Semantic change in word formation (1).
The present article seeks to provide an answer to the following question: according to which mechanisms may a pattern of word formation develop a new meaning? In order to keep the task to a manageable size only changes will be considered, the result of which stays within the same type of pattern (affixation, compounding, etc.). Hence, we will not discuss how affixes develop out of compounds or similar phenomena. The only scholar who, to the best of my knowledge, has addressed this issue in a systematic and comprehensive manner is Jaberg (1905), who claimed that semantic change in affixation is always the result of semantic change in individual words' plus reanalysis. Our study will reveal, however, that, though this is in fact the most common scenario, there is yet another, hitherto ignored mechanism of semantic change in word formation where no lexical change is involved. This" mechanism, which will be called "'approximation," allows a mismatch to arise between a word formation pattern and a neologism formed according to it, if the distance is bridged by metaphor or metonymy.
1. Jaberg (1905): semantic change in word formation as lexical change plus reanalysis
Semantic change may not only affect single words, but also patterns (2) of word formation. Two classic examples, one from Germanic and one from Romance, may suffice to illustrate this fact.
The first one is the emergence of verb-noun compounds in German, which Osthoff (1878: 15) was the first to explain as due to a reanalysis as verb-noun compounds of noun-noun compounds whose first noun could also be interpreted, both formally and semantically, as a verb stem. This explanation, which is now commonly accepted among students of German word formation, is summarized by Henzen (1965 ) as follows: Ausgangspunkt bilden zweifellos Worter mit einem Substantiv an erster Stelle, das auch als Verbalstamm aufgefasst werden konnte. Schlafkammer, ahd. slafkamara, ist, nach Ausweis der ubrigen Bildungen mit slaf-, ein altes Substantivkompositum; dennoch fassen wir den ersten Tell als Verbum, und in der gleichen zwiespaltigen oder unentschiedenen Lage befinden sich noch zahlreiche Falle wie Werktag (engl. workday), Baustein, Ruhebett, reisefertig. Diese Entwicklung hat im Ahd. ihren Anfang genommen. (Henzen 1965 : 69-70) (3)
Thus, according to this analysis, OHG slafkamara, (4) originally meaning 'room (kamara) for sleep (slaf),' has been reinterpreted as 'room for sleeping (slaf).' The semantic change here is very subtle, almost imperceptible on purely extensional grounds, as is generally the case between a verb and its corresponding action noun: the difference is normally characterized as one between an activity viewed as such in the case of the verb and the same activity viewed as an entity in the case of the action noun.
Somewhat more concrete but also more intricate is the semantic evolution that led Latin -aticus, a suffix originally forming relational adjectives like herbaticus 'relating to grass' (from herb- 'grass'), to end up in French as a collective suffix (cf. plumage 'plumage,' from plume 'feather') and as an action noun suffix (cf. lavage 'washing,' from lay- 'to wash'). According to Fleischman (1977), both changes presuppose the use of-aticus in Late Latin as a nominal suffix designating taxes, a new function which arose through the ellipsis of the noun meaning 'tax' (census, tributum, etc.) in noun phrases of the form census, tributum, etc. + relational adjective in -aticus/um. Thus, for example, as early as 722, herbaticum is attested with the meaning 'payment for right to pasture stock' (Fleischman 1977: 29). The change from 'tax' to collective meaning is explained in the following terms by Fleischman:
Taxes on goods and agricultural products were frequently paid in kind rather than in money. From the meaning 'tax paid in a particular commodity' to that of 'the commodity itself' is but a short step semantically, and given the fact that the majority of such commodities were, grammatically speaking, collectives or mass nouns, a number of Old French tax designations thus acquired an additional collective meaning, cf. cortillage ([left arrow] cortil 'small enclosure') 'tax on garden produce,' 'vegetables, garden produce'; [...]. (Fleischman 1977: 91)
The same kind of explanation is given for the rise of the action meaning:
In addition to commodities, work activities were also taxed, notably those in which the subordinate utilized the facilities belonging to the lord or for which he had to obtain the lord's permission. The passage from 'tax on a given activity' to 'the activity itself' to 'verbal action signifying such an activity' (and eventually to 'verbal action per se') is a logical semantic transition and one which probably resulted in a number of tax designations' coming to function concurrently as action nouns (especially in cases of deverbal derivation). (Fleischman 1977: 92)
Every morphologist could supply many more examples of this kind from any language familiar to him. However, though the phenomenon in itself has been well-known since the nineteenth century and pertinent observations on single cases abound in the literature, it is surprising that, to the best of my knowledge, no scholar seems to have tackled the problem in a systematic and comprehensive manner. No systematic treatment of our problem may be found in general handbooks such as Paul (1975 ), Breal (1924 ), Sturtevant (1917), Bloomfield (1984 ), Anttila (1989), Hock (1991), or Lehmann (1992), nor in Malkiel's (1966) monographic article or in collective volumes on historical semantics (and word formation) like Fisiak (1985) or Blank and Koch (1999). The same is true for language-specific histories of word formation like Henzen (1965 ), Marchand (1969 ), Nyrop (1908), or Meyer-Lubke (1966 ), the most complete treatment being that of Leumann (1963: [section] 167).
The only exception seems to be Jaberg's (1905) substantial review of Roediger (1904), the first monograph in Romance linguistics devoted to the semantic development of an affix. Jaberg's theory of semantic change in word formation is stated in such a succinct manner that it will be best to quote it verbatim here. He starts out with a premise on the nature of the meaning of suffixes: (5)
Was wir 'Bedeutung eines Suffixes' nennen, ist nicht ein selbstandiger Begriff; es ist bloss die konstante Modifikation verschiedener Grundbegriffe. Fur das Sprachbewusstsein bilden Stamm und Suffix einen einzigen Begriff. (6) (Jaberg 1905: 459)
From this subsidiary role of the meaning of suffixes with respect to the meaning of words, Jaberg then derives the following consequences:
Das Suffix als solches kann somit seine Bedeutung nicht verandern, es verandert sie nur in Verbindung mit dem Stamm. Sobald nun aber eine Anzahl von Wortern, die mit demselben Suffix gebildet sind, ihre Bedeutung nach derselben Richtung hin verandern, so verandert sich auch die Funktion des Suffixes; wir sagen: es hat eine neue 'Bedeutung' erhalten. Dies aussert sich darin, dab mit dem Suffix in neuer Bedeutung neue Ableitungen gebildet werden. Es ergeben sich aus dem Vorhergehenden folgende methodische Forderungen: 1) Die Bedeutungsfinderung eines Suffixes ist aus dem Bedeutungsubergang einzelner Worter zu erklaren. 2) Es ist ein prinzipieller Unterschied zu machen zwischen Bedeutungsubergangen einzelner Worter und Neubildungen auf Grund einer neuen Bedeutung des Suffixes. (7) (Jaberg 1905: 459)
The second one of these methodological claims is undoubtedly correct. What I would like to investigate in this article is whether this is also true of Jaberg's first claim. (8) This claim has the merit of being easily falsifiable in principle: it will be sufficient to find cases where the new meaning of a pattern of word formation is not explainable as the result of semantic change in individual words followed by a reanalysis which attributes the semantic features resulting from the change to the pattern itself.
In studies on linguistic change it is customary now to distinguish between mechanisms and paths of change (cf., e.g., Jurafsky 1996; Traugott and Dasher 2002: 1). The process of change itself is usually divided into three phases: the individual act of innovation, the diffusion of the innovation through the speech community (also called "conventionalization"), and the resultant state in the language system. And with respect to innovation, at least two aspects are usually distinguished: the mechanisms through which the innovation comes about and the motives behind the innovation. Relying on this terminology, we can now define the scope of this article more precisely: we will restrict ourselves to the investigation of the mechanisms of innovation.
