brood size, predicts future mortality of the brood and hence its reproductive value (the “brood success” hypothesis). Earlier studies have not attempted to distinguish between these alternatives. We tested the hypotheses in the precocial, nidifugous common goldeneye Bucephala clangula, a species with uniparental female care. Maternal effort was measured as the time spent by the female in rearing the brood. We found that brood size itself is not associated with maternal effort, but that females modify their maternal effort according to the mortality already experienced by the brood, supporting the prediction of the brood success hypothesis. We also found that brood mortality varied considerably between broods and that previous mortality predicts future mortality within broods, basic assumptions of the brood success hypothesis. Received: 30 January 1996 / Accepted after revision: 27 October 1996     

Antler asymmetry and immunity in reindeer

Fluctuating asymmetry (FA) measures an individual's ability to undergo identical development in bilaterally symmetrical characters and may indicate sensitivity to environmental stress. FA in ornamental characters is often positively related to parasite intensities, which are important environmental stressors. Parasites affect and are affected by several parts of the immune system, and the ability to resist parasites may be signalled via FA in ornaments. In this study we examined reindeer antlers, which show FA, demonstrated to be caused by parasite infections. We measured antler FA, immune parameters (i.e. densities of different classes of leukocytes, IgG levels and abomasal lymph node numbers) and intensity of abomasal nematodes in free-ranging 1.5-year-old male reindeer slaughtered in the early part of their rutting period. We found a relationship between parasite intensity and immune parameters suggesting that our measures of immune activity reflect density of current parasite infections. More important, these immune parameters were associated with FA in both the main beam length and numbers of antler tines. The immune parameters were, however, only weakly correlated to antler size. This indicates that FA, but not size, of antlers grown during exposure to a multitude of environmental stressors may reveal information about individual immunity that can be important for host-parasite interactions. Antler FA may therefore communicate an individual's quality during the rut in reindeer. Received: 30 March 1998 / Accepted after revision: 15 August 1998     

Weapon size versus body size as a predictor of winning in fights between shore crabs, Carcinus maenas (L.)

Relative body size (carapace width) and weapon size (chela length) were used as indicators of resource holding potential (RHP) in the agonistic behaviour of male shore crabs, Carcinus maenas (L.). Weapon size was found to be a more reliable predictor of the outcome of pairwise fights than body size. Crabs with longer chelae than their opponents were more likely to win fights than crabs with relatively larger bodies. Body size had less influence on the outcome of fights. Relative body and weapon size did not influence initiation of contests but did affect the likelihood of winning; however, this was significant only for weapon size. Winning crabs had heavier claws with greater surface area than losing crabs. There was no relationship between relative size and fight duration. The frequency of cheliped display increased with chela length and win- ners performed significantly more displays than losers. Received: 5 February 1997 / Accepted after revision: 20 May 1997     

Female mate choice in the non-territorial lizard Ameiva plei (Teiidae)

Mate choice by females has been documented in a variety of taxa. Female mate choice in species lacking male resource control or paternal care might occur if preferred males provide protection from harassment. Female mate choice was investigated in a natural population of the non-territorial lizard Ameiva plei (Teiidae). Consort pairs were allowed to form naturally. Consort males were significantly larger than non-consort males. After removal of consort males, the “abandoned” female's reaction to the first male who approached her was recorded. Females rejected all small males. Female preference for large males was significantly higher than preference for small males. Large males may be better equipped to guard the females from harassment and behavior of large males is less harassing than behavior of small males, thereby affording the female increased foraging time. Received: 21 June 1996 / Accepted after revision: 28 December 1996     

Inter- and intra-sexual patterns of fluctuating asymmetry in the red-billed streamertail: should symmetry always increase with ornament size?

