Nest Failures in the Fulmar: the Effect of Observers

NEST FAILURES IN THE FULMAR: THE EFFECT OF OBSERVERS

BY JANETC. OLLASONAND G. M. DUNNET

INTRODUCTION Numerous seabirdstudies have dealt with the factorsaffecting breeding success(e.g., Coulson,1900; Mills, 1973;Davis, 1970; Brooke, 1978; Ollasonand Dunnet, 1978). Complementaryto thesestudies is the in- vestigationof causesand timing of breeding failures. Unfortunately it is impossiblein practiceto examine natural lossesof eggsor chicksin isolationfrom unnatural ones becausethe observer'spresence is likely to increasethe losses,i.e., causedisturbance. However, the stagein the breeding cycle (egg or chick) at which lossesare more likely to occur and their timing in relation to date can be determined. In the Fulmar (Fulmarusglacialis) Dunnet et al. (1903) suggestthat mostfailures occurat the egg stage:of all egg lossesin the first 9 days after laying,71% occurwithin the first 3 days.Mougin (1907) showstwo peaksin egg losses,one shortlyafter laying,the other around hatching. Ollasonand Dunnet (1978) demonstratethat late-laid eggsare more likely to fail. Breeding successin the Fulmar can vary widely from year to year, e.g., on the islandof Eynhallowin Orkney, Dunnet et al. (1979) quote a range of 16% to 52% over 28 years,although no known changeoc- curred in conditions on the island. This does not account for variations betweenyears in observereffort, which will now be consideredin detail. The aim of this paper is to determine for the Fulmar the most vul- nerablestage of the breeding cycle,to describethe pattern of egg losses in relation to date, and to examine the effects of observersstudying breeding biology on the responseof breeding adults and their subsequent breeding success.Causes of failure for a generalizedseabird will be describedtheoretically, and the Fulmar fitted into this framework.

METHODS In 1950, R. Carrick and G.M.D. began color-bandingbreeding Ful- mars caught from their nestson the small uninhabited island of Eyn- hallow, Orkney. Observationsduring the breeding seasonhave contin- ued annually to the present time. General methodshave already been described(Dunnet and Ollason, 1978). The initial aims of the study were to collectdata on breeding and survival.Only recentlywas it re- alized that we have useful data on the effects of disturbance. The data analyzedhere were collectedover 21 years from 1958 to 1978. Three annual vistswere taken to Eynhallow:in May (to count eggs),in July (to determine hatchingsuccess), and in August (to determine fledgingsuccess). 40] j. c. Ollasonand G. M. Dunnet J.Field Ornithol. Winter 1980

70--

60--

E 30--

20--

10-- I

14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 1 May Date June

FIGURE1. Average number of eggslaid in 1960, 1961, 1962, and 1978.

Visit in May In 1960, 1961, 1962, and 1978 the May visit lasted for about three weeksduring laying.Nests were checkedas often as possible(usually daily) to obtain laying datesand the number of eggslaid (apart from any eggslaid and lost betweenconsecutive checks of nests).The mean laying datesin theseyears ranged from 21.6 May in 1961 to 23.5 May in 1962 (Variance Ratio F = 7.61, df = 3.498, P < 0.001). However, it was felt that combiningthe data from all four yearsshould produce a distribution of laying dates typical of an average year (Fig. 1). In the remaining 17 years (1958, 1959, and 1963 to 1977), Eynhallow was visited during the fourth quarter of the laying distribution for 2 to 6 days beginning between 26 May and 8 June. The number of eggs ob- servedin eachof theseyears is therefore lessthan the total laid (because some egg losseswould have occurred before the first annual visit); it consistsonly of the number of eggspresent on the first day of the visit, plus any laid subsequently.A correction therefore needs to be applied to the observedtotals of these 17 years to allow for egg lossesbefore the first day of the May visit. Using the data from 1960, 1961, 1962, and 1978, the pattern of eggslaid and eggslost during laying has been determined for an averageyear: Figure 2 shows(a) the number of eggsto be laid after a particular day, expressedas a pro- Vol.5l, •4o.1 ObserverEffect on Nesting Fulmars [41

