The influences of encountering a food object on subsequent searching behaviour in Gasterosteus aculeatus L

Anim. Behav., 1974,22,941-952

THE INFLUENCES OF ENCOUNTERING A FOOD OBJECT ON SUBSEQUENT SEARCHING BEHAVIOUR IN GASTEROSTEUS ACULEATUS L. BY G. THOMAS

Zoiilogisch Laboratorium

der Rijksuniversiteit

te Groningen, Haren, Netherlands

Abstract. The influences of discovery of a food object upon searching behaviour in the three-spined stickleback were investigated by comparing these to behaviour in the absence of such a food object. Changes in searching behaviour after acceptance concurred with previous findings in that there was a reduction in the tendency to move away from the site of eating plus an immediate increase in the intensity of searching, leading to ‘area restricted searching’. After rejection, however, there was an increased tendency to move directly away from the site of the prey plus an initial decrease in intensity of search, leading to what I have called ‘area avoided searching’. Some inferences are drawn regarding the possible fun&on of this to the predator. De Ruiter (1967), defines food searching as that part of feeding behaviour that is performed in the absence of stimuli that inform the predator as to the exact location of the prey. According to this definition searching is terminated upon the discovery of the potential food object, but may be reinstated subsequent to rejection or ingestion of that object. In this paper the question will be considered whether or not discovery of a food object results in modifications of food searching behaviour and, if so, what these modifications consist of. This question has only partly been dealt with by previous workers. Several authors have demonstrated the influence of the acceptance of a food object on the predator’s subsequent behaviour. Laing (1937), Fleschner (1950), Banks (1957), Dixon (1959), and Mitchell (1963), all working with insects have pointed out that after capture of a prey the search path becomes more tortuous. Beukema (1968), working with the three-spined stickleback, showed that the probability of a turn occurring was 25 times greater after the discovery of a prey. Tinbergen, Impekoven & Frank (1967) indicated that many birds react to the discovery of a prey item by an intensified effort in the area around such a first find and termed this ‘area restricted searching’. Croze (1970), working with carrion crows supported Tinbergen’s argument and Smith (1971), making a detailed analysis of the search path and factors influencing this, corroborated Tinbergen’s conclusions. In these cases behaviour of the predator subsequent to acceptance of a food object is compared to the situation before discovery, i.e.

where the location of the food object was unknown. Such analyses of the search paths have revealed differences in various spatiotemporal measures of such paths after acceptance. These measures fall into two basic groups, contributing respectively to (1) the total distance covered per unit of time (track length, TL) and (2) the tortuosity of the track or path. Reduction in track length after acceptance was noted by Dixon (1959) and Mitchell (1963). Decreases in the step distance (Smith 1971) between turns, i.e. in length of moves between consecutive turns, has been demonstrated by Mitchell (1963) and contributes to changes both in track length and tortuosity. Changes in tortuosity may further result from increases in the frequency of turning (Laing 1937; Banks 1957; Dixon 1959; Mitchell 1963; Beukema 1968) or in the total angle turned (Banks 1957). Transition from an alternation between left and right turns to a bias for one or the other side has been shown by Banks (1957). Discovery of the food object does not, however, always result in its acceptance; it results in rejection if the prey is unacceptable to the predator at that instant of time. Behaviour subsequent to rejection has received little if any attention, due possibly to the difficulty of recognizing rejection in the field where the majority of studies on searching behaviour take place. For a rejection to be discernible, it is essential that it is known that discovery has actually occurred. As such then, rejection is recognizable only if the predator has first manifested an overt sign that discovery has ensued, such as an orientated movement with respect to the food object. If discovery is not 941

942

ANIMAL

BEHAVIOUR,

SAtype track

22,

4

assumption, in the absence of any experimental evidence either that BDR = BD,

(W

BDR < BDA

(2b)

or that

SEtype track

where BD, is the beeline distance of the search path subsequent to rejection (SR), per unit of time. Earlier experiments by the author on the other hand? and also Prop’s (1960) work, suggest a thrrd hypothesis, namely that BDR > BDA

Fig. 1. Hypothetical search tracks. See text for explanation of the figure.

