Journal of Chemical Ecology, VoL 7, No. 3, 1981
P A R T I A L M O L T I N G S Y N C H R O N Y IN THE G I A N T MALAYSIAN PRAWN, Macrobrachium rosenbergii : A Chemical Communication Hypothesis
NATHAN R. HOWE Biology Department, University of Houston 4700 Avenue U, Galveston, Texas 77550 (Received August 7, 1980; revised September 28, 1980) Abstract--Groups of 50juvenile specimens of M. rosenbergff were observed daily for the occurrence of molting for periods up to 50 days. Each animal in a group was physically isolated from the others in one of 50 chambers assembled in a flat chamber array that was immersed in a recirculated bath of fresh water (28 ~ C). While an average of 6.4% of the animals molted each day in each of three separate trials, molting occurred in significantly nonrandom peaks and valleys. There was no evidence that peaks in molting frequency occurred at regular intervals, nor were animals of the same size molting synchronously, although weight and molting frequency were significantly correlated. Based on intervals between molts predicted from animal weights, animals molting during a peak molted sooner than animals molting with few others. Molting frequency in a group of animals tended to rise until water was changed, at which time molting frequency dropped significantly. Within each array of animal chambers, animals molted in significant spatial aggregations that coincided with the measured pattern of water flow among chambers in the array. Those spatial aggregations were most pronounced during peaks in molting frequency. These results suggest that some waterborne stimulus, perhaps released by molting animals, is responsible for partial molting synchrony in groups of M. rosenbergii. The potential advantage of molting synchrony is discussed in connection with the behavioral ecology of M. rosenbergii. Key Words--Molting, Macrobrachium rosenbergii, chemical communication, crustacean.
INTRODUCTION
The act of molting (ecdysis) and the softness of newly formed cuticle expose crustaceans to an increased risk of mortality, a vulnerability that has been 487 oo98-o331/81/o5oo-o4875o3oolo 9 1981PlenumPublishingCorporation
488
Howe
shown to be reduced in some species by molt-associated behavioral adaptations (Passano, 1960). Part of the risk of molting for aggressive gregarious species is the danger of being killed or displaced by conspecifics. Reaka (1976) has described a novel and complex risk-reduction strategy for several species of mantis shrimps that are normally vulnerable to attack by conspecifics at ecdysis. That strategy includes not only reclusive behavior in molting animals but also control of the timing of ecdysis. In common with other aggressive crustaceans, the durations of molt-cycle stages in mantis shrimps are skewed to minimize the vulnerable (soft) period and to allow a relatively protracted premolt period, during which animals may molt in response to an appropriate stimulus (Reaka, 1975). Reaka's (1976) studies show that molting in mantis shrimp populations is significantly synchronous, with rhythmic molting episodes that are correlated with lunar or tidal cycles. Her results support the hypothesis that synchronous molting provides safety in numbers for molting animals. A comparable risk-reduction strategy would be adaptive for species with similar life histories; for example, the Malaysian prawn Macrobrachium rOsenbergii, a large, aggressive palaemonid native to the Indo-Pacific region. Adults ofM. rosenbergii live in bodies of freshwater as far as 200 km from the ocean (Ling, 1969). Mated females migrate into brackish estuaries to release larvae, and the planktonic larvae develop into benthic juveniles that migrate upstream into adult habitats (Raman, 1964). The behavior of M. rosenbergii in its natural habitat is largely unstudied. Because Maerobrachium has been successfully cultured for more than a decade (Ling, 1969), however, there is a growing literature on the behavioral ecology of captive animals raised at high density in large ponds. Aggressive behavior is conspicuous in captive M. rosenbergii and has been implicated as contributing to high mortality in some ponds (Forster and Beard, 1974; Peebles, 1977, 1978) and to markedly nonuniform growth rates in animals of the same age (Fujimura and Okamota, 1970; Malecha, 1977). Animals are subject to lethal attacks from conspecifics primarily at ecdysis (Peebles, 1977). Adult males are larger and have proportionally larger chelae than females (Ling, 1969). Animals forage at dawn and dusk and at other times defend home ranges that are smaller and more permanent for males than for females (Peebles, 1979). Animals that are soon to molt or that have recently molted tend to occupy less-preferred substrates, where a choice of substrate types is available. In the course of an unrelated experiment, some degree of molting synchrony seemed to occur among juvenile M. rosenbergii that were maintained in the same water supply. Because the social behavior of M. rosenbergii, at least in captivity, suggested that molting at the same time as potential conspecific competitors could be adaptive, the possibility of molting synchrony was examined in greater detail.