A 1'01 0000 0

0'8-- 0 0

0'4--

ß o

0'2•

0000 0

0• 0000

14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 1 May Date June

F•GURE2. Patternof egg-layingand egg-lossduring the laying period of an average year: (A) eggslaid after day x as a proportionof eggspresent on day x pluseggs laid after day x; (B) eggslost by day x as a proportionof eggspresent on day x. Data from 1960, 1961, 1962, and 1978. portion of those present on that day plus those laid subsequently,and (b) the number of eggslost by a particular day, expressedas a proportion of those present on that day. The totals were adjustedas follows: the observedtotal was multiplied by the proportion (a) referring to the first day of the May visitto give an estimateof the number of eggslaid after the first day of the visit.Subtracting this number from the observed total givesan estimateof the number of eggspresent on the first day, which is then multiplied by the relevant proportion (b) referring to the first day of the May visit to give the estimatednumber of eggslost by that day. Adding the three estimatesgives an adjustedtotal for each of the 17 yearswhich can then be comparedwith eachother and with the observedtotals of 1960, 1961, 1962, and 1978. No significantdifference was found betweenthe distributionsof the observedand adjustedtotals 42] J. c. Ollasonand G. M. Dunnet j. FieldOrnithol. Winter 1980

(Kolmogorov-Smirnovtest X22 = 5.29, NS) and the overallbreeding success(percentage eggs that fledge) calculatedfrom the observedtotals is highly correlated (% = 0.961, df = 17, P < 0.001) with that calculated from the adjustedtotals. The observednumber of eggspresent on the firstday is not significantlydifferent from the estimatednumber of eggs present on the first day (•: = 144.9 and 148.8; SD = 36.28 and 39.39, n = 16) and they are highly correlated(r = 0.988, P < 0.001). There- fore in all subsequentanalyses in this paper, the adjustedtotal number of eggshas been used for 1958, 1959, and 1963 to 1977. Visitin July In all years of the study, a secondvisit to Eynhallowof three or four daysoccurred in July, coveringpart of the hatchingperiod. By the end of'this visit most eggs had hatched.However, some eggs remained, a few of whichmay havehatched but failed beforefledging. Hatching success will therefore tend to be slightlyunderestimated. Visitin August The final annualvisit, of one day, occurredduring August,by which time almostall adults had left the area. All nestswith fledglingswere recorded,and therefore a good estimateof overall breeding success couldbe determined.Since the estimateof hatchingsuccess will be less accuratethan that of overall breeding success,all analysesinvolving hatchingsuccess have also been carriedout on overallbreeding success. In 1978, the responseof Fulmarson an egg or a chick to the stan- dardizedapproach of one particularobserver (J.C.O.) wascategorized as follows:"on"--bird remainedon the egg or chickall the time while the observerwas visible; "at"--bird left its egg or chickbut did not fly away and usuallystayed within a few inchesof its nest site; and "off"-- bird left its egg or chick and flew away. For each nest observedmore than five times,an index of remaining"on" the nestwas determined by observingthe number of times that the parent(s) remained "on" the nest and expressingthis as a proportion of the total observationsof parentsat that nest, while it had an egg or a chick. Nestswith fewer than six observations (55% of the total 215 nests) were eliminated to avoidthe biasesof smallsamples. This procedurewill alsohave removed nestswith birds very sensitiveto disturbance,i.e., neststhat failed after five or fewer disturbances.Thus the index of remaining "on" the nest cannot include the most sensitiveparents. The data werecoded and storedon magneticdisc in a database (using the AberdeenUniversity Data BaseManagement System). Most analyses were carried out using a Honeywell Level 66 computer.

RESULTS TheLaying Period: Eggs Laid and EggsLost The combinedlaying distributionsfor 1960, 1961, 1962, and 1978 are assumedto be typicalof an averageyear (• -- 22.7 May, SD = 3.40, Vol. 51, No. 1 ObserverEffect on Nesting Fulmars [43

T•,BI, F• 1. Mean percentagefailures before hatchingfor varioussets of years.

Mean SD n d P

(a) Years with long May visit = 66.90 3.289 4 1960 to 1962 and 1978 NS (b) Years with short May visit = 65.78 8.624 17 1958 and 1959, 1963 to 1977

(c) Yearswith large group in 71.79 7.007 8 July = 1971 to 1978 3.119 <0.01 (d) Years with small group in 62.42 6.115 13 July = 1958 to 1970

(e) Years with large group in 71.79 7.568 7 July excluding years with long May visit = 1971 to 1977 2.854 <0.02

(f) Yearswith smallgroup in 61.57 6.804 10 July excluding years with long May visit = 1958 and 1959, 1963 to 1970 NS

