1 This article is the written and extended version of a paper presented at the Inaugural Conference of the International Society for the Linguistics of English (ISLE-1), October 8-11 2008, Freiburg i.Br., Germany. AAA - Arbeiten aus Anglistik und Amerikanistik Band 35 (2010) Heft 1 Gunter Narr Verlag Tübingen New ways of investigating morphological productivity 1 Anne Schröder and Susanne Mühleisen The potential of a word-formation pattern to be exploited in the creation of new words is seen to be one of the defining characteristics of morphological productivity. However, most measures of morphological productivity as applied in recent studies deal with words attested in dictionaries and corpora and thus per definitionem with actual words. The present paper shows different approaches to the investigation of morphological productivity by presenting results from two recent studies on the productivity of English word-formation patterns in which the aspect of potentiality of word-formation was tested. These include a coinage test via an online survey and the exploitation of the World Wide Web to test the ratio of potential versus actual new creations. Although both studies presented therefore deal with expressive morphology and so-called creative coinages, which are frequently believed to be irrelevant for the investigation of morphological productivity, this contribution illustrates that such formations are decisive in gaining a deeper understanding of the multi-layered and complex phenomenon of morphological productivity. KEYWORDS: morphological productivity, word-formation, creativity, elicitation experiments, the Web as a corpus, verbal prefixation, noun-derivation 1. Introduction: on the notion of morphological productivity The question of productivity has always been considered a difficult one for modern linguistic theories in general, as, for example, pointed out by Aronoff (1980: 71), who claims it to have frequently been swept under some conve- Anne Schröder and Susanne Mühleisen 44 nient rug. Nevertheless, the question of productivity has become one of the central empirical problems for theories of word-formation, which has led Bauer to point out that “despite its importance, it is still poorly understood” (2003: 70). Therefore, it may not come as a surprise that literature on wordformation abounds with varied definitions of productivity. To our knowledge, the most exhaustive summary of possible definitions of productivity is provided by Rainer (1987: 188-190), taken up by Bauer (2001: 25). Most of these definitions take productivity as a quantitative notion in that they acknowledge the existence of a continuum, as for instance in Aronoff and Anshen’s (1998: 242-43) “the extent to which a particular affix is likely to be used in the production of new words in the language. On this view, productivity is a probabilistic continuum that predicts the use of potential words.” Thus, at the one end of this continuum, we find completely unproductive patterns, while on the other end, we find highly productive patterns with a number of cases ranging in-between (see also Kettemann 1988: 13). As a logical consequence, various ways of measuring the productivity of a particular pattern have also been proposed. However, most of these measures deal with words attested in dictionaries or corpora, i.e. they tell us, as Romaine (1983: 181) puts it, “which words are actual, but not which words are possible.” On the one hand, this means that measures of productivity depend to a large degree on the quality and size of the data base. On the other hand, this also means that these measures give no evidence of the potential of a word-formation pattern to be exploited in the creation of new words. This, however, is a serious flaw, as potentiality is one of the defining features of morphological productivity, as shown, for example, in the definition by Aronoff and Anshen (1998) above or in the following one by Plag (2006: 127): “The productivity of a word-formation process can be defined as its general potential to be used to create new words and as the degree to which this potential is exploited by the speakers.” In the present paper, we will try to remedy this by presenting results from two recent studies (Mühleisen 2006, Schröder 2008a) on the productivity of English word-formation patterns, in which the aspect of potentiality of wordformation was tested. These include a coinage test via an online survey and the exploitation of the World Wide Web to test the ratio of potential versus actual new creations. Before introducing these studies in more detail, we will first give a short overview on established and widely applied methods of measuring productivity. New way of investigating morphological productivity 45 2. Established methods of measuring productivity In the relevant literature, three principal ways of measuring productivity have been established. These measures are based on dictionary listings, on the analysis of corpora, or on psychological tests of native speakers’ intuition, i.e., elicitation tests (cf. Bolozky 1999; Plag 1999, on different types of productivity measurements, see Risto-Donevic 1999: 19-26). 