Toxins Mediate Glaucous-Winged Gull Predation on Bivalve Prey

SEQUESTERED PARALYTIC SHELLFISH POISONING TOXINS MEDIATE GLAUCOUS-WINGED GULL PREDATION ON BIVALVE PREY

RIKK G. KVITEK 1 Departmentof Zoology,University of Washington,Seattle, Washington 98195 USA

A•sllo,cr.--Glaucous-winged Gulls (Larusglaucescens) avoid paralyticshellfish poisoning by conditionedaversions developed after the regurgitationof contaminatedbivalve prey. In feeding experimentswith free-rangingGlaucous-winged Gull chicks,toxic (445 •g saxitoxin equivalentsper 100 g) butter clams(Saxidomus giganteus) were regurgitatedin <5 rain (n = 58), and nontoxic butter clamswere never regurgitated(n = 30). Chicks that had ingested toxicS. giganteusthen refusedto eat either toxicor nontoxicbutter clamsoffered later, but they acceptedother bivalve prey. In feeding experimentswith wild adult and juvenile L. glaucescens,gulls took significantlyfewer butter clamsat a site where S. giganteushas been chronicallytoxic for > 14 yr than they took at a nontoxicbut otherwisecomparable site. These preferenceswere species-specificand independentof prey toxicity or whether the live clams were intact or removed from their shells.The gulls showedno differencebetween sites in their preferencesfor three other bivalve species(Clinocardium nuttalli, Protothacastaminea, Tresuscapax), all of which containedeither no or very low levels of toxins at both sites.The aversionto intact but not to shuckedbutter clamswas more strongly developedin adults than in juveniles.Most of the gulls that ate butterclams at the toxicsite discarded the siphons (which accountfor the majorityof the toxicity in contaminatedbutter clams)of both toxic and nontoxicbutter clams,but they never discardedsiphons of other bivalves.Gulls at the nontoxicsite never discardedthe siphonsof bivalve prey. I suggestthat avian predators generallyare not at risk from paralyticshellfish poisoning toxin via bivalve vectorsbecause they can detectand avoid toxic prey. Becauseof the apparentinability of gulls to discriminate between toxic and nontoxic individuals of the same prey species,avian predation pressure on somebivalve populationsmay be greatly reduced.Received 16 [uly 1990,accepted 8 November 1990.

PARALYTICshellfish poisoning (PSP) results al. 1971, Shumway 1990). Toxic "red tides" from the consumption of food contaminated (bloomsof toxigenic dinoflagellates)recur glob- with highly lethal neurotoxinsproduced by ally, appear to be increasingin frequency, and toxigenic dinoflagellates (Protogonyaulaxspp.) result in numerous shellfishery closuresand (Halstead 1978). Paralytic shellfish poisoning human illnesses and deaths every year (Hal- toxins (PSPT) can move through marine food stead1978, Nishitani and Chew 1988,Shumway chains and have been implicated in the mass 1990). mortalities of planktivorous fish (White 1977, Although many marineand coastalbirds prey 1980, 1981a, b) and seabirds (McKernan and on bivalves (Oldham 1930, Recher 1966, Barash Scheffer 1942, Coulson et al. 1968, Bicknell and et al. 1975,Sanger and Jones1982, Vermeer 1982, Walsh 1975,Sasner et al. 1975,Armstrong et al. Bourne 1984, Nettleship et al. 1984, Richardson 1978, Hockey and Cooper 1980, Nisbet 1983). and Verbeek 1986, Roberts et al. 1989) known The most common vector of the toxins in cases to concentratePSPT (Quayle 1969, Prakashet of bird mortalities has been planktivorous fish, al. 1971),avian mortalities have only rarely been yet it is suspension-feedingbivalves that pre- associatedwith PSPTcontaminated bivalve prey sent the greatestparalytic shellfish poisoning (Bicknell and Walsh 1975, Sasner et al. 1975, risk to human health (Quayle 1969, Prakashet Hockeyand Couper 1980).Indeed, it is the rarity of PSP-relatedavian deathsthat is most sur- prising given the extremelethality of the toxins • Present address:Moss Landing Marine Labora- and their widespread occurrence. Paralytic tories, P.O. Box450, Moss Landing, California 95034 shellfish poisoning toxins are among the most USA. potent neurotoxins known and are fatal to most

