chapter 25 WORD-FORMATION friedrich ungerer 1. Introduction Word-formation is one of those linguistic terms that may be unsatisfactory on a more theoretical level, but that are immensely useful when one tries to survey processes of extending the lexicon. Loosely defined as ‘‘creating new words from existing words,’’ 1 word-formation ranges from prefixation and suffixation (where it overlaps with inflectional morphology in the use of bound morphemes; see Janda, this volume, chapter 24) to processes not even reflected in the phonological form of the item involved (e.g., conversion); there, word-formation borders on purely semantic processes of metaphor and metonymy (Lipka 2002: 108–9). Between these two extremes may be placed the many ways in which words can be combined, fused, and condensed (as in compounds, lexical blends, back-formations, clippings, and acronyms). Since English is one of the languages that makes use of all these pro- cesses, mostly English examples will be chosen for illustrative purposes, but it should be kept in mind that some of the processes, in particular affixation, are much more widespread and more differentiated in other languages. However, the relative sparseness of affixal processes in English has not kept structuralist linguistics from approaching English word-formation within the frame- work of a morphological analysis focused on the segmentation of free and bound morphemes (with the emphasis on the latter), adding zero forms to include con- version (dirty adj. ? dirty verb), back-derivation (baby-sitt(er) n. þ Ø ? baby-sit verb) and bahuvrihi compounds (laptop þ Ø ? laptop computer) and only grudg- ingly admitting that blends, acronyms, and other ‘‘irregular forms’’ did not re- ally lend themselves to this kind of interpretation. The relationship between the constituent elements of word-formation items (i.e., the items that result from word-formation processes) was first interpreted as a hierarchy of immediate con- stituents based on a modifier-head relationship; later, underlying syntactic struc- tures were used to explain derivatives, especially compounds, and when this method was found insufficient for many noun-noun compounds, semantic argument structure was introduced as an analytical tool (e.g., in Levi 1978). Another long-standing aim of word-formation research has been the establish- ment of a set of rules and constraints to explain productivity of word-formation patterns (Bauer 2001); in addition, attention has been attracted by the processual aspects of institutionalization and lexicalization (Bauer 1983: chapter 3; Quirk et al. 1985: 1522–29; Lipka 2002: 6, 110–14). The ‘‘rules and constraints’’ approach was primarily an offspring of Transformational Grammar; but far from being restricted to formal approaches, it permeated the discussion of basic semantic patterns of composition (Levi 1978) and productive argument constellations for -er-derivations (Rappaport and Levin 1992), to name just a few examples, and reemerged in more recent onomasiological approaches (Stekauer 1998). Although the general trend over the years thus favored semantic analysis, it remained safely tied to grammatical word classes and a syntactically oriented argument structure. No wonder, it was difficult to find a place for lexical blends and acronyms in such a framework or to integrate convincing explanations for concepts such as institu- tionalization or lexicalization. If this is an acceptable account of the general research situation, it should be no surprise that Cognitive Linguistics has the potential to stimulate word- formation research. Indeed, it can provide both the theoretical background and the empirical tools to complete a process that had already been set going: the semanticization of word-formation analysis. Starting from the axiom of the centrality of meaning (Langacker 1987: 12), cognitive linguists will treat all aspects of word-formation as meaningful: the concepts expressed by word-formation items and their constituents (whether they enjoy the status of morphemes or not), the structural patterns underlying derivatives and the restrictions imposed on them, and finally the processual aspects of word-formation (such as lexicaliza- tion). On the empirical level, cognitive analysis has developed ways of describing lexical concepts in terms of schemas, prototypes, and radial categories, including metonymic and metaphorical extensions, of analyzing prepositions in terms of Figure/Ground alignment, and of explaining argument structures as event schemas. Linguistic processes have been described as conceptual fusion, their iconic aspects as a matter of form-meaning isomorphism. All these approaches have been used to integrate word-formation into concepts like Cognitive Gram- mar, Conceptual Blending, and form-meaning iconicity or, on a more specific level, to provide more comprehensive and consistent descriptions of individual word-formation phenomena. word-formation 651 2. General Aspects 2.1. Word-Formation in Cognitive Grammar Since Cognitive Grammar, as developed by Langacker, does not distinguish be- tween lexical and grammatical units in the traditional sense of the term, word- formation is regarded as part of a unified grammatical description, and this shows up with respect to categorization, Figure/Ground alignment, and accommodation. Word-Formation as Semantic Extension Just like additional meanings of simplex lexical items, word-formation items can be understood as encoding extensions, based on category judgments, from a profiled linguistic unit. The only difference between simplex and word-formation items is that in the latter, additional meaningful components, both