accompanied by such an overt sign (as theoretically could be the case if decision to reject is simultaneous with discovery), it is obviously impossible to separate failure to discover from rejection. Yet, for a complete picture to emerge as to the effect of discovery of a potential food object on the predator’s searching behaviour it is essential that rejections are also taken into consideration. Acceptance of a food item, as indicated, results in a decrease in track length and/or increases in tortuosity of the subsequent search path. As Smith (1971) has pointed out, the net effect of these changes is a reduction in the rate of movement away from the location of the accepted prey. This is reflected in the beeline distance of the track per unit time as shown in Fig. 1, where S* represents the actual path of search in the absence of a prey over time interval t ’ - t and SE the search path over the same interval after a prey has been eaten at time t (equal speed of movement is assumed for the two cases). The broken lines in Fig. 1 are the beeline distances (BO) of the tracks. The data presented by the workers above then, lead to the conclusion that BDE < BDA (1) where BD, and BD, are the beeline distances of the SE and S’* tracks respectively, per unit of time. Before it can be concluded that searching is indeed ‘area restricted’ in these cases, it must be ensured that the reduction of BD, is not the result of a temporary cessation of searching, as would occur with a digestive pause (Holling 1965). The influence of rejection is, as stated, unknown although it is perhaps the general

oc>

Hypothesis (2a) would imply that the rejected food object has neutral properties with regard to influencing subsequent behaviour, hypothesis (2b) that the rejected food object has a similar influence as an accepted food object and hypothesis (2~) that the influence of a rejected food object is antipodal to that of a food object which is accepted. An experiment was therefore designed, using the three-spined stickleback as the predator, to determine which of the three possible hypotheses is preferable. Simultaneously it permitted the analysis of food searching behaviour after acceptance. Methods The five experimental animals were selected from a group of sticklebacks captured during their spring migration, and consequently their prior feeding experience is unknown. They were kept in the laboratory for at least 2 months before use and fed exclusively on tubifex, 4 hr daily, to ensure sufficient experience with the prey species. The light cycle during this period was held at 8 hr day, 16 hr night to prevent the fish from coming into breeding condition (Bagger-man 1957), as this has been shown to have a considerable effect on feeding behaviour (Beukema 1968). Water temperature, which also affects food intake (Beukema 1968), was maintained at 17°C f 1“C. Each fish received food items so selected that the predator/prey size ratio was equal for all fish. This was achieved by assuming that the daily maintenance requirement for each individual fish with this prey type (tubifex) was 12 per cent of its body weight per day (Beukema 1968). A daily ration of this size was given in the form of thirty equally sized tubifex. Apart from this no other selection of the prey was made. As stated above, discovery of a prey is difficult to observe under natural conditions, but precise records of this event are necessary

THOMAS:

STICKLEBACK

BEHAVIOUR

for the present study. This necessitated considerable restriction of the freedom of movement of the fish, and an almost one-dimensional environment was chosen. The width of the field (9 cm) however, was sufficient for the fish to turn around easily, so that tortuosity could well manifest itself. The experimental environment is illustrated by Fig. 2. A large, all-glass aquarium 93 x 27 cm and 29 cm high was divided using grey plastic partitions to form a home cell for each fish behind a large frontal searching chamber. The home cells were separated from one another by glass walls to obviate the effect of isolation. The access passage and holding chambers were entered and left via 4-cm2 grey plastic gates which could be raised and lowered by the observer sitting behind a screen in front of the tank. The searching chamber was subdivided into nine smaller search areas numbered as in Fig. 2. The floor was composed of eighty-one yellow plastic food cups (originally used for making ice blocks) 3 cm2 and 2.5 cm deep. Each of the search areas had nine such cups numbered as in Fig. 2. The bottom of each cup contained a thin layer of sand one to two grains thick, which was insufficient to allow the Tubifex to burrow out of sight and produced a high conspicuousness of the food object, as evidenced by the occurrence of only two non-discoveries. The grey plastic dividers between the search areas were equal in height with the food cups and functioned only to mark the areas visually, The water level was maintained to a depth of 9 cm above the top of the food cups. 82.5 cm . i

Fig. 2. Plan diagram of the experimental environment. Heavy lines indicate opaque walls, fine lines indicate glass walls. Two of the search areas are enlarged showing the individual food cups. LHC and RHC indicate the left- and right-hand holding chambers respectively; G indicates a gate.