381 The Auk 108: 381-392. April 1991 382 RIKKG. KVITEK [Auk,Vol. 108

vertebratesat extremely small concentrations 1984, Nettleship et al. 1984, Richardson and (Halstead1978). The World Health Organiza- Verbeek 1986). tion has set the maximum PSPT concentration Gulls (Larusspp.) commonlyprey on bivalves allowed in shellfishfor human consumptionat by dropping them from aloft to crack open the 80 •g STX equivalents/100g. (Paralyticshell- shell (Oldham 1930, Barash et al. 1975, Roberts fish poisoningtoxins are derivativesof saxitox- et al. 1989). In the Pacific Northwest, Glaucous- in [STX], one of the two most lethal PSPT, and winged Gulls (Larusglaucescens) frequently for- mouse bioassay results are expressed in STX age on infaunal bivalves (Protothacastaminea, equivalents.)In two studieswith birds,pigeons Tapesjaponica, Clinocardium nuttalli, and Saxido- (Columbasp.) were found to have a LD50of 90 musgiganteus) (Barash et al. 1975, Vermeer 1982, •g STX equivalents/kg(see Halstead 1978), and pets. obs.). Each of these prey species retains the lethal dose for EuropeanStarlings (Sturnus PSPTfor varying lengths of time (5 weeksto 2 vulgaris)was <490 •g STX/kg (Kvitek and Beit- yr) following exposure to a toxic "red tide" let 1988, 1989). Because bivalve toxicities are (Quayle 1969, Shumway 1990). Red tides are routinelywell abovethese levels in manyareas, commonalong the west coastof North America, often measuredin the thousandsof gg STX/100 with many beachesclosed to clamming each g (Quayle 1969, Prakash et al. 1971, Halstead year (Quayle 1969, Nishitani and Chew 1988), 1978, Shumway 1990), either avian survival or but there have never been reports of gull mor- their predationon bivalvesmust be profoundly talities associated with toxic shellfish. In Wash- influenced at these times, unless the birds are ington State alone, bivalves at >50 sites con- highly resistantto PSPT. tained PSPT concentrations above the public Immunity to PSPT is not the most probable health closurelevel (80 •g STX equivalents/100 explanation for the rarity of bivalve-related g) during 1989 (Washington State Department PSPTbird mortalitiesgiven that PSPThas been of Health unpubl. rep.). implicated in bird kills via other vectorsas well My primary purpose was to resolve the par- as bivalves in a few cases.A more likely hy- adox of seabirds that prey on bivalves but do pothesis is that avian predators may develop not die as a result of PSPT-contaminatedprey. aversionsto toxicbivalve prey. Tufts (1979)cites I conducteda seriesof feeding experimentswith anecdotalaccounts of residentgulls (Larusspp.) free-ranging wild Glaucous-winged Gulls to and eiders (Somateriaspp.) avoiding toxic shell- determine (1) if experienced and naive gulls fish that kill migratory Black Ducks (Anasrub- differ in their responseto toxic prey, (2) wheth- ripes),and of a duck that regurgitated force-fed er or not naive gulls develop a conditioned toxic mussels.Acquired aversionsto toxic prey aversionfollowing the ingestionof toxic prey, have been well documented for some insectiv- and (3) at what level (individual, species,genus orous birds (e.g. Blue Jays,Cyanocitta cristata, or family) conditioned gulls discriminate be- that encounter monarch butterflies; Brower and tween prey, if an aversion is formed. I chose Fink 1985). These specieshave learned to iden- butter clams as the toxic prey because of the tify and avoid prey that is toxic or likely to be availability of highly toxic individuals due to toxic, and in some cases to circumvent the this species'tendency to retain PSPTfor > 1 yr chemical defensesof their prey by selective re- (Quayle 1969, Shumway 1990). They are also jection of the most toxic tissues. Similar re- abundant in many intertidal habitats and thus sponseshave been demonstrated in sea otters frequently eatenby Glaucous-wingedGulls and (Enhydralutris; Kvitek et al. in press) and fish other birds (Barash et al. 1975, Vermeer 1982, (Leptocottusarmatus; Kvitek MS) fed PSPT-con- Vermeer and Bourne 1984, pets. obs.).I selected taminated bivalves. three sitesfor the experiments.One was a Glau- If it occurs, avoidance of bivalve and other cous-wingedGull breeding area that does not intertidal invertebratesby avian predatorsas a supportinfaunal bivalve prey asforage for gulls; result of sequesteredPSPT could have signifi- the second, an area heavily used by foraging cant ecological and economic implications. Glaucous-wingedGulls where butter clamsbut Shorebirds have been shown to have profound not all bivalves are chronically toxic; and the effectson infaunal prey abundanceand distri- third, a similar but nontoxic site as a control. In bution (O'Conner and Brown 1977, Evans et al. this way I was able to comparethe feeding be- 1979, Quammen 1984) and to prey heavily on havior of birds chronically exposedto PSPT- commerciallyimportant bivalve species(Bourne laden butter clams with that of naive chicks at April 1991] Gullsand Shellfish Toxins 383

TAnIll 1. Bivalve communitiesat ScatchetHead (nontoxic site) and Middle Ground (toxic site). Toxicity levels aregiven for bivalvesused in gull feedingexperiments; NDT = nodetectable toxin (< 37•g STXequivalents/ 100g). All sampleswere analyzed with mousebioassay, except the cockles,which were tested with high performanceliquid chromatography. Means _+ SD are givenfor bivalvedensity (n = 5 and 6 for Scatchet Head and Middle Ground, respectively).

Scatchet Head (nontoxic) Middle Ground (toxic) Toxicity Density Toxicity Density Species (ug STX/100g) (ind./0.25m 2) (ugSTX/100 g) (ind./0.25m 2) Butter clams 42 _+ 13 11 _+ 6 Whole NDT 870 Siphonsonly 2,299 Littleneck clams NDT 40 _+ 11 129 10 _+ 4 Horse clams NDT 5 + 6 43 13 + 9 Cockles 3 _+ 2 29 2 _+ 2

the breeding site and older gulls foraging on equivalents/ 100g), or PSPTconcentrations above clo- clams in a nontoxic area. sure level (80 •g STX equivalents/100 g) since Sep- tember 1983(Washington State Department of Health METHODS unpubl. data). Butter clams at Middle Ground have been chronically well aboveclosure level for > 14 yr, STUDY AREAS generally in the range of 300-1,000 •g STX equivalents/100 g (WashingtonState Department of Health TatooshIsland is a breeding site for >4,000 gulls unpubl. data). Butter clamsare the best indicator of (Paine et al. 1990) and is on the outer coastof Wash- site PSPThistory becauseof their extremely long tox- ington State at the mouth of the Strait of Juan de in-retention time, whereas the other species men- Fuca. The presenceof chicks on 29-31 August 1989 tioned above usually depurate PSPT in 4-11 weeks allowed me to test the responsesof naive birds to (Quayle 1969, Shumway 1990). Thus, even though PSPT-contaminatedprey. In addition,individual birds Middle Ground butter clams are dangerously toxic could be followed during the experimental period year-round, the cockles,littleneck clams, and horse becausemany chicks had not fledged, and family clamsare usuallyonly seasonallytoxic for a relatively groups remained on their territories. short time following a "red tide" (WashingtonState ScatchetHead, at the south end of Whidbey Island, Department of Health unpubl. data). Washington, and Middle Ground in Sequim Bay, Toxicity levels were determined for butter clams Washington,were selectedas nontoxic and toxicsites, and littleneck clams collected at both sites, and cock- respectively.The extensive intertidal gravel flats at les and horse clams collected at Middle Ground be- both sitesare usedheavily by gulls to forageon rich tween 26 April and 8 May, 1990. Mouse bioassays populationsof cockles(Clinocardium nuttalli), little- (AOAC 1984) were conductedby Washington State neck clams (Protothaca staminea), butter clams (Saxi- Department of Health on all samplesexcept the cock- domusgiganteus), and horse clams(Tresus capax) (pets. les,which were testedwith high performanceliquid obs.). Prey abundancewas sampled using randomly chromatography(HPLC) (courtesyof M. Beitler, Di- tossed0.25 m2 quadrats at ScatchetHead (n = 5) and vision of Aquaculture and Food Science,University Middle Ground (n = 6). All bivalves excavated from of Washington, Seattle) by the methodsof Sullivan the quadratswere identified to speciesand counted. and Wekell (1986) (Table 1). The gulls excavatethe shallowerprey (cockles, Gull populationsand demographicswere alsosam- littleneck clams, and small butter clams) themselves pled at both sitesduring the feedingexperiments. All and obtain deeperclams (horse clams and larger but- individuals in the foraging areas were counted and ter clams) from excavationsleft by clammers.Large agedby plumagecharacteristics (Table 2). Other gull numbers of both of these speciesare routinely un- species,if present,were alsorecorded. Although oth- coveredand left exposedby the numerousclammers er speciesof birds are known to forage on clams, in searchof the preferredlittleneck clams at both of Glaucous-wingedGulls are virtually the only avian thesesites (pets. obs.). For this reasonit is likely that residentsobserved to prey on bivalves at both sites. most,if not all, gullsat both ScatchetHead and Mid- Only crowsare occasionallyseen as well (pers. obs.). dle Ground have had frequent encounterswith all four speciesof bivalves in this study. FEEDING F•XPERIMENTS: GULL CHICKS At the time of the study (29 April to 23 May 1990), butter clams at Scatchet Head had not contained de- Responseof naive gull chicksto PSPT-contaminated tectable levels of PSPT since June 1985 (<37 •g STX prey.--Glaucous-wingedGull chicks were tested for 384 Rir,x G. KVITEK [Auk,Vol. 108