PARTIAL MOLTING SYNCHRONY IN GIANT MALAYSIAN PRAWN
METHODS
489
AND MATERIALS
The results of three experiments are reported. Since the design of the first differed from that of the latter two in several respects, they are described separately. All were performed in the laboratories of the National Marine Fisheries Service and the University of Houston in Galveston, Texas. The water used in all animal tanks was obtained from the Galveston Municipal Water Supply (mean hardness = 84 mg/liter) and was aerated before use to remove chlorine. Experiment l. M. rosenbergii postlarvae from a commercial source were held for three months in a 40,000-liter tank of heated fresh water under diffuse natural illumination. One hundred animals were then transferred to a 200-liter tank in the laboratory where they were maintained for 100 days at 28~ (+0.5 ~ C) with a light regime of 12 hr dim (500 1x) fluorescent light and 12 hr darkness. Fifty randomly selected animals were then blotted briefly, weighed (range: 0.17-2.84 g), and transferred to individual chambers, 6 cm on each side, arranged in a flat, 5 X 10-chamber array. Chambers in the array had rigid, opaque, white styrene walls, plastic screen bottoms and a removable plastic screen cover. The array was fitted over a shallow sand and gravel filter bed inside a fiberglass tank and covered with 40 liters of water. Six airlift pumps circulated water in through the tops of the chambers and out through the b o t t o m of the filter bed at approximately 1 liter/min. Light and temperature regimes were not altered. Distilled water was added daily to replace evaporative loss, but water was not exchanged except as noted in Results. Twice daily, 2 hr before the end of the dark period and 5 hr before the end of the light period, the tray was removed from the water, the positions of molted animals were recorded, exuviae (cast exoskeletons) and dead animals were removed, and food (Tetramin flake to maintain an excess) was provided to each animal. This observation period lasted 15 rain, during which room lights were on. Experiment 1 was terminated after the 50th day of observations, when 40 animals survived. Experiments 2 and 3. For each of these later experiments, two independent 50-chamber arrays were tested concurrently for shorter periods of time. These chambers were constructed of 3.8-mm-thick grey polyvinylchloride sheets and were smaller than in experiment 1:4 X l0 X 4 cm high. Each array was immersed in a shallow tray containing 16 liters of aged, aerated municipal water drawn from a c o m m o n tank. Water drawn from 24 evenly spaced drains in the bottom of each tray was pumped at approximately 10 liter/min through a filter cannister containing a glass-wool element. Filtered water returned to the chamber tray through a diffuser (Figure 1). At three-day intervals glass wool filter elements were replaced. At the same time 800 cm 3 of granular activated carbon was temporarily placed in the filter cannisters and removed after 1 hr. Regular charcoal filtration
4912
Howe |ilter
water level chamber
l~ " . . . . . ~.
inanifold
~ ~-] i_~
pl . . . . purnp
FrG. 1. Schematic diagram of water recirculating system for animal chamber arrays used in experiments 2 and 3. substantially reduced any progressive deterioration in water quality and its attendant effects on mortality; no animal died in either experiment. The two trays were placed side by side under identical dim fluorescent illumination (14L: 10D). Water temperature in the two trays ranged from 26~ C to 27 ~C in experiment 2 and from 26.5 ~ C to 27.5 ~ C in experiment 3. The two trays never differed in temperature by more than 0.2 ~C. During the experiments, 1 ml of water was exchanged between trays each day to minimize potential qualitative differences in tank microflora or fauna. In the course of experiment l, 96% of 146 observed molts occurred during the dark period. Therefore, animals were observed only once daily in experiments 2 and 3, 1 hr after the end of the dark period. In these later experiments, chamber arrays were only partly emersed during observations to reduce trauma to the experimental animals. Exuviae were removed and food added as before, except that the diet used was an experimental pelletized ration developed for penaeid shrimp by the University of Arizona. Experiments 2 and 3 were terminated after 27 and 15 days, respectively. Mean molting frequencies for groups of animals fed either diet did not differ from each other or from published molting frequencies, suggesting that the experimental diets were adequate to promote normal molting. RESULTS
Molting Peaks. The number of animals molting each day in each chamber array, expressed as a percentage of the number of animals surviving on that day, is shown for all experiments in Figure 2. In experiment 1 (upper trace) an average of 6.3% of the animals molted each day, but molting activity appeared to occur in peaks rather than continuously. That hypothesis was confirmed by fitting a Poisson distribution to grouped (8 classes) daily molt percentages and testing goodness of fit (for H ~ = no difference: X2 = 39.4, P < 0.005). As Figure 2 suggests, the frequency of days when no animal molted or when more than 14% molted significantly exceeded expectations. The results of experiments 2 and 3 are similar. Mean daily molting
PARTIAL
MOLTING
SYNCHRONY
IN GIANT
MALAYSIAN
491
PRAWN
Experiment 1 20 -,r o
i=
10
10
20
30
40
50
Day
Experiment 2
Experiment 3 ?