2.1 Dictionary-based approaches Dictionaries can be used in a number of ways to identify the productivity of a morphological process. For example, dictionaries which record new words only, such as the Oxford Dictionary of New Words (Knowles and Elliott 1997), can be employed to investigate whether a word-formation rule is still active. One can also search the Oxford English Dictionary (OED) for all very recent neologisms created on the basis of such a rule. In addition, one can compare the number of neologisms created using a particular word-formation rule during various periods of historical time and thus identify different degrees of productivity of the same rule over the course of time. However, one of the greatest flaws of dictionary-based investigations of productivity is that the quality of the results depends on the quality of the lexicographers’ work. In addition, dictionaries do not necessarily represent actual usage and may be influenced more by the linguistic norm rather than the linguistic system (cf. Bauer 2000: 838). As Booij puts it: “A dictionary is always lagging behind with respect to the use of productive morphological patterns because it only registers […] established words. Morphological productivity manifests itself most clearly in the appearance of complex words that never make it to the dictionary” (Booij 2005: 69). Most importantly, however, any dictionary can only list actual words. Dictionary-based approaches to productivity are therefore unable to include the aspect of potentiality to morphological productivity, which is considered central to the definition of productivity (see e.g. the definition by Plag 2006 given above). The same disadvantage also applies to the second type of data-base, viz. large computer corpora. 2.2 Corpus-based approaches The corpus-based approach has become increasingly popular for measuring morphological productivity. As Baayen and Lieber point out, the advantages of corpora over dictionary lists are threefold: firstly, corpora generally contain transparent words which are typically neglected in dictionaries; secondly, corpora contain only words actually used in naturally occurring speech whereas dictionaries may list words not attested in actual speech; thirdly, Anne Schröder and Susanne Mühleisen 46 corpora provide information on the frequency of words (Baayen and Lieber 1991: 803). There are a number of ways in which corpora can be used to measure productivity. The simplest method is to establish type or token frequencies: “the more words with some affix, the more productive the affix” (Baayen and Lieber 1991: 802). However, these types of rather unsophisticated measurement can, at best, only be used as a preliminary indicator of actual usage, because such a simple count may often be principally a reflection of past usage. Thus, a corpus, similarly to a dictionary, can only contain ‘existing words’ and will thus, theoretically, only yield results on the past productivity of word-formation patterns. However, a number of statistical methods developed by Baayen and his collaborators enable us to use corpora to compute “the probability of encountering a newly formed word of the relevant morphological category” (Plag 2000: 65; for further sources, see References). We need not discuss these statistical methods in detail, as they have been widely discussed and applied in e.g. Plag (1999), Scheible (2005) and Schröder (2007, 2008a). But it is worth mentioning that all of Baayen’s formulae have been subjected to severe criticism, especially by van Marle (e.g. van Marle 1992), although even he acknowledges that “[n]o one seriously involved in the study of morphological productivity can afford to leave Baayen’s work unread” (van Marle 1992: 152). 2.3 Elicitation tests Finally, elicitation tests can actually test the aspect of potentiality in wordformation and have therefore been described as a useful supplement to the other types of productivity measurement (e.g. Schröder 2008a: 41). There are two types of elicitation experiments: firstly, informants can be asked to judge the acceptability of existing or non-existing complex forms; or secondly, they can be induced to create words according to given word-formation rules. In the latter, the informants list all the words following a particular word-formation pattern they can think of. Acceptability experiments with novel or rare words, intended to reveal the potential of a given word-formation rule, have been conducted, for instance, by Aronoff (1980), Romaine (1983), Berman (1987), and Scheible (2005). To our knowledge, the only studies to include a coinage experiment (besides Schröder 2008a) are Anshen and Aronoff (1988) and Bolozky (1999). Given that elicitation tests are the only type of established productivity measurement which can actually test the aspect of potentiality in wordformation, it seems surprising that these tests are rarely carried out in studies on morphological productivity. In our discussion of new methods of investigating morphological productivity we will therefore pay special atten- New way of investigating morphological productivity 47 tion to this method by introducing Schröder (2008a), a study on the morphological productivity of verbal prefixation in English, in which new types of elicitation tests have been applied. 