TAnrE2. Density and age compositionof Glaucous- FEEDING EXPERIMF2qTS:NONTOXIC VS. winged Gulls at ScatchetHead (nontoxic site) and TOXIC SITES Middle Ground (toxic site). Populationswere not significantlydifferent with respectto population The following experimentswere designedto de- size (t-test, t = 2.52, df = 4, P = 0.09) or gull age termine whether gulls foragingon bivalvesat a toxic structure(X 2 = 8.19, df = 1, P = 0.09) during the site had learnedto identify and avoid toxic prey. All feeding experimentperiod. nontoxic butter clamsused as prey were collectedat Scatchet Head (hereafter nontoxicsite) and the toxic Juveniles Adults ones at Middle Ground (toxicsite) (Table 1). The other Date (1-2 yr) (->3 yr) n speciesused in the experiments (cockles,littleneck Scatchet Head (nontoxic site) clams,and horseclams) were locallyobtained at each 12 May 17% 83% 46 test site (Table 1). Sample sizesare given with the 14 May 30% 70% 40 results. 21 May 20% 80% 49 ExperimentI: shuckedbivalve prey.--The objectives Middle Ground (toxic site) of this experiment were to learn whether (1) free- 13 May 28% 72% 61 ranging adult and postfledgling juvenile gulls dis- 23 May 13% 87% 95 criminatebetween toxic and nontoxicprey of the same and different species,(2) the frequencyof toxic prey consumptionvaries with gull age,(3) gulls selectively eat butter clam tissues,avoiding the highly toxic siphon, (4) older gulls vomit ingestedtoxic prey as did the chickson Tatoosh,and (5) gulls at a toxic versus their responseto toxic prey and their tendency to a nontoxicsite differ in the aboveresponses. Nontoxic developconditioned aversions. Shucked, whole, fresh and toxic butter clamsand locally obtainedhorse and nontoxic and toxic butter clams were made available littleneck clamswere offeredto free-ranginggulls of to gull chickson their parents'breeding territoriesat all ages at both the nontoxic site and the toxic site. Tatoosh Island. Nontoxic clams had been collected at Shuckedclams were again used so that the bird that Mukilteo, Washington (no detectable toxin, Wash- actually consumed the prey could be more easily ington StateDepartment of Health unpubl. data);tox- identified and followed. Whole, fresh bodies of each ic clamswere collectedat Middle Ground(445 •tg STX specieswere tossedout in a random seriesto foraging equivalents/100 g, Washington State Department of gulls during low tide. I noted whether eachclam was Health unpubl.data). Shucked, rather than intact,prey (1) eaten or rejectedwithin 5 min, (2) eaten by an was used so that the gulls would not have to fly off adult (->3 yr) or juvenile (<-2 yr), (3) entirely eaten with the clamsto crackthem open,which would have or the siphon rejected,(4) entirely eaten and regur- made it virtually impossibleto keep track of individ- gitated within 20 min. Whenever possible,birds that uals. Butterclams were presentedto haphazardlyse- had eaten butter clams were tracked for 20 min, and lected family groupsby placing the shuckedclams the time to regurgitation (if it occurred)was noted. I within the territory of the adults. The clams were considered20 min to be a sufficientobservation pe- either eaten directly by the chicks,eaten by the adults riod based on previous feeding experiments. At the and immediately (<1 min) regurgitated to the beg- nontoxic site this experimentwas conductedon an ging chicks,or stolenand eaten by an intruding gull. area of beachca. 200 m away from where the other I observed88 ingestions(30 nontoxic and 58 toxic) testswere run, to minimize biasing future testsshould and followed individuals for 15 min or until I ob- the gulls develop PSPT-relatedaversions to butter serveda regurgitation.All chicksthat receivednon- clams.This waseffective because individual gulls ap- toxic clams also received toxic clams, but 28 chicks yearedto prefer foraging territoriesat the site. only received toxic clams. The regurgitation of an ExperimentII: three-speciespredation test.--To deter- adult to a chick was not consideredto be due to prey mine if there were species-specificdifference• in bi- toxicity. In addition, 14 of the chicksthat ate toxic valve predationrates related to prey toxicitiesat non- prey and 11 that ate nontoxic prey were timed in- toxic and toxic sites, live intact individual butter clams, dividually for 15 min or until regurgitation to deter- cockles, and littleneck clams were made available to mine the time required to detectand regurgitatethe foraging gulls at both sites. Groups of three clams toxic clam. each(one individual of eachspecies per group) were The day followingtheir firstexposure to toxicbut- laid out at 25-50 m intervals along the beachat low ter clams, I tested seven of the birds for aversion to tide while gulls were foraging.The clamswere com- butter clams. They were offered shucked nontoxic parablein sizeand weight, with the butter clamnever butter clams for 5 min followed by nontoxic horse the largestor smallestin the group. Eachgroup was clamsas a control. The samplesize was 7 individuals checkedfor missingclams after 5-10 min. becausethis was the number that could be unambig- ExperimentIII: prey preferencetests.--This experi- uouslyidentified as subjects from the following day. ment was designedto determine if there were age- April 1991] Gullsand Shellfish Toxins 385 and site-specificprey preferencesrelated to prey tox- at the nontoxic site contained levels of PSPT icity patterns.Prey preferenceswere determinedby within the detection range of the mouse bio- placingpairs of live intactclams on the beachamong assay(>-37 •g STX equivalent/100 g, Washing- foraginggulls at low tide. At ScatchetHead eachpair ton State Department of Health) (Table 1). At consistedof a locally obtained butter clam matched the toxic site, all speciestested contained some with a local cockleor littleneck clam of comparable size. At Middle Ground the setupwas the samewith PSPT,but only butter clamsand littleneck clams the addition of horseclams. Each prey pair was laid were above State Health Department closure out and watchedindividually for 5 min to seewhich level, and the butter clams were >6 times more clam specieswas taken first, if both were taken, and toxic than the littleneck clams (Table 1). PSPT the age of the bird taking each clam. The order of levels found in the toxic-site cockles with HPLC speciespaired with the butter dam was random. Horse were below the detection limit for the mouse clamswere not usedat ScatchetHead becausethey bioassay. were too large for the gulls to carryaway and could not be matchedby size with the smallerbutter clams. FEEDING EXPERIMENTS: GULL CHICKS STATISTICAL ANALYSIS AND ASSUMPTIONS Responseof naivegull chicksto PSPT-contaminated prey.--All gull chicks that ingested Unlessotherwise stated, I usedChi-square and con- shuckedtoxic butter clams(n = 58) regurgitated tingency table analysesto comparethe responsesof and abandonedthe clam body intact. The mean gulls in the experiments,and P -< 0.05 was the significance level for all tests. I assumed that each trial time (+SD) to regurgitation was 4.7 + 2.1 min (n = 14) for chicks fed toxic butter clams, where- involved a different gull and was thus independent, althoughthis cannotbe absolutelyassured because as chicksthat ingestednontoxic butter clams(n none of the birds in the populationswere marked. = 30) never regurgitatedthe prey. Of thosegulls However, becausemy investigationwas focusedon tested for aversion to butter clams (n = 7) the population-and community-leveleffects of seques- day following their ingestionand regurgitation tered PSPT, testing for differencesin the relative fre- of toxic butter clams, none would eat available quencyof prey speciesconsumption proved a good shucked nontoxic butter clams,but all readily measureof the differencesin gull responseto bivalves consumed shucked horse clams. at the two sites. As a precaution, I minimized the likelihood of testingthe sameindividual by timing and spacingeach trial appropriately.Whenever a gull FEEDING EXPERIMENTS: NONTOXIC VS. tookan intactclam from oneof my trials,it generally TOXIC SITES flew away from the test area and spent severalmin- utes repeatedlyflying up and dropping the clam to Gull populationsand demographics.--The Glau- break it. Also, individuals at both sites had discrete cous-winged Gull was the dominant gull spe- foragingterritories along the beach.Thus by restrict- cies observed at both sites during the feeding ing the trialsto only a few minutesand spacingthem studies.A single Glaucous-winged x Western morewidely than the apparentforaging territory size, (Larusoccidentalis) gull hybrid was seen once at and by watching to seethat the birds that took clams ScatchetHead. The only other avian predators left the area, I maximized the number of birds tested that I observed were a few (<5) Northwestern in each experiment.Feeding trials with shuckedclams Crows (Corvuscaurinus). Although they were were also spacedalong the beach to maximize the number of birds tested. occasionallypresent at both sites, a crow was observedtaking a clamfrom an experimentonly once, and this was omitted from the results. RESULTS Gull population sizes at Scatchet Head (45 + 4.6, n = 3) and Middle Ground (78 + 24.0, n = PRmC POPULATIONS AT TOXIC AND 2) were not significantlydifferent, nor were the NONTOXIC SITES age structuresof the two populations (Table 2). At the nontoxic site, the most abundant bi- ExperimentI: shuckedbivalve prey.--Glaucous- valve species was the butter clam (47% of all winged Gulls foraging at the toxic site, but not bivalves). At the toxic site, butter clams were at the nontoxicsite, displayed a significantaver- only slightly lessabundant (31% of all bivalves) sion to all butter clamsand butter-clam siphons, than horse clams (36%), the most common clam but no aversion to other species(Table 3). Gulls there (Table 1). None of the bivalves collected at Middle Ground ate significantly smaller per- 386 RxKKG. KVITEK [Auk, Vol. 108

