THE PREDATORY BEHAVIOR OF THE GOLDEN-WEB SPIDER NEPHILA CLAVIPES ( ARANEAE ARANEIDAE ) BY

INTRODUCTION Details o.f the role of wrapping behavior in the. predatory activities of Nephila clavipes (Linnaeus) were given by Robinson, Mirick & Turner (I969). Their account also gave broad details of the total behavior of this species. At that time, the publication of an .exhaustive account of the predatory behavior of N. clavipes was intended and anticipated. Since then, however, the senior author has carried out studies, of the behavior of other species of Nephila in Africa and Asia,, and with co-workers is currently engaged in a study of the behavior and ecology, o.f Nephila maculata (Fabricius) in New Guinea, It now seems appropriate for us to leave details of part of our work on N. clavipes for inclusion in a broad comparative paper and publish here those, aspects which relate most directly to the main points cited in outline by Robinson, Mirick & Turner (I969). This paper therefore presents a summary model of the predatory behavior of Nephila clavipes, based on the. investigations of the present authors in the summer of I968 and further experiments carried out by the senior author in I969. We give emphasis to the investigations and experiments that led to establishing ,some ot the major aspects of the model and leave detailed descriptions of behavior units and behavior sequences for inclusion in the projected comparative paper. We have also left consideration o.f the temporal aspects of the behavior sequences for inclusion in the later paper.


INTRODUCTION
Details o.f the role of wrapping behavior in the.predatory activities of Nephila clavipes (Linnaeus) were given by Robinson, Mirick & Turner (I969).Their account also gave broad details of the total behavior of this species.At that time, the publication of an .exhaustiveaccount of the predatory behavior of N. clavipes was in- tended and anticipated.Since then, however, the senior author has carried out studies, of the behavior of other species of Nephila in Africa and Asia,, and with co-workers is currently engaged in a study of the behavior and ecology, o.f Nephila maculata (Fabricius) in New Guinea, It now seems appropriate for us to leave details of part of our work on N. clavipes for inclusion in a broad compara- tive paper and publish here those, aspects which relate most directly to the main points cited in outline by Robinson, Mirick & Turner (I969).
This paper therefore presents a summary model of the predatory behavior of Nephila clavipes, based on the.investigations of the present authors in the summer of I968 and further experiments carried out by the senior author in I969.We give emphasis to the investigations and experiments that led to establishing ,some ot the major aspects of the model and leave detailed descriptions of be- havior units and behavior sequences for inclusion in the projected comparative paper.We have also left consideration o.f the temporal aspects of the behavior sequences for inclusion in the later paper.

Materials and Methods
Our basic observations and some of our experiments were car- ried out with captive adult emale spiders.The spiders were not confined to cages but were released in a large screened insectary at effects of experience we avoided successive presentations, of the same type of prey.In all cases at least two days (usually more) inter- vened between one presentation and the next presentation of the same type.Usually another type of prey, or several other types of prey, would be presented between presentations of the same type.
In addition we made presentations of all the prey types to a large number of free-living spiders as a check on the behavior of our captive spiders.
In the course of our initial observational work we presented the following prey items: moths (living & dead), grasshoppers (living), crickets (living & dead), Tenebrio beetles (living), dragonflies (liv- ing), Trigona sp.(living), pentatomids (living), and blowflies (living).Later when we attempted to elucidate the stimuli which the spiders were capable of detecting at various stages in the process o predation we used a number of experimental techniques involving modified insects in the form of 'dummies'.These techniques are de- tailed in the appropriate section below.
All the insects that we used were weighed and measured before being presented to the spiders.Where insects were presented dead they were killed by freezing to avoid contamination by chemical killing agents.Behavioral observations were supplemented by cine- photography and the duration of behavior units was recorded on a Rustrak multi-channel chart recorder.