Recent work on fluctuating asymmetry has suggested that ornaments should have higher levels of fluctuating asymmetry than (1) non-ornaments and (2) homologous structures in the non-ornamented sex. In addition, as both ornament size and symmetry should increase with individual quality there should be a tendency for ornament symmetry to increase with ornament size. In non-ornaments, a U-shaped relationship between symmetry and size is expected, with the individuals at the extremes being more asymmetrical than individuals around the optimum. We tested these predictions in the red-billed streamertail (Trochilus polytmus), a sexually dimorphic endemic Jamaican hummingbird. The lengths of four bilaterally symmetrical traits (first and second outermost tail feathers, tarsi and wings) in 43 adult males and 42 females were measured. The second outermost tail feathers of adult males (which are elongated into streamers) were absolutely but not relatively more asymmetrical than non-ornaments (including the homologous feathers in females). When character size was controlled for, wings were shown to be relatively more symmetrical than other traits. Symmetry did not increase with increasing trait size in any of the morphological traits measured. There was a U-shaped relationship between asymmetry and trait size for four traits (adult male streamers, adult male wings and female outer tail feathers). These results do not support any of the predictions made by fluctuating asymmetry hypotheses and suggest that stabilising selection may act on ornaments as well as non-ornaments. These predictions have been supported in swallows and peafowl but not in sunbirds; this may be due to differences in female perception of tail ornaments. Perhaps male tails do not convey information about quality in some species, or there may be inter-specific differences in the relative costs of tail ornaments and the benefit of marginal increases in tail length and symmetry.     

The structural significance of a vocalization that stimulates gaping in black-capped chickadees

In a variety of avian species, vocalizations used to stimulate gaping in nestlings during feeding share similar features such as broad frequency range, multiple spectral components, and sudden onset. It is currently not understood what functional significance, if any, these acoustic features have for close-range communication. One example of such a vocalization is the “squawk” of the black-capped chickadees (Parus atricapillus). The present study analyzes the significant features of the squawk by observing the response of chickadee nestlings to field playbacks of natural and computer-modified squawks. Of the features tested, frequency range appears most critical, but responses to frequency ranges change with age, such that low frequencies are most stimulating around the time of hatching but cease to elicit gaping after 5 days posthatch. Aperiodic stimuli and computer-modified squawks with gradual onsets tend to be less stimulating than natural squawks at some ages, but these differences are not significant at any age. The functional significance of the broad frequency range of the squawk is discussed, including the possible role of accommodating shifting frequency preferences related to ontogenetic changes in auditory sensitivity and elaboration of the vocal repertoire. Received: 16 January 1996 / Accepted after revision: 28 December 1996     

The role of turbidity as a constraint on predator-prey interactions in aquatic environments

Many of the world's most productive aquatic ecosystems usually contain turbid water. Paradoxically, many fish species that live in these habitats are also those that often rely on vision to detect their predators and their prey. For these fish, turbidity will reduce the distance at which predator-prey interactions occur, and there should be a reduction in the opportunity for behavioural modification to control the risk of predation. Under these conditions, we predict that most antipredator behaviour will become ineffective and that predator-prey interactions in turbid water will be primarily characterised by the direct effects of predator consumption of prey, rather than behavioural modification reducing the growth rates of prey. This hypothesis was tested in two laboratory experiments. The first experiment investigated how water turbidity, risk of predation, and their interaction affect habitat choice decisions by fathead minnows (Pimephales promelas). These data demonstrate that fathead minnows reduce their use of dangerous habitats, but that this effect is diminished in turbid water. A second experiment examined mortality patterns when these fish were preyed upon by yellow perch (Perca flavescens) in clear and turbid water. Absolute mortality rates were unaffected by visibility, but patterns of mortality were random in turbid water and skewed towards smaller individuals in clear water. Combined, these results support our hypothesis and suggest that the impact of predation risk will be reduced in turbid aquatic ecosystems. Received: 23 May 1996 / Accepted after revision: 18 November 1996     