20
2 0 84
!i i
r
o
lo
.?
,!~
~ ~o
i
0
20
10
Day
0
5
10
15
Day
FIG. 2. Number of animals (as % of survivors) molting each day in experiments 1 (upper trace), 2 (lower left), and 3 (lower right). Dashed and solid lines in the graphs for experiments 2 and 3 indicate the response of two independent groups of animals. See text for the significance of values marked by open circles. percentages for the two chamber arrays in experiment 2 were 6.5% and 6.2% and in experiment 3, 6.8% and 6.3%. None of these latter means differed significantly from another or from the mean in experiment 1. Experiments 2 and 3 were too short in duration (27 and 15 days, respectively) for the distributions of daily molt percentages in either experiment to differ significantly from a Poisson distribution. When all daily observations in the latter experiments were pooled, however, the distribution of pooled observations differed significantly (X 2 = 36.7, P < 0.005, 8 df) from Poisson expectations, again with both high and low values more frequent than expected. Periodograrn Analysis, To detect potential periodicities in the data from experiment 1, daily molt percentages were used to construct a periodogram (Enright, 1965) according to the modification suggested by Sokolove and Bushell (1978). The latter authors showed that the statistic, Qp, a measure of the degree to which hypothetical periods explain variation in a time series, has a distribution that closely approximates X 2 with degrees of freedom equal to one less than each integral period. Qp is plotted in Figure 3 together with the appropriate X20.005values for periods between 2 and 16 days. The significance level of X 2 was chosen to produce an acceptable probability of a type I error.
Howe
492
//
//
/
jJ
25 // / /
20
i///*
g
t L~
15
(.3 /
&
/
/
/
q
/
lO
I
5
10
15
Period (days)
FIG. 3. Periodogram (Sokolove and Bushell, 1978) for daily molting percentages in experiment 1. The dashed line indicates X2000s, a significance criterion. For critical X2 values at the 0.005 level and 14 test periods, the probability of one or more type I errors, i.e., false peaks for which Qp exceeds x 2 when the data are actually random, is approximately 0.07 (Sokolove and Bushell, 1978). Since no Qp value exceeds criterion, the data from experiment 1 do not support the conclusion that the peaks in daily molting frequency are periodic. Experiment 1 was too brief to permit testing for possible periods longer than 16 days, and the later experiments were too brief for any periodicity analyses. Effect of Animal Weight on Molt-Cycle Duration. In experiments 1 and 2, time between successive molts for a given animal (intermolt duration) was significantly correlated with its initial weight (r = 0.33, 71 df, andr = 0.54, 64 df, respectively; P < 0.01). In experiment 1 the intermolt duration for the hypothetical animal of mean weight (0.65 g) was 14.6 days and each additional gram of weight lengthened intermolt duration by approximately 3 days (leastsquares regression). For experiment 2, comparable values were 0.50 g (mean weight), l l.1 days, (mean intermolt) and 6 days (slope). No obvious explanation accounts for the difference in regression slopes between the two experiments. Experiment 3 was too brief for analysis. Although animals of the same weight tend to molt with equal frequency, molting peaks did not appear to reflect simultaneous molting in groups of like-size animals. The weights of animals that molted in five large, welldefined peaks in experiments 1 and 2 (open circles, Figure 2) were compared by X2 goodness of fit to the weight distributions of all animals used for those
PARTIAL MOLTING SYNCHRONY IN GIANT MALAYSIAN PRAWN
493
experiments. The probabilities that animal weights in the sample peaks could have been random samples of the overall weight distributions ranged from 0.3 to 0.9 (mean P > 0.5). Effect of Peaks on Intermolt Duration. The fact that intermolt duration and weight were strongly correlated in both experiments provided a method for predicting intermolt duration: for each animal that molted at least twice, an expected intermolt duration was generated from the appropriate regression equation. Actual intermolt durations for those animals were then compared to the predicted values, and animals were scored as "late" or "early" molters, accordingly. For each day on which second (or third) molts occurred, the total number of molting animals was counted. For experiment 1 the mean for that count was 4.5 animals; for experiment 2, 4.9 animals. Days on which four or fewer animals molted were therefore scored as "few"; those on which five or more animals molted were termed "many." Contingency tables (Table 1) were then constructed for each experiment based on those scored attributes. For both experiments the two attributes were significantly nonindependent (xzc = 4.1 and 5.2, respectively; P < 0.05), and the deviations from random expectations were in the predicted direction. In general, animals molting with few others molted later than predicted on the basis of weight, and animals molting with many others (that is, in peaks) molted sooner than expected. Effect of Water Exchange on Molting Frequency. To examine the possibility that a molt-accelerating factor accumulated in experimental water supplies, the water recirculated in experiment l was replaced at day 33. Potential long-term effects of that manipulation were detected by computing an 1 l-day moving average for daily molt percentage. The smoothed data are shown in Figure 4. In that figure open circles represent l l-day means that include only days before the water change, half-filled circles indicate means that include progressively more postchange days, and closed circles indicate means of solely postchange days. Smoothed molt frequency appears to rise slightly (although not significantly) until the water change. At day 29, as the TABLE l. NUMBER OF SECOND AND THIRD MOLTS IN EXPERIMENTS 1 AND 2, CLASSIFIED ACCORDING TO DURATION OF PRECEDING INTERMOLT PERIODS AND NUMBER OF ANIMALS MOLTING ON THOSE DAYS.
