observations on the nature and characteristics of sprat shoals

Observations on the Nature and Characteristics of Sprat Shoals.

BY J. Armitage Robertson, M. A.

With 5 figures.

Introduction. The adult sprat is caught in English coastal waters only during the

winter months, from November t,ill February or March, and during this winter immigration it neither feeds nor spawns ( R o b e r t s o n , 1938).

The sprats appear in shoals, and their gregarious habit is evident in the characteristic nature of the sprat catches. I t is a common occurrence, for instance, for some boats to make good catches, while others close to them take little or nothing. Sometimes, moreover, a large catch is made suddenly in a short time. I t also frequently happens, in the case of a fleet of drift nets, that some nets make much heavier catches than others, while even in a single net the catch may be enmeshed in unevenly distributed patches. This evidence points to a fish with gregarious habits moving about in parties, companies, or shoals, rather than to solitary fish distributed and dispersed evenly over the fishing ground.

Large flocks of predatory sea birds commonly accompany these gatherings of sprats, and while such a flock may cover an area of many acres its limits are usually fairly well defined. I t is probably not too large an assumption to suppose that the extent of a flock of birds bears some relation to the area occupied by the gathering of sprats on which the birds are feeding. In any case, on the Brightlingsea fishing grounds the fishermen try t o shoot their stownet,s downstream in the tidal track of such flocks of sea birds, in anticipation of a satisfactory catch of sprats.

In this paper, for the sake of convenience, the word ,,shoal" will be used to describe fair-sized gatherings of sprats; gatherings, that is,

Observations on the Nature and Characteristics of Sprat Shoals 57

which occupy an area of sea measurable in acres or fractions of an acre rather than in square feet or in square miles1). Small parties of a few. dozen or a few hundred fish may perhaps canveniently be called ,,~choo1s", but i t will be unnecessary to use any special term to desig- nate the whole body of immigrant sprats present in a fishing area, such as the mouth of the Thames 3r the Suffolk coast.

The shoaling habits of fish have been but litt!e studied (cf. P a r r 2927 and S p o o n e r 1930-31) and next to nothing is known of the causes of this habit in the open sea.

In spite of a general absence of actual observation the writer has encountered no lack of theories to account for this habit, more parti- cularly those which attribute gregariousness to a communal urge to spawn or to feed, neither of which are appropriate t o sprats caught in the winter months. There is also a highly ingenious hypothesis which seeks to define a shoal as a gathering of fish all of which are of like size or of like age (or both), and which makes the shoal's integrity depend upon the dissimilar swimming powers of fish of different sizes, so that larger and smaller fish are automatically separated. This hypothesis would no doubt explain the shoaling habit t o admiration if i t could be shown either that the sprat did in fact shoal by size or by age (for which there is no evidence), or if i t could be proved that all fish habitually swim a t or near their maximum pace (in which case alone would their differences in speed cause separation).

Other people even maintain that shoals do not exist a t all, or are extremely transitory, apparently on the grounds that the personnel of such gatherings is liable to change.

The writer was quite unable to discover any sufficient grounds for the conviction with which such views were held, or any concrete evidence in support of these various theories, least of all in so far as they could be applied to the sprat.

In the case of the sprat, on the contrary, the only evidence for the occurrence of shoals pure in point of size and age is found in the catches of whitebait sprats ( R o b e r t s o n 1938). Adult sprats taken in non-selective gear, either by commercial vessels or by the research ship ,,Onaway", whether in long hauls or in short, almost invariably showed a wide range of size and of age.

Since the aduIt sprats do not spawn or feed during their shoreward Thus every net in a fleet of 15 or 20 drift nets may take fish, or only some

nets, and adjacent vessels may or may not participate in the catch.

58 . J , hrmitage Robertson

immigration in winter i t may be that their shoaling habits at this time .bear a resemblance to the gregarious behaviour displayed by many other migrants (lemmings, birds, butterflies) and hibernating animals (certain species of bats and flies). Such gatherings characteristically exhibit a unison of behaviour and a tendency to remain in consort which appear to depend on psychological factors. They cannot readily be explained as mere fortuitous congregations, occasioned and specially localized by some necessity in common. P a r r (1927) and S p o o n e r (1930-31) have shown that fish also exhibit similar gregarious behaviour in captivity, and there is little room for doubt that sprats form gatherings of individuals swimming in consort, which vary in size and are here termed shoals.

The sprat is a shsrt-lived fish and the adult shoals only contain fish of age-groups I to IV, whereas the size range extends from about 5 to 15 cm. Since the sprats were separated into half-centimetre length- groups i t was easier to compare the length-distributions2) than t,he age-compositions of different samples.

All the samples quoted in this paper were taken either in stownets or in sprat-trawls, which do not release a large number of small fish as do drift-nets, and which are therefore relatively non-selective.

When the length-frequency distributions of the samples were plotted as curves and examined, it appeared that certain samples resembled one another and thus formed series of similar samples; series which might be short or long and either consecutive or interrupted by one or more dissimilar samples.

Such series of like samples were found in more than one season, which discounted the effects of pure coincidence and suggested the possibility that similarity between samples might imply a common origin, such as derivatian from the same shoal.

In order t o examine this possibility i t seemed desirable to express in quantitative terms the degree of resemblance detected by the eye between the length-curves of different samples. For this purpose Fisher’s modification of Pearson’s x 2 test (F isher 1934) has been used m e r e l y a s a .measure of s imi la r i ty . This measure is expressed by the value of the term P, which in this case does not connote the mathematical probability of any two samples coming from a common source, for in that case the two samples to be compared

2, Since the sprats were caught in the winter months growth does not com- plicate these comparisons.