lexical items and affixes, are added. This is illustrated in figure 25.1, which is based on Langacker’s notation and on some of his examples (Langacker 1987: 374–83, 451) complemented by ad- ditional ones. Figure 25.1 shows that the degree of morphological complexity is not decisive for the relationship of semantic extension from an assumed prototype unit (here tree), but that there may be compounds that are semantically closer to the pro- totype than some simplex extensions, as for instance apple tree compared with tree in the metaphorical sense of ‘family tree’. Word-Formation Items and Figure/Ground Alignment Although the Figure/Ground contrast (see Schmid, this volume, chapter 5) is em- ployed at all levels of categorization, an important application concerns the level of words or linguistic units and their interaction with other linguistic units (Lan- gacker’s term is valence relation). Figure and Ground, here called trajector and landmark, are not only embodied in lexical items such as verbs or prepositions, but also in affixes like the -er suffix. 2 Figure 25.2 shows how trajector and landmark are involved in creating the prepositional phrase above the table and the -er derivation climber. In figure 25.2a, the prepositional relation is characterized by the contrast be- tween the trajector, which is positioned in the upper section, and the yet unspec- ified landmark, which is placed below. Combining the prepositional concept with a lexical concept like ‘table’ provides the specification of the landmark and estab- lishes the composite structure, documented in the top box (see Svorou, this vol- ume, chapter 28). The -er suffix (figure 25.2b) also consists of a trajector and an unspecified landmark, but the meaning of the suffix is vaguer than the preposi- tional meaning and the processing is complicated by the fact that lexical concepts like climb already contain a complex trajector-landmark configuration spread over time (indicated by the changing position of the trajector). 652 friedrich ungerer This influences the integration of the two constituent units, which results in fusing the affix concept with the verbal concept, but does not prevent treating derivation and prepositional phrase in a parallel fashion (see figure 25.2b, top box). Word-Formation, Composite Structures, and Accommodation The process of -er derivation is just one instance of the many parallels between word-formation items and phrases emerging from Langacker’s approach. Another example is compounds that can be compared with adjective-noun combinations. Figure 25.3 makes it clear that the same schema is applicable both to modifier-head phrases and compounds. The parallels are not only a matter of structure in the more conventional sense, but include the semantic adjustment of the components and the addition of conceptual content in the composite item. This is indicated through the element X in the formula for the composite item in the schematic boxes on the left of figures 25.3a and 25.3b. The ‘‘accommodation’’ of the components, as Langacker (1987: 75–76) calls this semantic process, may be limited in the case of tall tree, where only a certain semantic adjustment of the prototypical meaning of tall as ‘height of human beings’ is required. Yet with regard to Christmas tree, it is obvious that a great deal of additional information, mostly of an encyclopedic kind, has to be added to permit a proper conceptualization of the item compared with the Figure 25.1. Word-formation items and simplex items as semantic extensions (based on Langacker’s notation; length of arrows indicates relative distance from prototype) word-formation 653 prototypical notion of ‘tree’. Postulating the necessity of accommodation for compounding and other word-formation processes implies a rejection of full compositionality, which means that the components of word-formation items can no longer be understood as building blocks of the composite structure. Instead, Langacker (1987: 452, 461) proposes a scaffolding metaphor to indicate that the components only trigger off or motivate the compounds, supply a certain amount of conceptual assistance, but are discarded when the compound is fully entrenched by frequent activation. 3 Compositionality and Analyzability Yet even the partial compositionality to which Langacker admits, and which is mirrored in the scaffolding metaphor, requires a more fine-grained analysis. While compositionality can be claimed to denote an objective relationship between the composite structure and its components, analyzability introduces the psycholog- ical perspective of the hearer (or reader). Composite structures, Langacker (1987: 457) explains, may, but need not be, analyzed by the hearer in the comprehension process, and we cannot be sure to what extent this process is carried out consci- ously or unconsciously. For instance, swimmer, mixer, and complainer are deriva- tives suggesting a strong awareness of the constituents while, barring exceptions, the compositionality of computer, propeller, and ruler is not normally realized by the language user (Langacker 1987: 297). Schematicity So far, word-formation has been mainly approached from the angle of semantic extension and composite structures. However, Cognitive Linguistics offers an al- ternative, but complementary, view in terms of schematicity. As explained in detail Figure 25.2. Trajector and landmark in a prepositional phrase and an -er derivation (from Langacker 1990: 25, figure 12a, and Langacker 1987: 311, figure 8.8, excerpt) 654 friedrich ungerer by Tuggy (this volume, chapter 4), the notion of schema (in Langacker’s definition) offers a flexible way of generalization