In recent years a fresh approach to the of grammatical rela tions (both at the clause lev l and at infraclausal levels, i.e. in phrases) has opened perspectives on typology which are relatively new. Recent work by Nichols has been particularly important; it has taken off from Russian work in linguistics, i.e. dependency rather than the constituency better known in the rest of the field. The de pendency grammar approach concentrates on how in linguistic struc tures the dependent part relates to the (Nichols 1978a; 1978b; 1979; and especially 1986 and 1992). Nichols concentrates on morphological as opposed to, say, by sequential order. But should be more precisely defined first (the way Nich ols understands that term). We may say, for example, that in many lan guages (like English, or Latin) the verb is marked for tense, or mood, or that in other languages (such as Russian) the verb is marked for as pect. This is what Nichols calls marking (it is internal in the sense that it does not signal any grammatical relation between and dependent), and Nichols is not concerned with those. Rather, what she calls are morphological markers such that they signal a gram matical relation. On this understanding of the point Nichols makes is that typologically we should distinguish between head marking (HM) and dependent-marking (DM) grammar. There are also mixed forms of HM and DM, viz. marking (some constructions in a language are HM; other constructions in the same language, DM) and marking (both and dependent are marked; let us symbolize both split and double as SDM). There are also con structions without any other than word order, as well as cases of neutral marking, i.e. which is neither HM nor DM, but I will not be discussing those here; for them, see Nichols 1986, p. 65.) Nominal case is a good example of DM (or SDM), for case invariably marks a dependent (e.g., accusative for object, or genetive for possessor). Typical forms of HM are found in verb agreement (for subject, and/or for other arguments). Indo-European languages, and also some Semitic lan guages like Arabic, are typically SDM, having both a case system and ver bal agreement.