Observations on the Nature and Characteristics of s p r a t SINJ~IS 59

must not be chosen for any reasn2n, such as their mutual resemblance, but must be taken for testing at random3). Fisher 's test compares the whole length-frequency distribution of one sample with that of another, and the term P has a maximum value of 1 when the two samples are identical. Fisher 's test is usefuI in that i t expresses in a simple quantitative manner (by the value of P) the likenesses detected by the eye in the frequency curves of different samples, and thus provides a better basis for subsequent argument as to their origins. In this paper it has no deeper or more abstruse significance.

Examination of the curves of the length-frequency distributions. Figs. 1 to 3 show the curves representing the percentage length-

frequency distributions of samples taken in stownets and sprat-trawls at Brightlingsea and Poole respectively, and also of stownet samples taken by M. V. ,,Onaway". Each sample is identified by its serial letters and number, but these numbers are not consecutive owing to the omission of intermediate drift-net samples, which are useless for the present purpose on account of net-selection.

In samples from non-selective gear (stownet and sprat-trawl) the range of size is wide (5 t o 15 cm), and although it is clear that the length-distributions of consecutive samples may differ radically, even within a few days, it is also apparent that pairs or short series of samples occur whose curves look alike. This is evident in the collections of more than one season.

Some of the more obvious instances of apparently similar and dissimilar consecutive samples are given here as examples :

Similar: Fig. 1. SH. 6 & I IB, SK. 6 & 8, SK. 36 & 37; Fig. 1A. SM. 38 & 41, SM. 43 & 44; Fig. 2. SJ. 5 to 9, SI,. 39 t o 48; Fig. 3. OSL. 10 & 11, OSM. 1 & 1A.

Dissimilar: Fig. 1. SH. 16 & 17;

SM. 46 & 47; SL. PI & P 2, SM. 28 & 32; OSK. 8 & 9, OSL. 3 & 5, OSM. 11 & 12.

Fig. 1A. ST.,. 14 et 15, SL. 23 & 27, SL. 50 & 56,

Fig. 2. Fig. 3.

3, The f i s h forming a sample were themselves, of course, taken a t random in every case.

60 J. Armitagc Robertson

There are many other less striking instances as well as cases of an intermediate or doubtful character.

Apart from particular cases of rather detailed resemblance in pairs or short series of curves (such as those just given) a more generalized type of likeness can be detected, usually involving several samples obtained within a few weeks of each other. The following are instances of longer series showing resemblance of this generalized character as estimated by the eye:

Fig. 1. Fig. IA. SL. 56 t o 65, SM. 31 ,33 , 38, 41 & 46, SM. 37, 43, 44 & 47. Fig. 3. Thus in samples SK. 14 to 23 there is some approach to a unimodal

curve, the principal mode being very pronounced and the subsidiary modes relatively inconspicums. In the last nine samples obtained from Brightlingsea during the fishing season 1933134 (SM. 31 to 47) the length distribution is diffuse and polymadal, but there seem t o be two general types of distribution which appear from late January t o mid March. Sample SbI. 31 represents a group with its principal mode at 10% t o 11 cm. and SM. 37 a group with its principal mode a t about 7 % cm., but both groups have a very similar range of size. Samples OSM. 4 t o 8 are alike in containing a majority af ,,whitebait" sprats of 4 to 6y2 cm., with very few fish larger than 7 cm. I t should, however, be remembered that sprats of such small size are probably subject t o net-selection even in a stownet, and the shape of the curve a t the smaller sizes may therefore be affected.

These peculiar sequences of samples which resemble each other on mere inspection of their curves, sometimes interspersed by dissimilar samples or by samples apparently belonging t o another sequence, do at least suggest the possibility that separate shoals are present on the ground, differing in length-frequency composition, but each retaining its particular and individual composition for an appreciable time and being therefore recognizable when sampled on subsequent occasions.

SK. 14 t3 23.

OSM. 4 to 8.

Application of Fisher's x 2 test solely as a measure of likeness. In applying Fisher's test the term P is taken to express the degree of

likeness between the length-frequency distributions of the two samples compared, and i t has a maximum value of 1 when the samples are identical.


As i t is necessary to make arbitrary distinctions between the values of P for the purpose of assessing similarity and dissimilarity, the following limits are proposed, an'd are used in this paper :

P = .99 - .50 great similarity .49 - .20 similar .19 - .05 slight similarity .049 - .01 probably dissimilar

< .01 dissimilar

As it is usual to give the ,,benefit of the doubt" to that aspect of any case which appears least likely, or to the opposite of any hypothesis which one is concerned to prove, i t is difficult to draw a strict line of demarcation between similarity and dissimilarity. When the former is suspected the critical value of P is taken as higher, and when dissimilarity is more likely i t is taken as lower than if the opposite were the case. Consequently tests with values of P from .02 to .09 are doubtful, though they might sometimes be decided by reference to other circumstances, such as the time elapsing between the capture of the two samples.

I t is perhaps worth mentioning that i t is quite helpful to regard the value of P as a percentage, ranging from less than 1% to 100% similarity.

Table 1 gives the values of P for some of the samples shown in Fi- gures 1 to 3, together with the time which elapsed between the capture of each sample of the tested pair.

Where the length distributions of two samples overlap only slightly, or not at all, and the bulk of each lies clear of the other i t is safe to assume that P would have a value of less than. .01 but where there is a considerable overlap of the two distributions i t is extremely difficult to assess the value of P merely by inspection of the curves, and i t is not always even safe to infer that i t lies on one or other side of the critical value separating like from unlike samples. In fact where two distributions are apparently alike the eye proves to be a singularly unreliable judge of the d e g r e e of similarity, though of course when samples appear to be almost identical (e. g. SM. 38 & 41, SM. 43 & 44, OSM. 1 & IA) its estimate is usually substantiated.