Molt timing Number of animals molting
Late
Early
Late
Early
Few Many
21 8
19 23
21 9
14 22
Experiment 1
Experiment 2
494
HOWE O
OQ 00000 O0
000000000 O00
0 .~.
9
~ l ) . t l .D
o E
Q" , l " ~..
9 ~O
f0
2'0
3'o
4'o
-
Central day of ll-day means
FIG. 4. Eleven-day moving averages of molting percentages in experiment 1, before (open circles), during (half-filled circles), and after (closed circles) a water change. The dashed line represents the least-squares fit to the half-filled circles. moving average begins to include postchange days, molt frequency drops markedly. A line fitted by least squares to the half-filled circles has a significantly negative slope (Student's t test, two-tailed, P < 0.01). By the seventh day after the water change, smoothed molt frequency begins to rise. Spatial Aggregation in Molting Peaks. The results of experiments 2 and 3 were examined for possible spatial aggregation in the locations of molting animals. The fiat arrays of chambers used in those experiments rested on a shallow plenum (Figure 1) from which water was removed at a maximum measured rate of 10 liters/rain. That rate of pumping resulted in a net flow of water downward through the chambers at a calculated velocity of from 5 cm/min when filters were clean to 3 cm/min just before the filter elements were replaced. Despite many regularly spaced drains from the plenum, that slow flow of water was not laminar, that is, water flowed laterally between chambers as well as vertically through each chamber. The magnitude of that effect was checked in a series of dye experiments. With the recirculating pump operating at 10 liters / rain, a small volume of dye solution was released into a selected chamber. Water samples (0.5 ml) were then removed from that cell and from neighboring cells at 30-sec intervals, and dye concentrations were measured photometrically. Dye concentrations were highest in nearby cells at the first sampling time (+30 sec), but even 1 min after dye release, concentrations in certain nearby cells remained well above those in more distant cells. Figure 5a shows dye concentrations recorded 60 sec after release as a percentage of the initial concentration in the central cell for two typical experiments. Two results are worthy of mention. First, dye released into corner cells dissipated more slowly than from more central cells. Second, as the geometry of the chamber array suggests, water flow between cells in the same column (centers 4 cm apart) is much greater than between cells in the same row (10 cm apart). By 3 min after dye release in most experiments, the dye was uniformly distributed in all cells in the array.
P A R T I A L M O L T I N G S Y N C H R O N Y IN G I A N T M A L A Y S I A N P R A W N
495
Based on the results of the dye experiments, an algorithm was developed to test the results of experiments 2 and 3 for spatial aggregation (autocorrelation) of molting within columns of chambers. That algorithm was modeled after the approach of Sokal and Oden (1978) and had the following properties: 1. In order to accumulate sufficient molts for each autocorrelation analysis, three successive days of molting records were combined into a single chamber matrix. 2. In the experiments analyzed, an intermolt period shorter than 7 days occurred only once. Accordingly, each calculation excluded animals that had molted six or fewer days before the end of the 3-day period. 3. Consistent with the dye experiments, an instance of molting in two adjoining column cells was assigned a high relative weight (1 join), and molting in two cells in the same column that were separated by a single cell was assigned a lower weight (1 / 2join). No other spatial relationship between cells in which molting occurred was given weight The number of joins expected with this weighting rule, assuming random locations for molts, was computed by the method of Sokal and Oden (1978).
