The Growth and Life-History Strategy of the SoftShell Clam , Mya arenaria

The growth and life-history strategy of the soft-shell clam, Mya arenaria were analyzed through the exanination of 25 populations spread throughout the species' geographic range. Growth rate was determined by length-frequency analysis and was analyzed using Gallucci and Quinn's~ parameter of the von Bertalanffy equation. Principal components analysis (PCA) was used to analyze environmental variability. Variations in growth were related to environmental differences. Growth varied inversely with latitude. This relationship was due mostly to temperature differences. PCA indicated that several environmental factors varied systematically with latitude. Growth was found to be negatively correlated to the sil tiness of the sediment and to the concentration of sedimentary hydrocarbons. Distinct shifts in growth rate occurred when environmental quality was significantly altered by a discrete pollution event. Growth was reduced in response to heavy metal effluents and oil spills. Fairly rapid r .ecovery followed heavy metal pollution abatement. Recovery at oil spill sites was variable but slow. The relationships between various life-history parameters were analyzed to elucidate the species' modes of adaptation and life-history strategy. Within the context of 1 ati tud in al variations in the environment an association was found between decreasing latitude and: faster growth, greater variation in juvenile mortality, larger maturation size, larger egg size, lower egg density and decreasing longevity. These relationships indicated that~· arenaria follows a bet-hedging life-history strategy. On a local scale a positive relationship was found between larger egg size and greater egg density. This trend appeared to be associated with the condition of the population and may represent a physiological response to local env ironm en tal conditions. The proportion of females (52%) was found to be significantly greater than the proportion of males. No evidence of a mechanism behind this disparity was detected. PREFACE This thesis was prepared according to manuscript format. Three manuscripts are presented and each is written in a style appropriate for submition to a particular scientific journal. The first manuscript will be submitted to Marine Biology; the second manuscript has been submitted to the Proceedings of the National Shellfisheries Association; and the third manuscript will be submitted to Oecologia. The thesis as a whole constitutes an analysis of data collected during a study of neoplasia in Mya arenaria and pollution in the environment. Two limitations were encountered. First, because the analysis was designed after data collection many of the measurements in the raw data were not entirely suited for their application in this study. The problems included no or insufficient data and too much variability in the measurements. These problems could have been solved at the time of sampling. Second, no attempt was made to investigate and support conclusions based on the data analysis througb further sampling or experimentation. Although such work would be a welcome addition, it is believed that this work would constitiute a thesis unto itself. It is felt that the present analysis can stand by itself with the questions remaining to be answered serving as an incentive for further study.

Principal components analysis (PCA) was used to analyze environmental variability. Variations in growth were related to environmental differences. Growth varied inversely with latitude. This relationship was due mostly to temperature differences. PCA indicated that several environmental factors varied systematically with latitude. Growth was found to be negatively correlated to the sil tiness of the sediment and to the concentration of sedimentary hydrocarbons. Distinct shifts in growth rate occurred when environmental quality was significantly altered by a discrete pollution event. Growth was reduced in response to heavy metal effluents and oil spills. Fairly rapid r .ecovery followed heavy metal pollution abatement. Recovery at oil spill sites was variable but slow.
The relationships between various life-history parameters were analyzed to elucidate the species' modes of adaptation and life-history strategy. Within the context of 1 ati tud in al variations in the environment an association was found between decreasing latitude and: faster growth, greater variation in juvenile mortality, larger maturation size, larger egg size, lower egg density and decreasing longevity. These relationships indicated that~· arenaria follows a bet-hedging life-history strategy. On a local scale a positive relationship was found between larger egg size and greater egg density. This trend appeared to be associated with the condition of the population and may represent a physiological response to local env ironm en tal conditions. The proportion of females (52%) was found to be significantly greater than the proportion of males. No evidence of a mechanism behind this disparity was detected.

PREFACE
This thesis was prepared according to manuscript format.
Three manuscripts are presented and each is written in a style appropriate for submition to a particular scientific journal. The first manuscript will be submitted to Marine Biology; the second manuscript has been submitted to the Proceedings of the National Shellfisheries Association; and the third manuscript will be submitted to Oecologia.
The thesis as a whole constitutes an analysis of data collected during a study of neoplasia in Mya arenaria and pollution in the environment. Two limitations were encountered. First, because the analysis was designed after data collection many of the measurements in the raw data were not entirely suited for their application in this study. The problems included no or insufficient data and too much variability in the measurements. These problems could have been solved at the time of sampling. Second, no attempt was made to investigate and support conclusions based on the data analysis througb further sampling or experimentation. Although such work would be a welcome addition, it is believed that this work would constitiute a thesis unto itself. It is felt that the present analysis can stand by itself with the questions remaining to be answered serving as an incentive for further study. The growth of Mya arenaria, the soft-shell clam, has been studied by many investigators (Wilton and Wilton, 1929;Newcombe, 1936;Swan, 1952;Brouseau, 1979; and others) and much work has been done in assessing the importance of various environmental factors to the growth process. These factors include water current and quality, food, temperature, salinity, various edaphic parameters, and pollution. In the past, investigators were obliged to study these factors individually even though i t was realized that many were interrelated . Researchers often disagreed on the relative importance of each of these factors.
With the advent of high-speed computers and the develoµnent of more sophisticated analytical techniques it is now possible to efficiently analyze complex ecological systems involving multiple factors.
The purpose of this study was to investigate the factors contributing to variations in the growth rate of soft clam populations. Principal components analysis was chosen for the mul tiv ar iate analysis of the environmental data and the von Bertalanffy model was chosen for the analysis of growth using the recently introduced growth parameter w of Gallucci and Quinn (1979).
This study represents one of the first applications of w for the investigation of growth variations. Of specific interest was a methodology which could establish the presence or absence of a latitudinal trend in growth and the factors responsible for it, since a definite relationship had yet to be demonstrated (Brouseau, 1979).

