In 1899, Mortensen initiated the systematic marking of wild birds with numbered leg bands (Wood 1945), but apart from the use of mod- ern alloys and color bands there have been few major developments in bird-band technology. Most types of bird bands used today still require the recovery of a dead bird or the recapture of a living bird in order to identify the band, even though these two sources of information have obvious disadvantages for the study of non-sedentary free-living birds. Many individual life histories are derived from banded birds being shot or found dead and hence for each bird there are only two records, and where an individual is retrapped at a particular colony during breeding seasons there are no data on where the bird went during the non- breeding season. In this paper we report results from the use of bands with large numerals which were developed to enable individual identification of free-living Australian Silver Gulls (Larus novaehollandiae) at distances up to 50 m. The effectiveness and durability of the bands are evaluated, the nature and frequency of errors detected in field observations dis- cussed, and the quality of the data returned is compared with that from other studies which used standard bands. MATERIALS AND METHODS Description of the number bands and color bands.--Carrick and Murray (1970) described the manufacture and specifications of the bands they developed to enable positive identification of free-living Silver Gulls. In brief, number bands were made from 1 mm thick aluminum alloy, and when flattened were 16 mm high x 25 mm. A unique 5-digit number was impressed into the band four times, the impressions were coated with black enamel paint, and the bands then sprayed twice with "Scotch- lite Finishing Clear." When a number band was fitted to the leg of a gull, the numbers read vertically (Fig. 1), and the whole sequence could be seen from any angle. A return address for the bands was impressed onto the inside. Color bands were made from 0.9 mm thick commercial aluminum sheet, cut to 13 mm high x 24 mm for 1-color bands or 19 mm high x 24 mm for 2-color or 3-color bands. "Scotchlite" reflective colors and "Scotchcal" non-reflective black or white were heat-bonded FIGURE 1. Examples of number bands recovered from Silver Gulls: 61942, taken from chick which died on nest; 15567, standard CSIRO 080 series band; 63180, recovered after 9 yr 7 mo; 61029, recovered after 12 yr 11 mo; 51264, from dead chick on nest; 59638, 8 mo; 65966, 9 yr 10 mo; 66192, 9 yr 10 mo; 63583, 11 yr 10 mo. The large 081 aluminum bands are 16 mm (high) x 25 mm, I mm thick, and were sprayed with Scotchlite Finishing Clear (c) after coating the number impression with black paint. These bands read from top to bottom. As can be seen, most wear occurs at the bottom, and it is recommended that bands with vertical number sequences should be placed on birds to read from bottom to top. to the outside of the bands in various color and sequence combinations (Fig. 2). (Scotchlite (c), Scotchcal (c) and Scotchlite Finishing Clear (c) are registered trademarks of the 3M Company.) White Scotchcal was heat- bonded to the inside of color bands to increase durability of the alu- minum. Gull banding.--Between June 1967 and November 1970, 17,410 Sil- ver Gulls were banded in South Australia, using the bands described above (Table 1). Some 15,950 gull chicks were banded on the nest at 15 island breeding colonies, 173 juveniles were clap-netted and banded in the Beachport area, and 1287 adult gulls were banded after nest- trapping or clap-netting: 142 on islands at Encounter Bay and 1145 in the Beachport area (Fig. 3). Each banded gull was given a unique number band on one leg and a color band on the other. Some 467 types of color bands were used, being various combinations of blue, green, red, yellow, black, and white, size of band, vertical, horizontal, or oblique color sequencing. One-color and 2-color combinations were used for mass- banding of groups (e.g., each of 3530 chicks hatched on Penguin Island in Spring 1968 was given a 13 mm high yellow band on the right leg, referred to in text as R-Y), while 432 unique 3-color bands were used FICuRE 2. Examples of color bands used on Silver Gulls. RWR, new; YGBk, 1 yr 1 mo; YG-(--YGR), 8 yr 11 mo; BkGBk, 9 yr; GBk-(-GBkB), 10 yr; Bare metal (--WRW), 11 yr 9 mo; vBkW, new; vBkW, 12 yr 10 mo; WG, new; WG, 12 yr 2 mo; Y, new; Y, 12 yr 9 mo; v, each color runs vertically; B, blue; Bk, black; G, green; R, red; Y, yellow; W, white. The large bands are 19 mm (high) x 24 mm; the two small bands are 13 mm (high) x 24 mm. The color materials are Scotchlite (c) or Scotchcal (c) heat- bonded to aluminum. on nest-trapped breeding adults (e.g., only one gull had the band black/ red/green on the left leg, and one gull had the same color combination on the right leg. These are referred to as L-BkRG or R-BkRG). Band resightings and field records.