NATURAL HISTORY
In Panama, and wherever we.have encountered Nel)hila clavil)es in South America (Colombia & Venezuela), the spider is most fre- quently found at the edges of orest clearings, alongside forest trails and across forest streams and watercourses.It thus apparently ex- ploits areas within the forest that are. in all probability flightpaths of insects.The structure, size, and siting of N. clavil)es webs is consistent with the view that the spider specialises on prey that are in flight above the.herb layer rather than moving about in it.Al- though the web may be sited with part of the prey capture area within the herb layer it is most frequently above this or stretched across gaps in the vegetation.The structure of the web has been described by Gertsch (948) and Peters (I954, I955).For its area, which is large, the web has a very fine mesh which is far less penetrable by small insects than the.much smaller but coarse-meshed web of Zlrgiope argentata (Fabricius), as was shown by Robinson, Mirick & Turner (969).The web of the adult spider is.not a complete orb but is U-shaped with the hub very close, to the upper bridge thread.Kaston & Kaston (953,p.I76) give an excellent figure of the N. claviI)es web.Webs of several adult spiders are often built in close, proximity and fairly large aggregations of the spider may occur in apparently favourable areas (Shear I97O, has commented briefly on this phenomenon).
The adult web is frequently equipped with a barrier web con- sisting of a complex of strong lines arranged in a non-symmetrical manner above and/or below the main plane of the orb.There is considerable variation in the structure of barrier webs.Robinson   & Robinson (97o) have suggested that they may function as early warning devices enabling the spider to detect the approach of pos- sible predators.Certainly the spider often responds to.manipulation of the barrier web by escape or other orms of defensive behavior.
Web renewal is not a daily occurrence, and in captivity the spider may only renew part of the web at a time.(This is also certainly the case with free-living N. maculata but we have not made exten- sive observations on web renewal by free-living N. clavil)es).
Our observations on the natural prey of N. clavil)es, although limited in scope, confirm the deductions based on web structure and siting i.e. that flying prey of small to medium size may be the Psyche September speciality of this species.We found flies, bees, wasps and small lepi- dopterans to be predominant in the prey found in webs and amongst the corpses that are occasionally suspended from the barrier web after they have been consumed by the spider.Figure is a summary diagram of the complex of possible se- quences of predatory behavior that we have observed to be given by adult female N. clavipes to a variety of prey presented in several ways.It has the same summary function that Figure 2 of Robinson (969) had in relation to the behavior of gtrgiope argentata and is not meant to be a flow diagram or relate to any cybernetic con- ventions.It does, however, illustrate the temporal sequence and relationships between behavior units, their degree of association, and the effects, of some stimuli on the course of the sequence.Like most models it is certainly much more simple than reality.
In the predatory behavior of N. clavipes alternative behavior units are available at several functional stages and the 'decision' to employ one or other of these is shown, on the diagram, to.be. the result of a discrimination.We recognise that in some cases our assumption of the basis (or bases) for the discrimination may be an oversimpli- fication; this matter is discussed at the appropriate place in the text.
The diagram .employssimple conventions.The behavior units are shown as circles connected by lines.The arrows on the lines show the direction of change from one behavior unit to the next.In the upper half o.f the diagram two conventions are used to denote 'choice' points.Some of the circles are divided into two halves by vertical lines so that two behaviors may follow the behavior shown by the circle.In addition, here and throughout the diagram, small square boxes on the lines connecting behavior circles represent places where behavior may be .switchedfroaaa one course to another.Dotted lines from oblong boxes suggest imputs to the system that are de- pendent on stimulus properties of the prey.
Behavior occurring up to first contact with the prey is shown in a much simplified form.The activity spider on hub (which could be amplified a.s 'spider at hub in predatory position waiting for prey') is represented as an ongoing activity by a vertical line beneath the behavior circle..If a prey item, or in fact any item :zbove a certain weight, strikes the web the spider is alerted (the square box has an imput t:rom prey strikes web and an output to spider alert, this represents a diversion from the ongoing activity spider at hub).The overt behavioral change following impact may be a momentary in- crease in the flexion of the spider's legs prior to an almost immediate movement towards the prey or a more sustained adoption of this alert position.