(g) Years with long May visit and 65.27 0.473 3 small group in July = 1960 to 1962 n = 502; Fig. 1). The proportion of eggslaid each day that fail to reach hatchingincreased significantly with laying date (rs = 0.759, df = 17, P < 0.001). No similar change occurred with laying date in the proportion of chickshatching that eventuallyfledged. During the layingperiods of 1960, 1961, 1962, and 1978, the day on which each egg was lost was known. The proportion of eggslost each day was not constant(Kolmogorov-Smirnov test X.,2 = 13.52, P • 0.01) but varied in an inconsistentway with date. Disturbance The presenceof observerson Eynhallowlikely causeddisturbance to the breeding Fulmars, possiblyresulting in lower hatchingsuccess and overallbreeding success. In the years 1960, 1961, 1962, and 1978, the first annual visit covered about three weeks during laying, whereas in the remaining years from 1958 onwards,the May visit covered only 2 to 6 days.In all years,2 to 6 people were presentat this time. Sincethe egg stageespecially is vulnerableto predation,the yearswith a long May visit might be expectedto have a higher failure rate. This wasthe case, but the mean failure ratesfor the two setsof yearswere not significantly different at hatching (Table 1, a, b) or overall (Table 2, a, b). 44] J. c. Ollasonand G. M. Dunnet J.Field Ornithol. Winter 1980

T^BLE 2. Mean percentageoverall breeding failures for varioussets of years.

Mean SD n d P

(a) Years with long May visit = 73.98 4.568 4 1960 to 1962 and 1978 NS (b) Years with short May visit = 72.54 8.038 17 1958 and 1959, 1963 to 1977

(c) Years with large groupsin 77.69 7.366 8 July = 1971 to 1978 2.560 (0.05 (d) Years with small groups in 69.82 5.890 13 July = 1958 to 1970

(e) Years with large groups in 77.27 7.854 7 July, excludingyears with long May visit = 1971 to 1977 2.211 (0.05 (f) Years with smallgroups in 69.23 6.645 10 July, excludingyears with long May visit -- 1958 and 1959, 1963 to 1970 NS (g) Years with long May visit and 71.77 1.429 3 small group in July -- 1960 to 1962

In the years1971 to 1978,groups of from 8 to 12 studentsplus 3 or 4 staff visitedEynhallow for 3 or 4 daysduring July, at the beginning of the hatchingperiod. In the earlier yearsof the study,the July visit wassimilar in date and duration, but the group consistedonly of 3 or 4 people.Comparsion of thesetwo setsof yearsshows that the hatching successand the overall breeding successwere significantlyworse in the yearswhen larger groupsvisited the island, regardlessof whether the years with long May visits are included (Table 1 and 2, c, d), or not (Tables 1 and 2, e, f). Sincethe larger groupsapparently had a significanteffect, this effect might be maskingthe effectof the long May visit (Tables1 and 2, a, b). The'•eforeusing only yearswith no large group in July, yearswith a long May visit were comparedwith yearswith a shortMay visitand found to havelower success (Tables 1 and 2, g, f), but differenceswere still not significant. From the data it is possibleto determine at what stagefailures oc- curredand approximatelyat whatdate (Table 3). Eggfailures occurring before the July visit, or in and after the July visit have been compared for (i) the short-May-visit-small-July-groupyears (1958-1970, excl. 1960-1962) and the short-May-visit-large-July-groupyears (1971- Vol. 51, No. 1 ObserverEffect on Nesting Fulmars [45

TABLE 3. Timing (in relation to date) of egg and chick failures.

(a) Failuresat egg stage May visit (i) Short (ii) Long July group size Small Large Small Large Years 1958-1970 1971-1977 1960-1962 1978 (excl. 1960-1962)

BeforeJuly visit 828 724 172 68 In and after July visit 121 236 42 82 x• 2 = 43.18 x• 2 -- 46.66 P < 0.001 P < 0.001

(b) Failures at chick stage May visit (i) Short (ii) Long Julygroup size Small Large Small Large Years 1958-1970 1971-1977 1960-1962 1978 (excl. 1960-1962)

In July visit 46 23 5 6 After July visit 62 55 16 13 ,)(12= 2.80 X? = 0.04 NS NS

1977) (Table 3a [i]) and (ii) the long-May-visit-small-July-groupyears (1960-1962) and the long-May-visit-large-July-groupyear (1978) (Ta- ble 3a [ii]). Both tablesshow significantdifferences between the setsof years,there being a much larger proportion of failures in and after July in the years with a large group present in July. Using the same setsof years,comparisons of failures at the chick stagein July or after July did not showsignificant differences associated with the large group present in July (Table 3b, [i] and [ii]). In this long-term study, adults need to be recaptured at regular in- tervals in order to ensure that color and metal bands remain legible. Most captures were made in July, and usually birds were caught off chicks.Perhaps the larger group of people presentin July allowed more birds to be caught and thereby decreasedbreeding successby increasing disturbance.However, captures, expressed as percentagecaptures of the number of breedingadults for eachyear, wasnot sigmficantlycorrelated with overall breeding success(rs = -0.058, df = 19) or with hatching success(rs = -0.111, df = 19). However significant negative correla- tions were found betweenhatching successand man-daysin July for each year (rs = -0.630, df = 19, P < 0.01; Fig. 3) and betweenoverall breeding successand man-daysin July (rs -0.538, df = 19, P < 0.02; 46] J. c. Ollasonand G. M. Dunnet j. FieldOrnithol. Winter 1980