Claims about mechanisms of semantic innovation in word formation put forward in this article are meant to be valid for all natural languages, even though examples will be drawn exclusively from Latin, Romance, German, and English. The reason for the choice of such a restrictive sample of languages is that these happen to be the only ones the author is sufficiently familiar with to be able to separate with some confidence the wheat from the chaff in the relevant literature. In fact, though descriptions of cases of semantic change in word formation abound, I have found only relatively few whose empirical foundation was sufficiently broad and whose metalanguage was sufficiently explicit to serve as valid examples for our purposes. One of the subsidiary goals of this article, thus, will also be to elaborate a more differentiated and better-defined terminology that might prove useful in future descriptions of semantic change in word formation. In this endeavor, I have tried to do justice to those linguists who first described the relevant phenomena by generally adopting their terminological proposals, though not always their definitions and their analyses. As far as the examples are concerned, quality is more important than quantity: in principle it would suffice to give one single example for each mechanism that will be postulated, but this one example should not be open to alternative interpretations. One way of challenging the present theory of semantic change in word formation will thus simply consist in pointing out a more plausible alternative interpretation for some example. This, however, would only be problematic for the theory if no other convincing example were to be given to illustrate a postulated mechanism. A more serious objection to my theory, given its universalist claims, would consist in adducing examples of semantic change in word formation that are not accounted for by the theory, a line of argumentation which I will adopt myself against Jaberg's theory. Last but not least, the theory is also open to purely conceptual criticism which would show that some of its mechanisms can be derived from other elements of the theory or some more general theory yet to be proposed.
2. Natural vs. artificial semantic change
As we have seen, Jaberg's hypothesis excludes that affixes may change their meaning independently of a change in meaning of at least one word they originally formed part of. Now, however, cases of this kind clearly do exist, in languages for specific purposes at least.
In chemistry, for example, as shown by Corbin and Paul (2000:55-58), the prefixes cis- and trans- have been given new meanings only partially deducible from their meanings in common language. In standard language, our two prefixes serve to locate the referent of the head noun of a noun phrase with respect to the referent of the base of the prefixed adjective, which is conceptualized as a frontier: cisalpine region, for example, means 'region located on this side of the Alps' (from the point of view of the speaker/writer), while transalpine region means 'region located on the other side of the Alps.' In order to understand the usage in chemistry, look at Figure 1 displaying the structure of cis-2-butene and trans-2-butene. As one can see, cis-2-butene and trans-2-butene have the same chemical formula ([C.sub.4][H.sub.8]), but a different disposition of the atoms. The use of cis- is due to the location of the hydrogene atoms on "this" side of the C=C barrier in cis-2-butene, while one of them is located on the "other" side in trans-2-butene (hence the use of trans-). What differentiates usage in chemistry and in standard language is the fact that the base noun (butene) of the adjective does not refer to the barrier (C=C), but constitutes the hyperonym of the resulting formation (both cis-2-butene and trans-2-butene are kinds of butane). The use of cis- and trans- in chemistry thus, though still reminiscent of standard usage, has been adapted somewhat artificially for terminological purposes.
[FIGURE 1 OMITTED]
The same holds for a prefix like mega-, from Ancient Greek megas 'big, enormous,' which, together with other metrological prefixes (or combining forms, if you prefer), was endowed with a more precise meaning, viz. 'one million x' (megahertz, megabyte, etc.), at the eleventh 'Conference generale des poids et mesures' in 1960. (9) The meaning in such cases obviously did not arise through semantic change in one or more words containing these prefixes followed by a reanalysis attributing the new meaning to the prefix itself, but through an intentional act of redefinition of the affix's meaning.
Even though examples of this kind could easily be multiplied, it would seem somewhat unfair to consider Jaberg's hypothesis to be falsified by them since Jaberg certainly intended his hypothesis to bear on natural semantic change only. I will thus follow Gloning's (1996: 255) view that metalinguistic semantic stipulations of the kind we have seen should not count as part of a theory of meaning but of a broader theory of communication, and consequently restrict the following discussion to natural semantic change only.
The main contention of my article will be that, apart from artificial stipulations, there are two fundamental types of mechanism of semantic change in word formation, which I will call "reinterpretation" (10) and "approximation." (11) Only the various kinds of reinterpretation, it will be argued, are compatible with Jaberg's hypothesis.
In order to understand why reinterpretation of word formation patterns is so widespread, it is necessary to recall some general properties of human communication and of word formation.
As far as human communication in general is concerned, we will view it as an activity where a speaker (or writer) tries to influence a hearer (or reader) with the aid of mainly verbal signs. Since the speaker's message is underdetermined by the verbal signs alone and has to be supplemented by all kinds of inferences on the part of the hearer, the danger of misinterpretation is a constitutive aspect of human communication (cf., e.g., Keller 1995:132-146). While this source of change is hearer-based, another important source is speaker-based, namely, the habit of many speakers to use linguistic signs in innovative ways in order to augment their impact on the hearer.
With respect to word formation, we will base ourselves on a conception of the lexicon not as "eine fertig daliegende Masse," (12) but as "ein fortgehendes Erzeugnis und Wiedererzeugnis des wortbildenden Vermogens" (13) (Humboldt 1836: 109-110; quoted in Marle 1990: 267). Under this conception, "existing words, both simplex and complex, are subjected to a constant, never-ending inspection on the part of the speakers of the language" (Marle 1990: 267), and in the course of such inspections all kinds of new relationships between words may be established, giving rise, in some cases, to new patterns of word formation.
Equally important is the distinction, traditional in German studies on word formation at least since the 1970s (cf. Rainer 1993: 132-133), between Wortbedeutung (lit. 'word meaning')--also referred to as Wortschatzbedeutung (lit. 'vocabulary meaning')--and Wortbildungsbedeutung (lit. 'word formation meaning'). The first of these terms refers to the lexical meaning of a complex word, including all idiosyncratic aspects, while the latter focuses on that part of a complex word's meaning that may be attributed to the pattern itself by which the word was formed. This distinction is most obvious in nominal compounds, whose word formation meaning is often said to be only 'an [N.sub.2] which has something to do with [N.sub.1],' while the word meaning of single compounds may additionally present all kinds of encyclopedic information or idiosyncrasies. Word meanings can thus often present a semantic surplus with respect to the word formation meaning, and this surplus may eventually be reinterpreted as part of the word formation pattern itself, as we will see, for example, in Section 3.1.3 with respect to German compounds with Atom- as a first constituent.
A third property which is crucial for understanding semantic change in patterns of word formation is the relational nature of the semantics of such patterns. Contrary to monomorphemic words, the semantics of patterns of word formation always presents a relation between a constant and a variable (in the case of derivation) or between two variables (in the case of compounding). Thus, agent nouns have the meaning 'a person who Vs' and noun-noun compounds, as we have seen, simply mean 'an [N.sub.2] which has something to do with [N.sub.1]' at the level of the pattern. But since in some words the abstract word formation meaning is intimately tied up with the word meaning, the latter may provoke an outright change in the semantic relation (cf., e.g., the mechanism of "regrouping" in Section 188.8.131.52) or at least a shift of features in the semantic representation (cf., e.g., the mechanism of "irradiation" in Section 3.2.1).
The dynamic nature of the lexicon, the existence of a semantic surplus, and the relational nature of word formation meaning, together with the general properties of human communication mentioned at the beginning, are thus the main factors responsible for the pervasiveness of reinterpretation in word formation. In the following discussion, I will argue, in the spirit of Jaberg's hypothesis, that the process of reinterpretation may be divided into two steps: the semantic change of individual words (lexical semantic change) and the subsequent rearrangement of formal and semantic elements.
3.1. Lexical semantic change
The study of lexical semantic change has been witnessing a revival over the last two decades, especially within cognitive semantics, culminating in the comprehensive typology of Blank (1997), which will serve as the basis for the following discussion. In Chapter V of his excellent monograph (cf. Blank 1997: 157-344), Blank distinguishes the following eleven (14) mechanisms of semantic change ("Verfahren des Bedeutungswandels"):
1. Metapher (metaphor)
2. Bedeutungserweiterung (extension of meaning)
3. Bedeutungsverengung (narrowing of meaning)
4. Kohyponymische Ubertragung (cohyponymic transfer)
5. Volksetymologie (folk etymology)
6. Antiphrasis (antiphrasis)
7. Auto-Antonymie (auto-antonymy)
8. Auto-Konverse (auto-converseness)
9. Analogischer Bedeutungswandel (analogical semantic change)
10. Metonymie (metonymy)
11. Ellipse (ellipsis)
Of these eleven mechanisms, those numbered 4, 6, 7, and 8 have first been introduced by Blank himself, while the others are already in Ullmann (1957), though not always with the same interpretation. The newly introduced types, according to the author, "fallen von ihrem reellen Aufkommen her kaum ins Gewicht neben den 'etablierten' Typen" (Blank 1997: 343). (15)
The question that we will try to give an answer to in this section is whether all types of lexical semantic change distinguished by Blank may give rise in principle to reinterpretation in word formation, and whether they do so to the same extent.