MATERIALS AND METHODS
The growth of Mya arenaria was studied at 25 sites located along the east coast of North America from Maryland to Nova Scotia ( Figure 1). The sites were initially chosen and sampled as part of a study investigating the relationship between environmental pollution and neopl asia in Mya arenaria  and as sue h they varied greatly in-their environmental qualtiy. Estimates of the following environmental parameters were obtained for each site: salinity, tidal position, tidal range, average annual temperature, sedimentary grain size, dispersion and skewness of grain sizes, percent silt-clay, percent organic matter, and total sedimentary hydrocarbons.
Salinity, at low tide, was measured using a refractometer; tidal position was estimated on a scale of 0-1 where O• subtidal (never exposed) and 1 • full exposure (never submerged). Estimates of the average annual temperature at each site were obtained from various literature sources; estimates of the tidal range were obtained from the National Ocean Survey ( 1 978).
Sediment samples were collected and analyzed to determine the amount of organic matter and the grain size distribution. Each 2 sample consisted of a composite of two surface cores (21 cm x 8 cm depth) which were kept frozen prior to analysis. Using a 74 µm mesh screen an aliquot of the sediment sample was wet sieved to separate the sand fraction from the silt-clay fraction ( Buchanann, 1971 ). The sand fraction was further analyzed by dry sieving (ASTM, 1962) usi.ng 7 screens (ranging from 4-0. 062 mm mesh) set at one phi The silt-clay fraction was analyzed using the hydrometer method (ASTM, 1962) with readings taken at 2, 15, 60 and 120 minutes. The particle size and the cumulative percent of the sample represented at eacb reading were tben calculated. The particle size distributions obtained from the analyses of the sand and silt-clay fractions were pooled and the cumulative frequency versus grain size (<j>) was plotted for each sample. From the graphs the following summary statistics were obtained: median grain size (M::l<J>), quartile deviation (QD<J> ), and skewness (Skq<J>) (Buchanann, 1971). The results were reported in phi notation rather than in millimeters as this scale is commonly used to describe grain size characteristics and because it allows for greater discrimination in the silt-clay range which may be more meaningful biologically.
A length-frequency histogram for each population was plotted at 1 lllll intervals and the modes on the resulting graph were broken down into a series of normal curves by the Petersen Method  using a Dupont 310 Curve Resolver. The curve resolver is an analog computer which allows one to break down a complex envelope into its basic components in a graphical fashion. The computer has ten function generator channels each capable of producing a normal curve on a cathode ray tube. The images of these curves can then be projected on to the 1 eng th-frequency histogram. The histogram can then be broken down left to right in the following manner. The first channel is switched on and the projected curve is positioned such that its location, width, and height correspond to the left edge of the histogram. The remainder of the histogram is then resolved by successively turning on channels and positioning the curves such that the envelope projected (formed by the summation of all the 'on' channels) matches the outline of the histogram. The optical output gives the observer immediate feedback, and repeated trials can be made quickly by varying the size, shape, position, and number of curves until it is felt that a reasonable 'fit' to the data has been obtained. At this point the output of each channel can be turned on and displayed independently, and its projection traced on the histogram. Figure 2 illustrates the results of this process. From the resulting graphs the mean and standard deviation of the normal curve representing each mode of the histogram can be obtained. The mean occurs at the peak and the standard deviation is the half-width at 61% of the height  Figure 2) . The curve resolver is also equipped with an integrater allowing one to determine the percentage of the whole sample under each component curve.
Each curve generated by the above procedure was assumed to represent the length distribution of a cohort. Several methods were used t-0 corroborate the ages assigned to each group. For samples from Maine, shell ring counts on a subsample of clams were used.
Other methods, used for all sites, included comparison of the data to published growth data, comparison of the data to adjacent areas, and inspection of the subsequent growth curves. The ages assigned were those of relative age rather than absolute age where the age beyond the last annual increment represented the fraction of expected yearly growth already obtained. For exanple, a clan in its fifth year, having reached 50% of its expected growth for that year would be considered 4. 5 years old. This process results in a smoother growth curve since it avoids the problems of seasonal variations in the growth rate which would otherwise necesitate the use of a more complex growth model (Cloern and Nichols, 1978).
Growth estimates for each population were derived using the age-length relationships developed by the above 1 ength-frequency analysis. Five populations, located at sites where an acute pollution event had occurred, were excluded from this process because reliable growth rates could not be obtained (see Appeldoorn, 1980). Growth was estimated by fitting a von Bertalanffy curve to the data. The growth curve is described by the equation: where t•time, L•length at time t, L 00 •maximum asymtotic length, K•growth constant, and t 0 •time when L•O. According to the methods of Gallucci and Quinn (1979) the model was fitted to the data by nonlinear least-squares regression using the NLIN procedure of SAS79 . The use of such a nonlinear regression procedure is adva11tageous since not only are estimates of K, L and t obtained but estimates of their asymtotic standard 00 0 errors are also available as well as the correlation coefficient of K and L, (corr(K,L )). Normal linear curve fitting techniques do not 00 co allow for estimates of the variance of the parameters and hence the parameters could not be compared statistically.
After obtaining estimates of the von Bertalanffy parameters the growth function was reparameterized to obtain the single growth parameter w of Gallucci and Quinn (1979). The parameter is estimated by w•K ·L 00 and its variance is given as: Var(K)+K 2 Var(L )+ 2KL corr(K,L ) Var(K) Var(L ). 00 00 00 co 00 Biologically, w can be interpreted as the instantaneous rate of growth at L . Using a single parameter to represent growth 0 simplifies the testing of growth differences between populations.
Normally when comparing two curves the null hypothesis is Ho: K 1 -K 2 and L 001 aI.. 002 • Interpretations become difficult if one equality is rejected and the other is not due to the negative correlation between L 00 and K, that is, L 00 and K represent some of the same growth properties. By representing growth with w one need deal with only one parameter thereby simplifying the comparisons. In addition w tends to be more robust than either K or L 00 to variations in the data (Gallucci and Quinn, 1979).
The environmental data listed in Table 1 (Morrison, 1971 ). These components can then be Communality -the amount of each variable's variation that is accounted for by the components.
Percent variance -the amount of variance in the observed variables accounted for by a component.
The analysis was run on the Pearson product-manent correlation matrix of the environmental parameters (to allow for standardization of the units of measurement) using the CORR, FACTOR and SCORE procedures of SAS79 . The factors obtained from PCA were then used to analyze variations in the growth rate w.
Variations in w were analyzed using a step-wise functional regression of log 10 (w) on the components generated by PCA. The analysis was made by calculating the functional regression of log 10 Cw) on Component 1, taking the residuals and regressing them on Component 2, and so on. The geometric mean functional regression was deemed more suitable than the usual predictive regression because of variability in both wand the components, small sample size, and uncertainties about the distribution of the data (Ricker, 1973).
Predictive regression yields a regression coefficient (slope) of b while functional regression yields a regression coefficient of v• b/r where r is the correlation coe ffic ien t. The standard error of v (SEv) is the same as the standard error of b (SEb). 95 % confidence limits on v are approximated by v+2SE (Ricker, 1973). v The logarithmic transformation of w was used because the plot of against Component 1 indicated a logarithmic relationship. Estimates 2 of b, r and SEb were obtained using the GLM procedure of SAS79 . These estimates were used to calculate v and its 95% confidence limits. The significance of the . egression is tested by seeing if the confidence limits bracket v•O.
If not, the null hypothesis Ho: v•O is rejected.