--During October 1968-August 1981, teams of professional ornithologists made extensive searches of south- eastern Australia to find and identify these banded gulls. During 1969, 1970, and 1971, the teams spent 848 days searching for banded gulls, but only 56 days during 1972-1981. About 60 occasional observers, scattered through South Australia and Victoria, also watched for banded gulls, and members of the public returned bands, found on dead or injured gulls, to the Australian Bird-Banding Scheme. The professional ornithologists concentrated on shores and coastal areas between Ade- laide, South Australia, and Cowes, Victoria (Fig. 3) with occasional forays to more distant areas where banded gulls had been reported. Banded gulls could be detected at distances up to about 100 m, al- though it was necessary to be much closer to identify color bands and read number bands. Usually, a flock of gulls was approached by 2-4 observers in a vehicle suited to the particular terrain. Gulls were not disturbed unduly by the slow movement of the approaching vehicle, but TABLE 1. Summary of Silver Gull bandings,June 1967-November 1970, and resightings, October 1968-August 1981, in south-eastern Australia. Gulls Gulls Total Banding site a banded resighted (%) resightings 1. ICI, St. Kilda, Adelaide 2178 40 1.84 77 1.93 2. Outer Harbour, Adelaide 1780 59 3.31 108 1.83 3. West Island 1331 113 8.48 188 1.66 4. Wright Island 731 70 9.58 121 1.73 5. Lake Mulgundawa 933 40 4.29 52 1.30 6. Seagull Is., Coorong 94 14 -- 46 -- 7. Lake Hawdon 2025 385 19.01 1373 3.57 8. Nora Creina 1084 240 22.4 1144 4.77 9. Lake George Islands 1154 388 33.62 1927 4.97 10. Cowrie Rock 536 203 37.87 3812 18.78 11. Pleasant Cove 63 58 92.06 1282 22.10 12. Penguin Island 4446 1629 36.64 12,363 7.60 13. Rivoli Bay 679 609 89.69 13,663 22.44 14. Southend Beach 121 100 82.64 1394 13.94 15. Harrington Rock 7 1 -- 20 -- 16. Cullens Rock 79 9 -- 54 -- 17. Mounce and Battye Rocks 151 17 11.26 78 4.59 18. Carpenters Rock 18 2 -- 72 -- Totals 17,410 3977 22.84 37,774 9.50 a Colonies and banding sites 9-14 were within the intensive study area of the resighting program. b g = mean number of resightings per resighted gull. were always wary of observers on foot. With favorable conditions, num- ber bands could be read with the unaided eye at distances of less than 2.5 m, with binoculars from 2.5-20 m, and with telescopes from 18- 50 m distance. Color bands could be identified at up to 60 m. These approximate maximum distances for legibility depended on quality and clarity of the band, weather and light conditions, behavior of the banded gull at the time of observation, the individual observer's visual acuity and experience with reading gull bands, and the quality of the optical equipment. If possible, bands were read and checked by several ob- servers. Where a digit or color remained in doubt after a gull had left the observation site, an appropriate coded comment, which remained with the record, was made in the field notebook (e.g., 6(5)047 signified that the digit 5 was doubtful). A resighting record was considered valid if the number band, color band, and leg sequence of bands recorded at the time of sighting corresponded with the banding records. Where discrepancies were found between observation records and banding rec- ords, a subjective decision was made whether to reject the observation, whether to accept it as possibly valid but requiring further field confir- mation, or to accept it as valid despite the discrepancy. This decision was based on consideration of the combination of number band and o 100km I J 3 5 6 9 10 11 12 ,- ................... 13 14 FAIRY 15 16 18 17 1. ICl, St. Kilda, Adelaide 2. Outer Harbour, Adelaide 3. West Island 4. Wright Island 5. Lake Mulgundawa 6. Seagull Island, Coorong 7. Lake Hawdon 8. Nora Creina 9. Lake George Islands 10. Cowrie Rock I 1. Pleasant Cove 12. Penguin Island 13. Rivoli Bay 14. Southend Beach 15. Harrington Rock 16. Cullens Rock 17. Mounce and Battye Rocks 18. Carpenters Rock FICURE 3. Banding locations of Silver Gulls in southern Australia, 1967-1970. color band observed, the place of sighting, and the nature of the dis- crepancy (Table 2). By 1978, banded gulls were much scarcer, and most number bands and all color bands were showing signs of wear. To provide further quantitative data on the effectiveness of these bands and the band re- sighting program, a sample of the surviving banded gulls (N = 237) was shot in 1978 and 1981. Although selection of the gulls to be shot was not based on formal randomization (for reasons of both public safety and local government regulations), the gulls were shot during all daylight T^mE 2. Examples of appraisal of incorrect sighting records of Silver Gulls. Num- ber of Correct record Incorrect record similar color possibilities* Decision Left Right bands Left Right Left Right leg leg used leg leg Comment leg leg 53724 GBG 1 GBkG 53724 This bird was re- accept as: 57358 GBG ported as found dead GBkG 53724 GBG 57433 at a village 40 km 57364 GBKG from banding place. The leg order was incorrectly report- ed by the finder, and one of the profes- sional team incor- rectly recorded GBkG as GBG. 53757 GYW 1 BkYW 53757 At least two errors reject record 53744 GYW involved. GYW 57411 vYG (53970) 66 vYG 55970 At time of record, accept as: observer noted un- vYG 55970 certainty about num- ber, possibly because the bird flew away before confirmation was possible. Confu- sion of 5 and 3 was a common error. (In this instance band 53970 was not a pos- sibility since it had not been used.) 55927 66 55927 vYG The bird was stand- accept as: ingon oneleg, which 55927 vYG carried the number band. Left and right legs were often con- fused when only one leg could be seen. L-55927 had been seen on 9 occasions at the place of the in- correct sight record. GWR 56787 1 56787 vGW At least two errors reject record GWR 57420 involved. 