3. New methods of investigating morphological productivity 3.1 Schröder (2008a): On the productivity of verbal prefixation in English This study is part of a larger investigation on verbal prefixation in English. The verbs investigated are of the type upgrade (1920), overachieve (1953), underperform (1976) and download (1989), all of which are comparatively recent creations as the first citation dates in the OED show. Given that verbal prefixation is considered to be unproductive in Modern English, these verbs - most of which were created in the second half of the last century - may come as a surprise. According to Mair, these neologism seem to suggest “that this particular word-formation process might […] be making a comeback after centuries of decline” (Mair 2006: 63), and this is what the study investigates: Do creations such as these indicate an increase of productivity of a word-formation pattern which is considered to be old and unproductive? The study consists of altogether three different types of empirical investigation; the first two apply the established methods presented in section 2 above. However, since this paper is concerned with the more innovative and recent types of investigations of morphological productivity, only the third kind will be featured here (for a discussion of the other two types of investigation, see e.g. Schröder 2007, 2008a, 2008b). This third part of the study consists of elicitation tests, both acceptability tests and coinage tests. The design of these tests differs considerably from that of previous studies (e.g. Aronoff 1980, Aronoff and Schvanefeld 1978) in that it was conducted online and thus made accessible to a large number and a wide variety of native and non-native speakers of English. The acceptability tests in Schröder (2008a) still resemble those carried out in earlier studies (e.g. Anshen and Aronoff 1989, Aronoff and Schvanefeld 1978), while the coinage tests apply a new approach: in earlier, comparable experiments (e.g. Anshen and Aronoff 1988) informants had been asked to merely reproduce as many existing words as they could. In Schröder 2008a, however, the informants were asked to invent up to five new words with the prefixes overand underand to give a short explanation of their intended meanings (see reproduction of the relevant part of the questionnaire in the Appendix). 189 informants chose to participate in this experiment. Of these 189, 146 (77.3%) were native and 43 (22.7%) non-native speakers of English. On Anne Schröder and Susanne Mühleisen 48 average, the informants each proposed 3.81 words beginning with underand 3.57 words beginning with over-, and thus altogether, 727 possible words beginning with underand 673 overwere proposed during the experiment. Of these proposed words, verbs not occurring in the 2002 version of the OED on CD-ROM were regarded as ‘new’. In addition, all proposed words for which their creator did not provide a definition or which were not given in the form required (some informants proposed participles or nouns instead of infinitives) were eliminated from the final list, as were verbs which were coincidentally given several times with the same definitions by several informants. Still, altogether 425 new words with the prefix overand 502 with the prefix underwere created during the experiment, and we believe that this number suggests that these two prefixes can indeed be considered to be productive in Modern English. The results of this experiment also show that the two prefixes were freely combined both with Germanic (e.g. under-sit, over-spit) and Latinate/ Romance bases (e.g. under-acquire, over-criticize), with no clear preference for either. In addition, a variety of morphological rules seem to have been applied in the coining of neologisms with the two prefixes under consideration, as the following list shows: 1. Prefix + simple verb: e.g.: to under-sit (‘to sit under’); to over-spit (‘to spit on something’); to under-acquire (‘to fail to meet acceptable levels of acquisition’); to over-acquire (‘to acquire too much’); 2. Prefix + complex verb: e.g.: to under-acclimatize (‘to not adjust to the weather enough’); to over-brighten (‘to make too bright’); to under-outdo (‘to outperform others by a less-than-expected margin’), to over-download (‘to download (from the Internet) more than your limit’), under-appraise (‘to set the value of a property or item too low’); to over-disappear (‘to disappear without a trace’); to underempower (‘to give somebody inadequate authority’); 3. Prefix+ zero-derived verb: e.g.: to under-table (‘to be put under a table’); to over-fortnight (‘to spend 14 days in a specified place’); to under-brown (‘to fail to cook something long enough to achieve a desired degree of