When the range of size is small and the distribution is correspon- dingly compact (SK. 14 to 23) it seems to the writer that inspection of the curves is apt to be less critical and is more likely to differ from

62 J . Armitage Robertson

' 35 39 48

0 20 30 34 43 20 30 34 43 10 14 23

4 1 3 9

20 28

8

the result of the x 2 test, while with a diffuse distribution the eye has greater opportunity for discrimination and is less likely to err.

In general, then, it may be said that it'is nearly always possible to detect dissimilarity by inspection of the curves, and that i t is also usually possible to detect similarity, but i t is exceedingly difficult to estimate its degree at all accurately by eye.

T a b l e 1. Value of P in x2 tests of Brightlingsea Samples.

* Tests with P = .20 or more, and with 20 or more days between the tested samples.

Samples

S e a s o n 1929/30 SH. 6 Ji SH. 11A SH. 6 & SH. 12B SH. 6 & SH. (1 1A& 1lBcombined SH. 11A & SH. 11B SH. 1 3 & SH. 16 SH. 1 3 & SH. 1 7 *SH. 1 3 & SH. 21

SH. 16 & SH. 1 7 *SH. 16 Ji SH. 21

S H . 1 7 & SH. 2 1 SH. (13 & 16 combined) & SH. 21

S e a s o n 1931/32 S K . 6 k S K . 8 S K . 6 k S K . 1 4 S K . 8 & S K . 14 SK. 14 & SK. 1 7 SK. 14 & SK. 2.1 SK. 14 & SK. 23 SK. 1 7 & SK. 21 SK. 1 7 & SK. 23 SK. 21 & SK. 23 SK. 32 & SK. 33 SK. 32 & SK. 35 SK. 32 & SK. 35A SK. 32 & SK. 36

Days be tween samples

li 1 !I

1 !t 0 8

15 36

7 28 21

-

4 1 7 1 3 8

15 22

7 14

7 7

12 12 32

P approx.,

.38

.43

.46

.02

.16 <.01

.41 <.01

.84 <.01

.54

.19 <.01 <.01

.31

.08 c . 0 1 <.01 <.01

.02

.25

.71 <.01

.07

Samples

S K . 32 & S K . 37 *SK. 32 & SK. 38

S K . 33 & S K . 35 S K . 33 & SK. 35A

*SK. 33 & SK. 36 S K . 33 & SK. 37

*SK. 33 & SK. 38 SK. 33 & SK. 40 SK. 35 & SK. 35A S K . 35 & SK. 36 SK. 35 & S K . 37

*SK. 35 & S K . 38 SK. 35 & SK. 40 SK. 35A & SK. 36 *SK. 35A & SK. 37 SK. 35A & SK. 38

*SK. 35A & SK. 40 SK. 36 & SK. 37 SK. 36 & SK. 38

*SK. 36 & SK. 40 S K . 37 & SK. 38 SK. 37 & SK. 40 SK. 38 & SK. 40

*SK. 32 & SK. 40

S e a s o n 1932/33 SL. 27 & SL. 37

*SL. 27 & SL. 47 SL. 37 & SL. 47

P :approx.)

.03

.4:1

.34

.26 <.01

.56

.13

.89

.08 <.01

.09

.02

.22

.05

.03

.38

.04

.40

.76

.17

.28

.18

.84

.23

c . 0 1 .46 .05

Observations on the Nature and Characteristics of Sprat Shoals 63 -

Samples

SL. 56 t SL. 62 *SL. 56 & SL. 64 SL. 56 & SL. 65 SL. 62 & SL. 6 4 SL. 62 & SL. 65 SL. 6 4 & SL. 65 SL. (56 & 62 combined) & SL. 6 4

Season 1933/34 SM. 1 0 & SM. 1 3 SX. 1 5 & SM. 17 SM. 17 t SM. 20 SM. 22 h SM. 27 *SM. 22 t SM. 31 SM. 22 & SM. 33 SM. 22 t SM. 37 SN. 22 t SM. 38 SM. 22 h SM. 41 SM. 22 h SM. 43 SN. 22 t SM. 44 *SM. 22 t SM. 46 SM. 22 & SM. 47 SN. 27 Jz SM. 31 SM. 27 t SM. 33 SM. 27 & SM. 37 SM. 27 & SM. 38 SM. 27 t SM. 4 1

SM. 27 & SM. 4 4 SM. 27 & SM. 46 SM. 27 & SM. 47 SM. 31 t SM. 33 SM. 31 & SM. 37

SM. 27 & SM. 43

S e a s o n 1930131

SJ. 5 & SJ. 7 SJ. 5 & SJ. 9 SJ. 7 & SJ. 9 S J . 5 h SJ. (7 & 9 combined)

Days between samples

1 3 21 29

8 1 6

8

-

7 8 7 8

20 27 34 35 42 48 55 62 7 1 1 2 1 9 26 27 34 40 47 54 63

7 1 4

P approx.)

.17

.45

.05

.39

.01 <.01

<.01

.08 <.01

.19 <.01

.86

. I 0 < . O l

.016

.01G <.01 <.01

.88 <.01 < . O l < . O l <.01 <.01 < .01 <.01 <.01 <.01 <.01

.69 <.01

Samples

SM. 31 & SM. 38 "SM. 31 & SM. 4 1 SM. 31 & SM. 43 SM. 31 & SM. 4 4

CSM. 31 & SM. 46 SM. 31 & SM. 47 SM. 33 & SM. 37 SM. 33 & SM. 38 SM. 33 & SM. 41 SM. 33 & SM. 43 SM. 33 & SM. 4 4

"SM. 33 & SM. 46 SM. 33 & SM. 47 SM. 37 h SM. 38 SM. 37 & SM. 41 SM. 37 & SM. 43 SM. 37 & SM. 4 4 SM. 37 & SM. 46 SM. 37 & SM. 47 SM. 38 & SM. 41 SM. 38 8 SM. 43 SM. 38 cPC SM. 44

"SM. 38 & S M . 46 SM. 38 t S M . 47 SM. 4 1 B SM. 43 SM. 41 Jz SM. 44

@SM. 4 1 & SM. 46 SM. 4 1 & SM. 47 SM. 43 & SM. 44 SM. 43 & SM. 46 SM. 43 & SM. 47 SM. 4 4 & SM. 46 SM. 44 & SM. 47 SM. 46 & SM. 47

Value of P in x 2 tests of Poole Samples.

