1. INTRODUCTION. (1) The Trans-New Guinea family is the largest language family on the island of New Guinea and one of the largest in the world, containing upward of400 languages (Pawley 2005). In spite of the fact that it has been the subject of historical-linguistic research for decades (cf. McElhanon and Voorhoeve 1970, Wurm 1975), there has been relatively little in the way of comparative reconstruction, and subgrouping based on the comparative method remains at a rather early stage (but see Pawley 2005 and Ross 2005).
One of the most well-established subgroups of Trans-New Guinea is Madang, which was first proposed by Z'graggen (1971, 1975a). Z'graggen conducted a survey of approximately 100 languages spoken in Madang Province, Papua New Guinea, collecting wordlists of around 300 etyma for each language (Z'graggen 1980a-d). Based on a lexicostatistical analysis of those data, as well as a study of typological similarities, he proposed an internal subgrouping of Madang. However, as his arguments were generally not based on shared innovations, they have been viewed with caution by the linguistic community. Nevertheless, as a first order approximation of genetic affiliation, and as a source of data, his work has been invaluable. (2)
Z'graggen (1975a) argues for a primary division among Madang languages into two stocks, which he terms Madang and Adelbert Range. (3) This basic division has been called into question (Pawley 1998a; Ross 2000, 2005), but that debate does not concern us at present. Z'graggen further divides his Adelbert Range stock into three branches: South Adelbert, North Adelbert, and Brahman. The latter subgroup has subsequently been disbanded (Ross 2000), and one of its languages, Kulsab, reclassified as a South Adelbert language; the rest were assigned to other groups. Z'graggen's original South Adelbert group contained 12 languages, and with the addition of Kulsab that figure has now been brought up to 13.
This revised South Adelbert subgroup, which includes Kulsab, is the maximal node in Z'graggen's tree that contains all eight Sogeram languages under study here. This paper presents a reconstruction of Proto-Sogeram and a subgrouping of these eight Sogeram languages that is based on the comparative method, making a small contribution to the grand project of building a reconstruction of Trans-New Guinea from the bottom up. The five South Adelbert languages that are not included in this study--which Z'graggen calls Osum, Pondoma, Ikundun, Moresada, and Wadagnam--probably form a second subgroup of South Adelbert that is coordinate with Sogeram. Z'graggen notes that nouns in all five "are obligatorily suffixed by a number marker" (1980d:ix; the suffixes can tentatively be reconstructed as *-[??]ka 'SG' and *-n[begin strikethrough]i[end strikethrough][??] 'PL'). Additionally, as Pawley (1998a:17) and Barrett (2005:7) note, in many of Z'graggen's correspondence sets these five languages lenite word-initial *k, usually to [g] or [??]. Based on these two apparent innovations, I will treat these five languages as being external to the Sogeram subgroup, but I will leave a more thorough investigation of their genetic unity to future research. The internal structure of the Sogeram subgroup, as I reconstruct it, is shown in figure 1. (4)
[FIGURE 1 OMITTED]
The discussion below is based on data I collected in the field in January and February of 2006, with the exception of the Atemple and Apal[begin strikethrough]i[end strikethrough] data. Atemple data come from Z'graggen's (1980d) wordlist, and Apal[begin strikethrough]i[end strikethrough] data were provided by Wade (nd.).
I begin the discussion by comparing my proposed subgrouping, based on the comparative method, with Z'graggen's original lexicostatistical classification (1.1), and then with other proposed subgroupings (1.2-1.4). I then present my reconstruction of Proto-Sogeram (2), and present evidence for the phonological innovations upon which I base my own subgrouping (3-5). …