TABLE3. Responseof Glaucous-wingedGulls to shucked,toxic, and nontoxicbivalve prey (ExperimentI). In feedingtrials at the toxicsite, gulls ate significantlyfewer shucked butter clams (but not fewer other prey)than at the nontoxicsite. The gulls'aversion to butter-clamflesh at MiddleGround was independent of clam toxicitybecause nontoxic and toxicbutter-clam tissues were equallyshunned. There was no sig- nificantfeeding pattern related to gull agein this experiment.Siphons of both nontoxicand toxicbutter clamswere discardedby gullsin a significantnumber of trialsat the toxicsite. Siphons account for the majorityof butter-clamtoxicity in contaminatedindividuals. Only toxicbutter clams were regurgitated at both sites. NDT = no detectable toxins.

Shucked,whole bivalve prey availableto gulls Butter clam Horse Littleneck Non- clam clam Gull response toxic n Toxic n Local n Local n Scatchet Head (nontoxic site) Preytoxicity (#g STX equivalents/100g) NDT 870 NDT NDT Total prey eaten(%) 100 20 94 16 100 16 100 15 Prey eatenby adults(% of total prey eaten) 60 20 44 9 44 16 20 15 Siphonsrejected (% of total prey eaten) 0 20 0 9 0 16 0 15 Prey regurgitated(% of total clamseaten) 0 20 83 6 -- -- Time to regurgitation(min, • + SD) 12.7 + 6.8 5 Middle Ground (toxic site) Preytoxicity (#g STX equivalents/100g) NDT 870 43 129 Total prey eaten(%) 47 19 55 29 100 16 92 13 Prey eatenby adults(% of total prey eaten) 78 9 44 16 69 16 75 12 Siphonsrejected (% of total prey eaten) 89 9 56 16 0 16 0 12 Prey regurgitated(% of total clamseaten) 0 3 75 4 -- -- Time to regurgitation(min, • _+SD) 14.5 + 5.0 2

centages of available shucked toxic and non- (83%) of those that ate entire toxic butter clams toxic butter clams than those of the other avail- regurgitated them (Table 3; X2 = 15.62, df = 1, able clams(Chi-square tests,P < 0.05). They did P = 0.0001). None of the birds that ate the non- not, however, appear to distinguish between toxic butter clams regurgitated during the 20- nontoxic and toxic butter clams (X 2 = 0.055, df min observation period. I found the same pat- = 1, P = 0.82) (Table 3). The percentagesof total tern at the toxic site for the few birds that ate prey eaten for all available speciesat the non- entire butter clams and were followed. A Chi- toxic site (ScatchetHead) did not differ signif- squaretest was not significant,perhaps because icantly, nor did gull consumptionof nontoxic of the small sample size involved. versus toxic butter clams (Chi-square tests,P > ExperimentII: three-speciespredation test.--Gulls 0.05). Gulls at the nontoxic site also ate signif- at the toxic site avoided live intact butter clams, icantly more butter clamsthan at the toxic site but not cockles or littleneck clams, when com- (Chi-square tests,P < 0.05), but there was no pared with gulls at the nontoxic site (Fig. 1). difference in the percent consumption of the Gulls at Middle Ground took a significantly other two species(P > 0.05). There was no ob- lower percentage of both nontoxic and toxic vious or significant pattern in the behavior of butter clams compared with gulls at the non- adults versusjuveniles in this experiment. toxic site (X2 = 79.56, df = 3, P = 0.0001). There Gulls rejecteda significant percentageof all was no significant between-site difference in butter-clamsiphons at the toxicsite (Chi-square the percentagesof cockles(X • = 2.74, df = 3, P tests,P < 0.05). Siphons of the two other species = 0.4) and littleneck clams (X: = 2.64, df = 3, P were never rejected at the toxic site nor were = 0.5) taken by gulls. There were alsono within- siphonsof any of the available prey rejectedat site differencesin the percentagesof nontoxic the nontoxic site. and toxic butter clams taken (Scatchet Head, X• Although gulls at the nontoxic site showed = 0.78, df = 1, P = 0.38; Middle Ground, X2 = no difference in their willingness to eat toxic 0.50, df = 1, P = 0.48). and nontoxic butter clams,a significant number ExperimentIII: preypreference tests.--Gulls took April 1991] Gullsand Shellfish Toxins 387