ON ENCOUNTERING

FOOD

943

Each fish was placed in its home cell at least 1 week prior to starting the experimental schedule. This consisted of a maximum of ten trials per day or as many trials as were possible up to ten in 90 min. At the start of the daily session, the fish was chased gently from its home cell into the access passage, from where it was allowed to enter the left-hand holding chamber where it was given a 2-min settling period. Meanwhile, the prey situation was set up and then the gate leading to the searching chamber was opened and the fish allowed to enter in its own time. After entry the gate was closed quietly, this marking the beginning of the trial. Each trial lasted a maximum of 5 min or ended 2 min after the last prey had been discovered, if this was within 5 min (5 min proved sufficiently long for the whole of the chamber to be searched). At the end of each trial the gate leading to the right-hand holding chamber was opened and the fish was allowed to leave or was chased out if the leaving time became overly long. From here the fish was brought back to the left-hand holding chamber through the access passage for the second trial. The intertrial interval was 2 min, allowing time to set up the new prey situation. Subsequent to the last trial for that day the fish was allowed to re-enter its home cell and usually fed there for a further 90 min, after which the food, also Tubifex, was removed. This daily schedule was maintained for each fish for 10 days. The fish appeared to be completely habituated to the experimental procedure by the end of the second day and the data discussed here refer only to the last 8 days. In order to minimize the possible effects of learning, the presentation of food items was made as unpredictable as possible with respect to their location and density. This was achieved by manipulation of the prey situation used on each trial. Four prey situations were presented: the search chamber might contain, in any one trial either (a) no food object; (b) one tubifex; (c) two tubifex; or (d) three tubifex. It was randomly decided at each trial which of the situations would be used and into which of the eighty-one food cups each of the food objects would be placed. However, no food cup received more than one tubifex. As there were indications that the behaviour after later discoveries within that trial might be dependent upon what happened after the first, in the cases of SE and SR tracks only behaviour subsequent to the first

944

ANIMAL

BEHAVIOUR,

discovery in each trial was used in the present analysis. Discovery was recorded when the fish, arriving over the prey-containing cup, adopted a nose-down posture and approached the tubifex. From this point on one of the two events may occur. Either the fish will eventually eat the prey, or rejection may occur, either before (visual rejection), or after (post-grasp rejection) grasping. The changes in searching behaviour after visual and post-grasp rejections were so nearly similar that they will not be differentiated between here. The slight quantitative differences will be discussed in a later paper. Zero time for SR was defined as that point in time where approach finally stops and movement away starts. Zero time for SE was defined as that point in time where eating the prey is complete. Zero time for SA was defined as the moment of entry into the second search area to avoid any possible effects from entry. SA includes all trials with the 0 prey situation and those portions of trials with 1, 2 and 3 food objects up to the first encounter with the tubifex. SR is selected from all rejects at the first discovery omitting those either by a second reject, followed or an eat on a second encounter with the same food object. That is, it includes only those rejects which are indisputable ‘clean’ rejects, in the sense that the fish does not even return to the site of the food object. SE is selected from all first discovery eats, omitting those preceded by a visual or post-grasp reject, that is, only indisputable ‘clean’ eats are used. It should be noted that there was no qualitative difference in the changes in subsequent behaviour between ‘clean’ rejects and other rejects, and the same was true for ‘clean’ and other eats. Hence, the selection does not invalidate the conclusion. There are, however, someinterestingquantitativedifferences, but these will be discussed in a later paper. In analysing the results all search paths were used only until the fish either came within two food cups of a food object, or until it reached either end of the searching chamber, that is, if it crossed the boundary between search areas 2/l or 8/9. These precautions were necessary as the proximity of the food object or of the physical limits of the system impose a change in behaviour; this will be shown later. Tracks turning back, if they thereby reached the food object site were, from that point on, also not used in the analysis.

22,

4

The linearity of the set-up facilitated not only observation of discoveries but also measurement of beeline distances. Since the width of the field was well within the scanning range (the reactive perceptual field of Holling 1965) of the fish, for all practical purposes linear displacement (LO) along the longitudinal axis of the set up may be substituted for BD. LD was determined by considering displacements along the left/right longitudinal axis of the searching chamber only, omitting the effects of lateral movements between the three rows of food cups. For example, a move between search area 3, food cup 1 to search area 3, food cup 8, would be counted as a linear displacement of one food cup. The initial direction of movement away from the discovery point was always considered a positive LD. This point applies particularly to SE tracks where 50 per cent of the tracks turned back at the site of discovery. This would have resulted in a zero LD, had this precaution not been taken. This did not affect SR tracks, none of which turned at the point of discovery, or SA tracks which must all start in the same direction. A record of each trial was made by calling out into a cassette recorder the spatial coordinates of the head of the fish at each 5-s interval. These co-ordinates consisted simply of the search-area number followed by the foodcup number. The accuracy of the record of the movement path is largely dependent upon the time interval chosen to record these spatial coordinates. If this is too big, large sections of the track will be lost depending upon the degree of tortuosity. The greater the tortuosity within that time interval, the greater the error in the estimation of the track. The second interval eventually chosen was short enough to make this error negligible by providing small, relatively straight pieces of track. The only exceptions were when the track reversed on itself within the interval, which would not affect LD but would produce an underestimate of track length. In these cases (less than 1 per cent) the actual track length in that interval was used. A glossary of the abbreviations used is given below for convenience : SA, search in the absence of the food object; SE, search after eating the food object; SR, search after rejecting the food object; LD, accumulative linear displacement of the track; Zd, linear displacement of the track per 5 s; LDA, LDE and LD,, accumulative linear displacement of SA, Sn and SR tracks respectively;