50--

66 69

67 63 71 • o o• 40-- 65 õ

64 77 ß o 72o o • 75 78 59 74

20--

10--

0 0 20 40 60 80 100 Man-days on Eynhallow in July

FIGURE 3. P•rcentagehatching success plotted against man-days spent on Eynhallow duringJuly for 1958to 1978.Closed circles: years with small July group. Open circles: yearswith largeJuly group. Underlined symbols:years with long May visit. Captures (as percentageof breeding adults) were most numerousin 1960, 1959, 1962, and 1958, respectively.

Fig. 4), but neither hatchingsuccess nor overall successcorrelated sig- nificantlywith man-daysin May (rs = 0.146, df= 19 and rs = 0.117, df = 19, respectively).If the yearsare arrangedinto thosewith a large grouppresent in July (1971-1978)and thosewith a smallgroup present in July (1958-1971), overallbreeding success correlates with man-days in July when the group waslarge (rs = -0.838, df = 6, P < 0.01) but not when the group was small (rs = -0.149, df = 11). Responseof BreedingAdults to Approachof an Observer The responseof a Fulmaron an egg or a chickto the approachof an observervaried considerably.Some birds allowedan observerto seeits color and metal bands and to remove an egg or chick for weighing withoutbecoming disturbed or leavingthe nest site. At the other ex- treme, birds left their nest site as soon as the observerappeared, and did not return while the observer was able to see the site. Vol. 51, No. I ObserverEffect on Nesting Fulmars [47

40--167 ß 63 71 __• ß 69 ß o •_•68ß ß7o 65 ß60

72 -- --62 64 0 75 77 0 0 • 76 20- 5e 0 • 78 59 74 -

•o-

o I I I I I I I I I I 0 20 40 60 80 1 O0 Man-days on Eynhallow in July FIGURE4. Percentageoverall breeding successplotted againstman-days spent on Eyn- hallow in July for 1958 to 1978. Closedcircles: years with smallJuly group. Open circles:years with largeJuly groups.Underlined symbols:years with long May visit. Captures (as percentageof breeding adults) were most numerousin 1960, 1959, 1962, and 1958, respectively.

Neststhat failed during the egg stagehad a significantlylower mean proportion of birds remaining "on" (see methods)than nestsin which the egg hatched or the chick fledged (Table 4). The variation in proportion "on" was significantlygreater for nests that failed at the egg stage(Table 4). Some evidence indicatesthat the proportion "on" increasesas the breeding experienceof the bird increases;e.g., for nestsin which the length of breeding experienceof the male was known, thosein which the male'sexperience was 1 to 3 years had a significantlylower mean proportion of parents"on" than thosein which the male'sexperience was over 3 years (Table 5a). A similar, but nonsignificanttrend is ap- parent in relation to the female's experience (Table 5b). If breeding successis held constantby partial correlation,the proportionof parents "on" still correlatessignificantly with length of experienceof the male (r = 0.242, df = 76, P < 0.033).

DISCUSSION PossibleCauses of Failurefor a GeneralizedSeabird Figure 5 is a schematicrepresentation of a bird breedingseason, from egg productionto fledging, showingvarious factors that may increase 48] J. C. Ollasonand G. M. Dunnet J.Field Ornithol. Winter 1980

TABLE 4. Mean proportion of occasionswhen parent remains"on" the nest at the approachof an observer,for nestssuccessful to (i) the egg stage,(ii) the chickstage, or (iii) fledging.

Stage which Mean nest reaches proportion "on" SD n (i) egg only 0.6375 0.311 40 (ii) chick only 0.7813 0.177 18 (iii) fledgling 0.7612 0.213 36