brownness’); to overfull (‘to put a lot of things in a small place’) 4. Other: e.g. to over-ambiguate (‘to turn a puzzle into a puzzlement’); to undermilkwood (‘digging up Dylan’); to under-ling (‘to treat someone as if they were inferior to you’), to under-cover (‘to work for the police in a secret way’) One might even argue that the verbs given as pattern 3 are actually examples of pattern 2, as to brown, to full and to green as zero-derived verbs can certainly be considered to be morphologically complex. Thus, from the analysis of the elicited neologisms one thing becomes clear: there are a number of verbs for which several patterns could be assumed or which defy any classification and hence have to be classified as ‘other’ in the list above. In addition, it is frequently impossible to decide for one of the available patterns with any certainty. New way of investigating morphological productivity 49 Finally, whether the informants actually employed any of the morphological rules listed above in the coining of the possible words remains subject of speculation. In fact, there are a number of possible words which have very clearly been created in analogy to existing verbs. Thus, to undereat (‘to diet’, ‘to not eat enough’) is very likely to be formed on the model of to overeat (‘to eat too much’) attested in the OED. That the formations are cases of analogy is sometimes made explicit by the informants themselves: for instance in the case of to under-exaggerate, which is defined as ‘opposite of to overexaggerate’ and to over-define, which is defined as ‘to define with more detail than necessary, antonym of underdefine,’ although the latter is not attested in the OED. Sometimes analogical coining is joined by conversion, as for example in to undercare (‘to care less than it should be’), perhaps built on the model of overcare, n. (‘undue or excessive care’), attested in the OED. The verb to overcare (‘to care too much’) is given by another informant. But analogical coinings are not always based on the semantic counterpart with the opposite prefix; they may also be formed according to the semantics of another, similar verb. Thus, to over-groom (‘to attend so assiduously to one’s personal grooming as to invite suspicion or comment’) may have been coined on the model of to overdress, to over-guest (‘to invite too many guests’) on that of to overbook, and to over-nod (‘to sleep too late, particularly of a morning’) and to over-slumber (‘to sleep too much in a slumbersome way’) on to oversleep. Again, informants sometimes explicitly stated that this strategy had been applied in forming the possible words, as, for example, in to over-respond, for which the informant gives as the intended meaning: ‘like to overreact.’ In the literature, the first type of analogical coining is usually referred to as “affix-substitution” (van Marle 1985: 256), the second type of this process could then be referred to as ‘base-substitution’. The neologisms discussed here illustrate that various morphological processes were applied in the coining of neologisms. This can be taken as an indication that speakers do not analyse these words into their morphemes or according to morphological rules, but rather on the basis of their form, i.e. the existence of one of the prefixes and the resulting word class. In addition, they show that analogical coining might be less marginal than assumed. This is why, as a result of the analysis of the elicitation test in Schröder (2008a), a morphological network approach is proposed to analyse the structure of morphologically complex words: Anne Schröder and Susanne Mühleisen 50 Figure 1: Illustration of two interlinked morphological networks (taken from Schröder 2008a: 232) The advantage of such an approach is that no one rule needs to be chosen over another as an explanation for a particular complex form. This resolves the problem that, as already mentioned, there are a number of verbs for which several patterns could be assumed. This elicitation test was designed to investigate the potentiality of wordformation patterns in an experimental fashion. But such a test can then also lead to a comparison between potential words and actual usage: approximately 75% of the creations of this experiment can also be shown to exist in the World Wide Web (August-October 2008) - a result which may be less surprising if we look at our second study (Mühleisen 2006) on innovative approaches to morphological productivity in more detail. 