T^BLœ4. Glaucous-wingedGull two-speciesbivalve prey preferencetest (ExperimentIII). In pairedprey choicetests at the toxicsite (Middle Ground), gulls took significantly fewer butter clams than at the nontoxic foragingsite (ScatchetHead). All other bivalve prey paired with butter clamsin choicetests (cockles, littleneckclams, and horseclams) were preferred(taken first) more often than butter clamsat the toxic site, but not the nontoxicsite. Adult gulls at the toxic site seldomtook butter clamscompared with the otherprey available,but tookall clamsat the nontoxicsite as often asjuvenile gulls.

Choicetest 1 Choicetest 2 Choicetest 3 Summary But- But- Little- But- But- ter ter neck ter Horse ter Site/gull response clam Cockle n clam clam n clam clam n clam n Scatchet Head % taken 100 100 10 82 82 11 91 21 % taken first 20 80 64 36 43 % taken by adults 50 70 55 36 52 Middle Ground % taken 50 100 10 66 100 12 29 100 7 52 29 % taken first 10 90 25 75 0 100 14 % taken by adults 20 100 17 67 14 86 12

significantlyfewer butter clamsat the toxicsite minutes after ingestion (<4 min for Tatoosh thanat thenontoxic site (all comparisonspooled, chicks,and < 15 min for all postfledglingsand X2= 7.37, df = 1, P = 0.007;Table 4). The gulls adults; Table 3). Thus the toxin was voided be- at the nontoxicsite showedno significantprey fore serious illness occurred. preference in either of the choice tests (Chi- Gulls appear to be protected further from square tests,P > 0.05). Middle Ground gulls, PSPTvia food aversion.This hypothesisis sup- however, preferred cockles(X 2 = 6.6, df = 1, P ported by two lines of evidence.First, all of the = 0.01) and horse clams (X2 = 5.25, df = 1, P = seven chicks on Tatoosh followed as individ- 0.02)over butter clams(Table 4). Although lit- uals for 2 days appearedto have developeda tleneck clams were also chosen first more fre- conditioned aversion to nontoxic butter-clam quently than butter clams at the toxic site, the tissue--butnot to horseclams--on the dayafter differencewas not significant(X 2 = 3, df = 1, P their first ingestionand regurgitationof toxic = 0.08). Of all the butter clams taken at the toxic butter clams. Second,gulls at the site where site,significantly less were takenby adultsthan butter clams are chronicallytoxic generally at the nontoxic site (X2 = 5.4, df = 1, P = 0.02). avoidedbutter clamsindependent of toxicity (Table 3, Fig. 1). I believe that once an inex-

DISCUSSION periencedgull eatsand regurgitatesa toxic individual, it will rejectthat prey specieson the Gu•. AVOIDANCE OF PSPT next encounter. The aversion thus appears to be generalizedto the prey speciesand not to Although paralytic shellfish poisoning toxin the toxicityof prey individualsencountered af- has been implicated in several seabird kills ter conditioning. My results, however, do not (Nisbet 1983,Hockey and Cooper 1980,Coulson precludethe possibilityof sociallytransmitted, et al. 1968, Armstrong et al. 1978, McKernan PSPT-mediatedprey aversionswithin a gull and Scheffer1942), my studysuggests that Glau- population. Becausebutter clams but not other cous-wingedGulls are not at mortal risk from species at Middle Ground have been chroni- bivalve-bornePSPT. None of the gulls fed toxic cally toxicyear-round for many years,it is con- butter clams died or showed symptoms dis- ceivable that a tradition of butter-clam avoid- played by other birds that becameill or died ance has arisen and been transmitted from following the ingestionof contaminatedbutter parent to offspringat that site. Suchavoidance clams (Kvitek and Beitler 1988). The absenceof behaviorwould be further reinforcedany time illness in the gulls was probably not due to a gull ate a butter clam at Middle Ground, be- insufficientprey toxicity;rather, inexperienced causealthough toxicitymay vary betweenin- birds regurgitatedthe toxicprey within a few dividual butter clams,virtually all butter clams 388 RIKKG. KVIT•C [Auk,Vol. 108