P A R T I A L M O L T I N G S Y N C H R O N Y IN THE G I A N T MALAYSIAN PRAWN, Macrobrachium rosenbergii : A Chemical Communication Hypothesis
NATHAN R. HOWE Biology Department, University of Houston 4700 Avenue U, Galveston, Texas 77550 (Received August 7, 1980; revised September 28, 1980) Abstract--Groups of 50juvenile specimens of M. rosenbergff were observed daily for the occurrence of molting for periods up to 50 days. Each animal in a group was physically isolated from the others in one of 50 chambers assembled in a flat chamber array that was immersed in a recirculated bath of fresh water (28 ~ C). While an average of 6.4% of the animals molted each day in each of three separate trials, molting occurred in significantly nonrandom peaks and valleys. There was no evidence that peaks in molting frequency occurred at regular intervals, nor were animals of the same size molting synchronously, although weight and molting frequency were significantly correlated. Based on intervals between molts predicted from animal weights, animals molting during a peak molted sooner than animals molting with few others. Molting frequency in a group of animals tended to rise until water was changed, at which time molting frequency dropped significantly. Within each array of animal chambers, animals molted in significant spatial aggregations that coincided with the measured pattern of water flow among chambers in the array. Those spatial aggregations were most pronounced during peaks in molting frequency. These results suggest that some waterborne stimulus, perhaps released by molting animals, is responsible for partial molting synchrony in groups of M. rosenbergii. The potential advantage of molting synchrony is discussed in connection with the behavioral ecology of M. rosenbergii. Key Words--Molting, Macrobrachium rosenbergii, chemical communication, crustacean.
INTRODUCTION
The act of molting (ecdysis) and the softness of newly formed cuticle expose crustaceans to an increased risk of mortality, a vulnerability that has been 487 oo98-o331/81/o5oo-o4875o3oolo 9 1981PlenumPublishingCorporation
488
Howe
shown to be reduced in some species by molt-associated behavioral adaptations (Passano, 1960). Part of the risk of molting for aggressive gregarious species is the danger of being killed or displaced by conspecifics. Reaka (1976) has described a novel and complex risk-reduction strategy for several species of mantis shrimps that are normally vulnerable to attack by conspecifics at ecdysis. That strategy includes not only reclusive behavior in molting animals but also control of the timing of ecdysis. In common with other aggressive crustaceans, the durations of molt-cycle stages in mantis shrimps are skewed to minimize the vulnerable (soft) period and to allow a relatively protracted premolt period, during which animals may molt in response to an appropriate stimulus (Reaka, 1975). Reaka's (1976) studies show that molting in mantis shrimp populations is significantly synchronous, with rhythmic molting episodes that are correlated with lunar or tidal cycles. Her results support the hypothesis that synchronous molting provides safety in numbers for molting animals. A comparable risk-reduction strategy would be adaptive for species with similar life histories; for example, the Malaysian prawn Macrobrachium rOsenbergii, a large, aggressive palaemonid native to the Indo-Pacific region. Adults ofM. rosenbergii live in bodies of freshwater as far as 200 km from the ocean (Ling, 1969). Mated females migrate into brackish estuaries to release larvae, and the planktonic larvae develop into benthic juveniles that migrate upstream into adult habitats (Raman, 1964). The behavior of M. rosenbergii in its natural habitat is largely unstudied. Because Maerobrachium has been successfully cultured for more than a decade (Ling, 1969), however, there is a growing literature on the behavioral ecology of captive animals raised at high density in large ponds. Aggressive behavior is conspicuous in captive M. rosenbergii and has been implicated as contributing to high mortality in some ponds (Forster and Beard, 1974; Peebles, 1977, 1978) and to markedly nonuniform growth rates in animals of the same age (Fujimura and Okamota, 1970; Malecha, 1977). Animals are subject to lethal attacks from conspecifics primarily at ecdysis (Peebles, 1977). Adult males are larger and have proportionally larger chelae than females (Ling, 1969). Animals forage at dawn and dusk and at other times defend home ranges that are smaller and more permanent for males than for females (Peebles, 1979). Animals that are soon to molt or that have recently molted tend to occupy less-preferred substrates, where a choice of substrate types is available. In the course of an unrelated experiment, some degree of molting synchrony seemed to occur among juvenile M. rosenbergii that were maintained in the same water supply. Because the social behavior of M. rosenbergii, at least in captivity, suggested that molting at the same time as potential conspecific competitors could be adaptive, the possibility of molting synchrony was examined in greater detail.
