Dynamic aspects of German -er-nominals: a probe into the interrelation of language change and language acquisition *.
The discussion of the interrelation of language change and language acquisition, which has been going on for more than a hundred years, has yielded a number of theoretical approaches but still lacks empirical substantiation. Theoretical positions range from parameter setting approaches that view children as the ultimate source of language change, to those approaches that stress correlations or similarities between language change and language acquisition without necessarily assuming a causal connection between them. Empirical studies that try to settle the issue are very rare, mainly for methodological reasons. In our article, we report on our research on the development of German -er-nominals. There are three tendencies to be observed in language change and language acquisition, namely a tendency toward more multilexemic versus monolexemic bases of the -er-suffix, a tendency toward more verbal versus nominal bases, and a tendency toward more PERSON concepts versus OBJECT concepts expressed by the -er-suffix. These tendencies have been corroborated on the basis of a corpus study of German newspapers between 1609 and 2000, a diary study of a monolingual German child (2;0-4;11), and four elicitation experiments conducted with 56 German preschool children between 3;0 and 6;5. The tendencies observed in the development of -er-nominals are discussed in terms of their morphological productivity that is constituted by a number of language internal and language external factors. It turns out that children strive for adaptation to the adult patterns of word formation that have developed historically.
The interrelation of language change and language acquisition has been under discussion for more than a hundred years. Beginning with Haeckel's (1866) work Generelle Morphologie der Organismen, it was claimed that ontogeny is nothing but a short and rapid recapitulation of phylogeny determined by the laws of heredity and adaptation. The transfer of this view to language was refuted by researchers such as Guillaume (1973 ), who stated "that it is not necessary to believe in some mysterious internal necessity which causes the individual learning language to go via the tortuous route of history." What is found, however, are similarities between language history and language acquisition that are due to "certain logical and psychological necessities for assimilation in a complex organism." This view is still popular in many evolutionist accounts of language change such as Haspelmath (1999) and Croft (2000).
It was Paul (1975 ) in Prinzipien der Sprachgeschichte who viewed children as the main source of language change, as becomes clear from his famous dictum: "Es liegt auf der Hand, dass die Vorgange bei der Spracherlernung yon der allerhochsten Wichtigkeit fur die Erklarung der Veranderung des Sprachusus sind, dass sic die wichtigste Ursacbe fur diese Veranderungen abgeben." What Paul had in mind was sound change through "abweichende Neuerzeugung" by children. The basic idea that children learn the adult grammar only imperfectly, thereby modifying the grammar constantly, was taken up in early generative work on phonology (Halle 1962; Kiparsky 1965) and was finally transferred to the acquisition of syntax (Lightfoot 1979, 1991, 1999).
Clark and Roberts (1993) combine an evolutionist approach with a parameter setting approach. Language change is regarded as a special case of language acquisition. If all children acquired their mother tongue correctly, languages would never change. For Clark and Roberts (1993), language learning is a kind of natural selection. The child's hypotheses about language are "bit strings" which are more or less "fit" with regard to their input data. Hypotheses that analyze the input in the best possible way have a selectional advantage. If, in a certain acquisitional context parameter, values cannot be fixed uniquely by the child, parametric change can occur. In this case, principles of universal grammar like the subset principle are activated.
However, as functionalists like Romaine (1989), Kerswill (1996), and Croft (2000) have pointed out, such an approach focuses on the actuation of a change and neglects its promulgation. To what extent children may be considered promulgators is far from clear. They do not have the social status to spread a change in society, and they are not only influenced by their parents, but also by their grandparents, young adults, elder siblings, and peers. It is unclear how a generation of children should ever converge on a change.
The main objection to an imperfect learning approach to language change is the observation that there is no different parameter fixing at all, because children "unlearn" their errors and converge on their target grammar in the course of acquisition (Bowerman 1988; Romaine 1989). Ungrammatical, overgeneralized forms like *sie gehte instead of sie ging or *she goed instead of she went are eliminated and have no chance of being passed on to the adult system. Romaine (1989: 218) is right in remarking that "in normal communities ... adults act as brakes on the innovations produced by children so that analogical and other deviant forms likefoots get corrected and do not persist."
A more cautious approach to the interrelation of language change and language acquisition should therefore proceed from the observation of similarities. A thorough explanation of these similarities is by no means a trivial task and has hardly ever been undertaken (among the rare studies are Baron 1977 and Ravid 1995). The history of many language phenomena is virtually unknown; there are language internal, cognitive, as well as social-interactional factors involved, and a simple derivation from maxims such as the ones proposed by Slobin (1977) (i.e. Be clear! Be processible! Be quick and easy! Be expressive!) will not do.