57351 GWR 57304 Bk-W 6 57304 BkRW Middle color strip accept as: (red) had faded or 57304 BkRW worn away. TABLE 2. Continued. Num- ber of Correct record Incorrect record similar color possibilities* Decision Left Right bands Left Right Left Right leg leg used leg leg Comment leg leg (573(7)6) BkGW 1 57316 BkGW This band number, accept as: 57376 BRBk and especially the 57316 BkGW BkGW 53751 fourth digit, was noted to be doubtful at the time of obser- vation. R-BkGW was a unique band/leg combination. Y 60296 3530 60296 Y Leg reversal. accept as: 6O296 Y 63745 W 750 63745 RW This incorrect re- accept as: 53780 WWW cord came 5 yr 5 mo 63745 RW WWW 65748 after banding. Pre- vious observations indicated that R fad- ed or wore off within 2-4 yr. 66681 Y 3530 GWG 66681 Unlikely that R-Y reject record was confused for L- GWG. More proba- bly 1 or 2 digits of the number were in- correct (for in- stance, could have been L-56581). Some 3530 chicks were banded R-Y, hence the bird sighted could be one of many possibili- ties. 69300 Nil 463 69300 WG This bird lost the accept as: color band within 4 69300 (WG) months. It had 41 subsequent confir- matory resightings. * B -- blue; Bk = black; G --- green; R -- red; W = white; Y = yellow; v --- an individual color runs vertically on the band, otherwise assume individual colors run horizontally around band (see Fig. 2). Hence, L-GYW indicates green over yellow over white on the left leg. 110 - 100- 90 "5 80 Q. 70- ß  60- o z,O- 3o- ß ß ,. 'o .,../. ., . . . ß ..-.:'.:. ...". . 0 ' Time between banding and last record(years) FlouRs 4. Correlation between the time Silver Gulls were known to be alive and the number of records for individual gulls, for a sample of 188 breeding adults banded at Beachport, South Australia, 1968 and 1969 (rxy = .385). hours and at many different sites in order to achieve a representative sample of survivors. Recovered number bands were washed, dried, and weighed to the nearest 0.001 g on an analytical balance. Where the serial number of a band could not be identified, bands were opened, flattened, washed in water, then 100% ethanol and finally xylene. After drying, the bands were swabbed with a solution of 30% hydrofluoric acid, 15% nitric acid, and 55% glycerine (Villella's solution). The alu- minum affected by stamping of the serial number at the time of band manufacture etched away faster than the surrounding metal, and the digits were revealed, even on number bands which had been worn smooth. Recovered color bands were carefully examined for remaining traces of Scotchlite or Scotchcal, but were not treated further. RESULTS There was a significant (r = .385; P < .01) positive correlation be- tween number of records per individual gull and length of time that individual gulls were known to be alive; however, there was great vari- ation in the frequency with which individual gulls were resighted (Fig. 4). Of the 17,410 Silver Gulls banded, 3977 (22.8%) were seen again, T^BLv. 3. Band recoveries from 237 Silver Gulls banded in 1967-1970 and shot in 1978 and 1981. Condition of bands No. % Both bands, legible Both bands, number only legible Both bands, color only legible a Both bands, neither legible a Number band only, legible Number band only, illegible a Color band only, legible b Color band only, illegible b 116 48.9 53 22.4 4 1.7 1 0.4 21 8.9 1 0.4 36 15.2 5 2.1 Totals 237 100 a All illegible number bands were readable after acid etching. b By 1978 most Scotchlite and Scotchcal had worn off the aluminum bands. Where all traces of color were gone, the color band was considered illegible. and of 6999 gulls banded in the intensive study area, Lake George to Southend, 2987 (42.7%) were seen again (Table 1). A total of 37,774 valid sightings was accepted to August 1981 (R = 9.5 resighting records per resighted gull): 35,189 resightings from the professional teams and 2585 from the occasional observers. Some 26 gulls had > 100 records and 61 gulls had histories spanning > 12.75 yr from banding. Loss and wear of bands.--Loss and wear data were derived: (1) from general observations on the condition of bands while resighting banded gulls, (2) from specific observations on 246 adult gulls in the Beachport intensive study area, which were given unique 3-color bands in addition to their unique number bands, and (3) from a sample of 411 bands taken from 237 gulls shot in 1978 and 1981. Some band losses were discovered within 4 months of banding, al- though until the end of 1971 the total incidence of gulls known to have lost either band was considerably less than 0.5%. By 1978, 5.4% of known surviving gulls had lost color bands, and 10.0% had lost number bands. Of the 246 Beachport adults given unique color bands, 3 (1.2%) were known to have lost color bands and 5 (2.0%) were known to have lost number bands by August 1981, after 12-13 yr; however, many of these adults are thought to have died within a few years of banding. Of 237 banded gulls shot in 1978 and 1981, 41 (17.3%) had lost number bands and 22 (9.3%) had lost color bands (Table 3). Of the 196 number bands recovered, 160 (81.6%) were considered to have been still legible on the free-living gulls, 30 (15.3%) could be read only in the hand, and 6 (3.1%) required treatment with Villella's solution to be rendered legible. Of 215 color bands taken from the shot gulls, none was clearly identi- fiable on the free-living gulls, but 157 (73.0%) were identified in the hand. Some 58 (27.0%) had been worn to bare aluminum, with no traces of color. T^BLE 4. Comparison of resighting records from Silver Gulls, shot in 1978 and 1981, found to have fast-wearing, average-wearing, and slow-wearing bands. Number Total of indi- number vidual of re- Wear rate gulls Location of resighting records a sighting of band in records group group Beach Swamp Town Dump Colony (Z n) Fast 30 186 (37.0) 61 (12.1) 145 (28.8) 44 (8.7) 67 (13.3) 503 Average 30 172 (32.2) 59 (11.0) 160 (30.0) 70 (13.1) 73 (13.6) 534 Slow 30 191 (34.1) 39 (7.0) 158 (28.2) 44 (7.8) 128 (22.8) 560  Numbers in parentheses are percentages of total number of observations (Z n) in wear rate group. All recovered number bands (n = 196 from shot birds + 13 others = 209) were used in the regression of band weight on time (Fig. 5). This calculation gives a biased estimate of band wear, because the sample does not include the fast-wearing bands which