7 1 4

7

-

.61

.46

.16

.35

Season 1931132 SK. 22 & SK. 24

Season 1932133

SL. 39 b SL. 46 SL. 39 h SL. 48


1 5 22 28 35 42 51

7 8

15 21 28 35 44 1 8

1 4 21 28 37

7 1 3 20 27 3 6

G 13 20 29

7 1 4 2 3

7 1 6

9

8

6 1 3

P approx.)

.05

.30 <.01 < .01

.85 <.01 < . O l

-08 .43

<.01 <.01

.60 <.01 < .01 <.01

.05 <.01 <.01 <.01

.85 <.01 <.01

<.01 < . O l <.01

.77 < .01

.58 c . 0 1 < .01 <.01 <.01 <.01

.68

<.01

.65

.08

64

7 11 20

9

22

J . Armitage Robertson

.04 < .01

.14 <.01

.31

Samples

SL. 46 & SL. 48 SL. P 1 & SL. 55 SL. P 1 & SL. 61 SL. 55 & SL. 61

Season 1933/34 *SM. 21 & SM. 32

r between samples (approx. Samples

~~~

SM. 26 & SM. 34 SM. 28 & SM. 40

S e a s o n 1934/35 S N . 7 k S N . 1 4 SN.7 & S N . 21 SN. 14 & SN. 21

between samples

<.01 < . O l

36 .12

Value of P in x 2 tests of ,,Onaway" Samples.

Samples

Season 1931/32 OSK. 16 ~k O S K . 17

Season 1932/33 OSL. 2 cyt OSL. 3 OSL. 2 k O S L . 5 OSL. 3 & O S L . 5 OSL. 6 k O S L . 9 OSL. 10 QL. OSL. 11

Season 1933/34 OSM. 1 C OSM. 1 A OSM. 1 C OSM. 2 OSM. 1A & OSM. 2 OSM. 4 C OSM. 6 OSM. 4 & OSM. 7

*OSM. 4 C OSM. 8 OSM. 6 8z OSM. 7

*OSM. 6 C OSM. 8 OSM. 7 & OSM. 8

Places

Wallet (Brightlingsea)

Wallet (Brightlingsea) Wallet (Brightlingsea) Wallet (Brightlingsea) Deal Deal

Kessingland Kessingland Kessingland Kessingland Kessingland Kessingland Kessingland Kessingland Kessingland


1

1 2 1 2 1

0 6 6 8

42 43 34 35 1

NB.: The Gear used in all cases was the ,,Kessingland Stownet"

An examination of Table 1 shows that: 19 tests reveal great similarity

25 tests reveal slight similarity

P = .99 - .50 P = .49 - .20 P = .I9 - .05

22 tests reveal similarity

(exclusive of tests of combined samples)

P (approx.)

< .01

.09 < . O l < . O l <.01

.25

.64 <.01 <.01

.12 <.01

.23 <.01

.82 < . O l


In 22 of the tests showing similarity or great similarity the time which elapsed between the capture of the two samples concerned was 20 days or more, and for convenient reference these tests are marked with an asterisk. SM. 22 & 46 although taken 62 days apart had a value of P of .88, and SK. 32 & 38 although 46 days apart had a value of .43, but these are extreme instances.

I t may be argued that similar samples come (1) from the same shoal or from related shoals (e. g. offshoots of

the same parent shoal) or (2 ) from shoals which are entirely unrelated, separate and di-

stinct, but which are nevertheless identical or a t least have overlapping length distributions.

If similar samples came from the same or related shoals i t would not be surprising to find that they were taken within a short time of each other, for the shoal or its offshoots might be expected in some cases to remain for a time in the fishing area and be sampled more than once, but there is no reason why two entirely separate a6d distinct shoals, having similar length compositions, should o f t e n happen t o visit the area within a short time of each other.

Among the 40 Brightlingsea and Poole samples showing similarity or great similarity (P = .20 or more) there are no less than 13 pairs of consecutive samples with a mean interval of about 8.4 days. I t should be noted that consecutive samples from these ports were seldom sent at less (and sometimes a t more) than weekly intervals4), but samples OSL. 10 & 11 and OSM. 1 & I A , taken by M. V. ,,Onaway", shorn resemblance between catches taken a t much shorter intervals.

The frequent occurrence of mutual similarity in samples forming both long and short series also favours an explanation of their resemblance based on community of origin rather than upon coincidence.

Certain of the length distributions and the results of Fisher 's test may now be considered in somewhat greater detail.

1. B r i g h t l i n g s e a Sa,mples. (Figs. 1 B IA.) S e a s o n 1929130. Samples SH. 11A and IIB, caught a t different places on the same day, both

resemble SH. 6 but not each other. If combined in equal proportions6) they 4, This is a disadvantage due to the fact that the samples were obtained for

other purposes. s, The proportions are necessarily arbitrary and the resultant value of P

should therefore be accepted with reserve. Internat. Rev. d. Hydrobiol. 39. 3

66 J. Armitage Robertson

resemble SH. 6 slightly more strongly than do either of them separately. This may be due to the shoal from which SH. 6 came having divided into dissimilar daugther shoals.

Conversely SH. 13 and 16 resemble SH. 21 but not each other, and if combined as before they still resemble SH. 21. Thus SH. 21 may have come from an amalgamation of two shoals. SH. 1 7 is wholly unlike the other samples and would probably come from another shoal altogether.

U

Sample Date SH 6 5.12. 29 S H 11-4 19.12. 29 SH 1 1 B 19.12. 29 SH 13 2. 1. 30 SH 16 10. 1. 30 SH 17 17. 1. 30 SH 21 7. 2. 30

Total sample

NO of Fish 100 200 (Wallet) 197 (Thames)

89 Est. 7 1 70

100

SK 6 14.11. 31 90 SK 8 18.11. 31 191 SK 14 1.13. 31 159 S K 1 7 9. 12. 31 100 SK 21 16.12. 31 100 SK 23 23.12. 31 150 SK 32 22. 1. 32 100 S K 33 29. 1. 32 83 SK 35 3. 2. 32 61 (Wallet)

SK 35* 3. 2. 32 76 (Wallet) SK 36 23. 2. 32 100 SK 37 4. 3. 32 100 SK 38 8. 3. 32 100 SK 40 17 . 3. 32 100

Total sample

Figure 1. Length-frequency distributions of Brightlingsea samples from commercial

catches.

Season 1931/32. Samples SK. 14, 17, 21 and 23 look rather alike, but upon test only SK. 14

and 1 7 possess a satisfactory measure of resemblance. Samples SK. 32 to SK. 40 form a series whose relationships are given dia-