Colony life history and demography of a swarm-founding social wasp

Colonies of social insects are sometimes viewed as superorganisms. The birth, reproduction, and death of colonies can be studied with demographic measures analogous to those normally applied to individuals, but two additional questions arise. First, how do adaptive colony demographies arise from individual behaviors? Second, since these superorganisms are made up of genetically distinct individuals, do conflicts within the colony sometimes modify and upset optima for colonies? The interplay between individual and superindividual or colony interests appears to be particularly complex in neotropical, swarm-founding, epiponine wasps such as Parachartergus colobopterus. In a long-term study of this species, we censused 286 nests to study colony-level reproduction and survivorship and evaluated individual-level factors by assessing genetic relatedness and queen production. Colony survivorship followed a negative exponential curve very closely, indicating type II survivorship. This pattern is defined by constant mortality across ages and is more characteristic of birds and other vertebrates than of insects. Individual colonies are long-lived, lasting an average of 347 days, with a maximum of over 4.5 years. The low and constant levels of colony mortality arise in part from colony initiation by swarming, nesting on protected substrates, and an unusual expandable nest structure. The ability to requeen rapidly was also important; relatedness data suggest that colonies requeen on average once every 9–12 months. We studied whether colony optima with respect to the timing of reproduction could be upset by individual worker interests. In this species, colonies are normally polygynous but new queens are produced only after a colony reaches the monogynous state, a result which is in accord with the genetic interests of workers. Therefore colony worker interests might drive colonies to reproduce whenever queen number happens to cycled down to one rather than at the season that is otherwise optimal. However, we found reproduction to be heavily concentrated in the rainy season. The number of new colonies peaked in this season as did the percentages of males and queens. Relatedness among workers reached a seasonal low of 0.21–0.27, reflecting the higher numbers of laying queens. This seasonality was achieved in part by a modest degree of synchrony in the queen reduction cycle. Worker relatedness reached peaks of around 0.4 in the dry season, reflecting a decrease to a harmonic mean queen number of about 2.5. Thus, a significant number of colonies must be approaching monogyny entering the rainy season. Coupled with polygynous colonies rearing only males (split sex ratios), this makes it possible for a colony cycle driven by selfish worker interests to be consistent with concentrating colony reproduction during a favorable season.     

Neighbours, strangers and male-male aggression as a determinant of lek size

Interactions between males on leks may play an influential role in lek formation and the regulation of lek size. In this paper I present the results of a playback experiment that simulated de novo settlement at sites adjacent to currently existing display territories of the ochre-bellied flycatcher, Mionectes oleagineus. In the study population, males displayed both solitarily and at small leks. A large proportion of males held no display territory at all. A stranger's song was played to both solitary and lekking males from 10 m outside their territorial boundaries. In separate playbacks, lekking males were also played neighbour's song. Both lekking and solitary territorial males reacted to the playback by decreasing their song rate, approaching the playback speaker and, on occasion, attacking the model. Solitarily displaying males responded more aggressively to playback of stranger's song than did lek males. Lek males were able to distinguish between their neighbour's and a stranger's song and did so irrespective of whether it was played from the neighbour's territory or from outside the lek. In addition to distinguishing between neighbours and strangers, lek males modified their responses to these different playbacks depending on where the playback originated. These results suggest that male-male interactions can be influential in structuring leks. In M. oleagineus, interactions between males are aggressive and act to limit rather than augment lek size. Received: 6 March 1996 / Accepted after revision: 9 December 1996     

Control of reproduction in social insect colonies: individual and collective relatedness preferences in the paper wasp, Polistes annularis

Social insect colonies often have one or a few queens. How these queens maintain their reproductive monopoly, when other colony members could gain by sharing in the reproduction, is not generally known. DNA microsatellite genotyping is used to determine reproductive interests of various classes of colony members in the paper wasp, Polistes annularis. The relatedness estimates show that the best outcome for most individuals is to be the reproductive egg-layer. For workers, this depends on the sex of offspring: they should prefer to lay their own male eggs, but are indifferent if the queen lays the female eggs. The next-best choice is usually to support the current queen. As a rule, subordinates and workers should prefer the current queen to reproduce over other candidates (though subordinates have no strong preference for the queen over other subordinates, and workers may prefer other workers as a source of male eggs). This result supports the theory that reproductive monopoly stems from the collective preferences of non-reproductives, who suppress each other in favor of the queen. However, we reject the general hypothesis of collective worker control in this species because its predictions about who should succeed after the death of the present queen are not upheld. The first successor is a subordinate foundress even though workers should generally prefer a worker successor. If all foundresses have died, an older worker succeeds as queen, in spite of a collective worker preference for a young worker. The results support the previous suggestion that age serves as a conventional cue serving to reduce conflict over queen succession. Received: 3 May 1996 / Accepted after revision: 22 September 1996     

