PHYLOGENY
A first attempt to a phylogenetic reconstruction using cladistic methods was carried out by myself (Dempewolf, 2001). Morphological characters from immature stages and imagines of a selection of 54 (including outgroups) terminal taxa out of 23 genera were investigated. A part of these genera contain species being not of economic importance, which are therefore not treated on this CD-ROM. The main results will be briefly summarized here (AGRO tree.pct).

Generally the accordance of the reconstruction with the traditional classification is high. Both subfamilies and many genera could be confirmed as monophyletic. Their key apomorphies are mentioned in the descriptions of the higher taxa. Additionally, hypotheses on the relationship of higher taxa were supported. Although not all branches should be described as formal taxa, the Ophiomyia-group, embracing the genera Melanagromyza, Ophiomyia and Tropicomyia, are considered as tribe Ophiomyiini on this CD-ROM. There are so many common characteristics that it is practical to describe them in one taxon at once. The sister group of Ophiomyiini is clearly the genus Japanagromyza. Both taxa have in common the asymmetrical mandibles with the left mandible being larger than the right one (O galii Larva SEM1.pct). Both taxa together are closely related to Agromyza, which is the most ancient agromyzid genus.
The two rather isolated Nemorimyza and the cambium mining Phytobia species are the basal groups of the subfamily Phytomyzinae. Apart from Agromyza, Nemorimyza is the only taxon with species showing symmetrical mouth hooks that must be interpreted as plesiomorphic (Nemorimyza Larva SEM4.pct). Phytobia and all subsequent more derived groups have asymmetrical mandibles (the biological significance of asymmetry of mandibles can be found in the section on Immature Stages. The remaining taxa next to Phytobia are subdivided into two large groups, the Selachops- and the Phytomyza-group. Among others the Selachops-group (named after a genus, which is not on this CD-ROM) comprises among others the genera Amauromyza, Cerodontha, Calycomyza and Liriomyza. As apomorphic characters can be interpreted the ventral position of the surstyli in male genitalia (Lir chenopodii epandrium2.pct) and the shape of the larval posterior spiracles (L pusilla.pct). However, in many species these character states are further transformed into other states. The Phytomyza-group basically contains the genera Chromatomyia, Phytomyza, Napomyza and Ptochomyza among others, which are not mentioned on the CD-ROM. Phylogenetically, this is the most problematic group of Agromyzidae. Both the monophyly of the whole group and of some of the genera, especially Phytomyza, is uncertain. Cladistic investigations yielded in some parts inconsistent results, the picture (AGRO tree.pct) shows only the apparently most reasonable cladogram (obtained by Successive Weighting). Yet a probable apomorphic character might be found in the mandibles of first instar larvae (Dempewolf, 2001).

ANCESTRAL FEEDING HABIT
As reviewed in the section on Bionomics all agromyzid larvae are internal feeders of various living plant tissues, such as leaves (leaf miner.pct), stems (stem miner.pct, Mel sojae larva.pct, Oph phaseoli larva.pct) and - most spectacular - the cambium of trees (Genus Phytobia : ash tree.pct). All these different environments require morphological adaptations of the larvae. Even within similar feeding habits interesting morphological and behavioral differences were seen. Dempewolf, 2001 discussed this in detail but also in the section on Immature Stages some examples were given.
From these observations, the question of the hypothetical evolution of plant colonization within Agromyzidae arises. This subject has been already addressed by Nowakowski, 1962, Spencer, 1973 and Spencer, 1990. Both authors held the view, that the extant Agromyzidae must have been derived from a cambium-feeding ancestor. Nowakowski tentatively concluded the cambium-mining genus Phytobia to be the most ancestral. This remained the prevailing opinion in the literature and was not disputed. Later this idea was pursued and extended by Spencer (Spencer, 1990). He suggested a distinct evolutionary sequence: from the cambium the agromyzids invaded the stem pith, subsequently colonized external stem parts, and finally adopted the leaf as food. Following this scheme, in many genera the stem borers were seen as their most primitive members.
Apart from the "generally accepted" plesiomorphic condition found in Phytobia, Spencer (in Spencer, 1990) argued mainly on the basis of the discovery of Phytobia-like feeding traces at fossil trees from the tertiary (see SÜss and MÜller-Stoll, 1975, SÜss H., 1979, SÜss, 1980). Yet this argument is weak because of the general limited fossil record of Agromyzidae. Earlier attempts to connect the agromyzid evolution with that of their host plants (Spencer, 1973) were later withdrawn since no similar phylogenetic pattern between plants and the miners were detected (Spencer, 1990). Spencer concluded "there is really little logic in the hosts which have been colonized by the majority of genera." (p. 388). Indeed, some sort of coevolution of agromyzids and plants cannot be discovered by a comparison of phylogenetic relationships within both groups.
Phylogenetic reconstructions offer an approach to the determination of ancestral characters and the history of character transformation in monophyletic groups. To infer the ancestral feeding habit from the phylogenetic tree the feeding habit is coded as a character and mapped on the tree. The results are shown in AGRO tree.pct.
Evidence from this method suggests leaf-mining to be the ancestral feeding habit in Agromyzidae. Alternative feeding habits emerge independently many times within several genera. To some extent the opinion of Phytobia being a primitive genus could be confirmed, the cambium mining feeding habit emerges rather early, yet neither directly at the base of the Agromyzidae nor at the base of the subfamily Phytomyzinae.

Although the evidence was derived from all available characters under investigation (see Dempewolf, 2001), some of the larval characters are particularly good in illuminating the evolution from leaf miners to other feeding habits. Of these are mainly the shape of the larval posterior spiracles found to be primitive in many Agromyza-leaf-mining species (A albipennis Larva3 SEM.pct) and the plesiomorphic symmetry of the mouth hooks found in leaf-mining Agromyza (A albipennis Larva SEM.pct) and Nemorimyza (Nemorimyza Larva SEM4.pct). Those species, whose larvae initially feed on the leaves and later invade into the petiole and the stem can be seen as a model for the colonization of the host plant with the leaf as a starting point. There are several examples known in the literature, one of them is the bean fly Ophiomyia phaseoli. More on the morphology of larval mouthparts is presented in Immature Stages.

Some further evidence for an ancestral position of leaf miners in Agromyzidae can be obtained from the general character of the food source. Mesophyll cells are obviously less tough and easier accessible than tissue in stems and petioles, especially for potential ancestors, who dwell in moist environments. The development within stems should require a strong ovipositor that cannot assumed to be already developed in possible ancestors. For boring into a leaf, a less sophisticated ovipositor may be good enough. Larvae can also enter a leaf when the eggs were stuck on the leaf surface, although it may be a selective advantage for an individual to have an ovipositor.