RESULTS
For each population the mean length at age as determined through length-frequency analysis is given in Table 2. A von Bertalan ffy growth mod el was then fitted to these data ex cl ud ing the 5 sites acutely affected by pollution. The model parameters determined by the nonlinear curve fitting process plus the growth parameter w, are presented in Table 3. Using the 95% confidence limits around w statistically significant growth differences become  Table   4, and the results of the PCA are shown in Table 5. In order to simplify the table those loadings less than 0.30 have been left out although all variables contribute to all components to some degree.
The first five components have been retained and account for 88% of the observed variation. Of these, the first three were exanined in Table 2. Mya a renar i a.
The a g e (years) , leng th ±1 standard devia tion (mm), and perc ent of the population for e ach year cla s s at eac h si t e as d e t e rmi ned b y l eng th-fre quency a n alysis. The sample si z e for each site is given in parentheses.       Table 6. The results of the analysis are given in Table 7 and Figures 4, 5 and 6. As expected, growth was found to be negatively correlated with northness. The second regression showed a negative relationship between siltiness and growth. The last regression indicated that growth was negatively correlated with sedimentary hydrocarbons. Much of the deviation about this regression is due to the abnormally high values for Allen Harbor.

DISCUSSION
The use of many length-frequency analysis techniques for the determination of age structure and growth is not straightforward.  consider the lack of reproducibility as its major drawback. This results from the difficulty in properly distinguishing all the modes in a distribution, especially when multiple spawnings occur or sample size is low. As such, the results of analysis can depend, in large part, upon the experience of the observer. This is not only evident in the resolution of the distribution mixtures but also in the assignment of ages to the various modes. In this respect, however, even the more sophisticated curve resolving techniques are limited since independent estimates of the number of modes present and length at age are usually required.
However, length-frequency analysis bas proven itself to be a useful and informative technique and the results obtained here are generally reliable.  :r::   . Evidence for this trend in My:a arenaria is supplied by Newcombe (1936) who found an inverse relationship between growth rate and longevity. The determination of growth through length-frequency analysis is based on the assumption that there is little year to year variation in growing conditions.
Except in those polluted areas omitted from the analysis this assumption was reasonably approximated. The growth curves obtained are designed to estimate 'average' growth and in this sense they represent an integration of growth rates over slightly varing conditions.
The modeling of growth by the von Bertalanffy equation has been criticized for many reasons (see Roff, 1980). However, its general applicability and importance to many fishery models have led to its continued use. Dickey ( 1971)  is therefore not dependent upon possible inaccuracies in the assigmient of ages to modes of the length-frequency distribution.

32
The parameter w proved to be a suitable measure of growth in this study. By reducing the description of growth to a single parameter the discrimination and interpretation of growth rate differences was simplified. In addition, w easily lent itself to further analysis, using more sophisticated techniques, since its simplicity ma::l e it more tractable to manipulation and interpretation.
As with any parameter, the limits of ware dependent upon the quality of the orginal data. In the present study it was impossible to determine how much of the variation in w was due to inaccuracies in tbe data. However, after three regressions the variation in w due to errors of measurement was approaching a 'significant' proportion of the remaining total variation and further regressions would have been suspect.
The observed relationship between latitude and growth is not surprising, especially considering the rang _ e of temperatures reflected in the data. Increasing growth would be expected at higher temperatures due to temperature's direct effect on metabolism and length of the growing season (Brouseau, 1979). In addition, with increasing temperature Mya is found lower intertidally or even subtid ally, thereby increasing its daily feeding period. However, , Dow and Wallace (1961), Newcombe and Kessler (1936), and Swan (1952) have stated that local hydrologic and edaphic conditions are more important than temperature in affecting growtb; and previous studies have failed to firmly establish such a latitudinal relationship. Newcombe (1936) noticed growth differences between three widely separated sites which he attributed to temperature differences. Turner (1948) made a similar observation also based on three areas. Brouseau ( 1978) showed a tendency for populations in Massachusetts to grow faster than more northernly populations but the relationship was not definite. Each of these studies suffered from t\ol) deficiencies: (1) small sample size and (2) limited geographical range. Under these conditions variations in growth due to local conditions can mask any latitudinal trends. The wide variation of points around the regression line in Figure 3 gives evidence for this.
Principal components analysis proved to be useful in the further detailed analysis of growth. PCA produced a smaller number of meaningful variables, which were easily interpretable, and when used for further analysis produced lucid and rational results.
The first component, northness, correlated well with growth.
Again this is not surprising. Temperature had the highest loading for this component and its influence on growth has already been discussed. Tidal position, as mentioned, is a secondary result of temperature variations. As wi tb many factors, the components produced by PCA represent an integration of effects and the correlation between growth and northness may also be dependent upon factors other than temperature. The increase in tidal range with northness, due to the large tides of the Gulf of Maine and the Bay of Fundy represents to some degree an increase in tidal current.  considered current the most important factor pertaining to growth. Current, in turn, can al so influence ed aphic conditions. Hence a coarse but variable grain size distribution is associated with northness. The coarser sediments found toward the north are al so a reflection of their glacial orig ins; and they are beneficial in their own right since they allow for ample water percolation, drainage and exchange (Dow and Wallace, 1961;Swan, 1952). Within northness, then, there are two sets of opposing conditions which influence growth. Temperature is positively associated with growth while current and sediment characteristics are negatively associated with growth. Since northness is itself negatively correlated with growth it must be concluded that the effects of temperature are overriding and dominant.
The relationship between northness and w was logarithmic in nature. Since temperature daninates the relationship this result is reasonable. The effect of temperature on metabolic systems is known to be exponential in nature (Gunter, 1957). What requires explanation is the linear relationship between growth and latitude.
It is felt that the addition of the two Maryland sites in the growth-latitude graph unjustly extends the -relationship since the large change in latitude may not be accompanied by equally large changes in ecological conditions. In this case northness should be a more proper representation of the geographical variation in conditions than is latitude.
The effect of the second component, sil tiness, on growth al so represents an integration of processes. In small quantities silt and clay help to stablize surface sediments (Kellogg, 1905) but in large quantities they become detrimental. Sanders (1958) found that the distribution of filter feeders, in general, was limited by the silt-clay content of the sediment. Studies with Mya have shown that excessive .siltation can reduce feeding, through clogging of the gills , or lead to complete .smothering and death (Wilton and Wilton, 192 9;Dow and Wal ace, 1961 ) • Silty .sed iment.s al so tend to be fairly consolidated and reduced growth has been ob.served in such .sediments (Swan, 1952;Dow and Wallace, 1961 ). Silt can easily become trapped between the shell and mantle, with a con.sequent interruption of growth . Small grain size is al.so an indication of a poor current regime, it.self a contributing factor to reduced growth. The negative correlation found between .sil tine.s.s and growth is, therefore, logical and con.si.stant with previous reports.
Sedimentary hydrocarbons, the third component derived from the PCA, was al.so negatively associated with growth. Many studies have shown that the growth of Mya is adversely affected by the presence of petroleum hydrocarbons (Dow, 1975;Dow and Hur.st, 1975; Gil fill an Appeldoorn, 1980  growth. An attempt is made to relate the severity and persistence of the pollution effect on growth to the degree of deflection in the age-length curve. A method whereby pre-pollution growth can be estimated is presented and applied to two populations.