Psyche [September
After the spider has been alerted two things may occur.If the prey is in a state ot: sustained vibration the spider almost always makes an immediate approach to it.If on the other hand, the prey is immobile or merely making spasmodic movements the spider usually plucks the web.If the prey has remained in the web until the spider plucks, this behavior is likely to be followed by an approach to the prey (during which the plucking movement may be repeated).
When the prey escapes after the initial alerting impact, the spider usually returns to its pre-alerted state after plucking.
As examples o the simplication involved in our model we can cite the ollowing examples that occur in the airly complex ,section that we have traced so far.Thus a urther alternative behavior that can occur on the impact o prey is not shown.Instead o( the spider being alerted it may show an immediate escape response.This usually takes the (orm o the spider running upwards rom the hub, out of the web and onto the support lines (or even onto nearby vegetation).Similarly, i it did not result in an illegibly complex diagram, we could show at least three distinct approach-to-prey behaviors.There is a rapid, unhestitant, approach that is made to rapidly vibra.tinginsects and a hesitant, much interrupted by plucking, approach to non-vibrating insects.In addition very large or very heavy insects are approached in a slow 'deliberate' manner, in which legs I & II are flexed ar back over the prosoma in a very characteristic 'cautious' gait.
Ater arrival at the prey the spider may immediately attack with- out a perceptible pause, or may touch (with the tarsi) and palpate (with the pedipalps) the prey, before attacking.We have not shown these 'investigative' stages in our diagram.
There are three basic attack behaviors, all o which involve the use o the chelicera.e.In no case have we ever seen Nephila clavipes (or any other species o( Nephila, or that matter) use the strategy o attack wrapping.This matter is extensively discussed by Robin- son, Mirick & Turner (969).The three basic :orms of biting attack are as follows: . An attack similar to the seize and pull out beha.vior of Argiope argentata (see Robinson & Olazarri, 97).This is given to very small or light prey.
2. A long bite in which the bite is not immediately followed by pulling out movements but is sustained in sit;u.This long bite may be accompanied by a special posture in which legs I & II a.re raised off the web and the opisthosoma is raised at its apex so that the body presents a concave dorsal aspect.
3. The third attack strategy we call bite & back off.It consists of a rapid forward lunge, a short-duration bite and a rapid with- drawal to a distance at which the prosoma i.s well away rom the prey.We have shown this behavior as a repeatable unit since the lunge, bite, retreat sequence may be repeated a large number of times (we have records of over 2 repetitions) before a sustained long bite ensues.In many cases, when this i:orm of attack is em- ployed, the ,spider can be seen to open the chelicerae until they are almost' held horizontally, before making the rapid forward lunge that terminates in the bite.
The type of attack strategy that the spider employs seems to be largely determined by parameters o.f weight and/or size of prey; our experimental analysis of these factors is described later.We have not been able to determine what actor or factors mediate the decision to cease repetition.sot: bite & back-off and ,commence sus- stained biting.This process is not dependent on reduction of activity by the prey (that might be consequent on a series of bites) since it occurs in the spider's behavior towards large dead prey.
Behavior ollowing the initial attack phase is .somewhatmore complex: than is the case with araneids that are efficient at enswathing their prey in silk.Attack is almost always terminated by pull-out movements.The body o the spider is lo.wered, on flexed legs, during the biting attack and pulling out consists o stro.ngextensions o the leg pairs.These result in the spider pushing do.wn on the web and pulling up on the prey.Very small prey, adhering to a small area ot: viscid spiral, are quickly reed, as are lepidopterans which do not adhere strongly because of their loose wing scales (see Eisner el: al 964).Other prey may be .subjectedto repeated pull out m6vements before being reed.Prey that are not readily reed by pulling movements are wrapped in the web and then cut out by the ide.r.Robinson, Mirick & Turner ( 969) called this wrapping at the capture site, Type z.We have been able to show that the spider can be induced to wrap prey (in this way) if its pulling out attempts are blocked experimentally (see later).It seems possible that the pull out movements ena.ble the spider to gauge the degree of adhesion o the prey to the web and that the 'decision' to continue pulling, or to wrap in situ and then cut out, is influenced by this information.A urther complication arises rom the act that the spider may wrap, at the capture site, prey that have already been reed rom the web by pulling (Post-immobilization wrapping at the capture site, Type 2, of Robinson, Mirick & Turner I969).
Psyche [September (The functional interpretation of these behaviors is that pulling prey from the web results in less web damage than wrapping in situ and subsequent cutting out.This can easily be substantiated by com- paring the web damage resulting from the two techniques of prey- removal.Wrapping after removal from the web has a trussing effect which reduces the bulk of the prey and may facilitate its transportation to the feeding ,site.) After the prey is freed from the web, and trussed in some cases, it is transported to the hub, of the web.Here the spider employs one of two techniques.It may either carry the prey held in the chelicerae, perhaps supported by one or both of the fir.st legs, or it can carry the prey package on a silk thread hanging from the spin- nerets and supported by one or both of the fo.urth legs.Prey carried to the hub in the jaws are wrapped on arrival at the hub and then suspended from the hub silk as the spider turns to assume its preda- tory posture.Prey that are carried suspended en silk are not wrapped on arrival at the hub but are suspended.Nephila clavipeg does not store prey at the capture site but carries all prey to the hub where it is hung until previously caught prey are consumed.Very small prey, carried to the hub in the jaws, may not be wrapped on arrival (but prey as small as stingless bees o-3omg in weight are regularly wrapped on arrival at the hub).
The return to the hub from the capture site is carried out in a iorwards direction after the spider has turned to face the hub at the capture site. (Nephila maaulata requently backs slowly up the web, without turning, when carrying prey in its jaws.)When the spider has assumed its normal head-down predatory position it may undertake more or less extensive grooming activities before taking up the prey in its jaws and anterior legs.These are very similar to those that Robinson & Olazarri (97) described for Argiope argentata.Prior to the commencement of feeding the spider often carries out extensive manipulations of the prey during the course of which small bites are given to region after region o.f the prey body.
We have a few records of prey being wrapped after removal from the web and then being transported in the jaws.Most prey that are trussed in this way are then carried suspended on silk behind the spider.Heavy prey are also carried in this wa.y,but as in the case of drglope argentata, the weight threshold for the changeover from carry in jaws to carry on silk varies from individual to individual, and from time to time within individuals (see Robinson I969, p. I70-I).

