Institut de Lingüística Aplicada

Third Mediterranean Meeting on Morphology (MMM3)


Anke Lüdeling and Tanja Schmid (IMS - University of Stuttgart)
Does Origin Determine Combinatory Properties of Affixes and Stems in German?

It is frequently assumed, implicitly or explicitly, that neoclassical and native word formation in Germanic languages form two separate word formation components - each with its own elements and combinatorial rules. Our argument in this paper is that word formation is not constrained by any such distinction and that, instead, stems and affixes must be marked in the lexicon for the processes they undergo.

Contrary to Selkirk (1982), Scalise(1986), ten Hacken (2000) and many others origin does not determine the combinatory property of stems. The combinability of stems is determined by usage rather than morphology (cf. Bauer 1998, Lüdeling e.a. 2001). An example from German is the (bound) stem -itis "inflammation", which in the medical context combines only with neoclassical stems but which penetrates into everyday language where it combines with native stems (Gründeritis "compulsive founding", Unterbrecheritis "compulsive interrupting"). While -itis compounds are stylistically marked, there is no stylistic effect with many neoclassical stems that are not scientific but everyday words (Autobahn "highway", Kosmetikstift "cosmetic pencil", Audiogerät "audio equipment").

The stronger claim is that the origin of an affix determines its phonological properties and - correlated with that - its combinatory properties with respect to the bases it combines with and affix ordering. Neoclassical suffixes are said to combine with their bases before stress assigment whereas native suffixes combine with their bases after stress assignment (Giegerich 1985). But stress assignment does not always correlate with origin (Wiese 1996, Jessen 1999). It is also sometimes claimed that native affixes combine only with native stems and neoclassical affixes only with neoclassical stems. But also this correlation does not hold for all affixes in German (Wiese 1996:122). A simple division into native and neoclassical affixes hence does not make the correct predictions with respect to stress assignment and combinatorial properties. Booij (1995, 2001) suggests a tripartite division of affixes for Dutch: s.c. cohering affixes that form a phonological word with their bases and non-cohering affixes that form a phonological word of their own. Cohering affixes are subdivided into native and non-native; non-cohering affixes are all said to be native. Booij (2001) claims that non-cohering affixes combine with native and non-native bases while cohering native affixes only combine with simplex native bases and simplex or complex non-native bases. However, Booij's analysis does not lead to the correct predictions for German.

The following table lists a number of German affixes, their phonological class, and the bases they (productively) combine with. The suffix -bar "-able", which combines with transitive verbs to form adjectives, behaves according to Booij's classification: it is native, non-cohering, and non-selective with respect to its bases. The next two lines show non-cohering native affixes that are selective with respect to their bases. Both combine with native bases only; -sam even requires simplex native bases. -phob does not fit Booij's classification either because it is non-native non-cohering. It combines with neoclassical bases only. -ung and -isch combine not only with simplex but also with complex native bases. -isch (and other suffixes such as -ei; (Wiese 1996:119, Jessen 1999)) are perceived as native (never mind their etymology) but still influence stress and thus cannot be neatly classified as native or neoclassical. The cells in the table which violate Booij's predictions are marked by italics. Finally, Booij's tripartite structure also makes predictions for affix ordering: neoclassical affixes should precede cohering native affixes and these should precede non-cohering affixes. The last column shows that this is not borne out for German either. We list all productive affix combinations containing the affix in question.

The cells in italics show that origin and phonological properties are not sufficient in any combination to achieve the required constraints.

Affix cohering / non-cohering origin combines with stress affix combinations
-bar non-cohering native native, s
non-native, s+c
neutral -barkeit, ierbar,-isierbar
-keit non-cohering native native, s+c neutral -barkeit, -samkeit
-sam non-cohering native native, s neutral -samkeit
-schaft non-cohering native native, s+c neutral -schafter, -schaftler, -schafteln
-phob non-cohering non-native non-native, s+c attracting -phobie
-ung cohering native native, s+c
non-native, s+c
neutral -ierung, -isierung
-isch cohering native native, s+c
non-native, s
influencing -istisch, -atisch, -orisch
-ier- cohering non-native native, s+c
non-native, s+c
attracting -ionier-, -ierung, -ierbar, -ivier-

To sum up: origin and phonological propertiesof an affix or stem are not sufficient to determine its combinatory and ordering properties. We claims that ordering and combinatory properties cannot be exclusively rulebased but rather are to a large degree lexical properties that must be learned for each element. This is in line with recent psycholinguistic approaches (Pinker 1999, Baayen e.a. 1997) that show that many more complex words and patterns than had been previously thought must be stored in the lexicon.