INTRODUCTION
The need for more information on the effects of discharging pollutants into marine ecosystems has long been recognized. However, only recently has significant progress been made toward this end.
Early investigators studied only acute 1 ethal effects, and the variability in the number and reliability of the methods involved led to much confusion . With improving methodology there has been an increase in the interest of investigating chronic and sublethal effects . This has been coupled with the recognition that research should be concerned with effects on population processes rather than on individuals . Notable studies involving the long term monitoring of populations following a pollution event are those of the West Falmouth oil spill, the Chedabucto Bay oil spill, and studies of pulp mill effects in Sweeden (Rosenberg, 1976).
One major problem in studying the effects of sudden environmental cbanges is the availability of reliable control data from either measurements made prior to the change or from a suitable control area.
Recently an investigation into the status of soft-shell clan The purpose of this paper is primarily to present age-length curves of soft clan populations from the sample sites where a pollution event occurred. Based on a few assumptions these curves can be used to represent growth. It will be shown that a sudden change in the env ironm en tal qual ti y resulting from the onset or abatement of pollution is reflected by a shift in the age-length curve. In addition, a method will be presented whereby growth prior to a pollution event may be estimated.

METHODS
The growth of clans was studied at six sites where a discrete pollution incident (either onset or abatement) occurred. Five of the sites were affected by spills of various types of oil. The sixth site was exposed to the effluent from an intertidal heavy metals mine. Table 1 lists the sampling sites and gives a brief characterization of each area. Initial estimates for the extent of pollution are given in Table 2.
Each site was sampled once with the exception of Searsport which was sampled quarterly in 1977 and 1978. Clams were dug using a standard clam hoe. All clans excavated were measured for length to the nearest millimeter using vernier calipers. For Searsport, length data for clams setting after the spill were obtained from Dow ( 1 978, Coarse sand to very consolidated sand with rock & she 11 in a patchy distribution 1 All hydrocarbon concentrations are for the sediment (µg/g dry weight by GC). Samples were taken at the time of clam collection. Source : C. Brown (personal corrvnun i cation) . . 2 originally reported as JP-5 (Gilfillan et{!_. , 1977) .  (1977) reported that some of the oil was still present in certain ma rsh areas, but little evidence of oil was found at sampling.
A sediment concentration of 590 µg/g (dry wt.) was reported at Site II of the WHOI studies. This site is located up a tidal creek (Wild Harbor River) just below the present sampl i ng location. Later concentrations steadily declined reaching 1/3 of the initial value after 2 years (Sanders, 1970).
The site, located near the inner n10st of 3 main culverts where oil entered the cove, iS equivalent to Station 12 of Mayo et al. (1975) who found an initial concentration of 58 µg/g (dry wt . ). At two adjacent stations (1 1 and 13) sAmpled the following year the concentration had increased by an order of magnitude. Further contamination was due to oil leaching from saturated sediments upslope.
Crude oil and detergents init ially covered the flat . Li ttle evidence of contaminat ion was found at san¥J1ing.
The lagoon was initially covered by 30 cm of oil . Much of the oil has remained and is periodically .remobil i zed. Measurements taken 6 years l ater by Thomas (1978) and Gilfillan & Vandermeulen (1978) showed average concentrations at thousands of µg/g (dry wt . ). However , these measurements tended to vary by four orders of magnitude . At sampling oil was still abundant, and a sl i ck would form on any depression made on the flat.
During mine operat ions record levels of 8 metal s (Mn 341, Cd 1. 7, Cr 29.5, Ni 4.1, Zn 195 , Pb 55, Fe 2471, Co 1. 5 ppm) and extremely high Cu 1eve1 s were found in soft cl ams near the outflow. Levels typically ranged 1 t o 2 orders of magnitude above those found in control clams (Dow & Hurst, 1972) .
.i::. length-frequencies were plotted a 1 mm intervals and the modes on the resulting graph were broken down into a series of normal curves by inspection (Peterson Method)  using a lA!pont 310 Curve Resolver. The curve resolver is an analog computer which allows the investigator to break down a complex envelope into its basic components (in this case normal curves) in a graphical fashion. It utilizes function generator channels (10) capable of generating normal curves on a cathode ray tube. The images are then projected on to the length-frequency histogram drawn for each population. The histogram is broken down from left to right (young to old) in the following manner. One channel is swi tc bed on and the projected curve is positioned such tbat its location, width, and height correspond to the left edge of the histogram. The remainder of the histogram is then resolved by successively turning on the channels and positioning them such that the envelope projected (formed by the summation of the outputs of all the 'on' channels) matched the outline of the histogram. The o_ptical output gives the observer immediate feedback, ·and repeated trials can be made quickly by varying the size, shape, position, and number of curves until it is felt that a reasonable 'fit' to the data has been obtained. At this point the output of each channel can be turned on and displayed independently, and its projection can be traced on the histogram.
The result of this process is exanplified in Figure 1. From the resulting graphs the mean and standard deviation of each distribution can be obtained . The mean occurs at the peak and the standard deviation is the half width at 61 % of the height (See curve 4, Figure 1). The curve resolver is also equipped Each curve generated by the above procedure was assumed to represent the 1 eng th distribution of a cohort. Several methods were used to corroberate the ages assigned to each group. For the Searsport sample shell ring counts on a subsample of clcrns were used as well as following changes in the modal pattern of the length-frequency distribution over time. Other methods, used for all samples, included comparison of the data to published growth data, comparison of the data to growth data from nearby areas (unpublished data) and through inspection of the subsequent curves. The ages used were those of relative age rather than absolute age. The time beyond the last yearly increment represents the percent of expected yearly growth already obtained. Hypothetically, if a clan first set in the beginning of April and was collected in November three years later its relative age would be 4 rather than 3. 6 because i t would no longer be expected to grow significantly during the rest of its third year; the size obtained by November would roughly equal its size at age 4. This process results in a smoother growth curve since it avoids the problems of seasonal variations in the growth rate which would otherwisw necessitate the use of a more complex growth model ( Cloern and Nie ho ls, 1978).
F'or three sites, West Falmouth, Searsport, and Janvrin Lagoon, a sufficient number of year classes were represented to allow a von Bertal anffy growth curve to be fitted to the data. Only po st-spill age classes were used to fit the curve which reduced the number of points available for the analysis. The growth curve was fitted by nonlinear regression according to Gallucci and CNinn ( 1 979) using the NLIN procedure of SAS 76 . This procedure yielded estimates of the parameters for the von Bertalanffy growth equation