Non-adaptive adoptions of nestlings in the colonial lesser kestrel: proximate causes and fitness consequences

We examined the causes, costs and benefits of adoption in the altricial lesser kestrel Falco naumanni. Specifically, we tested the intergenerational conflict hypothesis, proposed to explain adoption in some birds. Adoptions involved 76% of the nests and 51% of the nestlings at a mean age of 25 days (12 days before fledging). Nest-switching nestlings were not in poorer body condition, more parasitized or younger than their siblings, and body condition and prey delivery rates of their parents did not differ from those of other parents. In the foster nest, adopted nestlings did not benefit from higher feeding rates or a prolongation of the nestling period. They did not have fewer nest-mates or achieve higher rank within the new brood. Thus, adopted nestlings did not improve their body condition and survival. Adult lesser kestrels seemed unable to finely discriminate beween their own and alien chicks. Foster parents bore the cost of an increase of prey delivery rates, although it did not affect their survival or subsequent reproductive performance. Therefore, our results do not support the intergenerational conflict hypothesis, and suggest that adoption in this species is non-adaptive. Traditionally, the lesser kestrel bred in cliffs where movement among nest-sites was restricted. Nowadays, about half of the colonies are in tiled roofs which facilitate nest-switching by nestlings. The high rate of adoptions may thus be explained as reproductive errors associated with the recent occupation of a new breeding habitat. Received: 3 May 1996 / Accepted after revision: 19 January 1997     

Infanticide in seasonally breeding multimale groups of Hanuman langurs (Presbytis entellus) in Ramnagar (South Nepal)

In the seasonally breeding langur (Presbytis entellus) population of Ramnagar, South Nepal, where multimale groups prevail, 25 attacks on 11 infants (including one actual killing) by seven adult males were witnessed in five groups by six observers between 1990 and 1996. Circumstantial evidence also indicates three additional attempts at infanticide and in seven additional cases infanticide was presumed or likely. Infanticide presumably accounted for 30.8–62.5% of infant mortality in the first 2 years of life. Most attackers (91.4%) were residents of the infants' group and had immigrated after the infants had been born (75.0%) or conceived (25.0%). Thus, they were not related to the victims. The interbirth interval was shortened if an infant died either prior to September of its 1st year of life (mean = 1.2 years), or its 2nd year (mean = 2.0) and even its 3rd year (mean = 2.4). All attackers remained in the group at least until the next mating season; high-ranking males maintained their dominance rank and lower-ranking males rose in rank. Since rank and mating success were correlated and rank and reproductive success might be correlated, all attackers had a good chance of siring the next infant of the victims' mothers and could thus have benefited by their action. Infanticide seems to be a male reproductive strategy at Ramnagar. Infanticide has never before been reported among seasonally breeding langurs living at such low densities. This is also the first detailed report of infanticide as a male reproductive strategy in a seasonally breeding primate population. Received: 19 December 1996 / Accepted after revision: 7 June 1997     

Correlates of extra-pair fertilization success in hooded warblers

We examined correlates of extra-pair fertilization (EPF) success in the hooded warbler (Wilsonia citrina), a species where females pursue extra-pair matings. The good genes hypothesis predicts that females choose extra-pair mates with morphological or behavioral traits that reflect differences in male genetic quality. EPFs were common, as 35.3% (42/119) of broods contained extra-pair young (EPY) and 26.7% (95/356) of nestlings were the result of EPFs. There was a strong skew in male EPF success which increased the variance in annual male mating success 2–3 fold. However, male morphology did not predict EPF success, as extra-pair males were not older or larger than the males they cuckolded. Likewise, there were no significant correlations between the proportion of extra-pair young in a brood and male size or age. The good genes hypothesis predicts that high-quality males will be consistently preferred as genetic mates, but the number of young sired by a male with his social mate was not consistent from one year to the next. There was a significant negative correlation between female age and proportion of EPY produced, which could result if older females obtain higher-quality social mates. We found no strong evidence that females choose extra-pair mates for good genes, but females may use behavioral rather than morphological cues to assess relative male quality. Received: 31 May 1996 / Accepted after revision: 17 October 1996     

