Abstract An interspecific partial genetic linkage map of Coffea sp. based on 62 backcross hybrids is presented. F1 hybrids were generated by a cross between the wild C. pseudozanguebariae and the anciently cultivated C. lib- erica var. dewevrei (DEW); progeny were then derived from a backcross between F1 hybrid and DEW. The map construction consisted of a two-step strategy using 5.5 and 3.1 LOD scores revealed by simulation file. The map consisted of 181 loci: 167 amplified fragment length polymorphism (AFLP) and 13 random fragment length polymorphism (RFLP) loci. The markers were as- sembled into 14 linkage groups, each with 4–31 markers covering 1,144 cM. Segregation distortion was observed for 30% of all loci, in particular 3:1 and 1:3 ratios equal- ly favouring each of the two parents. The existence of such ratios suggests genetic conversion events. This map also represents an initial step towards the detection of quantitative trait loci.morphism markers (AFLP, Vos et al. 1995).The AFLP technology has two main advantages: (1) more reliability and consistency than random amplified polymorphic DNAs (RAPDs); (2) a large number of markers on a sin- gle high-resolution sequencing gel. Consequently, a high-density map can be obtained from a small number of primers and minute amounts of DNA. AFLP markers are typically dominant, and most fragments correspond to unique positions on the chromosome and can be ex- ploited in genetic and physical mapping (Thomas et al. 1995; Meksem et al. 1995). RFLP (restriction fragment length polymorphism) markers do not have such advan- tages; the number of markers obtained from a gel is low- er and higher amounts of DNA are required. In contrast, they are co-dominant, which allows between-map com- parison. Coffee trees belong to the genus Coffea sub-genus Coffea, family Rubiaceae and are mostly present in trop- ical and subtropical regions of the world. Most species (up to 80 taxa) are diploid (2n=22), except C. arabica, which is tetraploid (Leroy 1980; Bridson 1987). Two species are currently cultivated world-wide: C. arabica and C. canephora. As wild species constitute the base of genetic resources for breeding, interspecific crosses have been realised to test breeding possibilities by introgres- sion (Louarn 1992). Those that include East African spe- cies and a West African species constitute an extreme situation, which is illustrated by the cross between C. pseudozanguebariae, native to Kenya and Tanzania, and C. liberica var. dewevrei, native to Central Africa. In the investigation reported here we used the C. pseudo- zanguebariae×C. liberica dewevrei cross for genetic mapping for two main reasons: (1) the availability of F1 hybrids as well as backcrossed hybrids and (2) the low genetic distance between C. liberica dewevrei and C. canephora (Carvalho and Monaco 1967; Charrier 1978; Louarn 1992), which facilitates relating the results to cultivated species. Distorted marker segregation has been reported in several interspecific linkage analyses (Grant 1975). This systematic deviation from an equal representation of al-