LaL
(1-exp(-K(t-t )) where t•time, Lalength at time t, max o L a maximum asymptotic length, K•growth constant, and max t •time when L•O. 0 Using the calculated von Bertalanffy curve the growth rate prior to pollution was estimated. This analysis was based on the assumption that growth follows a fixed schedule or pattern. Growth prior to pollution may be different (i.e. have its own growth schedule) from growth after pollution. It was assumed that the rost-pollution growth schedule was adequately modeled by the calculated von Bertalanffy curve. The pre-pollution growth schedule was then approximated in the following manner. The length (L 1) of the last year class to set prior to pollution was found. Then the age correspoding to this length on the von Bertalanffy curve was determined. One year was subtracted from ttiis age and its corresponding length (L2) on the growth curve was determined. Next the length (L3) corresponding to an age equal to (age at L1)-1 was found. The difference between L2 and L3 represents the extra growth experienced by clans having one year's growth on the pre-pollution growth schedule. This difference was then added to the expected length at year one on the po st-pollution curve ( von Bertal anffy curve) to obtain the expected length at year one on the pre-pollution curve. The second point on the pre-pollution schedule was found by applying the above procedure to the year class that set two years prior to pollution. This process was repeated for all available pre-pollution year classes. An approximation of this tecbnique could be used when few post-pollution points exist by drawing an approximate growth curve by eye. However, it is felt that the variability introduced would reduce the meaningfulness of the results and no such approximations are attempted here.

RESULTS
For each area the mean length and standard deviation for each age group as obtained from the length-frequency analysis are shown in Table 3. These data are plotted in Figures 2-7. For West Falmouth, Searsport, and Janvrin Lagoon the calculated von Bertalanffy curve is also plotted. Table 4 describes the parameters for these curves. In addition, pre-pollution growth approximations for Searsport and Janvrin Lagoon are plotted. For the remaining three areas approximate curves have been drawn by eye to ::mooth out the age-length relationship and to accentuate its cb.ange following a pollution event.
The figures danonstrate that changes in the incidence of pollution are reflected by changes in the growth rate. Only West Falmouth fails to show a significant change. The breaks in the curves clearly indicate that pollution has an adverse effect on growth and they reflect the degree to which growth can be reduced.
Growth was severly affected at Searsport, Janvrin Lagoon, and Goose C-Ove. At Goose C-Ove growth improved following pollution abatement.
At West Falmouth t he lengths of the year classes existing prior to the spill fail to differ significantly from the lengths expected on the basis of po st-spill growth. In would appear that the spill had   The circles are as in Figure 1. The solid line is as in Figure 5.   Figure 2. This area was used as a control site by Thomas ( 1 978) and by Gil fill an and  in their studies of Chedabucto Bay. In the latter study it was reported that soft clam growth at Janvrin Lagoon and Potato Isl and were similar prior to the spill. The estimate of pre-spill growth calculated here agrees remarkably well with the age-length determinations for Potato Island.
The parameters of the von Bertalanffy curve for Searsport appear anomalous in comparison to the other values shown in Table 4.
This probably resulted from sampling errors (note the standard deviations in Table 3) associated with a small sample size (N•15) and from successive improvments in post-spill growing conditions (see discussion below). The 1 atter would tend to increase the initial slope of the age-length curve, thereby increasing K.

