— Wilson Bull., 108(1), 1996, pp. 72-79 INTERSPECIFIC VARIATION IN THE CALLS OF SPHENISCUS PENGUINS Nina N. Thumser,' - Jeffrey D. Karron,' and MlLUCENT S. FiCKEN' Abstract. We compared the vocal repertoires of Jackass (Spheniscus demersus), Hum- boldt (5. humboldti), and Magellanic (5. rnagellanicus) penguins. Discriminant and cluster analyses of the bray call indicate that Jackass and Magellanic penguins are more similar to each other than either is to the Humboldt penguin, and all three are distinct from the Rock- hopper penguin (Eudyptes chrysocome). The congruence of the vocal analyses with phylog- enies based on allozyme data suggests that differences in vocalizations reflect gradual di- vergence over time, not character displacement. Received 1 Jan. 1995, accepted 20 Sept. 1995. Vocalizations frequently are used to assess taxonomic relationships in birds and the use of song in avian systematics has been thoroughly re- viewed by Payne (1986). Vocalizations also have been used to determine phytogenies of non-passerine species such as the caledrine sandpipers (Miller et al. 1988). Jouventin (1982) found calls to be the best behavioral criterion for classifying penguins. The degree of variation in calls was used to infer subspecies and species status in island populations of nine penguin taxa. Vocalizations have been shown to be of primary importance in the communication of many penguin species (Pettingill 1960, Stone- house 1960, Boersma 1974, Spurr 1975, Jouventin 1982). Although be- havior has been studied in all of the Spheniscus penguins, only prelimi- nary information exists concerning their vocalizations (Boersma 1974, 1976, Eggleton and Siegfried 1979, Jouventin 1982, Scolaro 1987). Species status and phylogenetic relationships in the genus Spheniscus are not clearly defined. There is insufficient detail in the fossil record to distinguish among species (Simpson 1976). Morphological studies led Clancey (1966) to classify Jackass penguins (5. demersus) as a subspecies of Magellanic penguins {S. rnagellanicus). O’Hare (1989) used 22 mor- phological characters to clearly differentiate Spheniscus from other pen- guin genera. However, he was unable to determine the taxonomic rela- DNA-DNA tionships among species within the genus. Utilizing data from hybridization, Sibley and Monroe (1990) proposed that S. demersus be viewed as a superspecies containing demersus, rnagellanicus, and the Humboldt penguin (S. humboldti). Recent allozyme analyses suggest that ' Dept, of Biological Sciences, P.O. Box 413, Univ. of Wisconsin-Milwaukee, Milwaukee, Wisconsin 53201. ^ Present Address: Dept, of Biological Sciences. 150 W. University Blvd. Florida Tech. Melbourne, Flor- ida 32901-6988. 72 Thumser et al. • SPHENISCUS VOCAL COMPARISON 73 Humboldt penguins form a distinct species, but that Jackass and Magel- lanic penguins are closely related (Grant et al. 1994, Thumser and Karron 1994). We quantitatively analyzed the vocalizations of three Spheniscus spe- cies (Jackass, Humboldt, and Magellanic penguins) and one outgroup (Rockhopper penguins, Eudyptes chrysocome). Their vocal repertoires were compared to determine if species consistently differ in the acoustical structure of their calls. A resulting phylogeny was compared to an inde- pendent phylogeny based on protein polymorphisms (Thumser and Kar- ron 1994). METHODS Vocalizations of 21 Humboldt penguins were recorded at the Milwaukee County Zoo in Wisconsin (February 1986—May 1987, February—March 1988), the Brookfield Zoo in Chi- cago, Illinois (November 1987-April 1988), and the St. Louis Zoo in Missouri (October 1988). Recordings of 12 Jackass penguins were made at the Henry Villas Zoo in Madison, Wisconsin (January-April 1988), the Knoxville Zoo in Tennessee (May 1989), and the Racine Zoo in Wisconsin (February-March 1990). Seven Magellanic penguins were record- ed at the Cincinnati Zoo in Ohio (April 1988) and by Jim Klinesteker at the John Ball Zoological Gardens in Grand Rapids, Michigan (Spring 1989). Eleven Rockhopper penguins were recorded at the Cincinnati Zoo (April 1988) and the St. Louis Zoo (October 1988). This study was performed exclusively on captive penguins. Although the majority of recorded Jackass penguins were born in captivity, most of the Magellanic, Rockhopper, and Humboldt penguins were born in the wild. The birds comprising these captive populations may have been drawn from a limited number of wild populations. However, the