Comparisons: F ratio P t or d P

(i) with (ii) 3.09 <0.015 2.23 <0.030 (ii) with (iii) 1.45 NS 0.35 NS (i) with (iii) 2.13 <0.025 2.04 <0.046

the probabilityof breedingfailure. The factorscan be classifiedas those due to either or both birds and those due to external causes. Their relativeimportance will vary dependingon the speciesand the parental contributions of the male and female. The pasthistory of both parents(age, breeding experience, and pairbond) is known to affect breeding success(e.g., in Kittiwake, Coulson, 1966; Red-billed Gull, Mills, 1973; Arctic Skua, Davis, 1976; Manx Shearwater, Brooke, 1978; Fulmar, Ollason and Dunnet, 1978). These factorsare presumablyeffective in termsof the physiologyand behavior of the parentsduring the breedingseason. Thus physiologicalcondition (Fig. 5a) of both birds is probablyimportant during egg production: poor qualityeggs are lesslikely to resultin fledgedchicks; e.g., a small egg producesa smallchick (fowl, Wiley, 1950; Brown-headedCowbird, Nolan and Thompson, 1978; Laughing Gull, Ricklefset al., 1978), with a lower survivalrate (Herring Gull, Parsons,1970); or the egg may be

TABLE 5. Mean proportion of occasionsthat parent remains "on" the nest at the approach of an observer,for parents with 1 to 3 years breeding experience,or over 3 years breeding experience.

Length of breeding Mean experience, proportion years "on" SD n t P

(a) Male 1 to 3 0.6368 0.262 13 2.08 <0.042 over 3 0.7769 0.209 56 (b) Female 1 to 3 0.6570 0.264 19 over 3 0.7335 0.213 47 1.23 NS Vol.5 l, No.1 ObserverEffect on Nesting Fulmars [49 infertile; or it may have constituentsin the wrong proportions.Possibly undue stressat this time might result in resorption of the egg, but this is unlikely to be detected.If physiologicalcondition is not good at the beginning of breeding, there may be effects later in the season;e.g., nutritional state of the female may affect overall breeding success(Red Grouse, Moss et al., 1975). If an egg is laid (Fig. 5b), both birds need to be efficient incubators (Fig. 5c; Baerends,1959; Drent, 1973) and communicatewith each other to maintain a shared incubation schedule. Poor incubation behavior may result in eggsbeing left uncovered:although eggsmay have some chilling resistance(Manx Shearwater, Mathews, 1954) and effects of weather are not alwayspredictable (Ad•lie Penguin, Ainley and Le- Resche, 1973), the probability of killing eggs will be increasedby ex- posure to excessiveheat (Herring Gull, Hunt, 1972), cold (Wilson's Storm Petrel, Beck and Brown, 1972) or wind (Common Tern, Goch- feld, 1978). Uncovered eggsare also vulnerableto predation (Lesser Black-backedGull, Davisand Dunn, 1976). Similarlyany disruptiveex- ternal factorsthat causethe incubatingbird to leavethe nestwill expose the eggsto weather and predators(gulls, Harris, 1964; Herring Gull, Hunt, 1972; Double-crestedCormorant, Kury and Gochfeld, 1975; WesternGull, Robert and Ralph, 1975; Double-crestedCormorant, El- lison and Cleary, 1978). If the egg hatches(Fig. 5d) the parents need sufficientforaging skill to obtainfood for themselvesand their young(Fig. 5e; Ad•lie Penguin, Ainley and Schlatter, 1972; Brown Pelican,Orians, 1969 and Blus and Keahey,1978) and they mustbe ableto feed them efficiently.The chicks also need protection from predators and weather, and disturbanceof the parents may reduce or remove this protection (Ad•lie Penguin, Reid, 1968;Glaucous-winged Gull, Gilletet al., 1975).However, in some speciesadults brooding chickshave been shownto be more tolerant of disturbancethan thosewith eggs(Double-crested Cormorant, Kury and Gochfeld, 1975), or chickshave becomeless frightened after repeated disturbance(Western Gull, Robert and Ralph, 1975). The vulnerability of the egg and chick stagestherefore varieswith the species. Thus the probabilityof breeding failure is increasedif the parents are in poor condition or if their behavior is inadequate.External disturbance,whether intentional or not, exaggeratesthe lossesby altering the behaviorof the parents.Some species may be lesssusceptible, since their successis improvedwhen they nest under cover (gulls,Brown, 1967; Lesser Black-backed Gull, Davis and Dunn, 1976; Common Tern, Nisbet, 1975). In othersit is shownthat adultsnesting on the periphery of the colony fled more easilyat the approachof an observerthan those nesting in the center (Ad•lie Penguin, Tenaza, 1971). Also disturbance in one year can have an effect on breeding in the followingyear (Ad•lie Penguin, Oelke, 1975 and Reid 1968; Fulmar, Ollason and Dunnet, 1978). 50] J. C. Ollasonand G. M. Dunnet j. FieldOrnithol. Winter 1980

(a) EGG PRODUCTION Malein Femalein [Malein Femalein AND poor poor I / good good FERTILISATION conditioncondition I /condition •dition

Egg of poor quality Egg of good quality

(b) LAYING EGG LAID

Egg poorly incubated Egg well incubated

(d) HATCHING EGG HATCHES

(e) CHICK BROODING AND FEEDING IMalefeeder poor} {feederFemale poor{{ Malefeeder good Femalefeeder good[

Chick develops slowly Chick develops normally

(f) FLEDG ING CHICK FLEDGES

FIGURE5. Schematicdiagram of a generalizedseabird breeding seasonfrom egg production to fiedging.Disturbance refers to any factor external to the parentswhich altersthe behaviorof the parents.In general,factors which cause the flow to be down the left sideof the diagram will increasethe probabilityof breedingfailure.