grammatically in Fig. 4. The series may be divided into Group A (SK. 35A, 36, 37 & 40) with a mean value of P among its constituent members of .45, and into Group B (SK. 32, 33, 35 and 38) with a mean value of -46. Between the members of Group -4 and those of Group B the mean value of P is only. .13. The


individual values of P are given (by arrows) in the diagram and i t will be observed that, with certain exceptions, those within each group do not fall below .22, whereas those between members of different groups do not exceed .IS. This may perhaps be explained by assuming that two distinct shoals represen-

Sample Date SL 1 4 18.11. 32 SL 15 23.11. 32 SL 1 9 1. 12. 32

SL 23 6. 12. 32 SL 27 14.12. 32 SL 37 3. 1.33 SL 47 11. 1.33 SL 50 19. 1. 33 SL 56 1. 2. 33 SL 62 14. 2. 33 SL 64 22. 2. 33 SL 65 2. 3. 33

Total sample

SM 10 21. 11. 33 SM 13 28. 11. 33 SM 15 6. 12. 33 SM 1 7 14.12. 33 SM 20 19./21. 12. 33 SM 22 3. 1. 34 SM 27 11. 1. 34 SM 31 23. 1. 34 SM 33 30. 1. 34 SM 37 6. 2. 34 SM 38 7. 2. 34 SM 41 14. 2. 34 SM 43 20. 2. 34 SM 44 27. 2. 34 SM 46 6. 3. 34 SM 47 15. 3. 34

Total sample

No. of Fish 150 152 151

151 146 150 146 100 152 138 190 170

91 119 144 177 151 275 169 213 150 150 150 181 309 228 131 292

Figure l a .

Length-frequency distributions of Brightlingsea samples from commercial catches.

ted by the samples of Groups A and B were present in the area a t the same time, and that a certain amount of intermixture occurred, quite possibly on more than one occasion, thus producing a mean value of P of . I 3 between the two groups. Further, the higher mean values of P within the two distinct groups

5*


would perhaps indicate a tendency for the shoals to retain their individuality despite intermixture, possibly due to a disposition to return to their original length-composition by segregation. The low value of P 1.03) between SK. 35a and 36 could be explained by assuming that they really came from different shoals (A, and A, respectively) and that these shoals combined later to form one shoal (A, + A,) from which samples SK. 37 and 40 were taken. This would a t least explain why both SK. 37 and 40 resemble SK. 35A and 36 more strongly than the latter pair resemble each other.

I t may be remarked that while SK. 33 and 38 look alike and have a value of P = 3 9 , neither SK. 32 nor SK. 35 have the appearance of close resemblance to them, yet both the latter samples are connected to the two former by reasonably high values of P. Likewise, a t a casual glance, S K . 40 does not look very similar to other members of Group A, though its P values with them are quite satisfactory.

Season 1932/33 .

Samples SL. 27 and 47 are alike (P = .46) although taken 28 days apart, but neither resembles the intermediate sample SL. 3 7 .

Samples SL. 56 and 62 both resemble SL. 64 more strongly than they resemble each other, which may be due to amalgamation of two shoals, although SL. 56 and 62 combined arbitrarily in equal proportions show no resemblance to SL. 6 4 .

Season 1933/34 .

Samples SM. 17 and 20 show some measure of similarity (P = .19) . Samples SM. 22 to SM. 47 were tested and their relationships are given i n

Fig. 5. They may be divided into Group A, consisting of samples SM. 43 and 44 which

are very similar, and Group B comprising SM. 2 2 , 31 , 33, 3 8 , 41 and 4 6 which appear to resemble each other. There is a third heterogeneous category, Group C, including samples SM. 27, 37 and 47, which are dissimilar to each other and to all members of Groups A and B. Groups A and B are, moreover, entirely dissimilar to eachother (P = < .01 in all cases). Within Group B the mean value of P is .48 and the values of P of each individual member of the group show a strong bond with two or more other members in every case, although there are low values of P between certain members of the group.

The suggested explanation is that two shoals (A and B), represented by the samples of Groups A and B, were present on the fishing grounds, of which one (Shoal B) was sampled several times and the other (Shoal A) only twice. Three other distinct shoals represented by SM. 27, 37 and 47 respectively, were also sampled a t times during this period, but appear to have had no influence on the length composition of Shoals A or B, which likewise had no influence upon each other. Both Shoals A and B appear to have largely retained their individual length compositions during the time they were under observation, though from the low values of P found in certain instances within Group B there seems reason to suspect that some unknown process of combination or segregation occurred within Shoal B. In view of the fact that three (out of five) of these


low values of P refer to SM. 38 when tested against earlier samples, i t seems likely t h a t this alteration of the length composition occurred in Shoal B some- time between the dates of capture of SM. 33 and SM. 38 (i. e. between 30th January and 7th February 1 9 3 4 ) . And further, since SM. 38 and the earlier samples (to which i t has but sligth resemblance) are alike i n resembling t h e la ter samples of Group B, there is some reason to think that this alteration of length composition was subsequently modified sufficiently to allow of strong resemblance between the early and late samples of the group, b u t was not eliminated so suddenly as to cause dissimilarity between those la ter samples and SM. 3 8 wherein the altered lenght composition had first become apparent.