3.2 Mühleisen (2006): Of Lessees, Retirees and Beseechees Heterogeneity and polysemy in word-formation were also the starting point for this exploration of a particular morphological pattern in English (Mühleisen 2006) - the -ee suffix of the type interviewee - which has sometimes been described as “verb-derived passive human noun” (cf. Katamba 1994: 65). Except for the fact that -ee words are nouns, none of the other criteria are able to stand a test of closer scrutiny: -ee words are often noun-derived (e.g., biographee, aggressee), often not passive or patient (e.g., retiree, elopee), and sometimes not even human either (e.g., possessee - ‘technical New way of investigating morphological productivity 51 term in linguistics’ or milkee - ‘animal’). The heterogeneity of -ee words has led to a number of syntactic and semantic descriptions that have elaborated on the constraints of -ee word formations (Bauer 1983, Barker 1998, Baeskow 2002, Lieber 2004). In her investigation, Mühleisen (2006) includes a diachronic dimension and highlights how new formations can be based on a number of syntactic and semantic possibilities which have arisen in the 600 years of the development of -ee words through an interplay between analogical coining and rule-governed production. One part of the study is concerned with the question of whether or not this suffix is “semi-productive”, as it has sometimes been labelled in the literature (e.g., Portero Munoz 2003: 130). At the time of research in 2005, approximately 500 -ee words were attested in dictionaries and collections of chance finds by linguists - not all that many cases altogether considering the history of this suffix over six centuries. As a logical consequence, Mühleisen (2006) focuses on neologisms, i.e., -ee words that have actually been produced but have not been listed or attested before in any of the established sources. The approach for measuring the productivity of -ee words in Mühleisen (2006) is thus corpus-based rather than built on native speaker introspection as in Schröder (2008a). The serious disadvantages of using any of the established large corpora, especially for the investigation of recent and very recent words, have already been made clear in section 2.1. Mühleisen (2006) therefore resorts to the one large body of text collections which is unbeatable with regard to size, diversity of “Englishes,” varieties of users (i.e., here producers of text), text types and recentness: the World Wide Web, a corpus which has gained in importance in innovative linguistic research in the last few years (cf. Kilgarriff and Grefenstette 2003, Hundt et al. 2007). While there are also shortcomings in the uses of the Web for particular types of linguistic analyses, the advantages easily outweigh them. Next to choice of database, the selection of the search criteria is of primary importance in the search for neologisms. Because of the high frequency of occurrence of ‘double e’ that is not a suffix in English, a simple search in uncoded Web material for the actual suffix -ee would result in an unmanageable amount of rather chaotic data. The alternative option is the search for specific lexemes, their frequencies and their contexts of occurrence. For prospective neologisms, the problem is that the searchable items cannot be found in any dictionaries or word lists, which would obliterate the whole point of the endeavour. Therefore, a set of 1,000 potential search words - non-existent in common dictionaries or scholarly articles - was created as a test pool for -ee neologisms. Each one of these was then tested for its actual realization by speakers/ writers in any of the English language webpages. The criteria for the list of potential -ee words were Anne Schröder and Susanne Mühleisen 52 2 C stands for “corpus-item”, the number attached to it stands for the position (from 1-1,000) of the test word in alphabetical order. 1. No occurrence in any of the established sources, such as the OED, selected other dictionaries, or in scholarly texts on -ee words. 2. Recognizable derivation of either a contemporary English verb (the majority) or -er noun (less frequently). The choice of bases was, apart from the above criteria, alphabetically ordered random choices from the OED. Examples of test words starting with the letter b, for instance, can be seen below: Table 1: 25 out of 1,000 test words on potential versus actual formations with -ee on the World Wide Web C 2 54 badgee C 63 besmearee C 71 boastee C 55 baitee C 64 besmirchee C 72 bouncee C 56 bangee C 65 bewailee C 73 boxee C 57 barkee C 66 bewitchee C 74 bruisee C 58 batee C 67 blamee C 75 brushee C 59 beggee C 68 blessee C 76 burglaree C 60 bellowee C 69 blowee C 77 buskeree C 61 bereavee C 70 bluffee C 78 butcheree C 62 beseechee One might say that this procedure is, in some ways, close to the good old Aronovian idea (1976) of relating potential to actual words, a notion of productivity whose advantage lies in the fact that it takes into account the number of bases the affix can attach to. What is new here is that by mining the Web for meaningful realizations of potential words, we are able to capture many more recent neologisms across a wider range of text types than ever before. It is evident that results from such a procedure have to be refined following a rigid quality criteria checklist in order to exclude cases not falling into the intended category: a) Misspellings or typos were not counted. b) The -ee word had to be part of a coherent English-language text. c) Personal names (e.g. desiree), nicknames (robbee) or brand names were eliminated from the list. New way of investigating morphological productivity 53 