THOMAS: STICKLEBACK BEI-IAVIOUR ON ENCOUNTERING FOOD NDA, nose down approach; TL, accumulative track length; tl, track length per 5 s; qt, quotient of straightness of the track. AU statistical tests were taken from Siegel (1956) except where otherwise indicated. Results and Discussion The Net Effect of the Food Object on Searching The accumulative mean LD for S*, SR and SE is given in Fig. 3, plotted accumulatively to allow an easier visual grasp of the effects. This method of plotting LD as pointed out by Smith (1971), has a drawback in that the displacement at the jth interval is very much dependent upon the displacements at j - 1, j- 2, etc. For this reason attention should be confined to the first interval displacement and particularly to the comparative slopes of the curves for St, SR and SE. Figure 3 Indicates clearly that LD, in the fist interval is very much greater than LD,. Conversely, LD, in the first interval is much smaller. The trend that LD, > LD, > LD, was highly significant for each fish (Kruskal/ Wallis test; de Jonge 1963, p. 343 ; for each fish P < 0.01). This implies that, when a food object is discovered and then rejected, the immediate tendency to move away from the site at which the food was rejected is much stronger than it would

0

5

10

15

20

25

30

35

40

45

50

55

60,

time in sets.

Fig. 3. Accumulative mean linear displacement (LO) in food cups of the searchpaths; In the absenceof a food object, S* (187 tracks); after acceptanceof a food object, Se (101 tracks); after rejection of a food object, ,P (32 tracks).

945

have been in the hypothetical case that there had been no food object there. In contrast, the reverse is true with a food object that is discovered and then eaten; the tendency to move away from the immediate vicinity of that site is reduced. This tendency for the predator to move away from the reject site is still present up to 20 s later, as can be seen from the slope of the graph. The behaviour after this 20-s period is influenced by the barrier at the end of the system and will therefore be discussed later. The reduced tendency of the predator to move away from a site where a prey has been eaten decays with time and/or distance, reverting gradually to S* type displacements after approximately 30 s. This suggests therefore that contact with a food object produces different changes in searching behaviour, dependent upon whether that contact resulted in a rejection or an eat. This interpretation may be objected to on two grounds. Firstly, that the data presented reflect differences in motivational state of the fish rather than differential effects of contact with food objects. Secondly, that the rate of LD is not a reliable measure of searching behaviour. First then, we must ensure that the differences in the rates of LD are not due to some previously existing difference in motivational state. For instance, a rapidly moving fish might be more inclined to reject and a slowly moving fish to eat. Figure 4 gives the mean linear displacement before and after reject or eat of those SR and SE tracks where it was possible to obtain at least five intervals prior to discovery. It is clearly seen that there is practically no distinction between SR and SE tracks before discovery. The change in motivational state then is dependent upon discovery and not upon the previous state of the fish. The slight increase in linear displacement from - 10 to 0 s indicates that discovery probably occurred before arrival over the food cup containing the tubifex. The increase in LD over this period could then be due to rapid orientated approach towards the tubifex. The stricter operational definition of discovery used in the experiment was to avoid any uncertainty regarding the event. Each track used in Fig. 4 was tested to determine any differences between the SA (before discovery) and the SR or SE portions. The linear displacements over individual 5-s intervals used for this purpose will be indicated

ANIMAL

BEHAVIOUR,

22,

4

Table L Comparisonof the Valuesof Id Eefore and After Relect

LD 12, 11

Intervals after reject (s)

1c1. 9 8 7 6 5

3

-25

-?O -15

-1cl

-5

15

20

Id after < before

0

1

3

2

Id after = before

2

s

1

2

Id after > before

15

9

10

6

Total no. of cases

17

15

14

10

Table IL CorneM:

2

sets.

10

Wilcoxon Test, on&ailed, P < O-001 that Id after > before for eachof the first four intervalsafter reject.

4

-30

5

I,.....**’ P 5 -1 ‘2,

l ..** 10

15

20

Valuea of Id Before and Intervals after eat (s)

25 30 sets.

5

10

15

20

Id after c before

26

25

21

16

Id after = before

4

6

5

9

Id after > before

1

0

4

4

31

31

30

29

Total no. of cases -6 -7 -8. -9 -10LD Fig. 4. Accumulative mean linear displacement.,of SR (17 tracks) and Ss (31 tracks) before and after discovery of the food object. (Zero = over prewontaining cup).

by lower-case abbreviations (Zd) to differentiate them from the accumulative displacements (LO). The individual Id’s after the event were compared with the mean Id in the same path for intervals - 5, - 4 and - 3, corrected to the nearest whole integer, before discovery. The results for SR are given in Table I and for SE in Table II supporting the conclusions reached with Fig. 3. The second objection highlights the difficulty produced by utilizing a pattern of behaviour, namely locomotion, which is common to many behavioural systems (Smith 1971). Attempts to include other measures of intensity of searching have done this in one or two ways: (1) By using the number of prey attacked (see Holling 1965, for a review). This is obviously dependent