In this article, we concentrate on the change and acquisition of German -er-nominals. After a short outline of the properties of -er-nominals in history and present-day German (Section 2), we show that there is evidence for certain structural and semantic developments that call for explanation and empirical substantiation in diachrony and ontogeny (Section 3). In Section 4, we report on our findings on the historical development of the -er-nominals in New High German (NHG) and acquisitional data of a diary study and elicitation tasks. We conclude our analyses by arguing that the similarities we found can be analyzed in terms of the notion of morphological productivity (Section 5).
2. Outline of -er-nominals in German
2.1. History of -er-nominals in German
Historically, the -er-suffix can be traced back to two different origins: first the Latin suffix -arius, which generated PERSON-denoting nouns out of nominal bases, and second, the Germanic suffix *-warja, which created names of origins out of geographic names. (1) Even though the contrary is often assumed, the Latin suffix -arius was not restricted to nominal bases. An estimated 4% of Latin -arius-nominals had verbal bases (Weinreich 1971).
(1) Latin: -arius > OHG -ari > MHG -oere/-ere > NHG -er Germanic: *-warja
Due to phonological processes and functional proximity, both suffixes had fused even before the Old High German (OHG) period. Today, the properties of both suffixes can still be perceived in NHG -er-nominalization.
One may conclude that the semantic variety of the NHG -er-suffix and the range of bases it may take arise from its inherited properties. A suffix that from the start was restricted neither to a narrow semantical concept nor to a single morphological base facilitates the attribution of new concepts and bases or the loss of old ones; in short, it encourages language change to occur.
2.2. -er-nominals in present-day German
In present-day German, the -er-suffix takes verbs and nouns as its base. There are also other possible bases, such as numeral bases (e.g. Drei+er 'a bus with the number three (three+er)'), but we will restrict ourselves to the discussion of verbal and nominal bases in the following. We assume that deverbal -er-nominals are connected with the global concepts PERSON, OBJECT, and EVENT, whereas denominal -er-nominals are connected with the concepts PERSON and OBJECT only. (2) The concept labels roughly correspond to the traditional morphological notions of agent noun, instrument noun, and resultative noun, respectively; however, we prefer to speak of concepts to signal that denominal -er-nominals are included.
There are not only -er-nominals with monolexemic bases, but also with multilexemic ones (Mahdrescher 'combine harvester' or Handballer 'handball player'). Second, ambiguities are often to be found as in Piepser 'cheeper/bleeper' that may be assigned to the concepts of PERSON (someone who is cheeping), OBJECT (something that is cheeping/ bleeping), and EVENT (result of cheeping/bleeping), respectively. Third, there are dual category cases of the kind Fischer 'fisherman,' where the category of the base cannot be determined uniquely (either fischen 'to fish' or Fisch 'fish'). It goes without saying that the conceptual categories PERSON, OBJECT, and EVENT must be further specified: for example, a Zubringer ('feeder road (towards-bringer)') is an OBJECT and, more specifically, a kind of road (Meibauer 1995b).
In the following, we will present evidence for three tendencies that can be observed in the change of -er-nominals. We will contrast data from OHG, Early New High German (ENHG), and present-day German in order to sketch the development of -er-nominals according to their base, morphological structure, and the concepts they represent.
3. Three tendencies in the change of German -er-nominals
In this section, we will briefly summarize data on -er-nominals extracted from various earlier studies that concentrated on different aspects of historical word formation. Linking these hitherto scattered sources of information allows us to distinguish three tendencies in the formation of -er-nominals.
3.1. Tendency I: from nominal bases to verbal bases
The formation of nouns with nominal bases, as opposed to verbal bases, is a structural option that is very old. Usually, it is presupposed that, in accordance with Latin, the OHG loan suffix -ari was primarily used for the formation of denominal PERSON nouns (2a). It is assumed that the function of -ari was later extended to verbal bases due to semantical and morphological analogy with the Latin -tor-suffix, which formed deverbal PERSON nouns (2b) (Weinreich 1971; Erben 1993 ).
(2) a. nominal base: Latin telon+arius OHG zol(l)an+ari 'customs collector' b. verbal base: Latin vena+tor OHG iag+ari 'hunter'
For Gothic, the claim that nominal bases outweigh verbal ones can be affirmed (Weinreich 1971). But Weinreich found significantly more deverbal -ari-nominals than denominal ones in his OHG corpus (cf. Table 2 (a)). In NHG, the proportion of deverbal -er-nominals is even higher than in OHG (cf. Table 2 (b)). Thus, we can identify the increase of verbal bases and the decrease of nominal bases as the first tendency in the development of -er-nominals.