The answer to the first part of the question seems to be positive, with one caveat: in order for a lexical semantic change to be able to give rise to reinterpretation, it is necessary that the change in question do not blur the relational nature of the pattern. Such a process of demotivation has taken place, for example, in the change from Latin panarium 'bread basket' (derived from pan- 'bread') to French panier 'basket,' where, due to the extension of meaning, speakers are no longer aware of any relation with French pain 'bread,' since 'bread' is no longer part of the definition of panier, even though the ending is still motivated by other words in -ier referring to containers, like cendrier 'ashtray' (from cendre 'ash'), encrier 'inkpot' (from encre 'ink'), etc. Through extension of meaning, panier has thus lost its relational character, that is, it is no longer paraphrasable by 'a container that has to do with x (bread),' destroying, by the same token, the basis for a possible reinterpretation of the suffix -ier on the basis of this word.
Apart from such cases, all types of lexical semantic change distinguished by Blank seem to be able to give rise to reinterpretation in principle. The analysis of a large number of cases of semantic change in word formation in Romance and Germanic (German, English) suggests, however, that not all types do so with the same frequency. (16)
3.1.1. Metonymy. The most important type of lexical semantic change giving rise to reinterpretation, according to my sources, is metonymy, which may simply be a consequence of the fact that contiguity, according to Blank (1997: 344), is the most important relation for semantic change in general. In cases of metonymic change, connotational or denotational meanings are added to single complex words through pragmatic inferences, which may then, in a second step, be directly linked to the word formation pattern. In this way, for example, diminutives and augmentatives often acquire pejorative or meliorative overtones, which eventually may become associated with the respective patterns themselves (cf. Blank 1998: 12-13). Thus, for example, the early Romance augmentative suffix going back to the Latin relational suffix -aceus turned into an exclusively pejorative suffix in Italian (cf. appartamentaccio 'ugly flat,' from appartamento 'flat'), while in present-day peninsular Spanish it is mainly augmentative-meliorative (cf. apartamentazo 'big, beautiful flat,' from apartamento 'flat'). The starting point for these divergent semantic changes must have lain, in the Italian case, in contexts where a big exemplar of the referent of the base is generally judged to be inferior to a normal-sized or small one, and vice versa in the Spanish case. Two other examples of reinterpretation based on metonymic lexical change have already been dealt with at the beginning of Section 1, where Fleischman's account of the change of Old French -age from the meaning 'tax' to collective and action meaning has been reported. Further examples of inferences enriching the meaning of complex words will be seen in Section 3.2.1 in the history of Latin -aster.
3.1.2. Metaphor. Metaphor, though extremely important in lexical semantic change in general, only very rarely seems to give rise to reinterpretation of patterns of word formation. The reason for this striking asymmetry may probably be found in the fact that it is rarely possible to reanalyze a metaphorically-used complex word in such a way that the meaning of the base remains intact while the new semantic features may be associated with the (formal exponent of the) pattern. It is difficult to see, for example, how the meaning 'muscle' of Latin musculus, a shape-based metaphorical extension of the primary meaning 'little mouse' (from mus 'mouse'), could be divided into the features 'mouse' + 'X,' such that 'X' could come to be associated with the diminutive suffix -culus and so eventually give rise to a new meaning of this suffix.
The standard example in the literature is the supposedly metaphoric extension of agent nouns to designate instruments. One of the first statements of this type of reinterpretation is the following one from MeyerLubke's (1890) Italian grammar:
Auf Grund einer oft eintretenden Metapher kann das Werkzeug, mit welchem eine Handlung ausgefuhrt wird, als der Trager oder als der Ausfuhrende, also personlich gedacht werden, und so konnen mit den Suffixen, die eigentlich lebenden Personen zukommen, auch Sachbezeichnungen geschaffen warden. (17) (MeyerLubke 1890: 272)
It is not quite clear as to how exactly Meyer-Lubke imagined this passage from the agentive to the instrumental use of the Italian suffix -tore. If he had in mind a Jabergian scenario, we would have to hypothesize that in a first step some individual agent nouns were used metaphorically as instrument nouns, and that in a second step the instrumental meaning was attached directly to the pattern itself. The passage quoted, however, also seems to allow a different reading, where the act of metaphorization would have occurred at the very moment of the creation of the first instrumental noun according to an originally agentive pattern. In this latter scenario, which would correspond to what I will call "approximation at the pattern level" in Section 4.1, there need not have been at any moment in the history of Italian a set of nouns presenting at the same time an agentive and an instrumental reading. Though, thus, the empirical predictions of the two possible analyses are quite distinct, there is no way at present to take a principled decision on this matter, since neither MeyerLubke nor other pertinent work on Italian word formation (cf. especially Kremer 1996) is detailed enough to present crucial evidence. (18)
3.1.3. Absorption (ellipsis). A very important type of lexical semantic change in Romance which gave rise to the reinterpretation of patterns of word formation is absorption. The term "absorption" was introduced into linguistics by Darmesteter (1904 : 54-60), who described the process in the following terms: "Il arrive [...] que de deux mots primitivement associes l'un est supprime Cette ablation fait que le terme qui reste [... ] en absorbe le sens." (19) Darmesteter's notion of absorption resembles the somewhat fuzzy notion of contagion in Breal (1883:133 140). In Breal (1924 : 205) contagion is defined as "phenomene [...] qui a pour effet de communiquer a un mot le sens de son entourage." (20) Absorption, accordingly, may be defined as a process where one linguistic unit takes over--"absorbs"--the meaning of another unit which has disappeared through ellipsis.
Thus, for example, Latin ferrarius, originally a relational adjective with the meaning 'related to iron' derived from ferr- 'iron,' absorbed the meaning of faber 'worker, esp. smith' when the term faber ferrarius was reduced to ferrarius through ellipsis. The absorbed element 'worker' was then later directly associated with the suffix -arius, which in this way acquired the new meaning 'worker' in addition to its original relational meaning. The mechanism of absorption has been of great importance in the history of Latin and Romance due to the abundant use of relational suffixes in Latin (cf. Ludtke 1995): thus, the Italian instrumental suffix -ale of ditale 'thimble' (from dito 'finger') and similar formations goes back to the Latin relational suffix -alis, the Spanish locative suffix -ar of melonar 'melonfield' (from melon 'melon'), and similar formations to Latin -aris, the Italian locative suffix -ile of canile 'kennel' (from cane 'dog'), and similar formations to Latin -ilis, Old French -age in the meaning 'tax,' as we have already seen in Section 1, to Latin -aticus, to mention but a few well-known cases. It seems plausible to assume that the more parsimonious use of such suffixes in Germanic (21) might be responsible for the fact that absorption seems to play a minor role in semantic change in Germanic word formation.
But Germanic, on the other hand, possesses a type of absorption absent from Romance, which is due to the tendency, repeatedly observed by students of German word formation, to drop the middle part of overly long compounds, whose meaning may then be absorbed by the first constituent of these compounds. Thus, for example, Matussek (1994) shows that Atom 'atom' as a first constituent of compounds has developed the meaning 'nuclear power station,' probably through the ellipsis of Kraftwerk 'power station' in compounds like Atomkraftwerksgegner 'antinuclear activist,' lit. 'atom power station opponent.' In the same vein, an Atomgegner 'antinuclear activist' is not opposed to atoms, but to nuclear power stations, just as an Atomminister is responsible not of atoms in general but of nuclear power stations only. We thus see that the frequency of certain types of reinterpretation is dependent on the presence or absence of certain structural features in the languages in question (importance of relational adjectives, compounding, etc.).