Experimental investigations
To investigate the possibility that certain aspects of the predatory pattern are responses to relatively siniple stimuli we carried out a number of investigations using 'modified' insects as dummies.\Ve manipulated the size and weight of some insects and also modified such parameters as the strength of their adhesion to the web.We were able to compare the responses of the spiders, at various stages in the predatory process, to such dummies, using adequate controls or using their response to unmodified insects as a baseline.
(a) The bite & back-off attack behavior Our observations on the responses of N. clavipes to a range of prey items (s(e page 123, above) showed that the bite & back-off attack behavior was only given to large and heavy prey.Since all of these that were presented to the spiders were alive there was a possibility that the response could be to size, weight or specific activity of large heavy prey, or a combination of any of these factors.Preliminary tests showed that the response was given to dead (immobile) prey so that although it remained possible that activity could enhance the response it was not the important stimulus.We then decided to manipulate the parameters of size and weight.Using.small acridiids (25-3omm, 4oo-55omg) as prey, we added lead shot to some to double the weight (approximately) increased the length of others by inserting a wooden tooth pick in line with the long axis, and with the third group we increased both the length and the vceight.The dead dummies were presented at right angles to the radii of the web.This ensured that the weight of prey was dis- tributed over as wide an area of web as was covered by the length of the insect (or the insect + toothpick).In t:act it meant that the maximum dimension of the prey was at right angles to the spider as it approached across the web.The insects were vibrated elec- trically at 25o cps until the spider left the hub on its predatory excursion.At that stage the vibrator was .switched off so that the prey was motionless when the spider came in contact with it.The results of this experiment are shown in Table I.In addition to the orm of biting attack we noted whether the spider raised legs I & II off the web during the attack (see page 127, above) and whether the prey was wrapped in situ, or free wrapped, after the attack.There was a significant increase in leg raising during the attack in the case ot: the weighted insects.There was a slight numerical, but not a statistically significant, increase in the number of attacks on the long dummies that involved leg rai.sing.There were three attacks out of ten, on heavy insects, that involved the bite and back off behavior.Ten spiders chosen at random from wild population receive A, ten captive spiders B ten wild spiders C. A is acridiid with weight added and lengthened (total weight ca.1000 mg length 70 mm).B is acridiid lengthened to 70 mm weight 400-550 mg.C is unmodified length 25-30 mm weight 400-550 mg.