PARTIAL MOLTING SYNCHRONY IN GIANT MALAYSIAN PRAWN
METHODS
489
AND MATERIALS
The results of three experiments are reported. Since the design of the first differed from that of the latter two in several respects, they are described separately. All were performed in the laboratories of the National Marine Fisheries Service and the University of Houston in Galveston, Texas. The water used in all animal tanks was obtained from the Galveston Municipal Water Supply (mean hardness = 84 mg/liter) and was aerated before use to remove chlorine. Experiment l. M. rosenbergii postlarvae from a commercial source were held for three months in a 40,000-liter tank of heated fresh water under diffuse natural illumination. One hundred animals were then transferred to a 200-liter tank in the laboratory where they were maintained for 100 days at 28~ (+0.5 ~ C) with a light regime of 12 hr dim (500 1x) fluorescent light and 12 hr darkness. Fifty randomly selected animals were then blotted briefly, weighed (range: 0.17-2.84 g), and transferred to individual chambers, 6 cm on each side, arranged in a flat, 5 X 10-chamber array. Chambers in the array had rigid, opaque, white styrene walls, plastic screen bottoms and a removable plastic screen cover. The array was fitted over a shallow sand and gravel filter bed inside a fiberglass tank and covered with 40 liters of water. Six airlift pumps circulated water in through the tops of the chambers and out through the b o t t o m of the filter bed at approximately 1 liter/min. Light and temperature regimes were not altered. Distilled water was added daily to replace evaporative loss, but water was not exchanged except as noted in Results. Twice daily, 2 hr before the end of the dark period and 5 hr before the end of the light period, the tray was removed from the water, the positions of molted animals were recorded, exuviae (cast exoskeletons) and dead animals were removed, and food (Tetramin flake to maintain an excess) was provided to each animal. This observation period lasted 15 rain, during which room lights were on. Experiment 1 was terminated after the 50th day of observations, when 40 animals survived. Experiments 2 and 3. For each of these later experiments, two independent 50-chamber arrays were tested concurrently for shorter periods of time. These chambers were constructed of 3.8-mm-thick grey polyvinylchloride sheets and were smaller than in experiment 1:4 X l0 X 4 cm high. Each array was immersed in a shallow tray containing 16 liters of aged, aerated municipal water drawn from a c o m m o n tank. Water drawn from 24 evenly spaced drains in the bottom of each tray was pumped at approximately 10 liter/min through a filter cannister containing a glass-wool element. Filtered water returned to the chamber tray through a diffuser (Figure 1). At three-day intervals glass wool filter elements were replaced. At the same time 800 cm 3 of granular activated carbon was temporarily placed in the filter cannisters and removed after 1 hr. Regular charcoal filtration
4912
Howe |ilter
water level chamber
l~ " . . . . . ~.
inanifold
~ ~-] i_~
pl . . . . purnp
FrG. 1. Schematic diagram of water recirculating system for animal chamber arrays used in experiments 2 and 3. substantially reduced any progressive deterioration in water quality and its attendant effects on mortality; no animal died in either experiment. The two trays were placed side by side under identical dim fluorescent illumination (14L: 10D). Water temperature in the two trays ranged from 26~ C to 27 ~C in experiment 2 and from 26.5 ~ C to 27.5 ~ C in experiment 3. The two trays never differed in temperature by more than 0.2 ~C. During the experiments, 1 ml of water was exchanged between trays each day to minimize potential qualitative differences in tank microflora or fauna. In the course of experiment l, 96% of 146 observed molts occurred during the dark period. Therefore, animals were observed only once daily in experiments 2 and 3, 1 hr after the end of the dark period. In these later experiments, chamber arrays were only partly emersed during observations to reduce trauma to the experimental animals. Exuviae were removed and food added as before, except that the diet used was an experimental pelletized ration developed for penaeid shrimp by the University of Arizona. Experiments 2 and 3 were terminated after 27 and 15 days, respectively. Mean molting frequencies for groups of animals fed either diet did not differ from each other or from published molting frequencies, suggesting that the experimental diets were adequate to promote normal molting. RESULTS
Molting Peaks. The number of animals molting each day in each chamber array, expressed as a percentage of the number of animals surviving on that day, is shown for all experiments in Figure 2. In experiment 1 (upper trace) an average of 6.3% of the animals molted each day, but molting activity appeared to occur in peaks rather than continuously. That hypothesis was confirmed by fitting a Poisson distribution to grouped (8 classes) daily molt percentages and testing goodness of fit (for H ~ = no difference: X2 = 39.4, P < 0.005). As Figure 2 suggests, the frequency of days when no animal molted or when more than 14% molted significantly exceeded expectations. The results of experiments 2 and 3 are similar. Mean daily molting
PARTIAL
MOLTING
SYNCHRONY
IN GIANT
MALAYSIAN
491
PRAWN
Experiment 1 20 -,r o
i=
10
10
20
30
40
50
Day
Experiment 2
Experiment 3 ?