Wilcoxon Test, one-tailed, P < 0@06 that Id after < before for each of the Grstfour intervalsafter eat.

upon the prey density as well as internal factors, and in any case is unsuitable here where we are dealing with that part of searching behaviour between consecutive food objects. (2) An alternative method is to use overt behavioural acts which are specific to searching behaviour. ‘Scanning’ movements have been used for this purpose (Banks 1957; Fleschner 1950; Smith 1971) but these are open to the same objections as locomotion, being common to exploratory behaviour in general. In the feeding situation, however, ‘scanning’ can lead to further inspection of potential food objects as was shown by Beukema (1968). He also showed that the frequency of grasps at inedible objects was directly correlated with hunger and therefore can be considered as related to food searching behaviour. Unfortunately, it was impossible to observe these grasps directly in most cases, although it was possible to identify the nosedown approach (NDA) which occurs before the grasp. This was therefore the measure selected as an indicator of the intensity of searching. The motor components of NDA are identical

THOMAS: STICKLEBACK BEHAVIOUR ON ENCOUNTERINO FOOD to

those of discovery,the differencebeing that

NDA occurs over empty food cups. Figure 5 givesthe frequencyof NDA before

and after discoveryin SE and SR tracks respectively.BeforediscoveryNDA is for interval - 3 to avoid includingthe effectsof discovery. It is clearly seenthat the frequencyof NDA after eating the prey increasesimmediately to almost double that before discovery (x2, Morony 1951;P < OWl) and remainsat this level for at least30 s. In contrast,the frequency of NDA after rejection drops to practically ml and remainslow for approximately15 s before recovering. In conclusion then, after acceptanceof a food object the fish increasesthe intensity of searchingimmediately.This, plus the fact that it movesaway more slowly from the site of the eaten prey, means that searchsubsequentto acceptanceis indeed ‘area restricted’. After

947

rejectionof the food object on the other hand, there is a rapid move away from the reject site plus an initial decreasein the intensity of searching.As a consequence,resumption of searchoccursafter a certain distancefrom the rejectsite hasbeencreated.Rejectionof a food object then, leads to what might be termed ‘area avoidedsearching’. Factors Contributing to the Net Efkct on SR The Id of the predatorin any time interval is

the direct result of the tortuosity of the track over that interval and the track length (t1). The increasein LDR shown above then may eitherresultfrom an increasein tl, or a decrease in tortuosity, or both. An approximation of the real track length was determinedby connectingthe points at which the fish found itself at the beginning and end of each 5-s interval by straight lines, as in Fig. 6. Then tl = J(ld2 + Iadz)

5,

time in sets.

SRtracks

time in sets. Fig. 5. Mean NDA frequency per 5 s before (- 15) and after discovery resulting in acceptance (SE tracks) or rejection (SRtracks). Samedata as for Fig. 3.

where lad is the number of food cups moved laterallybetweenthe threerows of food cups. Figure 7 givesthe mean tl for the SR and S* tracks consideredin Fig. 3. The tl’s of the predator searchingin the absenceof a food object(P) showa slightincreaseovertime which is the resultof a small increasein the proportion of fish which have turned and are retraversing cups already searchedat an increasedspeed. The meandistancecoveredis between1.3 food cups(3.9cm) and 1.9food cups(5.7cm) per 5 s. In the first interval after reject the mean tl increases to 4.8food cups(14.4cm) and remains larger than would be expectedfor searchingin the absenceof a food object for the following three intervals (Mann-Whitney U-test, one tailed; for the first interval, P < 040003). At least part of the differencebetweenLD,, and

Fig. 6. Method of estimating the track. The solid dots represent the spatial co-ordinatesat each 5 s. The solid liue representsthe ‘real’ path of the fish. The broken line connecting subsequent 5-s periods is the estimate of track length Or). The stippled lines represent ld, (the first ;thfee!ss would ail be positive,the fourth negative). e EJmdlc+ed for the first interval only by the broken/ Stlppkd vertical. (see text for further details).

ANIMAL

948

BEHAVIOUR,

22,

4

ii?ld

t’ 5- ., 4.

\

3.

\

Xlt

@ \

where qt is a quotient of the straightness or lack of tortuosity in the average track. A quotient of leOor - 1.0 represents a linear displacement equal to track length and a perfectly straight track. Negative quotients are the result of negative mean Id values, which are produced by a large proportion of fish turning back. As can be seen in Table III, the quotients for SR tracks indicate that there is an increase in straightness of the track after rejection in comparison to S* tracks. In conclusion then, it is apparent that the increase in LDR is due not only to an increase in TL, but also a decrease in the tortuosity of the track. That is, the fish swims both faster and straighter.