results from this study should be representative since seabirds usually have limited variation in their vocalizations, particularly at or below the species level (Pierotti 1987). Observations were made during breeding periods, mainly prior to andjust afteregg laying, since most of the calls occurred at those times. A microphone was placed inside the exhibit, but observations were made from outside tbe exhibit to minimize disturbance of the birds. Individuals were identified by tag color. General behavior was observed throughout the day (from 08:00 to 17:00 h CST). Notes were recorded on videotape (Hitachi HJ 5000) and by hand. Vocalizations were recorded throughout the period on a cassette recorder (Aiwa HSJ 500) using a Nakamichi (CM 100) microphone. Whenever possible the identity of the caller was noted. Peak periods of vocalization were simultaneously videotaped and tape-recorded. The recorded calls were analyzed at 150 Hz bandwidth using a Kay 7800 Digital Sona- Graph and digitized using a Sigma Scan (1988) Program. For the bray call, the number of syllables per call, total duration of the call, sum of the inter-syllable intervals, duration of the longest syllable, and minimum, main, and maximum frequency of the longest syllable were recorded. The main frequency represented the darkest band in the sonagram of the call. These seven variables were selected to assess the acoustic structure of the bray call based on both frequency and temporal components. In addition, these parameters were selected because they could be measured precisely. The vocalizations were analyzed using discriminant and cluster analysis in SYSTAT (Wil- ANOVA kinson 1990). The bray call was selected for analysis since nested of individuals within populations within species indicated significant differences at the species level for more than one parameter (Thumser 1993). The data set included all recorded bray calls of the three Spheniscus penguins and the Rockhopper penguins. There were 109 calls from 74 THE WILSON BULLETIN • Vol. 108, No. 1, March 1996 Table 1 Discriminant Eunction Showing the Correlations between Conditional Dependent 1 Variables and Dependent Canonical Eactors in Eour Penguin Species Factor Variable 1 2 3 Number of syllables -0.289 -0.710 -0.024 Total duration 0.337 -0.652 0.401 Inter-syllable interval 0.270 0.018 -0.345 Duration of longest syllable 0.622 -0.229 0.559 Maximum frequency 0.008 -0.635 -0.075 Minimum frequency -0.145 0.464 0.083 Main frequency -0.292 0.257 0.544 Chi-square 496.75 213.17 39.61 df 21 12 5 P <0.001 <0.001 <0.001 Correlation 0.764 0.644 0.339 Humboldt, 77 calls from Jackass, 38 calls from Magellanic, and 106 calls from Rockhopper penguins. Each of the bray call variables was standardized by converting its values to z- scores prior to analysis. In discriminate analysis known groups were used to generate linear models which gave the best fit for that grouping. The data were also analyzed to determine how well the model predicts the actual groupings. Another multivariate technique, cluster analysis, was used to detect natural groupings in data with no prior expectations. In this case, Pearson’s distance measures and the single-linkage method were performed by cal- culating the mean of each of the standardized variables for each species. RESULTS For the first discriminant function. Factor 1 arranged the four species primarily on the basis of the duration of longest syllable and the total duration of the call, while Factor 2 was primarily based on the number of syllables, the total duration of the call, and the maximum frequency of the longest syllable (Table 1). Overall, the analysis correctly catego- rized 86% of Humboldt, 82% of Jackass, 52% of Magellanic, and 79% of Rockhopper penguin calls. These vocal parameters clearly separated the Spheniscus penguins from the outgroup, Rockhopper penguins (Fig. A). Therefore, the outgroup was removed from the analysis and a second 1 discriminant analysis was run to increase the spread among the Spheniscus penguins. In this discriminant function. Factor A arranged the three spe- cies primarily on the basis of syllable number and maximum frequency of the longest syllable, and Factor B was based primarily on the duration and main frequency of the longest syllable (Table 2). The analysis cor- rectly predicted 91% of Humboldt, 71% of Jackass, and 61% of Magel- lanic penguin calls. Within the Spheniscus penguins, there was consid- Thumser et al. • SPHENISCUS VOCAL COMPARISON 75 FACTOR(I) Fig. 1. (A) A scatterpiot of the similarity between the bray calls of three Speniscus species (Jacka.ss, Magellanic, and Humboldt) and an outgroup (Rockhopper penguins). (B) A scatterpiot of the similarity between the bray calls of the three Spheniscus species. Both scatterplots have ellipses around 50% of the data points for Humboldt (dark star), Jacka.ss (open .square), Magellanic (dark circle), and Rockhopper (cross) penguins. 