3.1.4. Folk etymology. The fourth type of lexical semantic change which I will mention here because of its great importance for semantic change in word formation is folk etymology. Folk etymology--the term was loan translated by M. Muller in 1864 from German Volksetymologie, launched in 1852 by E. Forstemann (cf. Hasenkamp 2002: 592)--is traditionally defined as secondary motivation of an unmotivated or no longer motivated word: Bussmann (2002: 741), for example, says that "[d]urch diesen sprachhistorischen Prozess werden unverstandliche Worter (sekundar) motiviert," (22) and Olschansky (1996: 107), in her fundamental monograph, also considers the input of folk etymology to be "ein synchron isoliertes und als solches unmotiviertes Wort bzw. eine solche Wortkonstituente." (23) Two phenomena in the semantic development of word formation, I think, may be ascribed to this mechanism of change.
184.108.40.206. Adaptation. The first phenomenon will be called "adaptation," adopting a term coined by the Viennese sanscritist Alfred Ludwig (cf. Ludwig 1873; Delbruck 1884:66-73) and defined in the following way by M. Bloomfield (1891: 1):
The term adaptation is used here to designate the infusion with some definite grammatical or lexical value, of a formal element originally either devoid of any special functional value, or possessed of a value which has faded out so completely as to make this infusion possible. (24)
A perfect example of adaptation in word formation is provided by Jaberg's (1965) study on ordinal numbers and fractions in Romance. As Jaberg shows, several western Romance languages, especially in their older stages, have adapted the Latin distributive suffix -(e)nus (cf. noveni 'nine ... each,' etc.) as a suffix for the formation of ordinal numbers (cf. Spanish noveno 'ninth,' etc.). With respect to the reasons for this surprising reinterpretation, Jaberg notes:
Zwei Momente haben, glaube ich, diese Entwicklung befordert: -ENUS ist, weil seine eigentliche Funktion vulgarsprachlich mit syntaktischen Mitteln wiedergegeben wird, uberflussig, sozusagen beschaftigungslos geworden und kann nun fur andere Funktionen verwendet werden [...]. Dazu kommt nun noch ein anderes. Novena ist in der ganzen katholischen Christenheit seit dem Mittelalter ein vielgebrauchtes Wort fur eine neuntagige Andacht, sei es zu Ehren eines Heiligen oder zur Erinnerung an einen Verstorbenen (vgl. Du Cange, s.v.). Sollte nicht dieses kirchlich ausgezeichnete Wort dem Suffix -ENUS ein besonderes Relief gegeben und seine Verwendung in der Ordinalzahlenreihe befordert haben, dies um so mehr, als das Wort auch die Trauerfeier am neunten Tag nach dem Tode bezeichnen konnte (vgl. Levy, Prov. Suppl. Wb.) und die Neunzahl in Recht, Sitte und Brauch stets eine grosse Rolle spielte? (25) (Jaberg 1965: 167)
220.127.116.11. Regrouping. The second phenomenon I will call "regrouping," translating Leumann's (1973 ) term Umgruppierung. Folk etymology, as we have seen, is traditionally defined as secondary motivation of an unmotivated or no longer motivated word. I would like to propose now to slightly extend this definition and to define folk etymology more generally as any kind of better motivation of a word, motivated or unmotivated. Such a redefinition would allow us to consider as a type of folk etymology a kind of reinterpretation that is extremely frequent in the history of word formation. A case in point would be the reinterpretation, mentioned at the beginning of this article, of OHG slafkamara as 'room for sleeping' instead of 'room for sleep.' The difference with respect to adaptation is that there can be no question of speakers having had difficulties with the interpretation of slafkamara before the reinterpretation took place. But for some reason, the verbal interpretation of the first constituent must have appeared preferable to them. Interestingly Osthoff, who first proposed this analysis in his 1878 monograph, already considered the process to have occurred "gleichsam volksetymologisch" (Osthoff 1878:15), that is, due to "a kind of folk etymology." Once the reinterpretation had been completed, the new verb-noun pattern, as pointed out by Osthoff (1878: 15), could be extended through the mechanism of proportional analogy (cf. a word like Schiessubung 'shooting practice,' whose first constituent Schiess- 'to shoot' can only be a verb, the corresponding noun being Schuss 'shot'), giving rise eventually to a highly productive pattern of verb-noun compounds. This kind of reinterpretation has been well-known since the nineteenth century at least. In the older literature (cf., e.g., Darmesteter 1877; Collin 1918; but also Baldinger 1950) it is normally referred to as "(false) analogy," a misleading expression since what is irregular in such formations is not the analogy itself but the reinterpretation preceding the analogical extension of the new pattern. Leumann's term, therefore, better describes what is really happening in this process.
3.1.5. Other mechanisms. Metonymy, absorption, folk etymology (adaptation and, especially, regrouping) and, to a limited extent, metaphor have thus been found to be the most important types of lexical semantic change which may lead to semantic change in patterns of word formation. As already stated in Section 3.1, none of the remaining seven mechanisms distinguished by Blank (1997) should be excluded in principle from being able to give rise to new semantic patterns. For some, especially extension and narrowing of meaning, the reason for the absence may be that, as we have seen, they generally blur the relational nature of the semantics of complex words and so destroy a necessary condition for reinterpretation. With respect to the four rare mechanisms newly introduced by Blank, viz. cohyponymic transfer, antiphrasis, auto-antonomy, and auto-converseness, the absence of pertinent cases may simply be due to their great rareness; nothing, for example, should prevent in principle a diminutive pattern from turning into an augmentative one through extended antiphrastic use of diminutives, followed by a reinterpretation (e.g. by younger generations) that would directly associate the contextually construed augmentative sense with the morphological pattern itself.
As far as analogical semantic change is concerned, it certainly also gave rise to new patterns of word formation. Here the problem is that this mechanism is extremely difficult to pin down. In analogical semantic change, according to Blank (1997: 317-323), a given relation of polysemy is transferred from one word to another. So he argues with some plausibility that the polysemy of the Italian noun gruccia, meaning both 'crutch' and, metaphorically, 'coat hanger,' has recently been transferred to stampella 'crutch' through proportional analogy. The problem is that it is normally very difficult to decide whether the new word, in our case stampella 'coat hanger,' is really the result of proportional analogy and not rather of a parallel, independent semantic extension, in our case a metaphor that would look at a coat hanger as a kind of crutch. For the time being, I am not aware of any clear case where a new pattern should have arisen on the basis of analogical semantic change, but it is easy to imagine how this could happen. The meaning 'instrument,' for example, could be analogically transferred from a complex word which displays an agent/instrument polysemy to an agent noun formed according to a purely agentive pattern, and this new instrumental formation could then become the starting point of a new instrumental pattern.
3.2. Rearrangement of semantic and formal elements
We now come to the second step in the process of reinterpretation: once one or more complex words have undergone semantic change, some features of the new meaning may or, in some cases, even must become attached to the pattern itself. I would like to argue that there are, in fact, two different types of rearrangement to be distinguished, which will be called "irradiation" and "restructuring."
3.2.1. Irradiation. The term "irradiation" (26) was introduced into linguistics by Breal (1892: 20) with the following definition: "un suffixe de signification generale et vague a l'air de prendre une acception speciale et caracterisee, grace au sens du mot auquel il est joint." (27) In Breal's understanding, thus, the new meaning of a suffix in some cases "irradiates" from the stem to the suffix. Though I will retain Breal's term, I will not adopt his definition but rather that of Serbat (1983: 534), (28) who rightly points out that it is always the word meaning which induces the semantic change: "c'est le signifie D [i.e. the word meaning] qui bouscule les limites entre signifie S [i.e. the meaning of the pattern, or word formation meaning] et le signifie B [i.e. the meaning of the base], et qui opere un transfert semantique au benefice de S." (29) Irradiation may thus be defined as the transfer of a "floating" semantic feature, that is, a feature which has no immediate counterpart on the formal side, from the word meaning to the word formation meaning.