0-w.

2.

-\.

1. 5

10

15

I

20

54.

@

3. 2.

5

10

15

20

25

30

35

40

45

50

55

60

Factors Contributing to the Net Effect on SE As can be seen in Fig. 7(b) and (c) acceptance of a food object is followed immediately by a reduction in mean tl (Mann-Whitney U-test, one-tailed; for the first interval SE < S*; P < 040003). At least part of the reduction in LD, then is the result of a decrease in TL. As it has aheady been shown that the intensity of searching increases immediately (see Fig. 5), the temporary near cessation of movement can only be explained as a highly intensive search of the immediate locality of the captured prey. Alternatively, one could consider this as an emphasis of the perceptive elements of searching behaviour at the expense of the motile elements. It is a little more difficult to evaluate the distinction between tortuosity in SE and S* tracks (see Table III). The qt in the fist interval after eat is certainly smaller than that of the first interval in S*. However, qt decreases with time in S*. (It should be noted that qt never assumes negative values, which implies that there is always a tendency for ongoing searching.

5 4.

0

3. 2. 1. . . . . . . ..* l

.

.

5

10

**.-**

l

.,.,.....

l .-**

. . . . . . . . . . . . . . i . . . . . . . . l **

,

15

20

25

30

35

40

45

50

55

= qt

60

time in sets. Fig. 7. Mean track lengths (tr) for (a) P, (b) S* and (c) 278 tracks per 5 s. rl equals the number of food cups traversed. Same data as for Fig. 3.

LD, then, is the result of an increase in the mean accumulative distance covered (TL) after rejection. The influence of tortuosity upon LD was determined by taking one overall measure rather than attempting to measure specific effectssuch as the frequency of turns, bias of turns, etc. Tortuosity may be expressed as

Intervals

qt

1

SR tracks

O-984

0968

O-979

O-937

S* tracks

0.890

O-797

O-740

0.648

0.587

0.384

0.328

0.363

O-260

O-194

0.199

0.294

SE tracks

0.615

0528

O-712

0.658

O-739

O-457

0.544

0.450

0.359

0.111

O-094

o-493

2

3

4

5 -

6 -

7 -

8 -

9 -

10

11

12

-

-

-

THOMAS: STICKLEBACK BEHAVIOUR ON ENCOUNTERING FOOD That is, searching in the absence of a prey is not random.) Any statement that is made regarding tortuositv after eat is dependent upon which interval of Sn is chosen for the comparison. A more valid comparison is possible by selecting only those SE .:acks starting from the same point as SA tracks. With twenty-four SE tracks starting in search area two, qt’s for the first three intervals were 0.578, O-474 and O-560 respectively. These when compared with the qt’s for S* in Table III indicate that tortuosity does indeed increase subsequent to eat. In conclusion, the decrease in LDE is due to both a decrease in track length and an increase in tortuosity. Longer Term Effects of the Food Object on SR and SE

It was stated that movement away from the rejected prey represented by LDR in Fig. 3 was, from approximately the twentieth second on, influenced by the barrier at the end of the chamber. The subsequent course of the graph is the result of two factors. Firstly, a number of tracks crossed the partition into the ninth search area and therefore had to be dropped. A moments reflection will show that this results in an underestimate of the rate of increase in mean LD. This could not, however, cause the decrease in mean LD that actually took place. This was due to a second factor, namely an increase in the frequency of tracks turning back. This turning may be due either to the increasing distance from the food object or to the proximity of the border. Increasing distance from the food object has little influence, if any, within the spatial limitations of these expenments. Of sixteen tracks starting at least ateen food cups away from the ninth search area only three tracks turned back before this area was reached. Of these three all went at least twelve food cups from the reject point and two turned back within one food cup of the ninth search area. To verify the influence of the border, the mean LDR or those SR tracks starting at least fXteen food cups away from the ninth search area, was determined by extracting these tracks from the total SR tracks. The result is given in Fig. 8 and shows clearly that there is a distinct influence of the physical barrier. This led to the consideration of a further point. Up until now we have been considering only the movement of the predator up to the barrier. Subsequent to this the predator must necessarily

949

LD IE 14 1: 12 11 1c 9 8 7 6 5 4 3 2 1 1

5

10

15

20

25

30

35

time in sets. Fig. 8. Mean accumulative linear displacement (LO) for all (32) SR tracks (0) and for 16 SR tracks with at least fifteen food cups before searcharea 9 (I-J). return back through the system, assuming that movement continues. In view of the strong avoidance exhibited by the predator to the location of the rejected food object it seemed of interest to examine the retraversal of the system. Figure 9 therefore presents the continuation of the SR, S* and SE tracks, after turning near the end of the chamber. The zero point for all three is the point at which the fish turned nearest to the barrier, with the stipulations that this must occur in the eighth or ninth search area, and that there was no discovery of a second food object prior to reaching the barrier.