d) Unambiguous diminutive meanings were not counted. e) The -ee lexemes had to be recognisable as nouns. f) The -ee word had to appear in a meaningful context. g) Word plays are often part of creative coinages. However, no word plays with a meaning completely detached from the base of the derivation were counted (e.g., Eye-rackee - ‘Iraqui’). h) Repetitions of a word in the same website were counted as only one occurrence. i) -ee words which appeared as part of a compound were counted as valid (e.g. party ruinee, blue rinsee, lap dancee) The result of this potential to actual words test was well beyond expectations: out of the 1,000 test words, 748 had been realized by speakers/ writers in a meaningful way, i.e. the number of known -ee words has suddenly jumped from about 500 words (before the study) to approximately 2½ times this size (after the study). This includes relatively frequent words like accusee, circumcisee, discriminatee, harassee, spankee or touchee (altogether 120 occurrences or 12.0 %, see Table 2 below), but it comprises also a high percentage of Hapax Legomena (21.4 %) and rare occurrences (26.7%, see Table 2 below): Table 2: Test word analysis according to frequency categories Category No. of Words Percentage Not found (= 0) 252 25.2 Hapax Legomena (= 1-2 occurrences) 214 21.4 Rare (= 3-20 occurrences) 267 26.7 Established (> 100 occurrences) 97 9.7 Frequent (< 100 occurrences) 120 12.0 Not quantifiable 50 5.0 Among the non-quantifiable category one can find such unlikely coinages as elevatee, faileee and grippee, which are given below in context. These examples should provide some evidence that, similar to the neologisms in Schröder’s study (2008a), the new formations here are not only based on morphological rules but rather that analogical coining, especially of the affixsubstitution type, seems to play a large role: Gripper/ grippee C 346: President Polk considered handshaking an art to be developed, both for his image and self-preservation. When a Anne Schröder and Susanne Mühleisen 54 man approached, he’d mentally assess his strength and quickly beat a strong man to the shake, ensuring that he was the gripper, not the grippee. President Hoover never did master the art of handshaking. Elevator/ elevatee C 238: With the Secret Floor Ballot, each employee in a large office building has a small remote control in his pocket, with which he can signal “3” without the elevator button lighting up. By the time the thing stops at the third floor, he is out the door before he can see the grimaces of the elevatee he left behind. Failee C 297: a) This will ensure the other trainees don’t get bored and the “failee” doesn’t get disheartened; you can return to this exercise later, […] b) “I think you can be the victim of someone else’s failure. So my parents are moderately middle-class people - my dad’s a lawyer - who haven’t quite had the life they expected. My Dad’s judgement led to him mistiming and misjudging a couple of crucial deals that would have meant he could retire at thirty. So I guess that makes my mother a failee: she’s living a certain degree of failure, as a result of the actions of someone else. It’s a good reason not to get married, too, methinks. The creation of grippee (C 346) seems to be triggered by the co-occurrence of the relational opposite gripper - a phenomenon which can be observed rather often if we look at the creation of a new -ee word in context. In fact, a collocation test in the study (Mühleisen 2006: 175-181) reveals that of all the 748 successful -ee lexemes, 552 items (73.8 %) occurred at least in one instance in “lexical solidarity” with an -er word, whereas 196 items (26.2 %) produced no such collocations. Analogical coining from an existing -er word even extends to instances where the -er word and -ee word have (usually deliberately) slightly incongruous meanings - as in the instrument versus person pairing of elevator - elevatee (C 238) above. But affix-substitution is not the only kind of analogical coining process that is visible here: similar to the ‘base-substitution’ observed in Schröder’s study above, in the first example of failee (C 297), it seems to be the semantics of another, similar noun - trainee - which is responsible for the creation process. It is one of the advantages of observing hapax legomena and rare examples of a new formation that one can get easy access to the context and possible influences of the creation. For the more frequent new -ee words in Mühleisen (2006), the picture is not quite as straightforward. It has been shown, however (see Mühleisen 2006: 181-186 for more detail) that the most frequent tokens are relatively rule-governed and comply with “prototypical” characteristics of -ee words, i.e. New way of investigating morphological productivity 55 1. verb-derived (i.e. verb exists), 2. with existing correlative -er noun, 3. in direct object relation to the verb, 4. sentient and probably human, 5. participant role 6. non-volitional and non-active part in the event 7. can be used in a legal as well as more general contexts This does