that is not understood as a fixed a priori rule, but takes account of salience based on frequency of use (see Schmid, this volume, chapter 5); ‘‘schemas are essentially routinized, or cognitively entrenched, patterns of experience’’ (Kemmer 2003: 78). In the lexical sphere, this is reflected in the net- works in which both schematic ‘superordinate’ schemas and prototypes and ex- tensions combine easily, as in figure 25.4. As shown by the boxes in bold, the salient elements of the network are not the most general schemas (entity, thing), but the lower level schemas (e.g., plant) and the prototype (tree). The notion of schema becomes even more helpful when, leaving lexical net- works, one looks at the ‘‘rules’’ (of whatever status) that have been postulated to describe compositional and derivational processes, not to mention back-formation, blending, and acronyming, and considers how all these rules are notoriously jeop- ardized by a host of exceptions. Understandably, then, the notion of schema has— in different ways—been applied to various word-formation phenomena, to the analysis of compounds (see section 3.4.) as well as the explanation of blends (see section 3.5.), while other phenomena still await ‘‘schematic’’ treatment. Further- more, Lakoff’s concept of image schemas (‘‘basic’’ schemas based on bodily expe- rience such a up-down, part-whole, container, path) and event schemas (first used by Talmy 1991) have been influential; image schemas have left their mark on the cognitive analysis of prefixation (see section 3.3), and event schemas on the treatment of conversion (see section 3.2). 2.2. Word-Formation and Conceptual Blending Although Tuggy (this volume, chapter 4) regards conceptual blending as still an- other variant of schematicity, it is preferable to treat it as a phenomenon in its own right, as the most successful cognitive attempt so far to come to terms with the online processing aspect of conceptualization. Most word-formation processes Figure 25.3. Relationship of word-formation items to semantic extension of prototypes and integration into composite structures in Langacker’s system (based on Langacker’s diagrammatic representation; Langacker 1987: 451, figure 12.2) word-formation 655 involve semantic combination or fusion, and this qualifies them for an analysis in terms of conceptual blending as proposed by Fauconnier and Turner (this volume, chapters 14 and 15, resp.), even if the conceptual blending analysis has taken its time to discover lexical blending and acronym formation (but see section 3.5 below) In particular, conceptual blending seems well suited to elucidate processes like lexicalization, perhaps even more convincingly than Langacker’s notion of ac- commodation, as shown in figure 25.5 for the compound wheelchair. 4 While the two input spaces reflect the conceptual content of the two coun- terparts (the constituents of the compound), the blended space contains the emer- gent structure of the compound, which is not provided by the two input spaces alone, but which, according to Fauconnier and Turner (2002: 48–49) involves al- together three types of conceptual process: composition, completion, and elabo- ration. While composition refers to the contributions made by the projections from the two inputs, completion is concerned with the addition of background knowledge (here represented by the notions of ‘invalid’, ‘hospital’, ‘engine’), and this permits the emergence of a new conceptual structure (indicated by the square within the blended space). Finally, elaboration may be envisaged as a test of the correctness and consistency of the conceptual content of the blended space. 2.3. Word-formation and Form-Meaning Iconicity One aspect that is not in the foreground of Langacker’s and Fauconnier’s con- siderations is the iconic relationship between phonological form (including a form’s graphic shape) and meaning (Langacker’s phonological and semantic phonological poles). The link between the two is taken for granted, as in many traditional ac- counts, and this is probably quite acceptable for one-word (lexical) schemas, but Figure 25.4. Lexical network combining schemas (entity, thing, plant, etc.), prototype (here: tree), and compound extensions (apple tree, etc.) 656 friedrich ungerer not necessarily for word-formation items, in which the form-meaning relationship is systematically violated. Coupled with this item-related isomorphism is a second type of iconicity, which is process-related in the sense that an item’s phonological form mirrors a particular word-formation process. While lexical blending can be seen as the word-formation enactment of conceptual blending (see section 2.2) and compounding as a clustering of two (or more) concepts, acronyming involves a reduction of conceptual content, and these processes of fusion or accumulation and reduction are duly reflected in the phonological (and graphic) shape of the items. As I showed in Ungerer (1999), the interaction of these two types of iconicity is not always beneficial for the word-formation process. As for compounds, the parallelism of two forms and two concepts is only achieved for the minority of fairly compositional, analyzable compounds (apple juice, silk jacket). For compounds that have undergone conceptual accommodation and expansion, the ensuing asym- metry between two forms and multiple meanings can only be resolved if full en- trenchment (or lexicalization) is achieved because the word form is no longer an- alyzed (as probably in holiday) or if the complex phonological form is reduced to a simplex form again (paper for newspaper). The situation for word-formation blends like skyjacking, stagflation,orinfo- tainment