were lost from gulls (41 of the gulls shot in 1978 and 1981 had lost the number band). However, it is clear that the mean band wear is at least .0411 g/yr. Inspection of recovered number bands indicated that properly closed bands were unlikely to drop off as a result of wear until the bands were lighter than about .45 g. Since unused number bands weighed .9556 g (___.0124 g S D), significant band losses (> 5 %) would have occurred from these Silver Gulls by 9 yr. By 1978, there was a marked variation in the weights of number bands, with some having lost almost twice as much aluminum as others (Fig. 5). These differences seem unlikely to be due to differ- ences in the quality of the metal, but may be due to differences in behavior of the individual gulls. For instance, on gulls habitually fre- quenting sandy beaches and shores, bands may wear faster than on gulls which frequent villages or dumps. To test this, 3 band groups were selected from the plotted regression of band weights on time: one group contained the 30 bands showing the fastest rates of band wear, one the 30 slowest rates of band wear, and one the 30 bands showing rates of wear most similar to the calculated overall mean rate. The age distri- butions of gulls in each group were similar. The location records for the gulls carrying these bands were compared. There is significant het- erogeneity among the 3 groups in the locations where the gulls were resighted (x22 = 37.7; P < .001). Inspection of the data (Table 4), how- ever, reveals no clear trends. In part this may be because most resighting records were made during 1968-1972, whereas much, perhaps even most, band wear occurred during 1973-1981. If the behavior of indi- vidual gulls is not consistent over years, the paucity of observations during 1973-1981 would confuse elucidation of any relationship be- tween gull behavior and rate of band wear. Even so, this comparison 1'0 0'9' 0.8 0-6- .o.5 , o.4, o.3- 0'2- 0-1 FIGURE 5. Time from banding to band recovery(years) Regression of weight of number band (n = 209) on time (years). Regression equation is y -- .9556 - .041 lx (years). does indicate considerable potential for field study of band wear and the reasons for differential band wear. With color bands, the problem was not so much one of when bands were lost, but rather when the color of the Scotchlite or Scotchcal was not identifiable. Red Scotchlite showed some fading in 9 months, marked fading by 2 yr, and appeared white by 3 yr. All other color materials showed deterioration and evidence of peeling from the aluminum base by 4 yr. By 1978, no color bands had sufficient color left to enable positive identification on the free-living gulls. In the hand, from the 1978 and 1981 shootings, the proportions of color bands which could be identified positively ranged from ( 10% for bands originally bearing some red (e.g., WR), 16% for unique 3-color bands (e.g., BBkG), 54% for Y, and 73% for WG. Red Scotchlite appeared to wear away faster than the other materials, in addition to fading much faster, but with the exception of that material, the durability of colors on the bands appeared to be a function of the actual sizes of material bonded to the aluminum base. The larger the individual pieces of color material, the longer color could be identified on the bands. TABLE 5. Problems detected in 37,774 resightings of number bands on Silver Gulls (68 unresolved errors not included). Problems a Correct digit detected 0 1 2 3 4 5 6 7 8 9 Digit not read 5 3 16 43 9 29 27 12 14 19 Digit read as: 0 -- -- 1 ..... 4 1 1 1 .... 1 1 39 -- 3 2 2 -- -- 5 .... 3 5 3 -- -- 2 -- 1 14 3 2 2 -- 4 -- 1 2 2 -- -- 1 1 1 -- 5 -- -- -- 9 -- -- 16 -- 2 -- 6 3 -- -- 5 -- 37 -- -- -- 6 7 -- 15 3 8 2 ..... 8 7 -- -- 4 -- 2 4 2 -- 3 9 2 3 4 1 -- 5 8 1 3 -- Recorded as dubious, but later checked as correct 18 26 17 30 28 75 67 24 25 44 Totals 38 48 45 107 40 163 127 81 54 81  Some 329 number bands were misread leading to an incorrect field record. In addition, 68 other resighting records contained possible digit errors in combination with possible color errors. These multiple errors could not be resolved, and the 68 records were rejected. Errors in resighting bands, recording and processing data.--Regular checks during field work and data processing indicated that almost all errors were made at the time of the original field observation, sometimes by the recorder, but usually by the observer. The accuracy and efficiency of most observers improved markedly with an experience of only a few days, but some observers never improved to the level of accuracy where their observations were accepted. Common early errors were reading digits incorrectly (e.g., reading 7 as 1), and reversing the sequence of two adjacent digits. Such errors sometimes went undetected when a bird flew away before the bands could be checked by a second observer, or when the bands were seen at such a distance that only one observer had sufficiently acute vision to read the bands. For the purpose of this section, all errors will be considered together, regardless of the source. The most common error was leg reversal. Some 328 observations (.87% of all resightings) had the incorrect leg sequence for bands. Leg reversal errors occurred mainly when a gull, standing sideways to the observer, was seen near the limit of an observer's ability to read a band, or when a gull was standing on one leg with the second leg hidden. Leg reversal errors also occurred when a gull was in the hand: e.g., L-GBkG, R-53724 was sighted and recorded correctly on 21 occasions over 2 yr, but on the last occasion it was found dead by a local villager who reported TABLE 6. Leg Bands for Visual Identification Number of problems detected per 10,000 readings of each digit. [299 Problems Correct digit detected 0 1 2 3 4 5 6 7 8 9 Digit not read 4.4 2.2 12.4 24.2 7.9 9.1 6.7 7.9 11.7 8.4 Digit read as: 0 -- 0.7 3.3 0.4 1 0.9 -- 0.3 0.2 25.6 2 1.8 -- 2.8 2.5 1.3 3 1.5 -- 0.9 4.4 0.7 1.3 1.7 2.2 4 0.7 1.5 1.1 -- 0.2 0.7 0.8 5 5.1 -- 4.0 1.7 6 2.6 2.8 11.7 -- 2.6 7 11.2 2.3 4.5 1.8 -- 8 6.1 2.3 0.6 1.0 1.3 -- 1.3 9 1.8 2.2 3.1 0.6 1.6 2.0 0.7 2.5 -- Recorded as dubious, but later checked as correct 15.8 19.5 13.2 16.9 24.6 23.6 16.6 16.4 20.8 19.4 Totals 33.4 35.8 34.7 60.3 35.2 51.3 31.4 53.9 45.0 35.6 the bands as L-53724, R-GBkG. The team member noting the recovery added a second error, by recording the bands as L-53724, R-GBG. Table 5 shows the frequency of digit errors, omissions, and uncer- tainties in field records. Of 37,774 records, 329 (.87%) were found to contain digit errors. A further 68 (. 