3.2. Tendency II: from monolexemic bases to multilexemic bases
In this article, we will distinguish two degrees of complexity with regard to the base of -er-nominals: monolexemic bases consisting of only one free morpheme (and possibly one or more bound morphemes, as in [ver.sub.Af]+ [fuhr.sub.V] 'seduce' in Verfuhr+er 'seducer'), and multilexemic bases which involve two or more lexemes. Nominals that are derived from multilexemic bases are considered to be structurally more complex than those formed out of monolexemic bases. (4)
Derived nominals have never been restricted to monolexemic bases but could have multilexemic bases from early on, as can be seen from Germanic examples like wein-drugkja 'wine drinker' and arbi-numja 'heir, (inheritance taker+er)' (Hirt 1932). The OHG suffix -(e)o, which was superseded by OHG -ari in the long run, was mostly used with multilexemic bases, whereas -ari is mostly found with monolexemic bases (Weinreich 1971).
Since exact counts do not exist for present-day German (Wellmann 1975), we compensate for this gap by data available for -er-nominals with N+V base in ENHG. These "synthetic compounds" increase during the ENHG period. Brendel et al. (1997), who investigated the period from 1360 to 1490, found only 17.5% N+V+er nominals (cf. Table 3 [b]). However, the corpus of Muller (1993), which concentrates on Durer's language in the early sixteenth century, contains almost twice as many N+V+er nominals (cf. Table 3 [c]).
The OHG data in Table 3 (a) show that almost every third -er-nominal is multilexemic. The crucial question is where to trace the borderline between composition and derivation. As our main interest concerns the complexity of -er-nominals and not the distinction between derivation and composition, we opted for a strictly structural notion of complexity. Therefore, we operate with the hypothesis that monolexemic-based -er-nominals consist of only one lexeme followed by the -er-suffix. By default, any nominal with a different structure is considered to be derived from a multilexemic base. As a result, we are not forced to pinpoint whether an -er-nominal like Klavierspieler 'piano player' is a compound (Klavier+Spieler 'piano+player') or has been derived as a whole ([klavierspiel]+er '[piano play]+er'). Hence, we treat Klavierspieler and similar cases straightforwardly as multilexemic-based nominals.
The data given for multilexemic-based -er-nominals in Table 3 do not allow us to recognize a clear tendency with regard to complexity. But since Gothic nominals like bok+areis 'librarian' had in fact only monolexemic bases, we lean toward the assumption that the career of -ari started with monolexemic bases. Other historical facts like the growing need for professional names (often with the element--macher, e.g. korber [right arrow] korbmacher 'basket-maker', cf. Joeres 1995) and the process of noun incorporation in ENHG (e.g. pier preuer [right arrow] pierpreuer 'beer brewer', cf. Meibauer 1998) are likely to result in an increasing number of multilexemic-based nominals. (5) We identify this rise of multilexemic bases as the second tendency with regard to the -er-word formation pattern.
3.3. Tendency III: from PERSON nouns to OBJECT nouns
As stated above, both predecessors of the NHG -er-suffix, Latin -arius and Germanic *-warja, formed nominals denoting human beings (cf. Section 2.1). Therefore, the NHG data given in Table 4 are rather surprising, as only four out of five -er-nominals are linked with the concept of PERSON, whereas 13.3% denote OBJECTS, and another 3.5% EVENTS.
According to Henzen (1965 ), nouns denoting tools developed in accordance with agent nouns through a process called "personification." As early as in OHG, word formations expressing an OBJECT concept like hals+ari 'pillow' or zeig+ari 'forefinger' appeared. But they are still rare in ENHG. In the corpus of Brendel et al. (1997), six -er-nominals out of a sample of 510 (0.1[per thousand]) can be classified as denoting OBJECTS, whereas the word lists of Schottelius (1967 ) contain no OBJECT -er-nominal at all.
Reviewing these facts, we recognize a rising proportion of OBJECT-denoting -er-nominals and a decline of PERSON -er-nominals. We consider this change as the third major tendency in the development of -er-nominals from OHG to present-day German.
Neither for OHG nor for ENHG are -er-nominals denoting an EVENT concept reported. The Goethe corpus of Stricker (2000) gives five examples of EVENT nominals formed with the suffix -er out of a total of 1013 -er-nominals (i.e. 0.05%o). Since Wellmann (1975) reports 3.5% in NHG, we can observe a tendency towards more EVENT nominals, too. With respect to the small proportion, we will not discuss EVENT nominals in this article but focus on the development of the concepts PERSON and OBJECT.
4. An attempt at empirical substantiation
4.1. The Mainz research project on the dynamics of -er-nominals
Our research project attempts to substantiate empirically the three tendencies presented above for diachronic and ontogenetic change. Our data are collected from German newspapers, a diary study, and experiments conducted with preschool children.