Lack of detectable nepotism in multiple-queen colonies of the fire ant Solenopsis invicta (Hymenoptera: Formicidae)

Multiple-queen (polygyne) colonies of the introduced fire ant Solenopsis invicta present a paradox for kin selection theory. Egg-laying queens within these societies are, on average, unrelated to one another, and the numbers of queens per colony are high, so that workers appear to raise new sexuals that are no more closely related to them than are random individuals in the population. This paradox could be resolved if workers discriminate between related and unrelated nestmate sexuals in important fitness-related contexts. This study examines the possibility of such nepotism using methods that combine the following features: (1) multiple relevant behavioral assays, (2) colonies with an unmanipulated family structure, (3) multiple genetic markers with no known phenotypic effects, and (4) a statistical technique for distinguishing between nepotism and potentially confounding phenomena. We estimated relatedness between interactants in polygyne S. invicta colonies in two situations, workers tending egg-laying queens and workers feeding maturing winged queens. In neither case did we detect a significant positive value of relatedness that would implicate nepotism. We argue that the non-nepotistic strategies displayed by these ants reflect historical selection pressures experienced by native populations, in which nestmate queens are highly related to one another. The markedly different genetic structure in native populations may favor the operation of stronger higher-level selection that effectively opposes weaker individual-level selection for nepotistic interactions within nests. Received: 28 June 1996 / Accepted after revision: 6 October 1996     

Dominance and the dynamics of phenotype-limited distribution in common cranes

We studied the behavior of 13 radiotagged cranes dispersing from a communal roost over days when they changed their main daily foraging area between consecutive days during two winter seasons. Individuals went to a new foraging zone when on the previous day their morning food intake had fallen below their mean morning food intake measured over the whole winter. Food intake on the day before a change in foraging area was positively correlated with dominance rank. Dominant cranes changed to new zones with higher numbers of birds and food density, while subordinate cranes went to new zones with lower numbers of birds. As a result, all birds increased their food intake over that of the previous day. Dominant cranes remained more faithful to their most preferred foraging zone, where they spent 69% of the mornings, while subordinate birds were more mobile, switching among zones frequently. Dominant birds left the roost later than subordinate birds on the days they changed to a new zone, which could be used to track the main departing flows. The results suggest that the dynamics that led to a truncated phenotype-limited distribution were determined by social dominance and food abundance, with dominant cranes shifting to a new zone to maintain their high intake levels and subordinates changing more frequently whenever their daily intake did not reach the minimum metabolic requirements. Received: 16 December 1996 / Accepted after revision: 22 February 1997     

Lack of kin recognition in swarming honeybees ( Apis mellifera )

Honeybee colonies reproduce by colony fission and swarming. The primary swarm leaves the nest with the mated mother queen. Further “after-swarms” can leave the nest. These are composed of virgin queens and sister workers. Since all workers in the primary swarm have the same relationship to the mother queen, kin recognition cannot have any effect on the worker distribution in the swarm. Because of polyandry of the mother queen, the after-swarm is composed of super- and halfsister workers of the virgin queen. In this case kin recognition might affect swarm composition if workers increase their inclusive fitness by preferentially investing in a supersister queen. The distribution of workers in the mother colony, the primary and the after-swarm was analyzed using single-locus DNA fingerprinting in two colonies of the honeybee (Apis mellifera). The colonies were composed of 21 and 24 worker subfamilies because of multiple mating of the queen. The subfamily distribution in the mother colonies before swarming was significantly different from the subfamily frequencies in the primary swarm. This indicates different propensities for swarming in the various subfamilies. The subfamily distribution was also significantly different between the mother colony and the after-swarm. There was however no significant difference between the subfamily composition of the primary and the after-swarm. The average effects of kin recognition on the distribution of the subfamilies in the two after-swarms were less than 2%. We conclude that colony-level selection sets the evolutionary framework for swarming behaviour. Received: 22 May 1996 / Accepted after revision: 2 November 1996     