18%) contained errors which could not be explained, so those records were rejected totally and are not included in the total of 37,774 records analyzed. Table 6 shows relative occurrences, corrected for the different frequencies with which digits occurred on the bands actually used, and then standardised per 10,000 digit resightings. The most common errors were sighting or recording 7 as 1, 5 as 6, and 1 as 7 (Tables 5 and 6). Considering the relative frequencies of digits on the resighted bands, the digit 3 caused the most problems, followed by 7 and 5 (Table 6). With color bands, the most common errors were to read green or blue bands as black, when the bands were in shadow, and dirtied white bands as yellow. The most common error made while recording color bands was to write blue (B) instead of black (Bk). Color bands were recorded incorrectly 109 times when the bands were known to be in good condition (.3% error). Once the color bands showed signs of wear or fading, color band errors and color uncertainties increased markedly. From 1973 onwards, 841 records had color band discrepancies which could be attributed to wear or fading (53.7% error). Even so, most of the worn or faded color bands retained enough identifying color traces TABLE 7. Comparison of data derived from Silver Gulls banded in 1967-1970 and recovered in 1978. Recovery method of band Non-recaptured Shooting resighting Number of birds Mean number of resighting records Range of resighting records Percentage of gulls with only one record after banding (including recovery record) Mean number of days between penulti- mate record and last record 171 125 17.5 19.0 0-104 0-134 10.5 11.2 1815 1501 to be useful for additional confirmation of the accuracy of number bands recorded (see Table 2). Comparison of records from a shot sample with records from a non-shot sample.--Shooting is a common, effective method for recovering bands from free-living birds; however, it has the obvious disadvantage of ter- minating the return of data. With Silver Gulls, it was particularly quick and easy to shoot a selected individual within a range of 6-60 m. In order to assess the effectiveness of the visual resighting program, the histories of the 171 gulls shot during 1978 were compared with the histories, prior to the start of the shooting program, of 125 gulls which were seen-only during 1978 (Table 7). The shot sample was taken con- currently with routine field searches, during otherwise standard obser- vation procedures. Excluding all records made on the seen-only gulls after the 1978 shooting program, the shot birds still had slightly fewer resighting records (2 = 17.5) than the seen-only gulls (2 = 19.0), but in all of the comparisons shown (Table 7) there were no significant differ- ences between the data sets (Fig. 6). We conclude therefore that, with respect to the number of individual gulls "recovered" by the two meth- ods over the particular trial period, resighting of these bands by trained observers was as effective as the shooting program. We note, however, that some of the gulls shot in 1978 had bands which were no longer readable on the free-living gulls. These were the fast-wearing bands, and had the intensity of sightings been greater in 1976-1977, it is highly probable that there would have been significant differences. By the end of the field program, the range of resighting frequency, for the 3977 resighted gulls, was 1-204. There were major differences in distances of dispersal and foraging movements between individual gulls, even considering gulls from a particular colony and the same age cohort, which in part accounts for the great range in resighting fre- quency. However, the proportion of gulls recorded in 1978 for the first time since banding (18/171 of shot gulls, 14/125 of seen-only gulls: 30- 25. 20 10. 15- u>,1o ,t-5 B 0 20 0 60 80 100 120 10 Number of records after banding FIGURE 6. A--Frequency of resighting records for 171 Silver Gulls shot in 1978; B-- Frequency of resighting records for 125 Silver Gulls seen but not shot in 1978. The gulls were banded in 1967-1970. 