ANIMAL

950

BEHAVIOUR,

It may be seenin Fig. 9 that there is no significant differencebetween the mean LDE and LDA. In fact the displacementsseemto be larger on the whole, than those in Fig. 3 for the sametracks. This is not unexpectedas the predator is now retraversingcups which have alreadybeensearchedThe meanLD,, however, is significantly smaller for the first interval

LD IC!l

22,

4

(Mann-Whitney U-test;P = 0.018)andremains so throughout.This, ascanbe seenfrom Fig. 10 and Table IV, is the result of a decreasein TL and an increasein the tortuosity of the tracks in comparisonto the SAand SEtracks. Not only doesthe predatormove rapidly and directly away from the location of a rejected food object,but evenwhen the onwardpath is blocked, as it is in this artificial environment, the predatorwill not readilyreturnto a location

12,.

tt 4

11I-

3

1Ct-

2 1

9 8 7 6

5I

4-

. 5

lo

15

20

25

v 30

3-

time

-0

in sets.

time in sees.

Fig. 9. Mean accumulative linear displacement (LO) of SR (14 tracks), S* (25 tracks) and SE (18 tracks) subsequent to turning after reaching the barrier at the end of the searchingchamber. Table IV. qt per 5-s Interval for P,

Fig. 10. Mean track lengths (to per 5-sinterval from the barrier back through the searching chamber. (a) SR tracks; (b) S* tracks; (c) Se tracks. Same data as for

Fig. 9.

S* and SE Trndrs !kom tbe Bedary samed8taa!sinF4&9

Back Through the Search@

Chamber.

Intervals qt

1

2

3

4

5

6

SRtracks

0.744

0.826

04351

0660

0.191

0.376

SAtracks

o-912

O-899

0948

0.891

0.513

0*344

SE tracks

0.914

o-946

O-965

0.612

0.634

-0,587

THOMAS

: STICKLEBACK

BEHAVIOUR

where the food object was rejected. This, it is felt, strengthens considerably the concept of ‘area avoided search’. Some Further Implications In this paper we have looked at the effect of acceptance or rejection of a food object upon the subsequent search path by comparing this both to the search path in the absence of a food object and to the search path prior to discovery. It was found that this path is dependent upon whether the prey was eaten or rejected. If the prey is eaten the fish tends to move away more slowly, and simultaneously it searches more intensively. This is consistent with the idea of ‘area restricted searching’ which seems to be a general response on the part of many predatory species (see introduction and Smith 1971, for a review). It is presumed that the response is functional in that it increases the efficiency of the predator in discovering further prey items of that species. However, this response can only have this effect if the distribution of many potential prey species is such that the probability of discovering a second prey item is greater near to, than far from, the first prey item. That is, that the distribution of potential prey species is more or less aggregated, as suggested by the work of Taylor (1961) and the arguments of Royama (1970) and Smith (1971). Southwood (1966) gives a more complete review. In contrast, a discovery resulting in rejection (generally ignored in the literature) produces clear-cut changes in searching behaviour which are antipodal to those of discovery resulting in eating. It produces a rapid direct move away from the site of the rejected prey accompanied initially by a decrease in the intensity of searching. This whole process has been termed ‘area avoided searching’. It is reasonable to assume that such a response wouldincrease the probability of leaving a locality in which a given type of prey was aggregated and thus increase the likelihood of encountering a different type. Such a response might then facilitate variation in the diet, or at least promote sampling of other prey available within the environment. While the idea is, in itself, attractive, it is at best speculative and must await further experimental evidence concerning questions arising from the results of the current experiment. This question of functionalism cannot be clarified until we examine the causality of the rejection process in more detail. Theoretically