The transferred floating semantic feature of the word meaning may of course ultimately come from the base. A good example of this is Bourquin's (1979) analysis of the emergence of the "ethnic" variant of the French suffix -itude. Up to the middle of the twentieth century, French-itude was essentially an unproductive suffix present in de-adjectival abstract nouns like exactitude 'exactness,' etc. Things changed when, around 1935, Leopold Senghor launched the term negritude, formed from negre 'negro, black' after the model of servitude 'servitude,' in order to refer to the peculiar traits of the black race and its social, cultural, and political situation. The word immediately established itself in intellectual circles, sparking off a whole series of similar neologisms derived mainly from ethnic nouns or adjectives and designating the peculiar traits of the social group referred to as well as its emancipatory aspirations (cf. corsitude, from corse 'Corsican,' feminitude, from femin(in) 'feminine,' etc.). In this case, the special connotations and associations (inferior status, etc.) of the new suffix -itude are clearly traceable to negre, the base of negritude. The example would thus be compatible with Breal's more restrictive definition of irradiation.
But there are also clear cases where the semantic change cannot have had its ultimate origin in the base. One convincing reconstruction of the change of Latin -aster from a suffix expressing resemblance to a pejorative suffix goes as follows (cf. Mutz 2000: 179-184): nouns like oleaster 'wild olive tree' (from olea 'olive tree') or filiaster 'stepson' (from filius 'son') originally simply referred to something similar to an olive tree or somebody similar to a son, while the pejorative connotation of -aster seems to have arisen as a consequence of the negative opinions current among Romans with respect to wild olive trees and stepsons. The semantics of oleaster and filiaster has thus been enriched through inferences typical of metonymic change before the new features were passed on to the suffix via irradiation. Note that here, the irradiation of the pejorative connotation cannot have originated in the bases, since both olea and filius had positive connotations in Roman society.
The meaning element that irradiates from the word meaning may also have its ultimate origin in a word dropped through the process of ellipsis. We have already seen an example of this in Section 3.1.1, viz. faber ferrarius > ferrarius, where the meaning 'worker' first passed from faber to ferrarius and eventually came to be associated directly with the suffix -arius, as the Romance languages show (cf., e.g., Spanish herrero 'smith,' from hierro 'iron').
A fourth source where semantic elements that irradiate from the word meaning to the pattern of word formation may ultimately come from seems to be blending. If telegram is blended with cable to give cablegram, the meaning of telegram may become definitively associated with -gram and so give rise to a new series of words with a second element -gram with the meaning 'telegram.'
3.2.2. Restructuring. Restructuring will be defined as change in constituent structure as a consequence of change in meaning. While irradiation may occur with a certain temporal delay, restructuring by necessity occurs simultaneously with the semantic change that triggers it. When, for example, medieval speakers of German began to prefer 'room for sleeping' to the reading 'room for sleep' as the best interpretation of slafkamara (cf. Section 1), they could not escape changing the constituent structure from noun-noun to verb-noun.
Just as the pervasiveness of reinterpretation was motivated at the beginning of Section 3 on the ground of general features of human communication and word formation, the existence of the second fundamental kind of mechanism of semantic change in word formation, viz. approximation, is also to be expected under the following two general premises. On the one hand, the coining of complex words is not understood as an algorithmic process combining a list of morphemes according to a closed set of rules, but rather as an essentially pattern-based process, where the pattern that serves as a model may lie somewhere between the two extremes of a single complex word and a relatively abstract, rule-like pattern. On the other hand, it is assumed that human communication, in order to be effective, does not require a 100% match between model and copy, pattern and neologism (cf. Gloning 1996:152). An approximation, in many cases, will suffice if the hearer is able to bridge by inference the distance between model and copy. This is especially the case if model and copy are linked by metaphor or metonymy.
Approximation will thus be defined as a process of word formation where the relation between a pattern of word formation and a neologism formed according to it is not one to one, but mediated by metaphor or metonymy. Note that metaphor or metonymy here intervene in the very act of forming a neologism according to a pattern, while in the Jabergian scenarios discussed in Section 3, metaphor and metonymy (among other mechanisms of lexical semantic change) were applied to existing lexical items prior to irradiation or restructuring (in this second case "prior" only in the logical, not the temporal sense of the word). Approximation will further be divided into two subtypes, according to whether the metaphoric or metonymic mediation is located at the level of the pattern as a whole or at the level of the base.
4.1. Approximation at the pattern level
Approximation at the pattern level may be illustrated by the sporadic temporal use of the locative prefix cis-, which we have already encountered in Section 2, described in the following terms in Marchand (1969 : 150): "the words cis-Elisabethan 1870 and cis-reformation (time) 1662 transfer the notion of place into that of time. The meaning here is 'belonging to the time after -, subsequent to -'." Note that this semantic change of the prefix cis- from its proper spatial meaning to a temporal one cannot be accounted for in Jabergian terms. It was not the case that some individual adjective of the locative type cisalpine underwent a semantic change from the realm of space to that of time--no such case is documented nor is it easy to imagine how such a change could come about with subsequent irradiation of the new temporal meaning to the prefix cis-; the temporal meaning must have arisen at the very moment of the creation of the adjectives cis-reformation and cis-Elisabethan. The speakers or writers simply used the pattern itself in a metaphoric manner, relying on the pervasive conceptual metaphor TIME-RELATIONS AS SPACE-RELATIONS.
Another clear case in point is constituted by the ludic variant of the suffix -itis. The proper meaning of this suffix is 'inflammation of x,' where 'x' designates a part of the body, as in tonsillitis, etc. In the second half of the nineteenth century (cf. Schweickard 1993), however, the suffix is beginning to be attached, with a clearly ludic intention, to words other than parts of the body in order to refer to some metaphorical "disease": fiscalitis (1903), telephonitis (1935), electionitis (1945), etc. This semantic change, again, is not amenable to the Jabergian mechanism of reinterpretation, since it was not the case that some existing medical term in -itis underwent a semantic change that eventually led to the irradiation on to the suffix itself of the new meaning 'excessive tendency to x.' Rather, we must assume that the first person to coin a ludic formation of that kind consciously overextended the pattern of nouns in -itis with the intention of assimilating some kind of excessive tendency to a disease. The gist of these formations precisely relies on the implicit metaphor EXCESSIVE TENDENCY AS DISEASE. (30) Contrary to the c/s-case seen before, the ludic variant of-itis in the meantime has become a productive pattern in most European languages. (31)
4.2. Approximation at the base level
In the cases just seen of approximation at the pattern level, the copy (neologism or pattern, if it becomes productive) and the model pertain to different semantic categories: space vs. time, inflammation vs. excessive tendency, etc. In the cases of approximation at the base level, on the contrary, model and copy as such are categorially identical, differing only in the semantic category of the base (and, possibly, the semantic relation between base and affix or the two bases, in the case of compounds). Again, this semantic change does not come about through Jabergian reinterpretation of an existing complex word, but at the very moment of the coining of the neologism. What happens here is that the speaker or writer, while forming the neologism, switches from the semantic category normally required by the base to a metaphorically or metonymically related category.