Reproductive dominance and differential ovicide in the communally breeding burying beetle Nicrophorus tomentosus

Communal breeding can be characterized by the division of reproduction among cooperating individuals and, if the distribution of reproduction is inequitable, by the mechanisms for achieving skewed reproductive success. The burying beetle (Nicrophorus tomentosus) is a facultative communal breeder. Unrelated adults, especially females, provide extensive parental care to broods of mixed parentage. The frequency and degree of reproductive skew between two females were examined experimentally. On medium-size carcasses, the proportion of eggs attributed to each female was not significantly different from random in 42% of the broods, skewed in 42% and not shared in 16%. Although reproduction was usually skewed in favor of the larger female, the relative sizes of the two females did not predict the degree of skew. On large carcasses, the proportion of eggs attributed to each female was not different from random in 87% of the broods and weakly skewed in 13%. Several mechanisms for biasing reproductive success were investigated. Females increase the proportion of their offspring in the brood by committing differential ovicide. Secondly, burying and preparing a carcass cooperatively stimulates ovarian development of the larger female and slows it for the smaller female, reducing or delaying oviposition by the subordinate. Thirdly, larger females are more likely to be dominant and are more fecund than smaller females. Received: 20 July 1996 / Accepted after revision: 30 November 1996     

An estimate of female mate searching costs in the lekking butterfly Coenonympha pamphilus

We estimated the cost to females of the lekking butterfly Coenonympha pamphilus of visiting males on leks instead of taking off and soliciting courtship from males passing them outside leks, as occurs in the non-lekking congener C. tullia. We followed released virgin females of C. pamphilus in the field until they mated. We estimated the time they lost by remaining perched and not approaching males that passed them, and observed how often and at what distance virgin females were passed by males in the field. We then calculated how much faster these females would have been detected if they had taken off and approached these males, using probabilities of detection as a function of passing distance derived from field observations on C. tullia females. C. pamphilus females on average lost 201 min by not approaching males on their way to leks. To estimate what this time loss meant in terms of fitness, we measured how the age at mating affected fecundity. Using field measures of fecundity and mortality, the time loss translated into an average 2.8% reduction in fecundity as a best estimate, and an average 1.3% reduction in fecundity as a lower estimate. This fitness cost is larger than has been reported earlier for a lekking bird, but is probably too small to eliminate the possibility of indirect benefits of mating with males on leks. Received: 15 August 1996 / Accepted after revision: 24 January 1997     

Relatedness, polyandry and extra-group paternity in the cooperatively-breeding white-browed scrubwren (Sericornis frontalis )

We used DNA fingerprinting to examine the genetic parentage and mating system of the cooperatively breeding white-browed scrubwren, Sericornis frontalis, in Canberra, Australia. Our analyses revealed a remarkable variety of mating tactics and social organization. Scrubwrens bred in pairs or multi-male groups that consisted of a female and two or more males. Females were always unrelated to the pair male or alpha (dominant) male. Among multi-male groups we found three different mating tactics. Firstly, when alpha and beta (subordinate) males were unrelated, they usually shared paternity in the brood. This resulted in both males gaining reproductive benefits directly. Secondly, when beta males were not related to the female but were related to the alpha males, beta males sired offspring in some broods. In this situation, beta males gained reproductive benefits both directly and potentially indirectly (through the related alpha male). Thirdly, when beta males were related to the female or both the female and alpha male, they remained on their natal territory and did not sire any offspring. Thus beta males gained only indirect reproductive benefits. Overall, when group members were related closely, the dominant male monopolized reproductive success, whereas when the members were not related closely the two males shared paternity equally. This positive association between monopolization of reproduction and relatedness is predicted by models of reproductive skew, but has not been reported previously within a single population of birds. Other cooperatively breeding birds with both closely related and unrelated helpers may show a similar variety of mating tactics. Finally, we found that extra-group paternity was more common in pairs (24% of young) than in multi-male groups (6%), and we discuss three possible reasons for this difference. Received: 21 May 1996 / Accepted after revision: 14 December 1996