DISCUSSION
The problems inherently associated with the estimation of population age structure and growth through length-frequency analysis were reviewed by  and others. A reiteration of these problems does not seem necessary here. It should be pointed out, however, that the growth being measured is that for a cohort of the population and not of individuals (see pp 217-218). The difference between the two arises from the fact that the older modes in the length-frequency histogram are usually composed of slower growing individuals. It has been shown for fish that individuals which grow rapidly tend to mature earlier, become senile earlier, and die earlier than slow growing individuals . For M. arenaria in general an inverse relationship has been found between longevity and the rate of growth (Newcombe, 1936), i.e. older clans are slow growers. A good exanple of this has been shown by Dow ( 1978) for clans growing at Searsport.
As clams grow their burrow depth increases. Faster growing clans were penetrating the buried stratum of oil polluted sediment at an ealier age whereupon mortality occurred. Hence only the slower growing individuals survived and they now constitute the bulk of the older age groups in the population.
The assumption that clans grow according to a fixed schedule (especially after a pollution incident) is probably not valid. For exanple Dow (1978) has shown successive improvements in the growth of M. arenaria for each year class following the Sear sport oil spill. This is due both to the the further weathering of the oil and through the further deposition of clean sediment over the oil contaminated sediment. However, at Sear sport and Janvrin Lagoon post-spill recovery has been slow enough to allow the use of the von Bertalanffy curve to generate pre-spill growth estimates. Since only approximate growth estimates have been obtained no effort was made to apply rigorous statistical analysis to the data. It is sufficient here only to illustrate the apparent gross responses of population growth.
The results of this study show that there is a response in the growth rate to environmental changes due to pollution. Thi:s response is characterized by a noticeable break in the age-length curve. In each case the onset of pollution was coupled with a reduction in growth. The exact mechanisms for the observed growth reductions at each site are unknown. The volune of 1 iterature on the effects of pollutants on marine organisms in general and on bivalves in particular is now vast but it is still difficult to relate specific effects in the 1 aboratory to responses observed in the field.
For M. arenaria other field studies have shown that the onset of oil pollution is generally followed by a reduction in growth and an increase in mortality. Dow (1975) found a 65% reduction in the annual growth rate of clans transplanted to a site polluted with Iranian crude oil. At Searsport he reported a reduction in the growth of so ft clans following the spill (Dow, 1978). That mortality at Searsport greatly increased when clans cane into direct contact with the oil seems to indicate the the oil has either a direct toxic effect or that it leads to ::mothering (Dow and Hurst, 1975;Dow 1978). ~othering was considered to be the main cause of the large soft clan mortality following the spill of Bunker C oil at Chedabucto Bay (Thomas, 1973). Gilfillan and Vanderm~ulen (1978) found a reduced carbon flux in soft ·clans from Janvrin Lagoon as compared to Potato Island. This was coupled with a calculated reduction in the rate of shell growth in Janvrin Lagoon clans following the spill. In an earlier study Gilfillan et al. (1975) found a 50% reduction in the carbon flux of soft clans polluted by No. 6 fuel oil. They concluded that for bivalves a reduction in the assimilation ratio was a general response to environmental stress which could be triggered by a number of factors including pollution.
For West Falmouth the age-length curve failed to show a break at the time of the spill. There are two possible explanations for this. The first explanation is sampling error. Because sampling took place 8 years after the spill it is possible that the age of the sample masked any true effect. Only 6% of the sample consisted of clans that had set prior to the spill. Such a small sample size could have led to underestimation of the mean lengths for each age class.
The second explanation is that the curve accurately reflects the spill' s true effect on growth. While this may be true studies made after the spill indicated initially severe effects. Bll!ller et al. (1970) reported large mortalities among the benthos, including shellfish, immediately following the spill. Site II was particularly devastated (Sanders, 1978). A hydrocarbon concentration of 69 µg/g (dry wt.) was found in oysters from the tidal creek (Bll!ller et al., 1970) one month after the spill. This value is above those reported for soft clams from other oil impacted sites .
It seems unlikely, then, that clan growth would have remained uneffected. If conditions improved, however, the effect might become unnoticeable. Sediment oil concentrations at Site II steadily decreased over time reaching 140 µgig after two years. This is only twice the level reported for indigenous sedimentary hydrocarbons within the area (Bl l!ll er and Sass, 1972). The degree of this decrease may be attributable to sediment characteristics at the sampling site.
Loose coarse, shifting sand should facilitate rapid depuration or burial of the oil. As a result growth may only have been affected during the first couple of years. Significantly improving conditions invalidate the assumption of a fixed post-spill growth schedule.
Hence the von Bertalanffy curve cannot be expected to approximate the growth of an affected popul at.ion. Given the sampling problems mentioned above and the 8 year time 1 age bet ween sampling and the spill any initial effect on growth would now be undetaectable by the methods used. This situation differes both from Bourne and Perry, sites where little oil was found but which were sampled shortly after the spill, and from Searsport and Janvrin Lagoon, site:s sampled several years after contamination but which :still contained enough oil to adversely affect growth.
The mining operations at Goose Cove could have led to a reduction in growth through three mechanisms: sil tat.ion, food destruction, and direct heavy metal toxicity. Dow and Hur:st (1972) suggested that much of the damage done by the mining operations was due to heavy sil tat.ion and mothering. This would definitely interfere with feeding by clogging the clams' filtering apparatus.
They also reported that the mine effluent wa:s highly toxic to phytoplankton, the main food source for soft clams, and that this could contribute to malnutrition and starv a.tion. Eisler ( 1 977) reported that _!:!. arenaria was fairly su:sceptable to heavy metal contamination. Many of the metal concetrations reported by Dow and Hur:st (1972) were higher than the lethally toxic concentrations determined in bioasssay studies dealing with pure (Eisler and Hennekey, 1977) and mixed (Eisler, 1977) metal :solutions.
The concentrations of metals in :soft clams at Goose Cove were still high at the time of sampling, four years after mining operations ceased (L. Fink, personal communication). From the graph in Figure 7 it can be seen that growth improved following pollution abatement, although it did not seem to have returned to its original rate. If starvation and smothering were the major contributing factors to reduced growth then growth should have dramatically improved upon the cessation of mining activities. This may have been the case. However, the exact degree of recovery is difficult to gauge in this case due to the variability of the data. These observations would lead one to conclude that smothering and starvation were major factors working in conjunction with direct toxicity to reduce growth during the period of mining operations. In addition, it appears that to some extent growth was still being adversely affected at the time of sampling possibly due to direct toxic effects.
The pronounced growth reduction at Goose Cove can be attributed to the variety of ways in which the mining effluent affected the clans and to the constant output of effluent during the period of mine operation. Once mining operations ceased recovery was fairly rapid. This is in contrast to recovery at oil polluted sites and reflects the persistence of oil in the .sediment, and the different mechanisms by which oil and mining effluent affect clans.
Major contributing factors toward reduced growth at Goose Cove such as siltation and food reduction were removed after mining operations ceased. On the other hand, oil itself is a major factor in growth reduction. Oil can be taken up through the siphons (Fong, 1976) and the leaching of oil from saturated sediments following a spill can result in a contaminated water supply for an extended period of time (Mayo~ al., 1975). Because oil can be detrimental upon contact (Dow, 1978), the effects of a spill can persist after burial of the Oiled sediment. In addition, Vandermeulen (1977) and Vandermeulen and Penrose ( 1 CJ78) found that significant quantities (40%) of oil remained in J:Olluted soft clans following three month exJX)sure to clean water. All these factors contribute to the persistance of an effect following initial hydrocarbon contamination.
In spite of these tendencies, some areas showed signs of recovery. No break in the age-length curve was observed at West Falmouth as discussed earlier. Bourne seems to be a similar case.
Little evidence of oil was found at the time of sampling, and the break in the curve ( Figure 5) appears 1 ike a short depression in an otherwise normal looking growth curve. This would seem to indicate that growth was disrupted only for a short period of time, on the order of a few years.
The techniques used here are considered valuable in assessing pollution effects. Primarily they are useful in detecting gross resJX)nses in growth due to changes in environmental quality and they allow one to estimate pre-JX)llution growth. This is helpful since measurements taken prior to a pollution evel'lt are rare and usually fortuitous. A nunber of studies have used shell growth bands to monitor, in detail, subtle environmental changes (e.g. Kennish and Olsson, 1975). However, these techniques are limited in their application and the methods are involved and costly. The techniques used here sacrifice detail but have more general applicability. For exanple they allow one to study populations of~· arenaria south of Cape Cod where annual ring formation is unreliable . The resJX)nses observed only directly reflect the effects on growth. They do not directly reflect changes in mortality, settl anent, or JX)pul ation age structure. As was observed at Searsport, however, continued size-dependent mortality may indirectly affect the resulting growth curve.

ACKNOWLEDGEMENTS 69
The author wishes to express his gratitude to those people and agencies who assisted in clan collection, in particular R The proportion of females (52%) was found to be significantly greater than the proportion of males. No evidence of a mechanism behind this disparity was detected, and its cause remains enigmatic. A major factor limiting the number of empirical studies is the difficulty in measuring the necessary population parameters, especially those involving reproduction. As a consequence, only a few studies have been conducted on benthic invertebrates and these have primarily dealt with brooding species . . This work is exemplified by an excellent study on the fresbwater mussel Anodonta piscinalis by Haukioja and Hakala (1979). Much less work has been done on broad cast spawner s. Age-fecundity rel ationsbips have been reported for the mussel Mytilus edulis (Thompson, 1979;Bayne, 1976) and for the so ft-shell clan Mya arenaria (Brouse au, 1978b). Further work on the fecundity and mortality schedules of Mya arenaria bas been attempted (Brouseau, 1978a , 1979) and further understanding of the species may prove useful to its management.
The purpose of this study is to establish relationships between various life-history parameters in Mya arenaria and to use those relationships to elucidate its life-history strategy. The parameters are estimated from data collected on 25 populations of Mya arenaria. These populations were initially sampled for a study on neoplasia and pollution . As such, the sampling design and the data obtained were not en ti rely suited for the present study. Samples were collected at different times and under different conditions. Many of the estimates made here are crude approximations and estimates could not be made on all parameters for all populations. This variability makes individual comparisons tenuous.
However overall trends can be educed. It has been shown that a considerable and consistant change in environmental parameters (temperature, tidal position, sediment grain-size distribution) exists with changing latitude (Appeldoorn, 1980). This study deals mainly with the response of Mya arenaria to this latitudinal gradient in its environment.