Metonymic approximation at the base level may be illustrated by Spanish -era in its use as a suffix forming designations of recipients (cf. Rainer 1993: 477-478). In its proper use, the base designates the thing(s) contained in the recipient: chequera 'checkbook' (from cheque 'cheque'), tabaquera 'tobacco tin' (from tabaco 'tobacco'), nevera 'refrigerator; cf. Am. ice box (from nieve 'snow'), etc. All actual formations and neologisms follow this pattern, except mariconera and fresquera. When, several decades ago, men's handbags came into use, they embarassingly reminded Spaniards of women's handbags and received the contemptuous name mariconera lit. 'bag (recipient) suitable for gay people' (from maricon 'gay,' a common insult). The base, in this instance, does no longer refer to the content of the recipient, as in the target pattern, but to its bearer, a metonymically related concept. This metonymic switch, on the one hand, may have been due to the ensuing ludic effect, but on the other, one should not forget that the content of a handbag is heterogeneous and thus not very salient, which could have been another reason for choosing as a base the more salient bearer of this kind of handbag, considered to be typically an effeminate person. A similar motivation is probably also responsible for the anomaly observable in fresquera, the Spanish ancestor of the refrigerator, that is, a cupboard with a grille located in a shady and droughty place of the house, which is derived from the adjective fresco 'fresh.' Again, the base does not designate the content of the recipient, but rather its function: it served to keep food fresh. The metonymic switch may again have been motivated by the heterogeneity, and hence low saliency of the content of this kind of recipient. In this first example of approximation at the base level, just as in the cis-case discussed in Section 4.1, the two neologisms remained isolated and did not spark off productive patterns of the type 'recipient typically used by x' or 'recipient serving to keep things x.' Whether this is the case or not, however, is a question concerning diffusion and not innovation. According to the theory defended in this article, innovation can happen in two ways, either via reinterpretation or via approximation, while the question of whether the neologism formed according to one of these two types of change later on acts as a leader word for the creation of a new productive pattern or not, is a different matter. (32)
It seems that approximation at the base level may also be of a metaphoric nature. A case in point is the Spanish suffix -uno forming adjectives from animal nouns: caballuno 'horse-(like)' (from caballo 'horse'), perruno 'dog-(like)' (from perro 'dog'), etc. As Malkiel (1959) has shown, this suffix was extended to human bases during the Renaissance in order to convey derogatory overtones: frailuno 'monkish' (from fraile 'monk'), lacayuno 'servile' (from lacayo 'lackey'), etc. Through the switch in semantic category (from animal to human), the speakers or writers here probably wanted to establish an implicit comparison between animals on the one side, and monks or lackeys on the other. It has been pointed out to me that a case of metaphoric approximation at the base level such as this one is not easy to distinguish from what we have called "irradiation" in Section 3.2.1. If we would like to analyze the present case as one of irradiation, we would have to say that one of the established formations, let's say perruno, has acquired negative connotations subsequently transferred on to the suffix -uno via irradiation. Neologisms such as frailuno or lacayuno then could have been formed on the model of perruno by simple proportional analogy. The decision between these two analyses is not obvious and hinges on whether one thinks that the coiners of these neologisms have had ludic intentions (implicit HUMAN AS ANIMAL metaphor) or not. The important point is to see the difference between irradiation and approximation at the base level. (33) The example further illustrates how important it may be to have very detailed information in order to establish the exact nature of a change.
As another case in point we may adduce the Spanish suffix -ezno, present in a handful of designations of young animals: lobezno 'cub' (from lobo 'wolf'), osezno 'cub' (from oso 'bear'), etc. In Old Spanish, as Yndurdin (1952) reports, this suffix was also applied to two bases referring to humans, probably with jocular or derogatory intentions, namely judezno 'Jewish child' (from judio 'Jew') and morezno 'Moorish child' (from moro 'Moor'). As one can see, the pattern remained stable, while the base was metaphorically extended to young children. It has been pointed out to me that this might just as well be a case of analogical expansion (or generalization). Since analogical expansion of the domain of a pattern of word formation in general affects semantically (and/or formally) similar bases and similarity is also crucially involved in metaphor, the interpretation may indeed remain unclear in some cases. In the present case, the fact that there is no objective similarity between Jewish and Moorish children and young animals that would not also obtain for Christian children would seem to me to speak in favor of a metaphorical account, which implies that the similarity was created purposefully by way of metaphorical approximation. That's also the way Yndurain interpreted these examples, but a final decision again is impossible at a distance of several centuries. (34)
Both metonymic and metaphoric approximation at the base level are quite common in word formation though, to the best of my knowledge, they have not yet received much attention, or even a name, in the literature. This may be due to the fact that its detection presupposes a thorough analysis of the restrictions of the patterns involved, which is yet another neglected area of studies in word formation.
By way of conclusion, let us briefly summarize the main results of this article. We started out with Jaberg's hypothesis, according to which all kinds of semantic change in word formation are due to the mechanism we have called "reinterpretation," that is, a combination of lexical semantic change in individual words plus irradiation/restructuring. We have found that this mechanism is in fact the most important, but not the only one. On the one hand, semantic change in word formation may also be due to redefinition, affecting directly the pattern itself, especially in languages for specific purposes, but such redefinitions were judged not to be directly relevant to the evaluation of Jaberg's hypothesis, since they have their place not in a theory of meaning but in a more general theory of communication. On the other hand, however, we have also found cases where the semantic change occurred at the very moment of the coining of a neologism through a mechanism which we called "approximation" and which we found to be fundamentally different from reinterpretation. Approximation, which may operate at the pattern and at the base level, effectively falsifies Jaberg's claim, but nicely fits into a holistic conception of word formation which does not require the copies to be one hundred percent identical to the models, but tolerates deviations, especially if mediated by metaphor or metonymy.
Received 19 November 2002
Revised version received
+18 June 2003
Vienna University of Economics and Business Administration
(1.) Earlier versions of this paper have been presented to audiences at the Austrian Academy of Sciences, the 10th International Morphology Meeting (Szentendre, Hungary), the universities of Tubingen, Graz, and Konstanz. For helpful comments I would like to thank, among others, Wolfgang U. Dressier, Peter Koch, Walter Waltereit, and Davide Ricca, as well as two anonymous reviewers. All remaining shortcomings, of course, are of my own responsibility. Correspondence address: Institut fur romanische Sprachen, Wirtschaftsuniversitat Wien, Nordbergstr. 15, A-1090 Vienna, Austria. E-mail: Franz.Rainer@wu-wien.ac.at.
(2.) Others would prefer the term "rule" instead of "pattern." The present article is meant to be agnostic with respect to the ongoing rule/analogy controversy in morphology.
(3.) Translation: "The starting point is constituted undoubtedly by words with, in first position, a noun which could also be interpreted as a verb-stem. Schlafkammer ['bedroom:' lit. 'sleep room'; F.R.], OHG slafkamara, is an old noun-noun compound, as the other formations with slaf- show; nevertheless, we interpret the first constituent as a verb, and the same ambiguous and undecided situation may be found with many more cases like Werktag (Engl. workday), Baustein ['building block,' the first constituent meaning both 'construction' and 'to build'], Ruhebett ['bed;' lit. 'rest bed', the first constituent meaning both 'rest' and 'to rest'], reisefertig ['ready (to leave),' the first constituent meaning both 'journey' and 'to leave']. This evolution began in Old High German."
(4.) Slafkamara, of course, only serves an illustrative purpose here. Osthoff (1878: 12-15) claims to have detected two gothic verb-noun compounds in Wulfila's translation of the bible, which would be the oldest extant examples, if his claim is correct.
(5.) Since Jaberg reviewed a study on the semantic development of the Latin suffix -mentum in French, his considerations are restricted to suffixation, but he probably meant it to be valid for word formation in general. In the present article, we will deal with all kinds of word formation.
(6.) Translation: "What we call the meaning of a suffix is not an independent concept; it is only the constant modification of different fundamental concepts. For linguistic consciousness stem and suffix form one single concept."
(7.) Translation: "The suffix alone thus cannot change its meaning, it only changes it together with the stem. But as soon as a number of words formed by the same suffix change their meaning in the same direction, the function of the suffix also changes; we say: it has received a new 'meaning.' This is manifest from the fact that new derivatives are formed with the suffix in its new meaning. From what we have said, the following methodological claims may be derived: 1) the semantic change of a suffix has to be explained on the basis of the semantic change of individual words; 2) there has to be made a principled distinction between semantic change of single words and neologisms on the basis of a new meaning of a suffix."
(8.) Jaberg's theory is explicitly endorsed by Collin (1918). In the later Romance literature, I have not found any explicit references to Jaberg's review, but his views seem to be implicit in most work on semantic change in word formation by Romance scholars of the 20th century (cf., e.g., Gamillscheg 1921; Wartburg 1923; Baldinger 1950; and various works by Malkiel).
(9.) I am indebted to Arnold Leitner, an expert in metrology, for this information.