Causesof Failure in theFulmar In the Fulmar, both parentsgo on a prelayingexodus (Dunnet et al., 1963), which is typical of Procellariiformes(Warham, 1964): on the average the female is away for 20 days;the male is away for 9 daysand Vol.51. •o. 1 ObserverEffect on Nesting Fulmars [5 1 returns before the female (Macdonald, 1977). Presumablythey go to distant feeding grounds to build up reservesfor the breeding season. The nutritional state and physiologicalcondition of both birds would therefore seemto be critical(Fig. 5a). The singleegg weighsabout 13% of the female and relaying has never been known to occur. Some un- publishedevidence suggests that, within the history of a female, larger eggsare more likely to be successfulthan smaller ones; i.e., larger eggs are of a better qualityand presumablyuse more of the female'sreserves. Possiblydisturbance during egg production might causeresorption, but this would be difficult to demonstrate, and was unlikely in this study sinceobservers were not normally presentat this time and the female is away from the colony. After the egg is laid, the male normally relievesthe female at the nest within a few hours, and then incubatesfor about 10 days. Thereafter the incubation stints are approximately equal (Dunnet et al., 1963). Therefore, for incubation to be successful the male and female must be in the right place at the right time and communicatesuccessfully with each other (Fig. 5c). The egg is white (i.e., not camouflaged)and no nest material is present,so that if the egg is left uncoveredit is vulner- able to predators as well as to extremesof weather. Late-laid eggsseem more at risk than early-laid ones, since fewer of the late-laid eggs reachedhatching. This is consistentwith the fact that more experienced Fulmars tend to lay earlier and are more successful(Ollason and Dun- net, 1978). The data presented here show some lowering of hatching success(and overall success)in the four years in which observerswere present during the whole of laying (Tables 1 and 2; a and b, g and f) although the differences were not significant, probably because the number of observerswas small. A significantincrease in breedingfailure was found in years in which a group of at least 12 people was present for 3 or 4 days during hatching (Tables 1 and 2; c and d, e and f), i.e., at a time correspondingto the secondpeak of egg lossesof Mougin (1967). Thus disturbance(measured in terms of man-days)during incubation(Fig. 5c) does have a significanteffect upon hatching success (and overall success). After the egg hatches,both parents feed and brood the chick. It is not known what food is obtained, nor at what distancethe parents for- age.Disturbance at this time (Fig. 5e) leavesthe chickexposed, although Fulmar chickscan protectthemselves to someextent by ejectingnoxious oil (Armstrong, 1951). In this study, most capturesof adult birds occurred when the chick was present--capture of breeding Fulmars can affect their subsequentbreeding success(Ollason and Dunnet, 1978). However, no relationshipwas found here betweenoverall breeding successof the population as a whole each year and capturesin that year. This is probably becausecapture only affects the successof the nest at which an individual wascaptured, and capturesnormally only occurred at about 10% of nestsin any one year. But it is interestingto note that 52] j. c. Ollasonand G. M. Dunnet j. FieldOrnithol. Winter 1980 in two years (1958 and 1959, Figs. 3, 4) in which successwas low and man-dayswere low, disturbancedue to captures was relatively high. Thus disturbanceat the chickstage can have someeffect, but at leastin this study,the chickis not as vulnerableas the egg. Responseto theApproach of an Observer How doesan observercause disturbance? Significant differences between birds are shown(Table 4) in the responseto a particular observer at neststhat subsequentlyfailed (64% remained on site) and thosethat were subsequentlysuccessful (76% remained on site). The differences are similar to thosedescribed for the Ad61iePenguin by Tenaza (1971), and the behavior of birds at unsuccessful nests is more erratic. Thus putting a bird off its nestan additionalonce or twiceout of 10 occasions could make the difference between a successful and an unsuccessful nest. The evidenceindicates that the behavior of birds to the approach of an observertended towardsstaying on the nest as breeding experience increased(Table 5), and this may be one of the factors that resultsin improved successof more experienced birds (Ollason and Dunnet, 1978), rather than being a consequenceof it. Effects of disturbance probably vary with species(e.g., Glaucous- winged Gull, Gillet et al., 1975; Double-crestedCormorant, Kury and Gochfeld, 1975), as well as with habitat (gulls, Brown, 1967; Lesser Black-backed Gull, Davis and Dunn, 1976; Common Tern, Nisbet, 1976) and frequencyof disturbance(Western Gull, Robert and Ralph, 1975). Perhapsfactors such as these should be consideredwhen designingfield projects.