is the reverse, but it is not satisfactory either. Given the schematic license (see section 3.5 below), the fusion is impressively reflected in the phonological and graphic form, which thus provides an example of processual iconicity; yet one cannot be sure that a simultaneous, complete, and entrenched fusion takes place on the meaning side resulting in a new isomorphic form-meaning relation- ship. With regard to acronyms, the iconic reduction mirrored in chains of initial letters of words is convincing, but its effect can be devastating and cause a Figure 25.5. Compounding as conceptual blending process: Example wheelchair (based on Fauconnier 1997: 151) word-formation 657 complete breakdown of the isomorphic form-meaning link. This is one of the reasons why acronyms are often remodeled on existing items with an identical phonological form and a related meaning; the intention is to use these ‘‘prop words’’ and their attested form-meaning isomorphism to promote the entrench- ment of the acronym. 3. Description of Individual Word-Formation Phenomena It is not surprising to find that more specific cognitive studies of derivation and composition have concentrated on the areas most widely discussed in traditional word-formation research: prefixation, -er suffixation, conversion or zero deriva- tion, and noun-noun compounding. To sketch the possibilities—and hopefully the superiority—of the cognitive view, each of these domains will be represented by two different approaches. In addition, lexical blending will be discussed because its cognitive treatment opens up interesting research lines. 3.1. Prefixation, Contrast, and Diminutives The fact that prefixation in English lends itself to an interpretation in terms of con- trastivity has stimulated several differing cognitive approaches toward prefixation. Profiling Contrast Although English prefixes may reflect different semantic concepts, Schmid (2005: 162–65) claims that they have one thing in common: they profile contrast, and thus express a specific interpretation of the basic Figure/Ground distinction. This no- tion of contrast, of being ‘different from X’, can take on various guises depend- ing on the semantic specification of the prefix, as shown in table 25.1, where the traditional semantic classification of prefixes is interpreted in terms of the notion ‘contrast’. The dominant role of ‘‘contrast’’ is supported by gaps in the productivity of prefixation. Such gaps can be observed with verbs that do not permit contrast (*unlive,*unsit,*unsleep,*unplay), as well as with most concrete nouns, which, apart from gender-sensitive pairs like man and woman, do not naturally invite semantic opposition. Adjectives, according to Schmid, are characterized by a one- dimensional semantic structure which is particularly well suited for the establish- ment of contrast, and they therefore feature prominently in prefixation. 5 658 friedrich ungerer Contrastivity as a Complex Category Although he is in agreement with Schmid on the significance of ‘‘contrast’’ for prefixation, Mettinger (1994, 1996) does not regard contrastivity as a unified cog- nitive principle, but as a complex category rooted in several image schemas. As for adjectives prefixed with un-, Mettinger holds that they involve two image schemas, the scale schema for the prefixation of polar adjectives (unimportant, uncertain, unhappy) and the container schema for the contradictorily negated adjectives (untrue vs. true). Following Taylor’s (1992) interpretation of polar adjectives, Met- tinger interprets the -un prefix as encoding the relation between a thing serving as trajector and a scale serving as landmark scale, whereby the trajector is positioned on the landmark scale below the assumed norm (see figure 25.6a). For contradic- torily negated adjectives, the container functions as landmark and the trajector is placed outside the landmark; the un- predication is constituted by the relation between the ‘‘extraposed’’ trajector and the landmark (see figure 25.6b). 6 Prefixation, diminutives, and Underlying Idealized Cognitive Models The reason why diminutives are treated together with prefixes is that they can best be understood if they are compared with scaling-down prefixes of degree creat- ing a contrast, such as maxi- or mini As such, the difference between kitchenette Table 25.1. Prefixes and types of contrast (based on Schmid 2005: 162–65) Prefix Type of Contrast (based on the Figure/Ground distinction) Examples Negative prefixes un-, in- different from X unhappy, uneven Reversative prefixes un-, dis- different direction from unwrap, disappear Locative prefixes extra-, intra- not outside, but inside X extracellular, intramural Temporal prefixes pre-, post- re- not during or after, but before X X again, in contrast to expectation prewar, postwar rebuild, reopen Prefixes of quantity (degree and number) ultra-, sub-, super- more than the norm for X, less than the norm for X ultraright,?? Prefixes of attitude pro-, contra-, ounter- pro X, not anti X, etc. pro-Palestinian, counterproductive word-formation 659 . 25.5 for the compound wheelchair. 4 While the two input spaces reflect the conceptual content of the two coun- terparts (the constituents of the compound), the blended space contains the emer- gent. adjustment of the components and the addition of conceptual content in the composite item. This is indicated through the element X in the formula for the composite item in the schematic boxes on the. by the boxes in bold, the salient elements of the network are not the most general schemas (entity, thing), but the lower level schemas (e.g., plant) and the prototype (tree). The notion of schema