Psyche
Statistical analysis--Fisher's exact probability: differences between A & B, A & C, B & C for Bite and back off are not significant; difference between A & B significance level 0.0.5, difference between A & C significance level 0.005, no signficant difference between B & C, all for Bite, legs raised; for wrap in web difference between A & B and B & C level of significance 0.05.

All levels one-tailed
Although this number is not statistically significant we regarded the occurrence o.f this behavior in attacks on the weighted insects as being highly suggestive.We then carried out a further series of experiments, using similar sized acridiids, in which we quadrupled their weight.
We found that very heavy dummies of small size often dropped out of the webs before the spider reached them.We therefore in- creased the length o.t: both experimental (very heavy) dummies and controls (normal weight) by adding the toothpick to the insect.This worked in a perfectly satisfactory way to distribute the weight over a greater number of web members.As before, we vibrated the dummies until the spider commenced its predatory excursion.The results are given in Table 2.There is a statistically significant effect of weight on the occurrence ot: the bite & back-off attack behavior.Note that we did not induce thi.s attack behavior in all the presentations; we consider that with large, heavy active insects there may be some heterogeneous summation.We attempted to test or the possible additive effect of activity, at the moment of contact Prey A weights 1700-1800 mg.Prey B weights 400-550 mg.
Statistical analysis--Fisher's exact lrobability: Difference between A & B for bite and back off, level of probability 0.025; difference between A & B for bite, legs raised, level of probability 0.01; difference between A & B for wrap in web not significant; difference between A & B for quit the web not significant.
All levels one-tailed with the prey, by tapping weighted and control dummies rom behind the web.(We did this when the spider was in tarsal contact with the prey and before it had attacked).This led to a numerical, but non statistically signiiqcant, increase in the number o bite & back of/ attacks.It is very difficult to standardize simulated prey move- ments.
A by-product o these expei'iments was the suggestion that the wrapping o prey in situ might be a response to the ailure o the spider to pull out the prey.The number o wrap in situ responses was significantly greater in the case of the artificially lengthened prey used in the first experiment.These were, to the observer, obviously a much greater problem to the spider at the pull out stage.We therefore decided to carry out some experiments to see i increasing the adhesio.no the prey to the web, or its 'apparent'   adhesion, would af-fect the spider's behavior at this stage in the predatory process.

(b) Pulling out and wrapping behavior
As a first simple experiment we used domestic crickets as prey.We simply presented 2o dead unmodified crickets and scored the Psyche [September

Wrap in web
Pull out alone Winged crickets 18 Normal crickets 19 * Spider disappeared before completion of pair.
Behavior to live dragonflies matched in weight with crickets.