Data Collection
Samples of soft-shell clams were collected at 25 sites along the northeast coast of North America (Table 1 and Figure 1). All individuals were measured for shell length, weighed, and sectioned for histological analysis. The sections were prepared using standard histological techniques (Brown et al., 1977), cut at 6 m, and stained with hematoxylin and eosin. These data were then analyzed to obtain estimates of some population life-history parameters.

Calculation of the Variation in Juvenile Mortality (VJM)
To adequately discuss variations in population parameters it is necessary to have some estimate of the stability of the various environments. Environmental stability can be inferred by a low variability in reproductive success or juvenile mortality (Stearns, 1976). To estimate the variation in juvenile mortality an adaptation of catch-curve analysis    (Haukioja and Hakala, 1978). In a regression of Log (N) on age e the standard deviation of the residuals is then a measure of the  value would indicate a lower standard deviation than expected and hence a low degree of variability in juvenile mortality. The data used in this analysis was taken from a previous study on growth (Appeldoorn, 1980). In that study the age structure of each population was determined using length frequency analysis.

Calculation of Reproductive Parameters
Central to the study of life-history strategies is the concept of reproductive effort, defined as "the proportion of resources diverted to reproduction, summed over the time interval in question" (Stearns, 1976). In the present study it was impossible to estimate true reproductive effort. However, some measurements were made that could indicate relative reproductive differences between populations.
Using histological sections 20 females from each population were randanly selected and exanined. Each individual was classified as to reproductive stage using the seven catagories of  and the terminology of Brouseau ( 1 978b) substituting "developing" for "active" and "indifferent" for "inactive". The seven catagories are: indifferent, early developing, mid-developing, late developing, ripe, partially spawned, and spent. As with previous studies (Ropes and Stickney, 1965;Brousseau, 1978b) the continuous nature of ganetogenic activity often led to classification problems.
Frequently clans at the spent stage contained new ova already developing for a _possible second spawn.
Egg diameters were measured on five ova per individual using an occular micrometer. These data were converted to the average egg size at each observed reproductive stage.
Triplicate egg density counts were made for each individual.
Egg density was defined as the number of eggs (with nucleus 2 visible) /microscope field • • 29 mm . Al though based on a single section, these measurements were assumed to be reliable since Brousseau ( 1 978b Growth rate estimates were taken as the parameter w (Gallucci and Quinn, 1979) of tbe von Bertal anffy growth mod el. These estimates were obtained from a previous detailed analysis of growth in these _populations (Appeldoorn, 1980). Longevity was crudely estimated by calculating the age of the largest individual collected at each site.
The size of sexual maturation was estimated as the size of the smallest female collected at each site. Size was measured instead of age because sexual maturation occurs when a clam reaches a certain size rather than a certain age Coe and Turner, 1938;. In addition, the concept of increased cost of early maturation is easier to comprehend from the standpoint of size, especially given the wide variations in growth that occur between M.
For some sites estimates for the pathological condition of the clam population were available. These were taken as the disease severity index (DSI) of Walker et al. (in press).

Calculation of Regressions
Geometric mean functional regressions were used in all cases due to the natural variability of both the x-and y-variates, and the small sample size (Ricker, 1973). In the regressions between population parameters no single parameter could be considered as the independent variable. As a standard, the growth parameter w was somewhat arbitrarily chosen as the x-variate in these regressions.
The rest of the parameters were then regressed against w. All regressions were run using the GLM procedure of SAS79 . The predictive regression coefficient (slope) b was converted to the functional regression coefficient v through v•b/r where r is the correlation coefficient. The significance of the regression can be tested by seeing if the 95% convidence interval around v ( • v +2 standard errors) brackets zero (Ricker, 1973).

RESULTS
Estimates of all the parameters used in the regressions are shown in Table 1.

Sex Proportions
From Table  it can be seen that in the majority of the populations females outnumbered males. The overall average proportion of females was 52. 3%. The null hypothesis of equal sex ratios among all samples was tested using the Wilcoxon sign rank test . The test showed that females * significantly outnumbered males (T •2. 874, P""· 004) so the null hypothesis was rejected.
VJM, Longevity, Maturation vs w Table 2 shows the results of the regression of VJM, longevity, and the size of maturation against growth. The regressions indicated that longevity was negatively related to gr?wth while the size of maturation and variation in juvenile mortality were positively associated with growth. The size of maturation showed the weakest correlation. This might be expected since the realized size of maturation depends upon the present nutritional condition of the population (Coe and Turner, 1938 individuals in this stage than in any other. The data in Table 1 gives the average values for the late developing stage individuals.
The regressions show a positive relationship between egg size and growth, and a strong negative relationship between egg density and growth.
Egg Diameter vs Egg Density Table 3 illustrates the relationship between egg diameter and egg density in a series of pair-wise comparisons. The sample pairs consist of two sites sampled proximally in both time and space.
Hence any apparant trends should represent differences due to variations in local conditions. The table shows evidence of a positive relationship between egg diameter and egg density. This relationship was tested using a binomial test , and was found to be significant at p•0.07. At only 2 of the 7 sites for which comparisons of the pathological condition could be made did a poorer condition associate with a 1 arger egg density and diameter. This relationship was similarly tested and found to become significant at p•O. 22.

Sex Proportions
In previous studies on Mya arenaria Shaw ( 1 965 (Pelseneer, 1926). Normally in these species the disparity is not present in the young but increases with age. It is assumed to be due to a higher mortality rate for the males (Pelseneer, 1926). In the present study, populations with similar growth rates were pooled, and trends in the sex ratio were examined. No trend with size was apparent. The cause for this disparity remains enigmatic.