10.8% of total sample) suggests that the behavior of these individuals had changed with time. Thus, of the two methods, resighting of readable band numbers on uncaptured birds has overwhelming advantages for Silver Gulls and similar species: the method leaves the birds unharmed for further re- sightings and, with trained observers, allows for convenient searches in areas of urban development. Shooting is quicker, and will always be effective at greater distances than a band can be read, but it destroys the potential for future information and is unsafe or prohibited in many areas of human activity. DISCUSSION Following Lack's classic field study of Robins (Lack 1939, 1940), re- sighting records have been used very successfully to establish the pres- ence of uncaptured, uniquely color-banded birds in many ecological and behavioral studies, for example, on corvids (Rowley 1971, 1973), a crac- ticid (Garrick 1960, 1963, 1972), larids (Garrick and Murray 1970; Garrick 1972; Mills 1972, 1973, 1979) and a procellariid (Garrick and Dunnet 1954). Attempts have been made to read standard number bands on uncaptured, free-living birds (Garrick et al. 1957; Kadlec and Drury 1968, 1969; Kadlec 1975; Butler et al. 1980), but most number bands in use today require the banded bird to be in hand for the unique serial number to be read. To our knowledge, the number bands de- scribed in this paper are the first developed to enable resighting of the bands to be used as the primary method of bird identification. Of 17,410 Silver Gulls given these aluminum number bands, 22.8% were seen again. These gulls had a mean of 9.5 resightings. Considering only those 1287 gulls banded as adults, 85.6% were seen again, with a mean of >20 resighting records per bird. These figures may be com- pared with returns from other seabird programs which used the con- ventional types of bands. Recovery rates range from about .1-20% (Hitchcock and Carrick 1958; Dunnet et al. 1977), with, typically, a single record after banding for any bird. Even with gulls, which are amongst the most accessible seabirds, returns from standard bands are usually about 5%, rarely approach 20% (Table 8), and, again, typically consist of just a single record after banding. Such limited data can represent an extremely poor return for the expense and time invested in the initial banding programs. Using the new bands on Silver Gulls, we now have many detailed life histories which show the movements of individuals after they left the banding locations. Some individuals were identified at their breeding or natal colony and then up to 450 km away in the non-breeding months, in 3 different years. Over the 14 years of resightings, 1.1% of number band records were found to contain digit errors or digit omissions, even with observers in the field checking each other. This emphasizes that only those actively involved in the details of such a study can make the final decision of validity of a recovery. Some 2.5% of all color-band resightings were also in error, but this includes errors due to wear and fading of colors after about 4 yr. Some .9% of records showed leg reversal errors. Investigation of the sources of errors revealed that almost all were made in the field by the observers, rather than at some later stage; hence, where possible all observations must be checked and rechecked by 2 or more people. This greatly minimizes the level of errors, and also identifies any observers who are prone to make errors. When errors were found in the data files, field records were compared with original banding records, and only 68 of 37,842 sightings (.2%) had to be rejected because the errors could not be resolved with a reasonable degree of certainty. Of concern was the thought that errors could be made which might not be detected. For in- stance, if bands L-62817 R-Y were incorrectly read or recorded as L-62811 R-Y, which was another valid band combination, the latter Vo. 55, o.  Leg Bands for Visual Identification [303 record would be accepted unless L-62811 was known to be elsewhere at the same time, or dead. We can only speculate that, in our study, undetected errors should be few, and at least an order of magnitude less than the number of errors detected, because of the large number of different types of color bands used. Clearly, undetected errors would be reduced to negligible levels by avoiding multiple use of color bands to denote cohorts and by using discontinuous number band sequences. With a banding program of 20,000 birds, it would be best to avoid a sequence such as 50000-69999 and instead use, for example, 20000- 24499, 45000-49999, 60000-64499, and 75000-79999. These number bands could be paired with one of the 216 three-color combinations of black, blue, green, red, yellow, and white, used randomly. The proba- bility of an erroneous sighting then resulting in another valid number band and color band combination is minute. Since most band wear occurs near the foot, we suggest that bands with vertical number se- quences should be placed on the bird's leg to read from bottom to top. A most urgent need is to continue the experiments of Kadlec (1975), and others, to find an alloy that will increase the durability and reada- bility of bands. For long-lived species of seabirds, exceptionally durable alloys are needed to last 30 yr or more with intermittent immersion in the sea, brackish water and mud, and abrasion by sand. Poulding's (1954) data suggested that Larus argentatus actively removed leg bands, result- ing in a 50% loss of bands in the first year. Paynter (1966) thought that band loss from Herring Gulls was probably about 5% per year. Kadlec and Drury (1968) reported that by the sixth year about 20% of standard aluminum bands were being lost per year from Herring Gulls. Later this estimate was amended to 56% loss by the fourth year (Kadlec and Drury 1969 ), although Kadlec's (1975) most recent figure was 3% loss to the end of 7 yr. Mills (1972) used various types of bands on the Red- billed Gull of New Zealand, Larus novaehollandiae scopulinus, and found significantly fewer lock bands were lost than butt-to-butt bands (24% versus 89% loss) over 10 yr, but lock bands wore more quickly because of their imbalanced weight. Spear (1980) found that 23% of number bands on Larus occidentalis were lost by 3.7 yr, and the loss rate thereafter was 16% per year. Clearly, the loss rate from gulls is very variable. Paynter (1966) considered that losses could be due to bands weakening with wear, bands being removed