SUMMARY Data were collected over 21 years as part of a long-term study of breeding Fulmars on the uninhabited island of Eynhallow in Orkney. The pattern of egg lossesis described.Eggs laid early in the seasonwere more likely to hatch than those laid late. The relationshipbetween number of observerseach year and subsequent breedingsuccess that year wasexamined. In four yearsin which 2 to 6 observerswere present for 2 to 3 weeks during laying (May), hatching successand overall successwere lowered, but not significantly so. No relationshipexisted between man-days present in May and breeding success.In eight years in which at least 12 observerswere present for 3 to 4 days during hatching(July), hatchingsuccess and breeding successwere significantlylowered. A significantnegative correlation was found betweenman-days present in July and breeding success. Capture of breeding adultshad no significanteffect upon subsequent breeding successof the population as a whole, probably becausecap- turesonly occurredat about 10% of nests.The responsesof sittingbirds to the approachof an observervaried: at nestswhich were subsequently Vol.•1, No. ! ObserverEffect on Nesting Fulmars [53 successful,76% of birds remainedon site: the figure for unsuccessful nests was 64% and their behavior was more erratic. The causesof failure for a generalizedseabird are described,and the Fulmar is fitted into this framework.

ACKNOWLEDGMENTS We are grateful to the late MissJ. M. Robertson,Kirkwall, for allowing us to observeFulmars and to use her houseon Eynhallow,Orkney. We thank Mr. A. Anderson, Aberdeen University, for his long-time assis- tance, and the numerous others who have helped at some time over the years. The useful criticismsof Dr. I. J. Patterson,Aberdeen University, were much appreciated.Aberdeen University Computer Centre provid- ed facilitiesfor the data analysis.

LITERATURE CITED

AINLEY, D. G., AND R. E. LEREsc•IE. 1973. The effects of weather and ice conditions on breeding in the Ad•lie Penguin. Condor,75: 235-238. AINLEY,D. G., ANDR. P. SCHLATTER.1972. Chickraising ability in Ad•lie Penguins.Auk, 89: 559-566. ARMSTRONO,E. 1951. Dischargeof oily fluid by young Fulmars.Ibis, 93: 245-251. BAERENDS,G. P. 1959. The ethologicalanalysis of incubationbehaviour. Ibis, 101: 357- 368. BECK,J. R., AND D. W. BROWN. 1972. The biology of Wilson's Storm Petrel Oceanites oceanicus(Kuhl) at SignyIsland, SouthOrkney Islands.Br. Antarct.Surv. Sci. Rep., 69: 1-54. BLUS,L. J., ANDJ. A. KEAHEY.1978. Variationsin reproductivitywith age in the Brown Pelican. Auk, 95: 128-134. BROOKE,M. DEL. 1978. Somefactors affecting the layingdate, incubationand breeding successof the Manx Shearwater,Puffinus puffinus. J. Anim.Ecol., 47: 477-495. BROWN,R. G. B. 1967. Breedingsuccess and populationgrowth in a colonyof Herring and LesserBlack-backed Gulls, Larusargentatus and L. fuscus.Ibis, 109: 502-515. COULSON,J. C. 1966. The influenceof the pairbond and age on the breedingbiology of the KittiwakeGull Rissatridactyla. J. Anim.Ecol., 35: 269-279. DAVIS,J. W. F. 1976. Breeding successand experiencein the Arctic Skua Stercorarius parasiticus.J. Anim. Ecol., 45: 531-536. DAVIS,J. W. F., ANDE. K. DUNN. 1976. Intraspecificpredation and colonialbreeding in Lesser Black-backed Gulls Larus fuscus. Ibis, 118: 65-77. DRENW,R. 1973. The natural historyof incubation.In BreedingBiology of Birds. D. S. Farher (ed.). Washington,D.C., Nat. Acad. Sci. DUNNET,G. M., A. ANDERSON,AND R. M. CORMACK.1963. A study of survival of adult Fulmars with observationson the pre-laying exodus.Brit. Birds, 56: 2-18. DUNNEW,G. M., ANDJ. C. OLLASON.1978. The estimationof survivalrate in the Fulmar Fulmarusglacialis. J. Anita. Ecol., 47: 437-450. DUNNET,G. M., j. c. OLLASON,AND A. ANDERSON.1979. A 28-year study of breeding Fulmars Fulmarusglacialis in Orkney. Ibis, 121: 293-300. ELLISON,L. N., AND L. CLEARY.1978. Effects of human disturbanceon breeding of Double-crested Cormorants. Auk, 95:510-517. GILLET,W. H., J. L. HAYWARD,JR., ANDJ. F. STOUT. 1975. Effectsof human activityon egg and chickmortality in a Glaucous-wingedGull colony.Condor, 77: 492-495. GOC•IFELD,M. 1978. Incubatingbehaviour in Common Terns: Influence of wind speed and direction on orientationof incubatingadults. Anita. Behav., 26: 848-851. HARRIS,M.P. 1964. Aspectsof the breeding biology of the Gulls Larus argentatus,L. fuscusand L. marinus.Ibis, 106: 432-456. 54] J. c. Ollasonand G. M. Dunnet j. FieldOrnithol. Winter 1980

