FISH and FISHERIES, 2002, 3, 213^224 Behaviour: an important diagnostic tool for Lake Malawi cichlids Jay R Stau¡er Jr1, Kenneth R McKaye2 & Ad F Konings3 1School of Forest Resources, Pennsylvania State University,University Park, Pennsylvania, PA16802, USA; 2Appalachian Laboratory,UMCES,301Braddock Road, Frostburg, Maryland, MD 21532, USA; 3Cichlid Press, PO Box13608, El Paso,Texas, T X 79913, USA Abstract Correspondence: Historically,the cichlid ¢shes of Lake Malawi, which probably represent one of the best Jay R. Stau¡er Jr, examples of rapid radiation of vertebrates, have been diagnosed with morphological School of Forest Resources, and genetic data. Many of the populations once thought to be conspeci¢c have been Pennsylvania State hypothesized to be separate species based on behavioural data. The use of behavioural University, data, as expressed in mate choice based on colour patterns or bower shapes, has been University Park, successfully used to diagnose both rock-dwelling and sand-dwelling cichlid species. Pennsylvania, PA Additionally,a combination of bower shapes and courtship patterns have been used as 16802, USA Tel.: þ1 814 863 0645 synapomorphies to diagnose genera within the Lake Malawi cichlid £ock. It is con- Fax: þ18148653725 cluded that taxonomists need to include behavioural data with morphological and E-mail:[email protected] genetic databases to diagnose species and to determine the phylogenetic relationships withinthis diverse assemblage of ¢shes. Received 5 Dec 2001 Accepted10Jun 2002 Keywords allopatry,behaviour, Cichlidae, mate choice, sexual selection, species con- cepts, species recognition, sympatry Introduction 213 Definition of Lake Malawi fish species 215 Role of behaviour in the evolution of Lake Malawi cichlids 215 Application of behavioural parameters to analysing the taxonomy of Lake Malawi cichlids 217 Conclusions 221 References 221 species have been formally described, and many Introduction undescribed species are being discovered and In the lakes of East Africa, ¢shes of the family Cichli- described from newly explored areas (Keenleyside dae have undergone an extraordinarily rapid and 1991; Stau¡er and McKaye 2001). It is estimated that extensive radiation. Within Lake Malawi, over 450 in Lake Malawi, there may be as manyas1500 cichlid # 2002 Blackwell Science Ltd 213 Behaviour and cichlid species J R Stau¡er et al. species (Stau¡er et al.1997;Turneret al.2001). Cichlid in which the covariance matrix is factored (Humph- ¢shes have been the focus of considerable ecological, ries et al.1981;Bookstein et al. 1985). These analyses evolutionary, and behavioural research (Fryer and ordinate the morphometric factors independently of Iles 1972; Keenleyside 1991; Seehausen 1996; Kawa- a main linear ordination (Reyment et al.1984).The nabe et al. 1997; Konings 1998, 2001; Barlow 2001; ¢rst sheared principal component examines size dif- Coleman 2001). Unfortunately, many of these ferences, while subsequent components are indepen- research e¡orts have been slowed and often confused dent of size (Humphries et al.1981;Booksteinet al. as a result of the uncertain systematic status of some 1985); thus, ratios are not needed to standardize the of the cichlids being examined. Certainly, the discri- size of small and large specimens. Simultaneously, mination among species of Lake Malawi cichlids can meristic data are analysed using principal compo- be di⁄cult because di¡erences among species may nent analysis (PCA), in which the correlation matrix be very small (Konings 2001)and because morpholo- is factored. Di¡erences among species or populations gical characters may be prone to convergence can be illustrated by plotting one of the sheared com- (Kocher et al. 1993). The existence of distinct species ponents against the ¢rst principal components of (e.g. Tropheops ‘GomeYellow’;Tropheops‘yellow gular’ the meristic data (Stau¡er and Hert1992). If the mini- (Konings 2001) that possess minimal morphological mum polygon clusters of di¡erent populations are di¡erences (Lewis 1982) and display the results of signi¢cantly di¡erent along one axis, independent of parallel evolution (Kocher et al.1993) probably consti- the other axis, a range test can be used to determine tutes the greatest problem to the taxonomist attempt- which clusters di¡er. If, in fact, the clusters are not ing to distinguish cichlid species (Stau¡er and signi¢cantly di¡erent along one axis independent of McKaye 2001). the other, then a manova, in conjunction with a Historically,most attempts to diagnose cichlid spe- Hotelling^Lawley trace, can be used to determine cies have used morphological characters (Stiassny whether the mean multivariate scores of the clusters 1991). Many species were described based on meristic formed by the minimum polygons are signi¢cantly characters such as number of gill rakers (Iles 1960), di¡erent (Stau¡er et al.1997). dentition of the lower pharyngeal bone (Gu« nther Although such techniques have greatly improved 1893; Trewavas 1931), the presence or absence of our resolution of slight morphometric di¡erences enlarged suborbital