20
2 0 84
!i i
r
o
lo
.?
,!~
~ ~o
i
0
20
10
Day
0
5
10
15
Day
FIG. 2. Number of animals (as % of survivors) molting each day in experiments 1 (upper trace), 2 (lower left), and 3 (lower right). Dashed and solid lines in the graphs for experiments 2 and 3 indicate the response of two independent groups of animals. See text for the significance of values marked by open circles. percentages for the two chamber arrays in experiment 2 were 6.5% and 6.2% and in experiment 3, 6.8% and 6.3%. None of these latter means differed significantly from another or from the mean in experiment 1. Experiments 2 and 3 were too short in duration (27 and 15 days, respectively) for the distributions of daily molt percentages in either experiment to differ significantly from a Poisson distribution. When all daily observations in the latter experiments were pooled, however, the distribution of pooled observations differed significantly (X 2 = 36.7, P < 0.005, 8 df) from Poisson expectations, again with both high and low values more frequent than expected. Periodograrn Analysis, To detect potential periodicities in the data from experiment 1, daily molt percentages were used to construct a periodogram (Enright, 1965) according to the modification suggested by Sokolove and Bushell (1978). The latter authors showed that the statistic, Qp, a measure of the degree to which hypothetical periods explain variation in a time series, has a distribution that closely approximates X 2 with degrees of freedom equal to one less than each integral period. Qp is plotted in Figure 3 together with the appropriate X20.005values for periods between 2 and 16 days. The significance level of X 2 was chosen to produce an acceptable probability of a type I error.
Howe
492
//
//
/
jJ
25 // / /
20
i///*
g
t L~
15
(.3 /
&
/
/
/
q
/
lO
I
5
10
15
Period (days)
FIG. 3. Periodogram (Sokolove and Bushell, 1978) for daily molting percentages in experiment 1. The dashed line indicates X2000s, a significance criterion. For critical X2 values at the 0.005 level and 14 test periods, the probability of one or more type I errors, i.e., false peaks for which Qp exceeds x 2 when the data are actually random, is approximately 0.07 (Sokolove and Bushell, 1978). Since no Qp value exceeds criterion, the data from experiment 1 do not support the conclusion that the peaks in daily molting frequency are periodic. Experiment 1 was too brief to permit testing for possible periods longer than 16 days, and the later experiments were too brief for any periodicity analyses. Effect of Animal Weight on Molt-Cycle Duration. In experiments 1 and 2, time between successive molts for a given animal (intermolt duration) was significantly correlated with its initial weight (r = 0.33, 71 df, andr = 0.54, 64 df, respectively; P < 0.01). In experiment 1 the intermolt duration for the hypothetical animal of mean weight (0.65 g) was 14.6 days and each additional gram of weight lengthened intermolt duration by approximately 3 days (leastsquares regression). For experiment 2, comparable values were 0.50 g (mean weight), l l.1 days, (mean intermolt) and 6 days (slope). No obvious explanation accounts for the difference in regression slopes between the two experiments. Experiment 3 was too brief for analysis. Although animals of the same weight tend to molt with equal frequency, molting peaks did not appear to reflect simultaneous molting in groups of like-size animals. The weights of animals that molted in five large, welldefined peaks in experiments 1 and 2 (open circles, Figure 2) were compared by X2 goodness of fit to the weight distributions of all animals used for those
PARTIAL MOLTING SYNCHRONY IN GIANT MALAYSIAN PRAWN
493
experiments. The probabilities that animal weights in the sample peaks could have been random samples of the overall weight distributions ranged from 0.3 to 0.9 (mean P > 0.5). Effect of Peaks on Intermolt Duration. The fact that intermolt duration and weight were strongly correlated in both experiments provided a method for predicting intermolt duration: for each animal that molted at least twice, an expected intermolt duration was generated from the appropriate regression equation. Actual intermolt durations for those animals were then compared to the predicted values, and animals were scored as "late" or "early" molters, accordingly. For each day on which second (or third) molts occurred, the total number of molting animals was counted. For experiment 1 the mean for that count was 4.5 animals; for experiment 2, 4.9 animals. Days on which four or fewer animals molted were therefore scored as "few"; those on which five or more animals molted were termed "many." Contingency tables (Table 1) were then constructed for each experiment based on those scored attributes. For both experiments the two attributes were significantly nonindependent (xzc = 4.1 and 5.2, respectively; P < 0.05), and the deviations from random expectations were in the predicted direction. In general, animals molting with few others molted later than predicted on the basis of weight, and animals molting with many others (that is, in peaks) molted sooner than expected. Effect of Water Exchange on Molting Frequency. To examine the possibility that a molt-accelerating factor accumulated in experimental water supplies, the water recirculated in experiment l was replaced at day 33. Potential long-term effects of that manipulation were detected by computing an 1 l-day moving average for daily molt percentage. The smoothed data are shown in Figure 4. In that figure open circles represent l l-day means that include only days before the water change, half-filled circles indicate means that include progressively more postchange days, and closed circles indicate means of solely postchange days. Smoothed molt frequency appears to rise slightly (although not significantly) until the water change. At day 29, as the TABLE l. NUMBER OF SECOND AND THIRD MOLTS IN EXPERIMENTS 1 AND 2, CLASSIFIED ACCORDING TO DURATION OF PRECEDING INTERMOLT PERIODS AND NUMBER OF ANIMALS MOLTING ON THOSE DAYS.