(10.) Up to this point, I have used the term "reanalysis." But since this term is used in different ways by different linguists, I have decided to avoid it altogether in the terminology of semantic change which I am going to propose. Many cases of what is commonly referred to as "reanalysis" will here be called "reinterpretation," but the terms are not coextensive. So, for example, the change from Latin roma-nus 'Roman' (from Roma 'Rome') to rom-anus, which gave rise to the suffix -anus, is commonly also considered as a case of reanalysis, but it is not a case of reinterpretation in my terminology, since no semantic change but simply a shift in morpheme boundary is involved. Reanalysis may, of course, continue to be useful as a generic term for a cluster of semantic and structural changes related in a yet to be defined way.
(11.) The latter term was first introduced in Rainer (2003).
(12.) Translation: "a ready-made mass."
(13.) Translation: "a continuous production and reproduction of the word-formational capacity."
(14.) To these eleven mechanisms, Blank adds two "secondary" mechanisms: Bedeutungsverstarkung ('strengthening of meaning') and Bedeutungsahschwachung ('weakening of meaning'). They do not alter the semantics proper of a word but rather its domain of application, for example, from expressive to neutral, from standard to archaic, from regional to standard, etc. These secondary mechanisms qualify as kinds of semantic change only if the characterization of the domain of application of a word is considered as part of its semantics.
(15.) Translation: "are marginal as far as effective frequency is concerned, compared to the 'established' types."
(16.) Estimates of the frequency of the different mechanisms are intuitive and not based on explicit statistics, since this would presuppose that every case of semantic change encountered in the main handbooks on historical word formation may be attributed to one or the other mechanism. Unfortunately, as I have already stated, many descriptions are too vague to allow us to do so with some confidence. The intuitive estimates, however, are made on a conservative basis.
(17.) Translation: "On the basis of a frequently occurring metaphor the instrument, with which an action is executed, may be thought of as the executor, that is, as a person, and in this way one may create designations of instruments with suffixes properly reserved for human beings."
(18.) Since actor and instrument show a metonymic relationship (the actor uses the instrument), one might also want to consider, as one anonymous reviewer has suggested, the possibility of a metonymical interpretation of the semantic extension of -tore. Note, however, that metonymic relationships between actor and instrument normally seem to work in the opposite direction: it is not uncommon for an actor to be referred to by the name of a characteristic instrument (cf. French violon, meaning both 'violin' and 'violinist,' and many similar examples), but the opposite extension does not seem to occur, at least with simple words. Detailed studies on the history of tore in Italian and of -dor in Spanish, undertaken after the completion of the present article (cf. Rainer 2004a, 2004b), suggest that no semantic change was involved in these two languages, and in Romance in general.
(19.) Translation: "Sometimes one of two originally linked words is suppressed. This suppression causes the remaining word to absorb its meaning."
(20.) Translation: "a phenomenon which has the effect of communicating to a word the meaning of its context."
(21.) The most common translation into German of a Latin or Romance noun phrase containing a relational adjective is a noun-noun compound. German relational adjectives, furthermore, generally pertain to the learned layer of the lexicon.
(22.) Translation: "through this process of linguistic change incomprehensible words are provided with a (secondary) motivation."
(23.) Translation: "a synchronically isolated and as such unmotivated word or a constituent of a word with these same characteristics."
(24.) This definition closely matches that of "exaptation" in Lass (1990: 80): "Exaptation [...] is the opportunistic co-optation of a feature whose origin is unrelated or only marginally related to its later use." Lass does not seem to have been aware of having reinvented adaptation.
(25.) Translation: "Two factors, I think, have supported this development: since its proper function was rendered with syntactic means in the vulgar language, -ENUS had become superfluous, unemployed so to speak, and hence can be used for other functions [...]. To this one has to add a second factor. From the beginning of the Middle Ages novena is a much-used word throughout Christendom for a prayer of nine days in honor of a saint or in memory of a deceased person (cf. Du Cange, s.v.). Should not this important ecclesiastic word have provided -ENUS with a special saliency and supported its use in the series of ordinal numbers, all the more as this word could also designate the funeral service on the ninth day after the death (cf. Levy, Prov. Suppl. Wb.) and the number nine played an important role in law, customs, and traditions?"
(26.) Different metaphors used in the literature to refer to the same or a similar phenomenon are Abfarben 'coming off (said of colors)' (cf., e.g., Spitzer 1910: 134; Baldinger 1950: 119-121) or tainting of suffixes (Jespersen 1922: 388).
(27.) Translation: "a suffix with a general and vague semantics seems to acquire a more specific meaning due to the meaning of the word it is joined to."
(28.) Cf. also Brachet (1999).
(29.) Translation: "It is meaning D [i.e. the word meaning] which overturns the limits between meaning S [i.e. the meaning of the pattern, or word formation meaning] and meaning B [i.e. the meaning of the base], and which causes a semantic transfer in favor of S."
(30.) A similar metaphorical extension, by the way, had already taken place about one hundred years before with words in -mania (cf. Hofler 1972). This metaphorical use of -mania may well have facilitated the metaphorical use of -itis.
(31.) I have not yet found any clear example of metonymic approximation at the pattern level. Whether this is a lacuna in my documentation or has some deeper significance, I cannot say at present.
(32.) A detailed discussion of a case where several cases of approximation at the base level have turned into productive subpatterns, viz. the Spanish suffix -azo, may be found in Rainer (2003).
(33.) In lexical semantic change, it is also impossible sometimes to decide which mechanism of change has been applied in some particular case. So we can be sure that French bougogne 'Burgundy (wine)' is either the result of a metonymical extension of the name of the homonymous region of France or of an ellipsis of vin de in vin de Bourgogne 'wine from Burgundy," but without more detailed historical information, we cannot be sure which mechanism has been at work. Such a state of indecision must not be viewed, of course, as proof that metonymy and ellipsis are one and the same thing!
(34.) Even a third possible analysis might be proposed. The above analysis presupposes that children and young animals form a unitary category, which is of course true from a biological perspective. Natural languages, at least Spanish and other European languages, however, treat them as distinct categories (cf. Spanish nino, crio 'child' vs. cria, cachorro 'young animal'). If -ezno were given the more specific meaning 'young animal,' the application to human bases could be seen as a case of metaphorical approximation at the pattern level.
Anttila, Raimo (1989). Historical and Comparative Linguistics, 2nd ed. Amsterdam: John Benjamins.
Baldinger, Kurt (1950). Kollektivsuffixe und Kollektivhegriff Ein Beitrag zur Bedeutungslehre im Franzosischen mit Berucksichtigung der Mundarten. Berlin: Akademie-Verlag.
Blank, Andreas (1997). Prinzipien des lexikalischen Bedeutungswandels am Beispiel der romanischen Sprachen. Tubingen: Niemeyer.
--(1998). Kognitive italienische Wortbildungslehre. Italienische Studien 19, 5-27.
--; and Koch, Peter (eds.) (1999). Historical Semantics and Cognition. Berlin and New York: Mouton de Gruyter.
Bloomfield, Leonard (1984 ). Language. Chicago: University of Chicago Press.
Bloomfield, Maurice (1891). On Adaptation of Suffixes in Congeneric Classes of Substantives. Baltimore, MD: Johns Hopkins Press.
Bourquin, Jacques (1979). Remarques sur la formation neologique recente de substantifs en-itude derives ou non. In Neologie el lexicologie. Hommage a Louis Guilbert, Rosine Adda et al. (eds.), 47-66. Paris: Larousse.
Brachet, Jean-Paul (1999). Les preverbes ab-, de-, ex- du latin: etude linguistique. Villeneuve d'Ascq: Presses Universitaires du Septentrion.
Breal, Michel (1883). Les lois intellectuelles du langage. Fragment de semantique. L'Annuaire de l'association pour l'encouragement des etudes grecques en France, 132142.
--(1892). De l'irradiation grammaticale. Memoires de la Societe de linguistique de Paris 7, 20-23.
--(1924 ). Essai de semantique, 7th ed. Paris: Hachette.
Bussmann, Hadumod (2002). Lexikon der Sprachwissenschaft, 3rd ed. Stuttgart: Kroner.