dividuals are found at comparabledepths and have been reported as important in the diets of 80 some Glaucous-wingedGull populations(Ba- rash et al. 1975, Vermeer 1982). Furthermore, 60 gulls at Middle Ground gain accessto even the 40 largestbutter clamsbecause this speciesis ex- 20 cavatedand left exposedon the beachin great numbersby clammerslooking for the more pre- 0 Butter clams Cockles Littleneck clams ferred (less toxic) littleneck. Thus, the high abundance of butter clams (Table 1), the shal- low burrow depth of smaller individuals, and the activity of clammersinsures that most gulls at Middle Ground undoubtedly have encountered butter clams. Indeed, if neophobia were an important factor in the aversivebehavior of seagulls toward any clam at Middle Ground, I would expectit to be most pronouncedtoward the horseclam. This speciesis a deeperburrow- er, even at small sizes, than the butter clam, and thus it is encounteredmuch less frequently by gulls excavatingtheir own prey. Furthermore, Butter clams Cockles Littleneck clams horse clams are morphologically the most dis- tinct of the four species(Morris et al. 1980). Nevertheless, horse clams at Middle Ground Live prey availablein 3-speciesgroupings were readily accepted as prey by the gulls, Fig. 1. Percentagesof the available live bivalve whereasbutter clams,although very similar in prey "taken" by Glaucous-wingedGulls at Scatchet appearanceto littleneck clams,were generally Head (A: nontoxic site) and Middle Ground (B: toxic avoided (Tables 3 and 4). site). Stippled bar representsprey grouped with non- Gull aversionto PSPT-contaminatedprey re- toxic butter clams(n = 28 groupsat ScatchefHead semblesclosely the behavior of insectivorous and 21 groupsat Middle Ground) and solid bar representsprey groupedwith toxic butter clams(n = 10 birds that shun unpalatable insectsas well as and 24). Both nontoxic and toxic butter clams were their nontoxicmimics and conspecifics(Brower taken significantlyless frequently at Middle Ground, and Glazier 1975, Brower and Fink 1985). Un- the toxic site (x 2 = 79.56, df = 3, P = 0.0001). like birds known to discriminate between toxic and nontoxic individuals (Calvert et al. 1979, Brower and Fink 1985), Middle Ground gulls generally shunnednontoxic as well astoxic but- will containbiologically significant amounts of ter clams (Tables 3 and 4, Fig. 1). Some of the PSPT(Quayle 1969).The tendencyfor juvenile gulls at the toxic site ate butter clams,but they gulls to chooseintact butter clamsmore often generallydiscarded the siphon(the organwhich than adult gulls do (Table 4) implies that dif- becomes most toxic in contaminated butter ferences in either experience level or age-re- clams,Quayle 1969) (Tables 1, 3, and 4). This lated learning ability (Evanset al. 1987)may be behavior also occurs in sea otters (Kvitek et al. important in the developmentof prey aversion in press)and is similar to that of insectivorous in gulls. birds that selectively eat only the least toxic Aversive behavior toward food in birds has parts of their prey (Brown and Neto 1976, Fink been associatedwith neophobia, the fear of and Brower 1981, Brower and Fink 1985). It is novel food (Evans et al. 1987). This is an un- not clear how gulls learn to avoid the siphons, likely explanationfor the responseof the gulls particularlysince they do not appearto distin- toward butter clams at Middle Ground. Butter guish between toxic and nontoxic butter-clam clamsare abundant, and secondonly to the horse tissues(Table 3). Once acquired, this behavior clam in abundance at Middle Ground (Table 1). couldconceivably be sociallytransmitted (Galef Although largebutter clamsare deeperburrow- 1987). Additional work will be required to de- ers than cockles or littleneck clams, smaller in- termine how the initial learning takes place. April 1991] Gullsand Shellfish Toxins 389

PSPT AND SEABIRD MORTALITIES contain only moderate amounts of residual tox-

Some seabirds die from PSPT-laden prey (McKernan and Scheffer 1942, Coulson et al. PSPT AND PREDATOR/PREY 1968, Bicknell and Walsh 1975, Sasner et al. 1975, RELATIONSHIPS Armstronget al. 1978,Hockey and Cooper1980, Nisbet 1983), while others, such as the Glau- Paralytic shellfish poisoning toxin may not cous-wingedGull, appearto be behaviorallyex- posea significantthreat to mostavian predators empt from risk. Rather than being species-spe- of bivalves, but it does appear to profoundly cific, differential susceptibilitymay well be affectGlaucous-winged Gull prey selection.The mediated via prey type. Most seabird kills at- experienced gulls in this study exhibited a tributed to PSPT involved planktivorous fish marked aversion to toxic species.As a result, rather than bivalves as vectors. If seabirds con- gulls took significantly fewer butter clamsat a sume small fish intact, without rupturing the chronically toxic site than inexperiencedgulls body wall, PSPTconfined to the fish gut may at a nontoxicsite (Fig. 1, Table 4). Yet there was not be released and detected until much later no difference between sites in the frequency in the digestionprocess and perhapstoo late to with which gulls took other nontoxic or low- regurgitate and avoid absorption. In contrast, toxicity prey. Becausethe aversionat the toxic birds break the shell of bivalve prey and con- sitewas species-specific and independentof prey sumethe visceradirectly. Therefore, even pred- toxicity, avian predation pressure on bivalve atorsthat lackchemosensory receptors for PSPT speciesthat sequesterPSPT may be reducedboth may still be able to detect the early symptoms beyond the period of time that the clamsremain of PSP shortly after the ingestion of toxic bi- dangerouslytoxic and perhapsat nontoxicsites valves, when it is still possibleto regurgitate at which conditioned birds also forage. the prey intact. The gull chicksat Tatooshwere Superficially, it appears that some bivalves exceptionallysensitive to ingestedPSPT-laden may sequesterPSPT, which acts as a chemical butter clams,and they regurgitatedthem in < 5 deterrent to avian predation and is analogous min, morerapidly than requiredfor PSPsymp- to the milkweed-monarch butterfly-Blue Jay toms to be detected in humans (Halstead 1978) relationship (Brower and Fink 1985, but see or for observablesymptoms to appear in star- Broweret al. 1988).Gulls, however, alwayskill lings (Kvitek and Beitler 1988).Indeed, it is pos- the individual bivalve. It is thus difficult to sible that Glaucous-wingedGulls have become imagine how PSPT retention may be selected "physiologicallytuned" to PSPT to a greater by avian predation without invoking group se- extent than other birds and are thus better able lection. Furthermore, unlike most of the birds to detect and avoid a lethal dose. known to avoid cardenolide-laden butterflies, Seabird susceptibilityto PSPT may also be the gullsin this studydid not rejecttheir prey mediatedby novelty and toxin level. The only before consumption,based on the presenceof reported massivemortality of seabirdsassoci- toxin (Fig. 1). ated with PSPT-laden bivalves involved ex- A more likely selectiveagent for the reten- traordinarily toxic shellfish (9,500 #g STX tion of PSPT in bivalveswould be "grazing" equivalents/100g) in a New England areapre- predators,such as siphon-cropping fish (Kvitek viously free of PSPT (Sasneret al. 1975). There and Beitler 1991),which do not kill their prey is also an anecdotalreport of resident avian but do develop a PSPT-basedaversion to toxic bivalve predatorsshunning toxic shellfishthat but not nontoxic butter-clam siphons (Kvitek killed migratingBlack Ducks (Tufts 1979).These MS). Siphon-croppingfish have been linked accountsare consistentwith my results.Naive with significantreductions in bivalve growth chicksat Tatooshand gulls at the nontoxicsite rates (Peterson and Quammen 1982, Zwarts readily consumedtoxic butter clamswhen first 1986), presumably becauseresources must be encountered.Perhaps if these clams had been shunted from growth and reproductionto si- sufficientlytoxic, the birds may not have been phon regeneration.However, retention of PSPT able to regurgitate their food in time to avoid in the butter-clam siphon may have been in- a lethal dose.Thus, birds foraging at a site that directlyselected for by avian predation,because is frequently toxic may have the advantageof a fishnipping off the tip of a siphoncauses the learning which prey to avoid when bivalves clam to reduce its burrow depth and this in- 390 RIKKG. KVITEK [Auk,Vol. 108