by the birds, and "an inconsistent or fluctuating loss" due to such factors as time immersed in saline water "or because the new bands are not of uniform strength." With the size and weight of our bands, it seems unlikely that even adult Silver Gulls (weighing ca. 400 g) could remove bands unless the metal had inherent defects or the banding technique was faulty. Removal of defective bands by the gulls would explain the low early losses noted. By 9-10 yr, 10% of surviving Silver Gulls were known to have lost number bands and 5% were known to have lost color bands. It is possible to recapture banded birds to replace worn or lost bands, but this sometimes results in bird mortality, which we wished to avoid. Number bands of the type we describe would have increased durability, and readability, on larger birds. Relatively increased in size for use on Herring Gulls, for instance, duration and distance of readability should be doubled on the figures we give for Silver Gulls. The lower incidence of loss of color bands, compared with number bands, was attributed to the Scotchlite and Scotchcal bonded to the outside and inside surfaces of the aluminum, protecting the metal from wear and corrosion. Even so, there were obvious differences in wear rates of similar color bands on different gulls, as for the number bands. We have not been able to explain this adequately, although we have suggested that it may be a consequence of individual gull behavior. Durability of number bands might be improved by cementing a plastic layer to the inside, as we did for the color bands, finding a durable clear coating that could be sprayed or painted on the outside of the bands before they are placed on the gulls, and increasing the band size and thickness when used on larger birds. We consider that the results of this program give an unequivocal indication of the effectiveness and usefulness of number bands which are readable on free-living birds, and also of the value of combining these number bands with clearly visible color bands. Our band return rates were far better than those which have been achieved by most other large-scale gull-banding programs (Table 8), approaching the returns that could have been achieved by actively seeking out and then shooting the banded gulls. A major and obvious advantage of "visual recapture" is that it is a non-destructive method of sampling: once banded, the gulls need not be handled or disturbed again. Another equally important advantage is that part-time observers can be trained and encouraged to read the bands, needing only binoculars and notebooks instead of trap- ping equipment or guns. Many valuable records came from part-time observers who were scattered throughout southern Australia, and who were prepared to search their home ranges on a regular basis looking for the banded gulls. Having color bands and number bands on the birds was especially essential here, because most of the part-time ob- servers were alone when they were reading bands, so the match of color and number with the banding record was the only check of validity. As colors faded or wore away, the advantage of having single bands with 3 colors, rather than 3 individual bands each with a single color, was demonstrated. Where one color of a sequence was lost from a 3ocolor band (e.g., L-BkRG becoming L-Bk/metal/G), there were far fewer original combinations possible (6) than if 1 of 3 single-color bands had been lost (e.g., if L-Bk/R/G became L-Bk/R, there are 18 possibilities for the original combination. The problem of loss of colors may be resolved by bonding a few layers of the color base to the band so that a fresh layer is revealed as each outer layer becomes abraded. Anderson (1980) has indicated that polyvinyl chloride (Darvic) color bands could be more durable, although at present these are not available commer- cially. SUMMARY Carrick and Murray (1970) described a new type of aluminum band that was designed to enable identification of large numbers of uncaptured, free-living birds. During 1967-1970, these number bands, used together with color bands on the other legs, were placed on 17,410 Silver Gulls in South Australia. The number bands had a maximum readable life of about 10 yr on these free-living gulls, although after this period the numbers could be revealed on smooth bands by etching the aluminum with Villella's solution. The color bands used had a maximum readable life of about 3-4 yr. Our results from these bands indicate enormous potential for studies using readable number bands, with returns far exceeding those from conventional bands. The nature and frequency of the errors detected in data files are discussed. Errors were made mostly during field observations, rather than subsequently, and the necessity is emphasized to continuously check and recheck all sighting records while the banded bird is still present at a location. Because all gulls had 2 bands, a number band and a color band, both of which were readable on the uncaptured birds, less than .2% of records were rejected as a consequence of unresolved errors. Undetected errors were probably few, but these could be reduced to negligible levels by the appropriate use of color bands to confirm the accuracy of reading of number bands. To our knowledge, this is the first study that has used resightings of number bands on free-living, unrecaptured birds as the primary method of data acquisition. ACKNOWLEDGMENTS We gratefully acknowledge the assistance of Mawson Institute staff with banding, field observations, data processing and logistical support: D. Creighton, D. Grey, D. Ferrett, S. R. Harris, S. E. Ingham, G. Jacobs, D. MacKenzie, and P. D. Shaughnessy. National Parks Rangers G. Har- rington and I. May also