The low values of P occurring between SM. 22 and 33, and SM. 22 and 4 1 , indicate tha t this disturbance of the length composition may have been gradual both in i ts appearance and i ts disappearance, affecting the samples immediately before and after SM. 38 in Group B.

An inspection of the curves of these samples (Fig. la) shows t h a t those belonging to Group B have a fairly strong general resemblance to each other, b u t it is noticeable that SM. 38 has a pronounced subsidiary mode at 13 cm. which is absent in earlier samples; this mode is rather less prominent in SM. 41 and is no longer present in SM. 4 6 . Perhaps the disturbance in the length composition evident in the tests of SM. 38 was caused by the entrance into Shoal B of a group of large sprats which later on gradually left the shoal.

I t is also possible to explain the low values of P within Group B in a somewhat different manner, by assuming tha t samples SM. 22, 31 and 33 really belonged to a different shoal (B,) from t h a t which provided SM. 38 (Shoal B,) and that these shoals later combined to form one shoal (B, + Ba) from which samples SM. 41 and 46 were taken. This offers some explanation of the fact tha t both SM. 4 1 and 46 resemble SM. 38 (Shoal B,) and the group SM. 22, 31 and 33 (Shoal B,) on the whole much more strongly than any resemblance existing between SM. 38 and the members of Shoal B,. This suggested explanation really-amounts to a slightly different and amplified aspect of tha t alteration of the length-composition and its subsequent modification, which was referred to in the preceding paragraphs. The chief difference is t h a t Shoal B has here undergone analysis into Shoals B,, B,, and (B, + B2).

The close resemblance between the curves of samples SM. 43 and 44 (Group A} has already been mentioned, and it may be observed tha t their length range i s very similar to tha t of the samples of Group B. I t is also evident t h a t the principal and chief subsidiary modes of samples SM. 43 and 44 lie in much t h e same position as the chief subsidiary and principal modes, respectively, of the samples of Group B. Nevertheless the values of P between members of Groups A a n d B are always less than .01.

Owing to the majority of i t s frequencies lying above 11 cm., the curve of t h e divergent sample SM. 27 differs in appearance from those of all the other samples of the series (SM. 22 to 4 7 ) , but the curves of the divergent samples SM. 37 and 47 bear quite a close apparent resemblance to those of SM. 4 3 and 4 4 (Group A), although the values of P show dissimilarity (subject to some slight doubt in the case of SM. 37 and 4 3 , where P = .05).


2. Poo le S a m p l e s (Fig. 2). Season 1930/31. Samples SJ. 7 and 9 both resemble SJ . 5 much more closely than they re-

semble each other, but when combined in arbitrary equal proportions they resemble SJ. 5 less closely than they do separately. This case may be a further illustration of a shoal, represented by SJ. 5, splitting into rather dissimilar daughter shoals.

Sample SJ 5 SJ 7 SJ 9

S K 22 S K 24

SL 33 SL 39 SL 46 S L 48 SLP 1 SLP 2 SL 55 SL 61

S M 21 SM 26 S M 28 SM 32 SM 34 SM 40

S N 7 S N 14 SN 21

Date 16. 1. 31 23. 1. 31 30. 1. 31

23. 12. 31 31. 12. 31

20. 1 2 . 32 4. . l . 33

10. 1. 33 17. 1. 33 20. 1. 33 23. 1. 33 31. 1. 33

9. 2. 33

2. 1. 34 10. 1 .34 17. 1. 34 24. 1. 34 30. 1. 34

8. 2. 34

20.11. 34 1. 1 . 3 5 6. 2. 35

No. of Fish 150 150 140

152 100

88 150 153 100

87 188 284 109

135 185 196 1 7 1 150 142

109 115 100

Figure 2. Length-frequency distributions of Poole samples from commercial catches.

S e a s o n 1931/32. Samples SK. 22 and 24 are dissimilar although the principal difference shown

by their curves is only the presence of a subsidiary mode a t 9% cm. in S K . 24.

Season 1932/33. The curves of samples SL. 39, 46 and 48 lbok somewhat alike but the test

shows that only SL. 39 and 46 have any appreciable degree of resemblance (P = .65).


Samples SL. PI, 55 and 61 do not look unlike each other, but, upon test, a

S e a s o n 1933/34. Samples SM. 21 and 32 were taken 22 days apart b u t resemble each other

S e a s o n 1934/35. Although taken at a n interval of 36 days, samples SN. 14 and 21 show a

slight similarity is found only between SL. PI and 61 (P = .14).

(P = .31).

value of P of .12. Their curves appear rather unlike since SN. 14 is bimodal.

3. S a m p l e s t a k e n w i t h a S t o w n e t b y M. V. , , O n a w a y " . (Fig. 3.) These samples, although they do not form a continuous or regular series

throughout a n y one season, are of some value as they often show the degree of resemblance found in samples taken within a short time of each other.

S e a s o n 1931/32. Samples OSK. 8 and 9 have very dissimilar distributions although both were

taken in the Wallet within a day of each other. OSK. 8, however, contains very few fish.

OSK. 16 and 17, caught in the same place, and also within a day of each other, do not look alike and are shown by the test to be dissimilar (P = < .01).

S e a s o n 1932/33. Samples OSL. 2, 3 and 5 were taken in the Wallet on three successive days.

OSL. 5 is obviously wholly dissimilar to OSL. 2 and 3, which are not unlike each other and upon test show some slight similarity.

Samples OSL. 6 and 9, taken a t Deal a t an interval of two days, appear to be rather unlike and prove to be dissimilar.

OSL. 10 and 11 were also taken a t Deal on consecutive days and appear to be rather alike, The test shows a moderate degree of resemblance between them.