Collin, Carl S. R. (1918). Etude sur le developpement de sens du suffixe -ata. Lund: Lindstedt.
Corbin, Danielle; and Paul, Jerome (2000). Apercus sur la creativite morphologique dans la terminologie de la chimie. La banque des mots 60, 51-68.
Darmesteter, Arsene (1877). De la creation actuelle de mots nouveaux dans la langue franraise et des lois qui la regissent. Paris: Vieweg.
--(1904 ). La vie des mots etudiee dans leurs significations, 7th ed. Paris: Delagrave.
Delbruck, Bertold (1884). Einfuhrung in das Sprachstudium, 2nd ed. Leipzig: Breitkopf & Hartel.
Fisiak, Jacek (ed.) (1985). Historical Semantics, Historical Word-Formation. Berlin and New York: Mouton de Gruyter.
Fleischman, Suzanne (1977). Cultural and Linguistic Factors in Word Formation. An Integrated Approach to the Development of the Suffix -age. Berkeley, CA: University of California Press.
Gamillscheg, Ernst (1921). Grundzuge der galloromanischen Wortbildung. In Beitrage zur romanischen Wortbildungslehre, Ernst Gamillscheg and Leo Spitzer (eds.), 1-80. Geneva: Olschki.
Gamillscheg, Ernst; and Spitzer, Leo (1921). Beitrage zur romanischen Wortbildungslehre. Geneva: Olschiki.
Gloning, Thomas (1996). Bedeutung, Gebrauch und sprachliche Handlung. Tubingen: Niemeyer.
Hasenkamp, Wiltrud (2002). Fr. etymologie populaire 'Volksetymologie' und seine Terminusgeschichte. Zeitschrift fur romanische Philologie 118, 587-597.
Henzen, Walter (1965 ). Deutsche Wortbildung. 3rd ed. Tubingen: Niemeyer.
Hock, Hans H. (1991). Principles of Historical Linguistics. 2nd ed. Berlin and New York: Mouton de Gruyter.
Hofler, Manfred (1972). Zur Integration der neulateinischen Kompositionsweise im Franzosischen. Tubingen: Niemeyer.
Humboldt, Wilhelm von (1836). Uber die Verschiedenheit des menschlichen Sprachbaus und ihren Einfluss auf die geistige Entwicklung des Menschengeschleehts. Berlin: Schneider.
Jaberg, Karl (1905). Review of Roediger 1904. Archly fur das Studium der neueren Sprachen und Literaturen 114, 458-462.
--(1965). Ordinal- und Bruchzahlworter. In Karl Jaberg, Sprachwissenschaftliche Forschungen und Erlebnisse. Neue Folge, S. Heinimann (ed.), 160-176. Bern: Francke.
Jespersen, Otto (1922). Language. Its Nature, Development and Origin. London: Allen & Unwin.
Jurafsky, Dan (1996). Universal tendencies in the semantics of the diminutive. Language 72, 533-578.
Keller, Rudi (1995). Zeichentheorie. Zu einer Theorie semiotischen Wissens. Tubingen and Basel: Francke.
Kremer, Robert (1996). Die Werkzeugbezeichnungen im Italienischen. Bonn: Romanistischer Verlag.
Lass, Roger (1990). How to do things with junk: exaptation in language evolution. Journal of Linguistics 26, 79-102.
Lehmann, Winfred P. (1992). Historical Linguistics, 3rd ed. London and New York: Routledge.
Leumann, Manu (1973 ). Gruppierung und Funktion der Wortbildungssuffixe des Lateins. In Probleme der lateinischen Grammatik, Klaus Strunk (ed.), 131-161. Darmstadt: Wissenschaftliche Buchgesellschaft.
--(1963). Lateinische Laut- und Formenlehre. Munich: Beck.
Ludtke, Jens (1995). Grundzuge der Entwicklung der Relationsadjektive vom Latein zum Romanischen. In Konvergenz und Divergenz in den romanischen Sprachen. Romanistisches Kolloquium VIII, Wolfgang Dahmen et al. (eds.), 138-150. Tubingen: Narr.
Ludwig, Alfred (1873). Agglutination oder Adaptation? Eine sprachwissenschaftliche Streitfrage. Prague: Calve.
Malkiel, Yakov (1959). Nuevas aportaciones para el estudio del sufijo -uno. Nueva Revista de Filologia Hispanica 13, 241-290.
--(1966). Genetic analysis of word-formation. In Current Trends in Linguistics. Vol. 3: Theoretical Foundations, Thomas A. Sebeok (ed.), 305-364. The Hague: Mouton.
Marchand, Hans (1969 ). The Categories and Types of Present-Day English Word-Formation, 2nd ed. Munich: Beck.
Marle, Jaap van (1990). Rule-creating creativity: analogy as a synchronic morphological process. In Contemporary Morphology, Wolfgang U. Dressier et al. (eds.), 267-273. Berlin and New York: Mouton de Gruyter.
Matussek, Magdalena (1994). Wortneubildung im Text. Hamburg: Buske.
Meyer-Lubke, Wilhelm (1890). Italienische Grammatik. Leipzig: Reisland.
--(1966 ). Historische Grammatik der franzosischen Sprache. Zweiter Teil: Wortbildungslehre. Heidelberg: Winter.
Mutz, Katrin (2000). Die italienischen Modifikationssuffixe. Synchronie und Diachronie. Frankfurt: Lang.
Nyrop, Kr. (1908). Grammaire historique de la langue francaise. Tome troisieme. Copenhagen: Nordisk Forlag.
Olschansky, Heike (1996). Volksetymologie. Tubingen: Niemeyer.
Osthoff, Hermann (1878). Das Verbum in der nominalcomposition im deutschen, griechischen, slavischen und romanischen. Jena: Costenoble.
Paul, Hermann (1975 ). Prinzipien der Sprachgeschichte, 5th ed. Tubingen: Niemeyer.
Rainer, Franz (1993). Spanische Wortbildungslehre. Tubingen: Niemeyer.
--(2003). Semantic fragmentation in word-formation: the case of Spanish -azo. In Explorations in Seamless Morphology, Rajendra Singh and Stanley Starosta (eds.), 197-211. New Dehli: Sage.
--(2004a). Del nombre de agente al nombre de instrumento en espanol: [??]como y cuando? Iberoromania 59, 97-122.
--(2004b). L'origine dei nomi di strumento italiani in -tore. In Analecta homini universali dicata. Arbeiten zur Indogermanistik, Linguistik, Philologie, Politik, Musik und Dichtung. Festschrift fur Oswald Panagl zum 65. Geburtstag, vol. 1, Thomas Krisch, Thomas Lindner, and Ulrich Muller (eds.), 399-424. Stuttgart: Akademischer Verlag Hans-Dieter Heinz.
Roediger, Max (1904). Die Bedeutung des Suffixes -ment. Doctoral dissertation, Berlin.
Schweickard, Wolfgang (1993). Zum ubertragenen Gebrauch des Suffixes -ite im Italienischen und in anderen Sprachen. Zeitschrift fur romanische Philologie 109, 136147.
Serbat, Guy (1983). Turibulum. Esquisse d'une theorie sur le signifie des suffixes de derivation. In Hommages a Robert Schilling. Hubert Zehnacker and Gustave Hentz (eds.), 525-536. Paris: Les Belles Lettres.
Spitzer, Leo (1910). Die Wortbildung als stilistisches Mittel exemplifiziert an Rabelais. Halle: Niemeyer.
Sturtevant, E. H. (1917). Linguistic Change. Chicago: University of Chicago Press.
Traugott, Elisabeth C.; and Dasher, Richard (2002). Regularity in Semantic Change. Cambridge: Cambridge University Press.
Ullmann, Stephen (1957). Principles of Semantics, 2nd ed. Oxford: Blackwell.
Wartburg, Walther von (1923). Review of Gamillscheg/Spitzer 1921. Zeitschrift fur romanische Philologie 43, 109-115.
Yndurain, Francisco (1952). Sobre el sufijo -ezno. Archivo de Filologia Aragonesa 4, 195200.