Wrap in web
Pull out alone 20 0 Time of persistent pull out attempts (before inception of wrap in situ.) averages.
Winged crickets Dragonflies 194.5 secs 68.7 secs No result 1" nmnber of them that were wrapepd in situ and the number that pulled out and subsequently wrapped at the hub.We then matched the sizes and weights of these crickets to a second set to which we attached thin paper 'wings' at right angles to the long axis of the body.These dummies were presented in such a way that the 'wings' greatly increased the surface adhering to the web.We then scored the number of wrap in situ and wrap at hub responses.The results are shown in Table 3.The increased adhesion resulted in a highly significant increase in the number of prey that were subjected to being wrapped and then cut out rather than being pulled out.The spiders made very persistent attempts to pull out the winged crickets, two succeeded and the remainder averaged I94.5 seconds of abortive pulling-out attempts.This is very interesting since the spiders started wrapping dragonflies (of lower weight) after only 69.7 seconds of pulling out attempts.(Testing these two sets of data-pull out times for dragonflies and winged crickets, with the Mann-Whitney U test, shows that the difference is significant ;p is less, than o.oo).
We also carried out a further experiment on this aspect of the predatory process.In this case we passed a thread through the thorax of the crickets and presented twenty crickets vith the thread hanging below the insect and twenty in which we passed the thread through the web and then held it from behind.In the second case we were able to prevent the spider from pulling the..cricket from the web by exerting a force in the opposite direction.The results are shown in Table 4. Again the spiders that were unable to pull the prey from the web wrapped it at the capture site and then Experimentals were held by string from behind the web.
it free.In this experiment the spiders made extended and persistent efforts to pull the prey before initiating the wrapping behavior.
The process of free wral proved to be much more difficult to elicit under experimental conditions.We had observed that thi.s behavior occurred most frequently in the treatment of butterflies and moths.These could be freed from the web by pulling but were, nonetheless, bulky and cumbersome insects.Other cumbersome in- sects .suchas dragonflies adhered .strongly to the web, were wrapped and then cut free of the web, and were thus largely trussed and packaged at this stage.They were not therefore suitable for experi- ments on free wrap behavior.We reasoned that free wrapping was a response to insects that .couldbe removed from the web by pulling but were too bulky to be transported without becoming entangled in the web.The spider makes movements during the pulling out process that could enable it to gauge the bulk of the prey as it is removed from the web.These movements involve legs I & II.The tarsi are passed along the adhering margins of the prey and ease them away from the viscid elements of the web.Our observations on the removal of butterflies and moths from the web suggested that the cumbersomeness of detached prey was, to a large extent, a function of apparently chance factors.Thus it was partially dependent on the point on the prey body at which the spider exerted the pull out movements (i.e. the point at which the prey was held in the jaws), and partially a result of the orientation of the wing and body sur- faces in relation to the sticky elements of the web.(Some idea of the complexity of these factors can be gained by visualizing the process of picking up a randomly cast down book, by the spine, from a sticky surface.A winged insect, like the book, may open in a variety of ways, depending on where it is seized and where it is stuck down.) Starting from this point we tried to present a series of moths to the spiders with the entire dorsal surface of their wings adhering to the web but at differing orientations to the radii (and therefore the hub and the spider).We hoped that these would be seized at different points on the body and that our sole experimental variable Psyche [September would be the 'apparent' bulk of the prey beneath the spider during the prey-removal process.Our aims proved to be extraordinarily difficult to achieve and the results, although suggestive, are not convincing.In an attempt to manipulate the bulkiness o a series of presentations of butterflies (Anartia sp.) we trimmed the wings o.f half of them down to small (5mm long) stubs.We then matched the weights of pairs of intact and mutilated specimens and presented the pairs, successively, in random order to the spiders.Only the intact insects elicited any free wrap behavior, and only 24% of these were so treated.This result is not statistically significant.If the number o free wrap occurrences is compared not with the total number of presenta.tions of intact butterflies, but with that number less the number wrapped in situ or subjectively scored as presenting below average bulk on removal from the web, the result is signifi- cant.However this depends on our subjective assessment of the bulk and is unsatisfactory.
These results do not enable us either to accept or reject, with confidence, the hypothesis that the free wrap response is related to the bulkiness of the prey after its removal rom the web.We have yet to design an adequate test for this.