S e a s o n 1933/34. Samples OSM. 1 and 1 A are of particular interest as they were taken in two

consecutive hauls of the stownet on the same day and at the same position off Kessingland. Considering the immense range of size, their curves appear to resemble each other very closely and they have a value of P of .64, which is confirmatory of their strong resemblance. OSM. 2, taken six days la ter at the same position, is dissimilar to both of them, although i ts curve resembles the middle and right side of their curves.

OSM. 4 , 6, 7 and 8 form a group of samples taken at irregular intervals during a very long period, on the Kessingland ground, All their curves are characterized by a large group of sprats of whitebait size with very few fish of sizes such as are normally caught on this ground with drift-nets.

The curve of OSM. 7 is differentiated from the other three by a marked subsidiary mode at 6 '/z cm. and the test shows t h a t i t is dissimilar to all of them.

OSM. 6 and 8, although separated by a n interval of 35 days, have almost identical curves and a very high value of P. The test shows t h a t OSM. 4 re-


sembles both OSM. 6 and 8 tolerably well, but i t is curious that i t should resemble OSM. 8, after an interval of 43 days, more strongly than i t resembles OSM. 6 after an interval of only 8 days. I t should be remembered, however, that the majority of the fish in these samples are 4 to 6 cm. in length, and i t is

Sample

OSK 8 OSK 9

OSK 16 OSK 17

OSL 2 OSL 3 OSL 5

OSL 6 OSL 9

OSL 10 OSL 11

OSM 1 OSM 1A OSM 2 OSM 4 OSM 6 OSM 7 OSM 8

OSM 9 OSM 15

OSM 11 OSM 12 OSM 13 OSM 1 6

Date No.of Place Fish

25.11.31 33 26. 11. 31 124)

14. 1. 32 108 15. 1. 32 108

30.11. 32 149 1. 12. 32 121 2. 12.32 68

4.12. 32 118, 6.12. 32 154

13. 1. 33 158 14. 1. 33 152

7.11. 33 209 7.11. 33 182

13.11.33 148 27. 11. 33 322

5.12.33 174 8. 1 .34 139 9. 1.34 180

31. 1 .34 7. 2 .34 100

3. 2. 34 79 5. 2.34 100 6. 2.34 100 8. 2 .34 115

Figure 3. Length-frequency distributions of samples taken by M. V. ,,Onaway" in Stow-

(To show samples taken a t short intervals.) nets.

probable that such small fish are subject to some net-selection even by a stownet, which is a relatively non-selective form of gear. This selection would probably smooth artificially the left side of the curve of the large group of small fish. The values of P in this group should therefore be accepted with reserve.

GROU

P 4

GIOUP

~9

SK

GIOUP

B

GROU

P A

GROU

P C

4d.4u

o c-

- # s W

c p t;

Figu

re 4.

Figu

re 5

. D

iagr

am o

f rel

atio

nshi

ps sh

own

by x

2 tes

ts o

f sa

mpl

es S

IC 3

2 D

iagr

am o

f rel

atio

nslii

ps s

how

n by

x2

test

s of

sam

ples

SM

22

#

v

to S

K 4

0. T

he a

rrow

s in

dica

te v

alue

s of

P.

(Sam

ples

arr

ange

d ve

rtic

ally

in o

rder

of

capt

ure,

and

hor

i-

zont

ally

in

grou

ps a

ccor

ding

to t

heir

aff

initi

es a

s sl

iow

n by

th

e va

lues

of P

in T

able

1.)

to S

M 47. T

he a

rrow

s in

dica

te v

alue

s of

P.

(Sam

ples

arr

ange

d ve

rtic

ally

in

orde

r of

cap

ture

, and

hor

i-

zont

ally

in

grou

ps a

ccor

ding

to t

heir

aff

initi

es a

s sh

own

by

the

valu

es of

P in

Tab

le 1

.)

-3 0


Samples OSM. 9 and 15 were obtained a t Deal a t an interval o [ 7 days. Their curves appear to be rather dissimilar and they were not tested.

Samples OSM. 12, 12, 1 3 and 16 were taken a t Folkestone a t short intervals over a period of 6 days. They were not tested, but their curves show very clearly the great variation in length-frequency distribution which is possible in the same area within a short space of time.

Comparison between individual samples and the sum of all samples

I t was suggested to the writer that the length-frequency distributions of individual samples might, after all, be merely random variations from that of a single common stock or huge shoal present in the fishing area through the season. In order to investigate this possibility each sample obtained from Brightlingsea (in the case of three seasons) was tested against a total sample, formed by the combination of all samples taken during the season on that ground (Total Samples, Figs. 1 and IA). The individual samples and their values of P relative to the corresponding total sample are given in Table 2 for the seasons 1931/32, 1932133, and 1933134.

The sum of all the samples obtained during a seasons) represents the nearest approach that can be made t o an estimate of such an hypothetical common stock. In 1931/32 only four of the individual samples showed a reasonable measure of agreement with this hypothetical stock, while ten showed complete dissimilarity. In 1932/33 one sample showed some similarity and eleven were dissimilar, while in 1933/34 three samples showed a slight similarity and thirteen were dissimilar to the total sample.

So far as the sampling is adequate throughout any season, the sum of all samples taken during that season (the total sample) must repre- sent to a greater 3r less extent the length composition of the stock of fish present in the area d u r i n g t h e season t a k e n a s a whole. On the other hand, in 34 cases out of 42 such a combination of samples is quite inadequate as an expression of the length composition of the stock a t a n y o n e p a r t i c u l a r t i m e d u r i n g t h e season.

In 1931/32 and 1933/34 the samples showing resemblance to the hypothetical stock were taken towards the end of the season. I t is possible that a process of gradual building up of the stock, due to the successive invasion of the area