not rule out analogical coining as part of the formation process of these frequent words. Neither does it mean that successful words have to comply with all criteria above to become established in usage. But it shows that there is a conglomerate of characteristics of -ee words which can be related to a successful contemporary formation (for historical characteristics of -ee words, cf. Mühleisen 2006: 78-112). We can therefore see that both frequent tokens and hapax legomena, can shed light on the types of processes - analogical coinage and rule-governed production - which are involved in establishing a new word. 4. Conclusion One might say that both of our new procedures of measuring productivity have produced what is often referred to as “creative coinages” - in the first procedure (Schröder 2008a) in an experimental and conscious way, in the second one (Mühleisen 2006) in “real life” but often in playful or humorous contexts. Indeed, many morphologists believe that words that draw attention to themselves are formed according to an unproductive pattern (e.g. Lieber 1992: 3), are marginal and “not necessarily to be taken seriously by a theory of word formation” (Lieber 1993: 3). However, we agree with Adrienne Lehrer that such an “attitude is unfortunate, because creative neologisms can tell us a great deal about important aspects of word-formation. Moreover, they are extremely productive! ” (Lehrer 1996: 64). Zwicky and Pullum (1987) find it necessary to differentiate between plain morphology, i.e. “the ordinary productive (and nonproductive) wordformation and word structure rules of a language” (Zwicky and Pullum 1987: 332) and expressive morphology, which is “associated with an expressive, playful, poetic, or simply ostentatious effect of some kind” (Zwicky and Pullum 1987: 335). And although they do not consider the latter to be a marginal phenomenon, they believe that “rules of expressive morphology are not subject to the same conditions as rules of plain morphology” (Zwicky and Pullum 1987: 338). The creations to under-milkwood and to under-fool about (see list above) are probably illustrative of this, especially as the former additionally does not Anne Schröder and Susanne Mühleisen 56 satisfy the criterion of transparency required of productively derived forms. They are probably the type of individual creativity “manifest in imaginative, ingenious neologisms” referred to by Bolozky (1999: 4), representing the part of an individual’s mental lexicon sometimes elicited in coinage tests. Zwicky and Pullum believe that there is not “simply a continuum from plain to expressive morphology” (1987: 338), but that the two should be viewed as two distinct phenomena. However, if we look at a detailed history of a word formation pattern - as for instance, suffixation with -ee - we can see that creative coinages were usually the trigger to initiate a new rule such as from noun-noun derivation to verb derivation in the 15 th and 16 th centuries, from indirect object formation (e.g. abandonee - ‘someone to whom something is abandoned’) to direct object formation (‘someone who has been abandoned’) (cf. Mühleisen 2006: 78-112). Bolozky (1999: 3) once pointed out that “precise measurement of word formation productivity […] would not seem to be a realistic goal”, and it is also not our goal to find the one conclusive formula for measuring productivity, but we hope with Bauer (2001: xiii) “that the work presented here will provide a stepping-stone in the development of a new deeper understanding”. We hope we have shown that there are more ways to include native speaker knowledge and production in investigating productivity than what we find in current literature, especially with regard to the feature of potentiality. 5. References: Anshen, Frank and Mark Aronoff (1988). “Producing Morphologically Complex Words.” Linguistics 26. 641-655. Anshen, Frank and Mark Aronoff (1989). “Morphological Productivity, Word Frequency and the Oxford English Dictionary.” In: Ralph Fasold and Deborah Schiffrin (eds.). Language Change and Variation. Amsterdam: Benjamins. 197-202. Aronoff, Mark (1976). Word Formation in Generative Grammar. Cambridge, MA: MIT Press. Aronoff, Mark (1980). “The Relevance of Productivity in a Synchronic Description of Word Formation.” In: Jacek Fisiak (ed.). Historical Morphology. The Hague: Mouton. 71-82. Aronoff, Mark and Peter Schvanefeld (1978). “Testing Morphological Productivity.” Annals of the New York Academy of Sciences: Papers in Anthropology and Linguistics 318. 106-114. Baayen, Harald (1991). “Quantitative Aspects of Morphological Productivity.” Geert Booij and Jaap van Marle (eds.). Yearbook of Morphology 1991. Dordrecht: Kluwer. 109-149. Baayen, Harald (2001). Word Frequency Distributions. Dordrecht: Kluwer. Baayen, Harald and Rochelle Lieber (1991). “Productivity and English Derivation: A Corpus-based Study.” Linguistics 29. 801-843. Baayen, Harald and Antoinette Renouf (1996). “Chronicling The Times: Productive Lexical Innovations in an English Newspaper.” Language 72: 1. 