76 THE WILSON BULLETIN • Vol. 108, No. 1, March 1996 Table 2 Discriminant Function 2 Showing the Correlations between Conditional Dependent Variables and Dependent Canonical Factors in Three Spheniscus Species Factor A B Variable Number of syllables 0.711 -0.020 Total duration 0.311 0.319 -0.136 -0.274 Inter-syllable interval Duration of longest syllable -0.225 0.735 Maximum frequency 0.441 -0.065 Minimum frequency -0.257 0.074 Main frequency -0.041 0.423 Chi-square 287.22 60.23 df 14 6 P <0.001 <0.001 Correlation 0.804 0.491 erable overlap in the vocalizations of Magellanic and Jackass penguins (Fig. IB). In fact, 37% of the Magellanic penguin calls were incorrectly classified as Jackass penguin calls, and 12% of Jackass penguin calls were incorrectly classified as Magellanic penguin calls. By contrast, there was less similarity in the bray call parameters of Humboldt and Jackass pen- guins (Fig. IB). Only 6% of the Humboldt penguin calls were incorrectly classified as Jackass penguin calls and 17% of the Jackass penguin calls were incorrectly classified as Humboldt penguin calls. Humboldt and Magellanic penguins were the least similar in their calls (Fig. IB). In both species 3% of their calls were incorrectly classified as the other species. These results were supported by the cluster analysis shown in Table 3 and Figure 2. The distance matrix and tree show the Magellanic and Jackass penguins clustering closely together, the Humboldt penguins more distant, and the Rockhopper penguins the most distant. Table 3 Pearson’s Distance Matrix Showing the Distance between Four Penguin Species Based on Seven Variables of the Bray Call Species 1 2 3 Humboldt Penguin 1 2 Jackass Penguin 1.304 3 Magellanic Penguin 1.436 0.260 4 Rockhopper Penguin 1.462 1.664 1.509 Thiimser et al. • SPHENISCUS VOCAL COMPARISON 77 A) Magellanic Jackass Humboldt Rockhopper I 1 1 1 0.00 0.50 1.00 1.50 DISTANCE B) Magellanic Jackass Humboldt Rockhopper King -+- 0.00 0.20 0.40 0.60 DISTANCE Fig. 2. (A) A tree using the single-linkage method based on Pearson’s distances of parameters ofthe bray call. (B) UPGMA tree based on modified Rogers distance ofallozyme data (Thumser and Karron 1994). DISCUSSION Overall, the Spheniscus penguins have retained a complement of calls that are similar in structure and function (Thumser 1993). The bray call is used to establish a territory and to advertise availability for pairing. The bird stands with its head pointing up and calls while slowly flapping its wings. This was the only call which showed sufficient species-level variation for phenetic analysis. The analyses of selected vocal parameters of the bray call clearly dis- tinguish Humboldt from both Jackass and Magellanic penguins. However, discriminant and cluster analyses often could not distinguish between the Magellanic and Jackass penguin calls. This may reflect the evolutionary relationships among the species or may have resulted from other factors. Since Humboldt and Magellanic penguins occur sympatrically in South America, another possible explanation for the differences in their breeding calls is character displacement. By contrast, the similarity of South Amer- ican Magellanic and African Jackass penguin calls is unlikely to result from convergence. In order to determine whether character displacement has occurred it is necessary to know which vocal characters are ancestral. It is difficult to root a tree based on vocalizations and determine the most ancestral species because vocalizations can be subject to strong selection. However, a comparison of allozyme variation enhances these results because protein markers are subject to weaker and different selective forces than those 78 THE WILSON BULLETIN • Vol. 108, No. 1, March 1996 influencing behavioral traits. Trees based on allelic characters are also more easily rooted using outgroup taxa. The allozyme data were consistent in two different studies (Grant et al. 1994, Thumser and Karron 1994). Grant et al. (1994) based