creases its vulnerability to bird predation on bivalves that sequesterthe toxin. This aver- (Zwarts 1986). Thus, a defensestrategy that re- sion appearsto be independent of the prey's ducespartial predation by fish will permit the current level of toxicity, and the protection it clam to retain its burrow depth and refuge from affordsmay be extendedin time and spacebe- birds. yond both the toxin-retention period and geo- Another reasonavian predation may not be graphic range of prey contamination.Although the most important mechanism selecting for experiencedgulls in this study avoided butter PSPT retention in butter clams is that burrow clams,siphon-nipping fish are a more likely depth increaseswith bivalve size (pers. obs.). selectivemechanism for the evolution of long- Larger individuals generally have a burrow term PSPT retention in Saxidomus. Neverthe- depth (10-30 cm, Morris et al. 1980, pers. obs.) less,once evolved, the broad spectrumof taxa that is below the foraging range of most birds susceptibleto PSPT widens its effectivenessas (Roberts et al. 1989, but see Vermeer and Bourne an acquired chemical defense wherever toxic 1984).Although gulls are frequentlyobserved bloomsof Protogonyaulaxspp. occur. eating large butter clams,this is mostcommon on beacheswhere clam diggers are excavating ACXNOWlm•Mr•w'rs and discardingor overlookingbutter clams,or where burrow depth is restrictedbecause of un- I am very grateful to D. Carney, R. T. Paine, C. Pfister, and T. Wootton for field assistance and for derlying clayor rock (pers.obs.). The mostfre- suggestionsthat substantiallyimproved this manu- quentlyobserved bivalve prey speciesthat gulls script. I am also indebted to M. Beltlet and G. Skow and crows excavate unaided are cockles and lit- for PSPT analyseswithout which this work would tleneck clams,which are usually just below the not have been possible.Special thanks go to R. T. sediment surface (Oldham 1930, Barash et al. Paine, who has generouslysupported this project 1975, Roberts et al. 1989, Richardson and Ver- through a National ScienceFoundation award (OCE beek 1986,pers. obs.). Both of thesespecies have 8614463).Additional supportwas provided by a NSF relatively short PSPT retention times (9 and 5 award to J. Oliver (DPP-8619394), and National Au- weeks,respectively) compared with butterclams dubon Graduate Student Research and American Mu- (> 1 yr) (Shumway1990). Furthermore, cockles seum of Natural History Lerner-Gray Fund for Ma~ fine Research awards to R. Kvitek. Research on Tatoosh and littleneck clams at Middle Ground, the site was made possibleby permissionof the Makah Tribal at which butter clamsare chronicallyvery toxic, Council and the U.S. Coast Guard. were far less toxic than the butter clams and were readily consumedby the gulls there (Ta- bles1 and 3, Fig. 1). Thereforethe value of PSPT retention in butter clams as a deterrent to avian ARMSTRONG,I. H., J. C. COULSON,P. HAWKEY,& M. J. predationmay be mostimportant only to small- HI3DSON. 1978. Further mass seabird deaths from er size classesor ancillary to its role asa defen- paralyticshellfish poisoning. Br. Birds71: 58-68. ASSOCIATION OF OFFICIAL ANALYTICAL CHEMISTS. siveagent against fish. Yet the extremelethality 1984. Official methodsof analysis(S. Williams, and rangeof taxasusceptible to PSPT(Halstead Ed.). Arlington, Virginia, Assoc. Offic. Anal. 1978)may extendthe usefulnessof thesetoxins Chem. as a chemical defense once the ability to se- BARASH,D. P., P. DONOVAl, & R. MYPaCK. 1975. Clam questerthem hasevolved. For example,the re- dropping behavior of the Glaucous-wingedGull tention of PSPT may profoundly influence pre- (Larusglaucescens). Wilson Bull. 87: 60-64. dation by scoters(Melanitta spp.), for which BICKNELL,W. J., & D.C. WALSH. 1975. The first "red butter clams make up the majority of prey at tide" in recordedMassachusetts history. Pp. 447- some winter sites (Vermeer and Bourne 1984). 457 in Proceedingsof the First International Con- Avian predatorsof bivalvesmay be at a much ference on Toxic Dinoflagellate Blooms (V. R. LoCicero, Ed.). Wakefield, Massachusetts,Mas- lower risk to PSPT than piscivorousbirds be- sachusetts Sci. Tech. Found. cause of the tendency of the former to regur- Boumq•,N. 1984. Clam predation by scoter ducks gitaterapidly the ingestedPSPT-laden prey. The in the Straitsof Georgia,British Columbia,Can- risk of poisoning is further reduced via a con- ada. Can. Tech. Rep. Fish. Aquat. Sci. 1331. ditioned avoidanceof previously regurgitated BROWER,L. P., & L. S. FINK. 1985. A natural toxic bivalve species.For this reason,PSPT may func- defensesystem: cardenolides in butterfliesversus tion as an effective deterrent to avian predation birds. Ann. N.Y. Acad. Sci. 443: 171-188. April 1991] Gullsand Shellfish Toxins 391