Life-History Strategy
Previous empirical and theoretical studies have identified two opposing life-history strategies: r-and K-selection, and bet-hedging (Stearns, 1976). Under conditions where environmental variability and density independent mortality affect the young more than the adults these two strategies yield contradictory predictions.
In a fluctuating environment r-and K-selection predicts rapid development, early maturation, semelpar ity, 1 arge reproductive effort, more young, and shorter life; bet-hedging predicts just the opposite. Although con tr ad ic tary results are predicted by the two theories it is often difficult to classify a particular species because of complicating factors (Stearns, 1976).
In a previous study on growth and environmental variation it was found that there existed strong and fairly consistant environmental changes with latitude and that growth was correlated to this latitudinal trend (Appeldoorn, 1980). Growth decreased toward the north. In the following discussion some simplifications are made. It is assumed that a decrease in latitude represents an increase in growth and all results will be discussed in a north-south context. This is done strictly for convience; to simplify the discussion of strategies exhibited by the observed trends in population parameters.
Mya arenaria produces vast numbers of planktotrophic larvae.
Estimates of its fecundity range from 120, 000/year by Brousseau (1978b) to 3 or 4 million/spawn 1 by  and 1to5 million/spawn by . This high fecundity is consistant with the strategy of larval dispersal . Concommitant with high fecundity is high mortality and as a species M. arenaria exhibits a basic pattern of high juvenile mortality and low adult mortality . TI-iis juvenile mortality is by and large density independent .
Overlying this basic pattern are two observable trends in For a 2. 5 inch clam Belding reported a fecundity of 4 million in the text, but 3 million in a summary of Mya' s life-history. Earlier publications of this work (Belding, 1907; gave no figures for absoluted fecundity. It remains unclear which figure is correct although the latter figure is most often quoted (e.g. Turner, 1948;Dow and Wallace, 1961;Brousseau, 1978b). mortality. Fir st, adult mortality increases toward the south. This is indicated by reduced longevity in the faster growing IX)pulations.
Evidence for this trend is also found in the literature, with a re!X)rted life span for ~ of 5 years in Chesapeake Bay (Pfitzenmeyer, 1972), 12 years in Massachusetts , and over 20 years in Nova Scotia . This trend is probably due to temperature effects. Mya arenaria is a boreal species (Laursen, 1966) and is under temperature stress in the southern part of its range (Pfitzenmeyer, 1972). Second, juvenile mortal tiy becomes more variable toward the south as evidenced by the i:;ositive correlation between w and VJM. TI-iis implies that the environment is more variable toward the south. In Chesapeake Bay Mya can be limited by high temperatures, low salinity, low d isolved oxygen (Pfitzenmeyer, 1972) and unsuitable substrate conditions (Pfitzenmeyer and Drobeck, 1963).  showed with Mytilus edilis than any delay in finding a suitable substrate for settlement may exhaust energy reserves and prevent normal metamorphosis; a problem compounded by rapid ·metabolism at high temperatures. In add it ion, with increased submergence, southern populations are exposed to more consistant predation pressure after settlement.
TI-1ese two opposing mortality trends would make any apriori prediction of a life-history strategy for Mya difficult. However, the trends observed in the present study reveal a consistant north-south pattern. (1) Growth rate increases. When the environment becomes more variable with respect to the juveniles the best strategy is to grow quickly and thereby escape the problems faced by small clans. With Mya, increased growth means an increase in burial depth and protection from short-term adverse environmental conditions and predation. (2) On the basis of size there is a delay in maturation. Al though not strong, this relationship is supported in the literature. The size of maturation is reported as 25 mm in Chesapeake Bay (Pfi tzenmeyer, 1972), 20 mm in southern New England (Coe and Turner, 1938), and 15 mm in Maine . Delayed reproduction allows more energy to be allocated toward early growth and also reduces the demand on resources already strained by high metabolic activity. Although a cost of reproduction has not been demonstrated for Mya, such a cost as measured by increased stress and mortality has been observed in Mytilus edulis (Bayne et al, 1978).
The diverting of productivity toward growth in early life, however, results in an increased reproductive potential in later life due to a general fecundity-size relationship (Brousseau, 1978b). (3) Egg density decreases. Assuming equal gonadal vol\.Jlles this would mean fewer eggs/brood. Since the possibility of a total reproductive failure from a single spawn increases towards the south, it would be advantageous to reduce the output/spawn but have more spawns. Ropes and Stickney ( 1 965) gave evidence for sue h a trend in spawning.
Northern populations spawned only once. Populations in tbe mid-range can spawn twice if conditions are favorable. The frequency of the second spawn increases towards the south. Two spawns are the rule in Chesapeake Bay (Pfitzenmeyer, 1962;Shaw, 1962;. (4) Egg size increases. The production of larger eggs may be interpreted as an attempt to increase survival CllJOung the larvae. Studies by Vance (1973) and  indicate that the planktotrophic larval strategy is optimized when the maximum nunber of eggs are produced subject to the constraint that each egg contains the minimum energy reserve necessary for successful larval develoµnent to the feeding stage. Consistant with the increase in the VJM, a larger egg size would mean that towards the south Mya needs more energy for successful develoµnent. The work of Bayne et al. ( 1975;  To summarize: with decreasing latitude there is evidence for an association between an increase in the variability of juvenile mortality, faster growth, delayed maturation, reduced fecundity, more broods/year, and larger eggs. All these traits are consistant with the predictions of the bet-hedging strategy. The only anomalous and confounding trait is the increase in adult . mortality already discussed. In variable environments bet-hedging predicts an increase in longevity coupled with a longer reproductive life. This clearly does not occur in Mya. However, the shortened reproductive life observed toward the south is somewhat offset by an increased frequency of spawning and a greater age-specific fecundity due to a larger age-specific size. A reduction in the total .number of spawning seasons should be associated with a reduction in the juvenile mortality/adult mortaltiy ratio (Stearns, 1976). In this sense the trends may be consistant since there is an increase in adult mortality toward the south while larger eggs may help reduce juvenile mortality toward the south.
The above discussion considers the life-history adaptations of populations to environmental conditions. These are adaptations to long-term variations and are therefore assumed to be genetic in nature. Direct evidence of genetic differences in Mya arenaria is scarce. Morgan et al. (1975) did report genetic differences between clans from Chesapeake Bay and Maine. Clams from Chesapeake Bay where found to have greater polymorphism and heterozygosity.

Effects of Local Environmental Variability
Organ isms ad apt genetically to long-term environmental variations. Within that genetic framework there can also be physiological response to short-term variations in the environment.
The evidence to date indicates that an increase in stress results in reduced reproduction. In the present study there was a correlation amoung proximal populations between egg density, egg size, and to a lesser degree pathological condtion indicating that when conditions are good more and larger eggs are produced. Brouseau ( 1978b) found a direct relationship between egg size and fecundity when comparing spring and summer spawns and she showed that total fecundity could vary between both seasons and years. However, no connection was made between reproduction and the conditon of the population. Coe and Turner (1938) reported that the number of young ovocytes developing to mature ova was dependent upon the nutritional condition of the individual. These findings for Mya are consistant with those reported by Bayne ( 1 972; and Bayne et al. ( 1 975; for Mytilus edulis. They found that in stressed individuals fewer, • smaller eggs were produced resulting in larvae with reduced viability. It is assumed that under local conditions of stress Mya responds in a manner similar to that observed for Mytilus.