their analysis on 15 polymorphic loci in 75 captive (Humboldt, Magellanic) and wild (Jackass, Rockhopper, and Macaroni [Eudyptes chrysolophus]) penguins. Thumser and Karron (1994) studied nine polymorphic loci in 165 captive (Jackass, Humboldt, Rockhopper, King [Aptenodytes pata- gonicus]) and wild (Humboldt, Magellanic) penguins. In both studies. Jackass and Magellanic penguins were very closely related and Humboldt penguins clearly formed a distinct species. There is a striking similarity of phenetic trees based on the allozyme data and the vocal analysis pre- sented here (Fig. 2). Cladistic analysis of the allozyme data confirmed that Spheniscus penguins form a monophyletic group (Grant et al. 1994, Thumser and Karron 1994). These findings suggest that differences in vocalizations between the Humboldt and Magellanic penguins are not due to character displacement, but rather reflect gradual genetic divergence of separate evolutionary lineages. Although the Humboldt and Magellanic penguins occur sympatrically, they have lower genetic identities and greater vocal differences than the more closely related Jackass and Mag- ellanic penguins. ACKNOWLEDGMENTS This research was partially funded by grants from the Ruth Walker Scholarship Lund, the Univ. ofWisconsin—Milwaukee Graduate School, and the Institute of Museum Services (IC- 10197-91). We thank the following zoos for allowing access and providing assistance in recording penguin vocalizations; Brookfield, Cincinnati, Henry Villas Park, Knoxville, Mil- waukee County, Racine, and St. Louis. We are also grateful to Jim Klinesteker at the John Ball Zoo for providing us with recordings of a Magellanic penguin population. P. D. Boers- ma, G. Miller, and an anonymous reviewer provided several helpful suggestions on an earlier version of the manuscript. LITERATURE CITED Boersma, P. E. 1974. The Galapagos penguin; a study of adaptations for life in an unpre- dictable environment. Ph.D. diss. Ohio State Univ., Columbus, Ohio. 1976. An ecological and behavioral study of the Galapagos penguin. Living Bird . 15;43-93. Clancey, P. a. 1966. The penguins Spheniscus clemersus Linnaeus and Spheniscus niagel- lanicus Forster. Ostrich 37;237. Eggleton, P. and W. R. Siegfried. 1979. Displays of the Jackass penguin. Ostrich 50; 139- 167. Grant, W. S., D. C. Duffy, and R. W. Leslie. 1994. Allozyme phylogeny of Spheniscus penguins. Auk 1 I 1;716-720. JouvENTiN, P. 1982. Visual and vocal signals in penguins, their evolution and adaptive characters. Adv. Ethology 24. Thumser et al. • SPHENISCUS VOCAL COMPARISON 79 Miller, E. H., W. W. H. Gunn, and B. N. Veprintser. 1988. Breeding vocalizations of Baird’s Sandpiper Calidris bairdii and related species, with remarks on phylogeny and adaptation. Ornis Scand. 19:257-267. O’Hare, R. J. 1989. Systematics and the study of natural history, with an estimate of the phylogeny of living penguins (Aves: Spheniscidae). Ph.D. diss. Harvard Univ., Boston, Massachusetts. Payne, R. B. 1986. Bird songs and avian systematics. Pp. 87—116 in Current ornithology. Vol. 3 (R. E Johnston, ed.). Plenum Press, New York, New York. Pettingill, O. S. Jr. 1960. Creche behavior and individual recognition in a colony of Rockhopper penguins. Wilson Bull. 72:209—221. PiEROTTi, R. 1987. Isolating mechanisms in seabirds. Evolution 41:559-570. SCOLARO, J. A. 1987. A model life table for Magellanic penguins {Spheniscus magellanicus) at Punta Tombo, Argentina. J. Eield Ornithol. 58:432-441. Sibley, C. G. and B. L. Monroe, Jr. 1990. Distribution and taxonomy of birds of the world. Yale Univ. Press, New Haven, Connecticut. Sigma Scan. 1988. Scientific Measurement Program, Version 3.90. Jandel Scientific, Corte Madera, California. Simpson, G. G. 1976. Penguins past and present, here and there. Yale Univ. Press, New Haven, Connecticut. Spurr, E. B. 1975. Communication in Adelie Penguins. Pp. 449-501 in The biology of penguins (B. Stonehouse, ed.). Macmillan, New York, New York. Stonehouse, B. 1960. The King Penguin (Aptenodytes patagonicus) of South Georgia. Talk. Isl. Dep. Surv. Sci. Rep. 23. Thumser, N. N. 1993. Phylogenetic relationships among Spheniscus penguins based on the analysis of vocal and allozyme data. Ph.D. diss. Univ. of Wisconsin-Milwaukee, Mil- waukee, Wisconsin. AND J. D. Karron. 1994. Patterns of genetic polymorphism in five species of penguins. Auk 111:1018-1022. Wilkinson, L. 1990. SYSTAT: the system for statistics. SYSTAT, Inc., Evanston, Illinois.