6, Since the fishing season occurs in the winter months the effects of growth

taken during the season (total sample).

may be neglected.


Table 2.

Season 1931/32

Season 1932/33

Season 1933/34

Sample

SK. 6 SK. 8 SK. 14 SK. 1 5 SK. 21 SK. 23 SK. 32 SK. 33 SK. 35 SK. 35A SK. 36 S K . 37 SK. 38 S K . 40 SL. 14 SL. 15 SL. 19 SL. 23 SL. 27 SL. 37 SL. 47 SL. 50 SL. 56 SL. 62 SL. 64 SL. 65

SM. 10 SM. 13 SM. 15 SM. 1 7 SM. 20 SM. 22 SM. 27 SM. 31 SM. 33 SM. 37 SM. 38 SM. 41 SM. 43 SM. 44 SM. 46 SM. 47

Value of P (approx.) relative to Total SamDIe

< .01 < .01

.01-.02 < .01 < .01 < .01 < . O l < . O l < .01

.28 < .01

.28

.06

.53

< .01 < .01 < . O l < . O l

.11 < .01 < .01 < .01 < .01

.02 < .01 < .01

< .01 < .01 < .01 < .01 < .01 < .01 < .01

< .01 < .01

.06

.079 < . O l < .01

< .01

.0176

.075


by distinct shoals, might be expected, towards the end of the season, to pro- duce by coalition something resembling a common population. In such a case the earlier samples would differ from the total sample because they would come from a stock still in process of formation. But it is more likely that the resemblance between these late samples and the hypothetical stock is due to the fact that in both seasons the former all belong to a large group of interrelated samples, which may have biassed the synthetic ,,common stock" in their own favour. I t is significant that in 1933/34 the late samples (SM. 4 3 , 44 and 47) which do not belong to the inter-related group do not resemble the hypothetical stock. In 1932/33 the later samples form only a comparatively small inter-related group and they also show no resemblance to the hypothetical stock.

In any case i t does not appear that the length-frequency distributions of individual samples are in fact mere random variations from that of a single, uniform, common population constantly present in its entirety.

Conclusions. I t does not seem unreasonable to assume, on the basis of the evidence

discussed in the introduction, that sprats move about in discrete gatherings or companies of variable size, which may conveniently he termed shoals.

I f it were t o be assumed, on the contrary, that the sprat population is present as a single continuously distributed and sedentary unit it is difficult to account for the capture of two (or more) similar samples and an intermediate one resembling neither of them [e. g.') samples SH. 16 (17), 21: samples SM. 22, (27), 31: samples SL. 27, (37), 47: and less clearly in samples SM. 31, 33, (37), 38, 411.

If such a dissimilar intermediate sample indicates a change in the length distribution of the ,,unit stock", this change is apparently mare or less exactly reversed in the following sample and the old length- distribution is restored.

Such occurrences can alternatively be attributed to the chances of random sampling, but i t is a t least as simple an assumption to suppose that one shoal was sampled twice in its wanderings and another shoal, with a different length-distribution, was sampled in the meantime. The known characteristics of sprat catches fit in far better with the theory that sprats move about in discrete shoals than with the assumption that the sprat population is one and indivisible. To go so far as to admit that this population may be divided into smaller local units

') Dissimilar intermediate sample in brackets.


(or populations) is to accept the underlying principles of shoal formation, but at the same time to deny the undoubted migratory powers of the sprat.

This paper produces evidence to the effect that sprats move about in discrete bodies or shoals, that these shoals contain fish with a wide range of size (and are therefore not ,,pure" shoals) and further, that there is reason to suppose that a shoal may possess a characteristic length-distribution. This length-distribution may be maintained by the shoal, and can be recognised in samples after a considerable lapse of time, but is liable to alteration probably occasioned by coalition with other shoais. There is also some reason to suppose that a shoal may divide into daughter shoals with different characteristic length- distributions within the same size range. The maintenance of a shoal's individual and characteristic length-distribution would occasion little surprise in an area where there are physical barriers to free dispersion and intermixture, such as are found in the Norwegian Fjord system, but on an open coast such integrity is probably dependent on the peculiarities of the sprat's gregarious behaviour.

Though little is known of the causes of gregarious behaviour in fish, in the case of the sprat which migrates to the English coastal waters in winter it is certain that i t is due neither to spawning nor to feeding, and it is exhibited by a mixture of fish of diverse age and size.

Ileferences. Fisher, R. A, ,,Statistical Methods for Research Workers." 5th Edition. Lon-

Perr, A. E., ,,A Contribution to the Theoretical Analysis of the Schooling Beha-

Robertson, J. A., ,,The Sprat and the Sprat Fishery of England." Min. Agric. S.

Spooner, G . M., ,,Some Observations on Schooling in Fish." Journ. Mar. Biol.

don. 1934.

vior of Fishes." Occ. Papers Bingham Oceanogr. Collect., No. 1, 1927.

Fish., Fish. Invest., Ser. 11, Vol. XVI, NO. 2, 1938.

ASSOC. U. K., V O ~ . XVII (N. S . ) , 1930-31.

observations on the nature and characteristics of sprat shoals

Documents