DISCUSSION
A number of eatures of the predatory behavior of Nelhila claviles are of interest from the comparative standpoint.The most im- portant of these, in our view, are the total reliance of the species on biting, as an attack strategy, and the fact that the spider does not store prey in the web at the capture site.Both these features represent marked differences from the behavior o araneids belonging to the genera Argiole, Araneus and Eriolbhora.Reliance on attack (immobilization) wrapping, as the principal means of attack, prob- ably extends to a much greater number o araneid genera.
These aspects of the predatory behavior of N. claviles have been discussed in some detail by Robinson, Mirick & Turner (969).
These authors suggested that "advanced" spiders would obtain at least two adva.ntagesfrom the addition of immobilization wrapping to their behavioral repertoire.They would be enabled to attack large and/or dangerous prey without closing to the potentially dangerous contact distance involved in biting, and also.could achieve a considerable economy in time spent at the capture site in .subduing the prey.The bite & back off attack behavior, that we have de- scribed for N. claviles, immediately suggests to the observer that it is a danger-avoiding device.Our experiments on the stimuli that I37 evoke this behavior gave results which are entirely consistent with this view.In t:act increasing the weight of the prey eventually results in the suppression o attack behavior and its supercession by escape.We obtained evidence that N. clavipes would lose some large prey by escapes, during the .prolongedprocess o a bite & back off attack.Prey of similar size presented to the much smaller Argiope argentata were not lost during wrapping attacks in any of our presentations.(The senior author .hasobtained similar results in comparing the performance of Nephila maculata and Argiope aemula (W'alckenaer), dealing with large acridiids and melo- lonthid beetles.)We have also recorded injuries inflicted on N.
clavipes during biting attacks on prey that had biting mouthparts and have one record o injury following the defensive kicking of an acridiid.The economy in time spent at the capture site that is a potential consequence o attack wrapping was examined in detail by Robinson, Mirick & Turner (I969), who suggested that a major factor in this economy was the ability of the spider to leave a,n attack-wrapped prey in .situafter delivering a .shortbite.These authors argued that once the prey was wrapped the spider could sat:ely leave it and not transport it to the hub until the bite had taken effect.With biting attacks, on the other hand, the bite could not be terminated until the prey had been saely subdued by its effects.The economy in time that results from this process is greatly exaggerated if one compares time ,spent in bite & back off attacks with time spent in wrapping attacks on similar prey by other spiders.(This comparison will be made in the projected comparative paper; it was not made by Robinson, Mirick & Turner (ibid) because the data for the attacks of ztrgiope species on very large prey was not then available).
It is interesting that although N. clavipes wraps prey at the capture site it does not store them there.Once this type of wrapping be- havior has evolved it would .seembut a short step to utilize it to enable the spider to interrupt the predatory process after the attack phase and defer the removal of the prey from the web, and its transportation, until later.Thi.s .step,according to Robinson, Mirick & Turner (ibid) would be advantageous in circumstances where large numbers of prey arrived in rapid succession, or where prey left at the hub during an attack might be in danger of being stolen by kleptoparasites in the absence of the spider.At the time these authors suggested that N. clavipes may transport all prey to the hub because the depredations of kleptoparasites might be more diffi- cult to detect if prey were stored at a number of capture sites in a Psyche [September very large web.The arguments dependent on the effects ot klepto- parasites are not easy to resolve.We have seen prey stolen irom the hub, during the ab.sence of the spider, on many occasions.We have also seen N. clavipes respond to, attack, and eat, kleptoparasites that were moving about near to the periphery .ot the web.Other hypotheses to explain the absence of capture-site ood storage by N. clavipes include the possibility that this spider has not evolved a sufficiently efficient wrapping technique to safely allow the storage of prey in situ, or that the presence ot prey packages in an already fine meshed web might render it more conspicuous, and hence avoidable, to flying prey.
It seems worth stressing the fact that N. clavipes can be induced to wrap prey in situ if these are made dicult to remove trom the web by pulling.This simple unction of post immobilization wrapping at the capture site may have been obscured by the fact that such wrapping can serve other unctions in the predatory strategy of other araneids.The function of simplitying the safe removal of prey from the web seems to us, on a priori grounds, to be basic and probably primary.Similarly the existence of free- wrapping behavior suggests that the trussing or packaging function ot wrapping is an important one in its own right, and not merely the useful by-product of a process serving another function.Both these opinions derive ro.m o.ur study of the behavior o N. clavipes, and, as far as we know, were not anticipated by earlier studies o more "advanced" araneids.
If we now consider the model shown in Figure our earlier comment that this represents a very considerable simplification can now be expanded.We have detailed some ot the behaviors that are not included in the model on pp.132-134.In our account o the ex- perimental side o.f our studies it is obvious that the investigation of the, effect of external stimuli on the course of predatory sequences is not complete.Although we have shown that some behavior units can be brought into play in response to simple stimuli we have not shown that these are the only effective stimuli in all cases.We have 'made no progress at all in investigating the effect of internal factors on the behavior of the spider.In all these respects the, model is inadequate, although it is already quite complex.
XVe have also ignored any discriminations that may be involved in the termination of acts of behavior after they are brought into play.The spider must, for instance, both .startand stop the process of bite & back off, and start and stop the process of pulling out the, prey.Peters (93, I933a, 933b), in elegant studies of the behavior o Araneus diadematus, was able to show that the change in stimula, tion brought about as a result o one behavior could be the trigger or the next behavior in the sequence.This approach has not yet been made in the case o N. clavipes.It should be productive.

Table 2 .
Acridiids weighted and unweighted all lengthened to 70 mm.