Molt timing Number of animals molting
Late
Early
Late
Early
Few Many
21 8
19 23
21 9
14 22
Experiment 1
Experiment 2
494
HOWE O
OQ 00000 O0
000000000 O00
0 .~.
9
~ l ) . t l .D
o E
Q" , l " ~..
9 ~O
f0
2'0
3'o
4'o
-
Central day of ll-day means
FIG. 4. Eleven-day moving averages of molting percentages in experiment 1, before (open circles), during (half-filled circles), and after (closed circles) a water change. The dashed line represents the least-squares fit to the half-filled circles. moving average begins to include postchange days, molt frequency drops markedly. A line fitted by least squares to the half-filled circles has a significantly negative slope (Student's t test, two-tailed, P < 0.01). By the seventh day after the water change, smoothed molt frequency begins to rise. Spatial Aggregation in Molting Peaks. The results of experiments 2 and 3 were examined for possible spatial aggregation in the locations of molting animals. The fiat arrays of chambers used in those experiments rested on a shallow plenum (Figure 1) from which water was removed at a maximum measured rate of 10 liters/rain. That rate of pumping resulted in a net flow of water downward through the chambers at a calculated velocity of from 5 cm/min when filters were clean to 3 cm/min just before the filter elements were replaced. Despite many regularly spaced drains from the plenum, that slow flow of water was not laminar, that is, water flowed laterally between chambers as well as vertically through each chamber. The magnitude of that effect was checked in a series of dye experiments. With the recirculating pump operating at 10 liters / rain, a small volume of dye solution was released into a selected chamber. Water samples (0.5 ml) were then removed from that cell and from neighboring cells at 30-sec intervals, and dye concentrations were measured photometrically. Dye concentrations were highest in nearby cells at the first sampling time (+30 sec), but even 1 min after dye release, concentrations in certain nearby cells remained well above those in more distant cells. Figure 5a shows dye concentrations recorded 60 sec after release as a percentage of the initial concentration in the central cell for two typical experiments. Two results are worthy of mention. First, dye released into corner cells dissipated more slowly than from more central cells. Second, as the geometry of the chamber array suggests, water flow between cells in the same column (centers 4 cm apart) is much greater than between cells in the same row (10 cm apart). By 3 min after dye release in most experiments, the dye was uniformly distributed in all cells in the array.
P A R T I A L M O L T I N G S Y N C H R O N Y IN G I A N T M A L A Y S I A N P R A W N
495
Based on the results of the dye experiments, an algorithm was developed to test the results of experiments 2 and 3 for spatial aggregation (autocorrelation) of molting within columns of chambers. That algorithm was modeled after the approach of Sokal and Oden (1978) and had the following properties: 1. In order to accumulate sufficient molts for each autocorrelation analysis, three successive days of molting records were combined into a single chamber matrix. 2. In the experiments analyzed, an intermolt period shorter than 7 days occurred only once. Accordingly, each calculation excluded animals that had molted six or fewer days before the end of the 3-day period. 3. Consistent with the dye experiments, an instance of molting in two adjoining column cells was assigned a high relative weight (1 join), and molting in two cells in the same column that were separated by a single cell was assigned a lower weight (1 / 2join). No other spatial relationship between cells in which molting occurred was given weight The number of joins expected with this weighting rule, assuming random locations for molts, was computed by the method of Sokal and Oden (1978).