ON ENCOUNTERING

FOOD

951

three categories of causal factors are available. Rejection may be the result of: (a) satiation; (b) interference from other motivational varrables; or (c) cue deficiencies of the food object at the current motivational level; (a) and (b) place a stress on motivational aspects, (c) on stimulus aspects. (a) Since the total intake over the experimental period was approximately half that needed for the daily maintenance requirement of sticklebacks, satiation seems to be an unlikely causal factor in this case. (b) As the experimental precautions were aimed at reducing all sources of interference as far as were feasible, interference also seems unlikely. However, as relatively little is known about interaction processes that may change motivation we cannot completely preclude this as a possible factor in our case. (c) The third alternative involving cue deficiencies of the food object becomes a distinct possibility and should therefore be investigated further. It must be determined whether it is that individual food object which is deficient in some way at that moment or whether this is applicable to all individuals of that species. There were some indications in the current experiment that the second alternative is the most likely case, as rejection of one individual enhanced the probability of rejection of subsequent prey items, at least of the same species. This is also supported by the results of an earlier experiment where fish showed an immediate tendency to search in the water column and to make grasping movements at small inedible pieces floating in the water after having rejected tubifex. As these fish had, prior to the experiment, fed on daphnia this could well be interpreted as searching for another prey type. Some of the questions arising here will be considered in a later paper. To complete the picture we should also consider the possibility of stochastic fluctuations in feeding motivation itself influencing the decision to accept or reject an individual food object after discovery. If such fluctuations are acting in this way one might expect that, for example, a reject is more or less ‘out of context’ with the behaviour that is expected towards cue properties of the food object which is present. Finally, no precise formulation of the function of ‘area avoided searching’ can be complete before an accurate picture of the distribution of prey species in the field is available, not only

952

ANIMAL

BEHAVIOUR,

22,

4

with respect to the tendency to aggregate, but also with respect to the relationship between distributions of prey species.

Hollmg, C. S. (1965). The functional response of predators to txey densib and its role in mimicrv Fadspiatibn regulation. Mem. Em. Sol.

Acknowledgments

Jonge, H. de (1963).Zrdeiding tot de Mediscbe Statistiek, Dee1 1, 2e druk, Verh. Inst. prev. Geneeak.,41, 1422. Laing, J. (1937). Host tiding in insect parasites. I. Observations on the finding of hosts by Alysia man&cator, Mormoniella vitripennis and Tricho-

Dr R. Drent, Professor dr. L. de Ruiter, Dr J.

Reddingius, and Dr. P. R. Wiepkema criticized the manuscript and offered many valuable comments. I am grateful to Dr J. Reddingius for his help with statistical tests and computer programming. Particular acknowledgment is due to Professor de Ruiter for his critical comments over the design and conception of the experiment. I would also like to indicate my thanks to Mr D. Visser for preparing the graphics and to Mrs J. Poelstra-Hiddinga for typing the manuscript. This work forms part of a thesis carried out at the University of Groningen. REFERENCES Baggerman, B. (1957). An experimental study on the timing of breeding and migration in the threespined stickleback (Gu.rterosteusactdeatus L.). Arch. neerl. Zool., 12, 105-317. Banks, C. J. (1957). The behaviour of individual coccinellid larvae on plants. Br. J. Anim. Behav., 5, 12-24. Beukema, J. J. (1968). Predation by the three-spined sttckleback(Gasterosteus aculeatus L.). Behaviour, 31.1426. Croze, I-I. (1970). Searching image in Carrion Crows. 2. Tierpsychol., Beiheft 5. Dixon, A. F. G. (1959). An experimental study of the searching behaviour of the predatory coccinellid beetle A&ha

decempunctata (L.). J. Anim. Ecol.,

28, 259-281. Fleschner,C. A. (1950).Studieson the searchingcapacity of the larvae of three predators of the citrus red mite. Hilgardia, 20, 233-265.

.

.

gramme evanescens. J. Anim. Ecol., 6,298-317.

Mitchell, B. (1963). Ecology of two Carabid beetles, Bembidion lampros (Herbst.) and Trechus qua&istriatus (S&rank). I. Life cycles and feeding behaviour. J. Anim. Ecol., 32,289-299. Morony, M. J. (1951). ‘Facts from Figures.’ Penguin Books Ltd., Middlesex. Prop, N. (1960). Protection against birds and parasites in some species of Tenthredinid larvae. Arch. n4erl. Zool., 13, 380-447. Royama, T. (1970). Factors governing the hunting behaviour and selection of food by the great tit (Parus major L.). J. Anim. Ecol., 39, 619-668. Ruiter, L. de (1967). Feeding behaviour of vertebrates in their natural environment. ZZanabook of Physiology, Section 6: Alimentary Canal, Vol. I: Food and Water Intake, Chapter 7,97-l 16. Washington, D.C. : PhysiologicalSot. Siegel, S. (1956). Nonparametric Statistics for the Behavioural Sciences. New York: MacGraw-Hi& Smith, J. N. M. (1971). Studies of the searching behaviour and prey recognition of certain vertebrate predators. D.Phil. Thesis, Oxford. Southw;$huzn R. E. (1966). Ecological Methodr. Taylor, L. R. (1961).Aggregation, varianceand the mean. Nature, Land.. 189, 732-735.

Tinbergen, N., Impekoven, M. & Frank, D. (1967). An experiment on spacing out as defenceagahrst predation. Behaviour, 28, 307-321. (Received 9 October 1973; revised 18 March 1974; MS. number: 1259)