--, & S.C. GLAZieR. 1975. Localization of heart numbers of dead seabirds on the Washington poisonsin the monarch butterfly. Science 188: coast. Condor 44: 264-266. 19-25. MORPaS,R. H., D. P. ABBOTT,& E. C. H^DERLIE. 1980. ß C. J. NELSON, J. N. SE•BER,L. S. FL•K, & C. Intertidal invertebrates of California. Stanford, BOND. 1988. Exaptationas an alternative to co- California, Stanford Univ. Press. evolution in the cardenolide-based chemical de- NETTLESHIP, D. N., G. A. SANGER, & P. F. SPRINGER. fenseof monarchbutterflies (Danaus plexippus L.) 1984. Marine birds: their feeding ecologyand againstavian predators.Pp. 447-475 in Chemical commercial fisheries relationships. Seattle, mediation of coevolution (K. Spencer, Ed.). San Washington,Annual Meeting of the PacificSea- Diego, Academic Press. bird Group. BROWN,K. S., & J. V. NEro. 1976. Predation on apo- NISBEr,I. C. 1983. Paralytic shellfish poisoning:ef- sematicithomiine butterfliesby tanagers(Piprae- fectson breedingterns. Condor 85: 338-345. ideamelanonota). Biotropica 8: 136-141. NISHITANI, L., & K. CHEW. 1988. PSP toxins in the CALVERt, W., L. E. HEDRICK, & L. P. BROWER. 1979. Pacificcoast states: monitoring programsand ef- Mortality of the monarch butterfly (Danausplex- fects on bivalve industries. J. Shellfish Res. 7: ippusL.): avian predation at five overwinter sites 653-669. in Mexico. Science 204: 847-851. O'CONNOR,R. J., & R. A. BROWN.1977. Prey deple- COULSON,J. C., G. R. POTTS,I. R. DEANS,& S. M. FRASER. tion and foragingstrategy in the Oystercatcher 1968. Exceptionalmortality of shagsand other Haernatopusostralegus. Oecologia 27: 75-92. seabirdscaused by paralytic shellfish poison.Br. OLDHAM,C. 1930. The shell-smashinghabit of gulls. Birds 61: 381-405. Ibis 6: 239-242. EVANS, D. L., N. CASTORIADES,& H. BADRUDDINE. PAINE,R. T., J. T. WOOTTON,& P. D. BOERSMA.1990. 1987. The degree of mutual resemblanceand its Direct and indirect effects of Peregrine Falcon effect on predation in young birds. Ethology 74: predationon seabirdabundance. Auk 107: 1-9. 335-345. PETERSON,C. H., & M. L. QUAMMEN. 1982. Siphon EvANs, P. R., D. M. HERDSON,P. J. K•GH'r, & M. W. nipping: its importanceto small fishesand its PmNKOWSKI. 1979. Short-term effects of recla- impact on growth of the bivalve Protothacastarn- mation of part of sealsands, Teesmouth, on win- inca(Conrad). J. Exp. Mar. Biol. Ecol.63: 249-268. tering wadersand shelducks.I. Shorebird diets, PRAKASH,A., J. C. MEDCOF,& A.D. TENNANT. 1971. invertebrate densities,and the impact of preda- Paralytic shellfish poisoning in easternCanada. tion on the invertebrates.Oecologia 41: 183-206. Fish. Res. Bd. Can. Bull. 177: 1-88. FINK, L. S., & L. P. BROWER. 1981. Birds can overcome QUAMMEN,M.L. 1984. Predationby shorebirds,fish, the cardenolide defence of the monarch butter- and crabs on invertebrates in intertidal mudflats: flies in Mexico. Nature 291: 67-70. an experimental test. Ecology 65: 529-537. GALEF,B. G., JR. 1987. Social influences on the iden- QUAYLE,D. B. 1969. Paralytic shellfish poisoningin tification of toxic foods by Norway rats. Anim. British Columbia. Ottawa, Canada, Fish. Res. Bd. Learn. and Behav. 15: 327-332. Can. Bull. 168. HALSTEAD,B.W. 1978. Poisonous and venomous ma- RECHER,H.F. 1966. Some aspectsof the ecologyof rine animals of the world. Princeton, Darwin migrant shorebirds.Ecology 47: 393-406. Press. RICHARDSON, H., & N. A.M. VERBEEK. 1986. Diet HOCKEY,P. A. R., & J. COOPER.1980. Paralytic shell- selection and optimization by Northwestern fish poisoning--a controlling factor in BlackOys- Crowsfeeding on Japaneselittleneck clams.Ecol- tercatcher populations?Ostrich 51: 188-190. ogy 67: 1219-1226. KvrrEK, R. G., & M. K. BErrLER. 1988. A case for ROBERTS,D., D. RrrrscHoF, D. J. GERHART,A. R. sequesteringof paralytic shellfish toxins as a SCHMIDT,& L. G. HILL. 1989. Vertical migration chemical defense. J. Shellfish Res. 7: 629-636. of the clam Mercenariamercenaria (L.) (Mollusca: --, & . 1989. An addendum to a case for Bivalvia): environmental correlatesand ecologi- sequesteringof paralytic shellfish toxins as a cal significance.J. Exp. Mar. Biol. Ecol. 126: 271- chemical defense. J. Shellfish Res. 8: 253. 280. --, & . 1991. Relative insensitivity of but- SANGER,G. A., & R. D. JONES. 1982. The winter feed- ter clam neurons to saxitoxin: an exaptation for ing ecology and trophic relationships of marine sequesteringparalytic shellfish poisoningtoxins birds in KachemakBay, Alaska. Final Reportsof as a chemical defense. Mar. Ecol. Prog. Ser. In Principal Investigators, Environmental Assess- press. ment of the Alaskan Continental Shelf, NOAA. --, A. R. DEGANGE,& M. K. BErrLER.In press. SASNER,J. J., JR., M. IKAWA, & B. E. BARRETT. 1975. Paralyticshellfish poisoning toxins mediatesea The 1972 red tide in New Hampshire. Pp. 517- otter feeding behavior. Limnol. Ocean. 523 in Proceedingsof the First International Con- McKERNAN, D. L., & V. B. SCHE•FER. 1942. Unusual ference on Toxic Dinoflagellate Blooms (V. R. 392 mKHG. KvrrEK [Auk, Vol. 108

LoCicero, Ed.). Wakefield, Massachusetts,Mas- eriesrelationships (D. N. Nettleship, G. A. Sang- sachusetts Sci. Tech. Foundation. er, and P. F. Springer, Eds.).Seattle, Washington, SHUMWAY,S.E. 1990. A review of the effectsof algal Annual Meeting of the PacificSeabird Group. blooms on shellfish and aquaculture. J. World WHITE,A. W. 1977. Dinoflagellate blooms as prob- AquacultureSoc. 21: 65-104. able causeof an Atlantic herring (Clupeaharengus SULLIVAN,J. J., & M. M. WFaC•L. 1986. The appli- harengus)kill, and pteropods as apparent vector. cation of high performance liquid chromatog- J. Fish. Res. Board Can. 34: 2421-2424. raphy in a paralytic shellfish poisoning moni- 1980. Recurrenceof kills of Atlantic herring toring program. Pp. 357-371 in Seafoodquality (Clupeaharengus harengus) caused by dinoflagel- determination (D. E. Kramer and J. Liston, Eds.). late toxinstransferred through herbivorouszoo- New York, Elsevier. plankton. J. Can. Fish. Aqua. Sci. 37: 2262-2265. TuFrS, N.R. 1979. Molluscan transvectorsof para- 1981a. Sensitivity of marine fishesto toxins lytic shellfish poisoning. Pp. 403-408 in Toxic from the red-tide dinoflagellate Gonyaulaxexca- dinoflagellate blooms (D. L. Taylor and H. H. vataand implicationsfor fish kills. Mar. Biol. 65: Seliver, Eds.). New York, Elsevier/North-Hol- 255-260. land. 1981b. Marine zooplankton can accumulate VERMImR,K. 1982. Comparison of the diet of the and retain dinoflagellate toxins and cause fish Glaucous-wingedGull on the eastand west coasts kills. Limnol. Ocean. 26: 103-109. of Vancouver Island. Murrelet 63: 80-85. ZWARTS,L. 1986. Burying depth of the benthic bi- ß& N. BOURNE.1984. The White-winged Sco- valve Scorbiulariaplana (da Costa) in relation to ter diet in British Columbia waters:resource par- siphon-cropping. J. Exp. Mar. Biol. Ecol. 101: 25- titioning with other scoters.Pp. 30-38 in Marine 39. birds:their feedingecology and commercialfish-