69-96. New way of investigating morphological productivity 57 Baeskow, Heike (2002). Abgeleitete Personenbezeichnungen im Deutschen und Englischen. Berlin: Walter de Gruyter. Barker, Chris (1998). “Episodic -ee in English: A Thematic Role Constraint on New Word Formation.” Language 74: 4. 1-33. Bauer, Laurie (1983). English Word-formation. Cambridge: CUP. Bauer, Laurie (2000). “System vs. Norm: Coinage and Institutionalization.” In: Geert Booij et al. (eds.). Morphologie/ Morphology. HSK 17: 1. Berlin: Mouton de Gruyter. 832-840. Bauer, Laurie (2001). Morphological Productivity. Cambridge: CUP. Bauer, Laurie (2003). Introducing Linguistic Morphology. Washington, D.C.: Georgetown UP. Bolozky, Shmuel (1999). Measuring Productivity in Word-Formation. The Case of Israeli Hebrew. Leiden: Brill. Booij, Geert (2005). The Grammar of Words. An Introduction to Linguistic Morphology. Oxford: OUP. Hundt, Marianne, Caroline Biewer and Nadjia Nesselhauf (eds.) (2007). Corpus Linguistics and the Web. Amsterdam: Rodopi. Katamba, Francis (1994). English Words. London: Routledge. Kettemann, Bernhard (1988). Die Phonologie morphologischer Prozesse im amerikanischen Englisch. Tübingen: Narr. Kilgarriff, Adam and Gregory Grefenstette (2003). “Introduction to the Special Issue on the Web as Corpus.” Computational Linguistics. Special Issue on the Web as Corpus 29. 333-347. Lehrer, Adrienne (1996). “Why Neologisms are Important to Study.” Lexicology 2: 1. 63-73. Lieber, Rochelle (1992). Deconstructing Morphology. Word Formation in Syntactic Theory. Chicago: The U of Chicago P. Lieber, Rochelle (2004). Morphology and Lexical Semantics. Cambridge: CUP. Mair, Christian (2006). Twentieth-Century English. History, Variation, and Standardization. Cambridge: CUP. Marle, Jaap van (1992). “The Relationship between Morphological Productivity and Frequency: A Comment on Baayen’s Performance-oriented Conception of Morphological Productivity.” In: Geert Booij and Jaap van Marle (eds.). Yearbook of Morphology 1991. Dordrecht: Kluwer. 151-163. Mühleisen, Susanne (2006). Of Lessees, Retirees and Beseechees. A Historical, Empirical and Sociolinguistic Analysis of a Word-formation and its Productivity in English. Habilitationsschrift, Universität Regensburg. Mühleisen, Susanne (2007). “Of Lessees, Retirees and Beseeches: a Corpus-based Analysis of the Productivity of a Heterogeneous Word-formation Pattern in English.” In: Sabine Volk-Birke and Julia Lippert (eds.). Anglistentag 2006 in Halle, Proceedings. Trier: WVT. 413-422. Plag, Ingo (1999). Morphological Productivity. Structural Constraints in English Derivation. Berlin: Mouton de Gruyter. Plag, Ingo (2000). “On the Mechanisms of Morphological Rivalry: A New Look at Competing Verb-Deriving Affixes in English.” In: Bernhard Reitz and Sigrid Rieuwerts (eds.). Anglistentag 1999 Mainz. Proceedings. Trier: WVT. 63-76. Plag, Ingo (2003). Word-formation in English. Cambridge: CUP. Plag, Ingo (2006). “Productivity.” In: Keith Brown (ed.). Encyclopedia of Language and Linguistics. Amsterdam: Elsevier. 121-128. Anne Schröder and Susanne Mühleisen 58 Portero Muñoz, Carmen (2003). “Derived Nominalizations in -ee: a Role and Reference Grammar Based Semantic Analysis.” English Language and Linguistics 7: 1. 129-159. Rainer, Franz (1987). “Produktivitätsbegriffe in der Wortbildungstheorie.” In: Wolf Dietrich et al. (eds.). Grammatik und Wortbildung Romanischer Sprachen. Beiträge zum Deutschen Romanistentag in Siegen. Tübingen: Narr. 187-202. Rainer, Franz (1993). Spanische Wortbildungslehre. Tübingen: Niemeyer. Risto-Donevic, Stefan (1999). Constraints on the Productivity of Affixation in English Word-formation. Graz: PhD Dissertation. Romaine, Suzanne (1983). “On the Productivity Word Formation Rules and Limits of Variability.” Australian Journal of Linguistics 3. 177-200. Scheible, Silke (2005). “‘Upgrading’, ‘Downsizing’ and Co: Revitalising a Moribund Word Formation Pattern in Present-day English? ” AAA: Arbeiten aus Anglistik und Amerikanistik 30: 1-2. 177-200. Schröder, Anne (2007). “Theoretical and Empirical Approaches to Productivity: Evidence from English Verbal Morphology.” In: Sabine Volk-Birke and Julia Lippert (eds.). Anglistentag 2006 in Halle, Proceedings. Trier: WVT. 399-411. Schröder, Anne (2008a). On the Productivity of Verbal Prefixation in English. Habilitationsschrift, Martin-Luther-Universität Halle-Wittenberg. Schröder, Anne (2008b). “Investigating the Morphological Productivity of Verbal Prefixation in the History of English.” AAA: Arbeiten aus Anglistik und Amerikanistik 33: 1. 45-65. The Oxford English Dictionary (2 nd ed.) on CD-ROM. Version 3.0. (2002). Oxford: OUP. Zwicky, Arnold M. and Geoffrey K. Pullum (1987). “Plain Morphology and Expressive Morphology.” Berkeley Linguistics Society 13. 330-340. Anne Schröder Institut für Anglistik und Amerikanistik Martin-Luther-Universität Halle-Wittenberg Susanne Mühleisen Fachgruppe Anglistik Universität Bayreuth New way of investigating morphological productivity 59 Appendix Reproduction of the relevant section of the online questionnaire (from Schröder 2008a):