assisted with field work. Our unpaid, part-time field observers included G. Bowker, C. Carrick, P. Fox, W. R. Garrett, I. Savage, V. Yeoman, and over 50 more. We thank them all. The help of D. Purchase, Secretary, Australian Bird-Banding Scheme, is acknowl- edged. Permission to conduct field work was given by the South Aus- tralian National Parks and Wildlife Service. Professor G. M. Dunnet, University of Aberdeen, gave very helpful criticisms of the manuscript. We especially thank Dr. F. Jacka, Director of the Mawson Institute for Antarctic Research, for support during the entire project. Grants and support came also from the Australian Research Grants Committee, the University of Adelaide, the Australian Commonwealth Scientific and Industrial Research Organization (CSIRO), Flinders University of South Australia, and the Utah Foundation. LITERATURE CITED ArmRso, A. 1980. The effects of age and wear on color bands. J. Field Ornithol. 51: 213-219. BUTLER, R. W., N. A.M. VERBEEK, AND R. G. FOOTTITß 1980. Mortality and dispersal of the Glaucous-winged Gulls of southern British Columbia, Canadaß Can. Field Nat. 94:315-320ß CARRICK, R. 1960. The Australian bird-banding scheme, 1953-1960. Australian Mus. Mag. 13:222-227ß ß 1963. Ecological significance of territory in the Australian Magpie, Gymnorhina tibicen. Proc. Int. Ornithol. Congr. 13:740-753. --. 1972. Population ecology of the Australian Black-backed Magpie, Royal Penguin, and Silver Gull. In Population Ecology of Migratory Birds: a Symposium. U.S. Dep. Int., Wildl. Res. Rep. 2:41-99. --, AND G. M. DUNNET. 1954. Breeding of the Fulmar Fulmaris glacialis. Ibis 96: 356-370ß , AND M.D. MURRAYß 1970. Readable band numbers and "Scotchlite" colour bands for the Silver Gull. Australian Bird Bander 8:51-56. --, W. R. WHEELER, AND M.D. MURRAYß 1957. Seasonal dispersal and mortality in the Silver Gull, Larus novaehollandiae Stephens, and Crested Tern Sterna bergii Li- chenstein, in Australia. CSIRO Wildl. Res. 2:116-144. DUNNET, G. M., W. R. P. BOURNE, J. C. COULSON, M.P. HARRIS, C.J. MEAD, H. MILNE, J. L. F. PARSLOW, C. M. PERRINS, R. C. STEELE, AND M. A. BOOTH. 1977. The report of a working group on ecological research on seabirds. Natural Environment Res. Council (U.K.) Publ. Series C, No. 18, 48 pp. HITCHCOCK, W. B., AND R. CARRICK. 1958. Fourth annual report of the Australian Bird- Banding Scheme, July 1957-July 1958. CSIRO Wildl. Res. 3:115-141. HOFSLUND, P. B. 1959. Fall migration of Herring Gulls from Knife Island, Minnesota. Bird-Banding 30:104-114. KADLEC, J. A. 1975. Recovery rates and loss of aluminum, titanium, and incoloy bands on Herring Gulls. Bird-Banding 46:230-235. --, AND W. H. DRURY. 1968ß Structure of the New England Herring Gull populationß Ecology 49:644-676. --, AND --. 1969. Loss of bands from adult Herring Gullsß Bird-Banding 40: 216-221. LACK, D. 1939. The behaviour of the Robinß I. The life history, with special reference to aggressive behaviour, sexual behaviour and territory. II. A partial analysis of aggressive and recognitional behaviourß Proc. Zool. Soc. 1091:169-200ß ß 1940. The behaviour of the Robinß Population changes over four years. Ibis 4: 299-324. MiLLS,J. A. 1972. A difference in band loss from male and female Red-billed Gulls, Larus novaehollandiae scopulinus. Ibis 114:252-255. 1973. The influence of age and pair-bond on the breeding biology of the Red- biiled Gull, Larus novaehollandiae scopulinus. J. Anim. Ecol. 42:147-162. . 1979. Factors affecting the egg size of Red-billed Gulls, Larus novaehollandiae scopulinus. Ibis 121:53-67ß MURRAY, M.D., AND R. CARRICK. 1964. Seasonal movements and habitats of the Silver Gull, Larus novaehollandiae Stephens in south-eastern Australiaß CSIRO Wildl. Res. 9:160-188. OLSSON, V. 1958. Dispersal, migration, longevity and death causes of Strix aluco, Buteo buteo, Ardea cinerea and Larus argentatus. A study based on recoveries of birds ringed in Fenno-Scandia. Acta Vertebratica 1:91-189. PARSONS, J., AND N. DUNCANß 1978. Recoveries and dispersal of Herring Gulls from the Isle of May. J. Anim. Ecol. 47:993-1005. PAYNTER, R. A. 1966. A new attempt to construct life tables for Kent Island Herring Gulls. Bull. Mus. Comp. Zool. Harvard Univ. 133:489-528. POULDING, R. U. 1954. Loss of rings by marked Herring Gulls. Bird Study 1:37-40. ROWLEY, I. 1971. Movements and longevity of Ravens in south-eastern Australiaß CSIRO Wildl. Res. 16:49-72. ß 1973. The comparative ecology of Australian corvids. I. Introduction, study areas, and methods of study. CSIRO Wildl. Res. 18:1-23. SMITH, W.J. 1959. Movements of Michigan Herring Gulls. Bird-Banding 30:69-104. SORENSEN, L. H. 1972. An analysis of Common Gull, Larus canus, recoveries recorded from 1931 to 1972 by the Zoological Museum in Copenhagen. Gerfaut 67:133-160. S'E^R, L. 1980. Band loss from the Western Gull on Southeast Farallon Island. J. Field Ornithol. 51: 319-328. S'rNCrR, R. 1971. Report on bird-ringing for 1969. Br. Birds 64:137-186. THRrLF^LL, W. 1978. Dispersal of Herring Gulls from the Witless Bay Sea Bird Sanctuary, Newfoundland. Bird-Banding 49:116-124. Woon, H. B. 1945. The history of bird-banding. Auk 62:256-265. (JRO) Mawson Institute for Antarctic Research, University of Adelaide, North Terrace, Adelaide, S.A. 5000, Australia (Current address: Department of Con- servation and Environment, 1 Mount St., Perth, W.A. 6000, Australia),' (RC) Mawson Institute for Antarctic Research, University of Adelaide, North Terrace, Adelaide, S.A. 5000, Australia (Current address: "Achandean," Meadows Road, Dornoch, Sutherland, Scotland); (MDM) Division of Tropical Animal Science, CSIRO, McMaster Laboratory, Private Bag No. 1, P.O., Glebe, N.S.W. 2037, Australia. Received 7 Oct. 1984; accepted 28 Mar. 1984.