HUNT, G. L., JR. 1972. Influence of food distribution and human disturbanceon the reproductivesuccess of Herring Gulls. Ecology,53: 1051-1061. KUR¾,C. R., ANDM. GOCHFELD.1975. Human interferenceand Gull predation in Cor- morant colonies. Biol. Conserv., 8: 23-34. MACDONALD,M. A. 1977. The pre-laying exodusof the Fulmar Fulmarusglacialis (L.). Ornis Scand., 8: 33-37. MATHEWS,G. V. T. 1954. Some aspectsof incubationin the Manx Shearwater,Procellaria puffinus,with particular referenceto chilling resistancein the embryo. Ibis, 96.' 432- 440. MILLS,J. A. 1973. The influence of age and pair-bond on the breeding biologyof the Red-billedGull (Larus novaehollandiae scopulinus ). J. Anita. Ecol.,42: 147-162. Moss, R., A. WATSON,AND R. PARR. 1975. Maternal nutrition and breeding successin Red Grouse (Lagopuslagopus scoticus) J. Anita. Ecol.,44.' 233-244. MoUC•N,J. L. 1967. -•tude•cologiquedes deux esp•ces de Fulmars--leFulmar Atlan- tique (Fulmarusglacialis) et le Fulmar Antarctique(Fulmarus glacialoides). œ'Ois. et la R.F.O., 37• 57-103. NISBET,I. C. T. 1975. Selectiveeffects of predationin a tern colony.Condor, 77: 221- 226. NOLAN,V., Jm, AND C. F. ThOmPSON.1978. Egg volume as a predictor of hatchling weight in the Brown-headedCowbird. WilsonBull., 90: 353-358. OEL•E,H. 1975. Breedingbehaviour and successin a colonyof Ad•lie PenguinsPygoscelis adeliaeat Cape Crozier, Antarctica.In The Biologyof Penguins.B. Stonehouse(ed.). London and Basingstoke,Macmillan. OLLASO•,J. C., AN• G. M. DUNNET. 1978. Age, experienceand other factorsaffecting the breedingsuccess of the Fulmar Fulmarusglacialis in Orkney. J. Anita.Ecol., 47: 961-976. ORGANS,G. H. 1969. Age and hunting successin the Brown Pelican(Pelecanus occidentalis). Anim. Behav., 17: 316-319. PARSONS,J. 1970. Relationshipbetween egg sizeand post-hatchingchick mortality in the Herring Gull (iLarusargentatus). Nature, 228.' 1221-1222. REID, B. 1968. An interpretation of the age structure and breeding statusof an Ad•lie Penguinpopulation. Notornis, 15: 193-197. RIngLEts,R.-E., D.C. HAHN, AN• W. A. MONTEVECCm.1978. The relationshipbetween egg sizeand chicksize in the LaughingGull and JapaneseQuail. Auk, 95: 135-144. ROBERT,H. C., ANDC. J. RaLph. 1975. Effectsof human disturbanceon the breeding success of Gulls. Condor, 77• 495-499. T•NAZA,R. 1971. Behaviourand nestingsuccess relative to nest locationin Ad•lie Pen- guins(Pygoscelis adeliae). Condor, 73: 81-92. WARnA•,J. 1964. Breedingbehaviour in Procellariiformes.In BiologieAntarctique. R. Carrick, M. Holdgate,and J. Prevost(eds.). Paris, Hermann. WILEY,W. H. 1950. The influenceof egg weighton the pre-hatchingand post-hatching growth rate in the Fowl. Poultry.Sci., 29: 570-574. CultertyField Station, Department of Zoolog?i,University of A berdeen, Aberdeen AB40AA. Received3 May 1979, accepted29 August 1979.