canals (Trewavas 1988), and the (Stau¡er et al. 1997), taxonomists are still stymied presence of an enlarged cleithrum (Stau¡er and because in many cases, distinct species within the McKaye 1985). Morphometric relationships such Lake Malawi cichlid species £ock are morphologi- as the size of the mouth expressed as percentage cally similar (e.g. number of lateral line scales, num- standard length (Boulenger 1901) were also used to ber of cheek scales, horizontal eye diameter, head diagnose species (e.g. Protomelas pleurotaenia (Bou- length). To attempt to diagnose these similarly lenger)). To further compound the problem of species shaped populations, colour patterns have also been diagnoses of Lake Malawi cichlids, many of the origi- used as morphological characters.The great diversity nal descriptions were based on one or two specimens of colour patterns within the haplochromine cichlids (e.g. Copadichromis eucinostomus (Regan), Metria- is phenomenal (see Konings 2001) and the existence clima zebra (Boulenger)) with no speci¢c locality of unique colour patterns has been recognized to information within Lake Malawi. Thus, the range in be a reliable character for distinguishing species morphological characters could not be estimated (Barlow 1974; Barel et al. 1977; Greenwood 1981, and additional topotypes could not be collected to 1991; Hoogerhoud and Witte 1981; McKaye et al. e¡ectively compare the variance of selected charac- 1982, 1984; Bowers and Stau¡er 1993; Stau¡er et al. ters among populations. 1995, 1997). Fortunately, within the last two decades, the ana- Genetic approaches are also problematic because lyses of morphometric data have evolved from the the Lake Malawi cichlids are speciating faster than use of ratios and univariate morphometric analysis, alleles are ¢xed within a species (Korn¢eld et al. to attempts to quantify the shapes of organisms 1985; Korn¢eld and Parker 1997). Attempts to use (Atch ley 1971; Hu mph r ie s et al.1981;Reymentet al. mtDNA haplotype frequencies have met with limited 1984; Bookstein et al. 1985). In e¡ect, distances are success (Stau¡er et al. 1995); however, the coales- measured between identi¢ed landmarks of the indi- cence of mtDNA haplotypes found within popula- vidual ¢sh (Stau¡er 1991). These data are analysed tions predates the origin of many species (Parker and using a sheared principal component analysis (SPCA) Korn¢eld 1997). A new method of using ampli¢ed 214 # 2002 Blackwell Science Ltd, F I S H and F I S H E R I E S, 3, 213^224 Behaviour and cichlid species J R Stau¡er et al. fragment length polymorphisms (AFLP) may be passes a wide array of possible species, both extant promising for identifying distinct species (Albertson as well as extinct. Mayden (1997) concluded that the et al.1999). evolutionary species concept was the only concept Trewavas (1983) pioneered the use of behavioural that was apropos to all organisms. data to characterize cichlids when she diagnosed The shortcomings of the evolutionary species con- three genera of tilapiine ¢shes depending on whether cept are that it relies heavilyon a well-resolved phylo- they were substrate spawners, maternal mouth broo- geny,and it is nonoperational (Mayden1997; Stau¡er ders, or paternal/bi-parental mouth brooders. In the and McKaye 2001). The reality is that in many sys- past several decades, authors (e.g. Holzberg 1978; tems, including Lake Malawi cichlids, a comprehen- Schro« der1980; Stau¡er1988; Stau¡er et al.1997)have sive and well-supported phylogeny is not available. used behavioural characteristics (i.e. bower shape, Hence, we are proposing several criteria to aid in the colour patterns) (McKaye et al. 1993; Stau¡er et al. de¢nition of Lake Malawi species. The ¢rst is repro- 1993) to distinguish among sympatric populations ductive isolation. This criterion has been tradition- and species.The lackof morphological di¡erentiation ally associated with the biological species concept, and the inability of allozyme data to distinguish spe- which Mayr and Ashlock (1991) de¢ned as ‘A repro- cies prompted Stau¡er and McKaye (2001) to con- ductively isolated aggregate of populations, which clude that a combination of genetic, morphological, can interbreed with one another because they share and behavioural data should be utilized to describe the same isolating mechanisms.’ Reproductive isola- new species of Lake Malawi cichlids. The purpose of tion is assessed through observations of mating this paper is to discuss how various behavioural stu- activities or analysis of gene £ow.We want to stress dies have proved invaluable in recognising species, that we are using reproductive isolation as an indica- diagnosing species, and estimating phylogenetic tor of the independence of evolutionary lineages and relationships. not as a species concept in and of itself. Second, we employ the morphological/behavioural similarity criteria that are typically associated with the mor- Definition of Lake Malawi fish species phological/phenetic species concepts to aid in spe- Delineation of species of Lake Malawi cichlids is chal- cies delineation.