The diachronic component of our project surveys four centuries of written language as represented in German newspapers starting with the year 1609, when the first newspapers worldwide, Aviso and Relation, appeared in Wolfenbuttel and Strassburg, respectively. From 1650 on, we selected measuring points every 50 years, with the last one being 2000. We chose supraregional newspapers that were of general importance for the period concerned. Consequently, newspapers from all over the German-speaking area are included in our corpus. Once the newspapers were selected, we extracted samples of about 100,000 word forms each (with a range of 98,000 up to 150,000 word forms), which were manually analyzed and fed into a database. In total, 1,031,000 word forms were scanned resulting in 2,092 different types of -er-nominals with 8,903 tokens (Scherer 2002).
Issues of the following newspapers constitute the Mainz newspaper corpus: Aviso Wolfenbuttel 1609, Wochentliche Donnerstags Zeitung including Ordinari Diengstags Zeitung and Appendix Hamburg 1650, Mercurii Relation Munich 1700, Stats= u. Gelehrte Zeitung Hamburg 1750, Allgemeine Zeitung Stuttgart 1800, Koniglich privilegierte Berlinische Zeitung yon Staats= und gelehrten Sachen Berlin 1850, Berliner Tageblatt Berlin 1900, Frankfurter Allgemeine Frankfurt/Main 1950, and Die Welt Berlin 2000.
In contrast to Aronoff (1976), we do not assume that an affix is limited to a syntactically and semantically uniquely specified class of bases. In case of the German -er-nominals listed in (3), Aronoff's (1976) unitary base hypothesis would result in having one different affix for each base class, thus distinguishing seven different homophonic -er-suffixes. In fact, we suppose that the German -er-suffix is polysemous to a certain degree and can act on different classes of bases as well. Therefore, we consider the examples given in (3) as occurrences of one single word formation pattern. Apart from deverbal (3a) and denominal -er-nominalizations (3b), with nominal bases including geographic (3c) and person names (3d), we found -er-nominals with adjectives (3e), numerals (3f), and even prepositions (3g) as a base.
(3) a. verb+er: Wanderer 'wanderer (wander+er)', Kleber 'glue (stick+er)' b. noun+er: Politiker 'politician (politics+er)', Schiffer 'mariner (ship+er)' c. geogr. name+er: Englander 'Englishman (England+er)', Champagner 'champagne (Champagne+er)' d. person name+er: Benedikter 'benedictine (Benedict+er)', Calviner 'devotee of Calvin (Calvin+er)' e. adjective+er: Barfusser 'barefooted monk (barefoot+er)' f. numeral+er: Dreissiger 'person at the age of 30 (thirty+er)', Neunziger 'nineties (ninety+er)' g. preposition+er: Gegner 'opponent (against+er)' (7)
Because names are usually considered a subclass of nouns, deonomastic nominals are generally incorporated in the class of denominal ones (e.g. Wellmann 1975; Fleischer and Barz 1995 ; Stricker 2000). As stated above, we treated denominal and deonomastic nominals in the same way, thus fusing (3b), (3c), and (3d). However, names do have specific properties that other nouns do not have (restricted inflexion, lack of article in certain contexts). Furthermore, there is a historical difference between -er-nominals generated out of nouns and toponyms. Whereas denominal -er-nominals originate in Latin -arius-formations, names of origin were derived from the Germanic suffix *-warja. Therefore, the share of names of origin in our data will be indicated on several occasions due to its large proportion and, as a consequence, its considerable influence on our results for denominal -er-nominals.
Nominals with ambiguous bases (either nouns or verbs), as in (4a), as well as nominals generated out of word groups or phrases (4b) can be found in any period. In contrast, nominals like Fanatiker 'fanatic,' created on the base of bound stems (4c) with no existing free base, first occur in 1700.
(4) a. ambiguous stem+er: Maurer 'bricklayer (to lay bricks/ wall+er)', Morder 'murderer (to murder/ murder+er)' b. phrase+er: Anrufbeantworter 'answering machine (call reply+er)', Zweimaster 'two master (two mast+er)' c. bound stem+er: Fanatiker 'fanatic (*fanatic+er)', Akademiker 'academic person (*academic+er)'
The structure of nouns like Anrufbeantworter in (4b) is still subject to discussion. There is neither a verbal compound *anrufbeantworten that could serve as a base for derivation nor an -er-nominal *Beantworter that could be used as the second constituent in a compound. This means that both possible binary analyses, [[[anruf.sub.N]+[beantwort.sub.V]].sub.V]+er and [anruf.sub.N] +[[[beantwort.sub.V]+er]].sub.N], are not fully convincing. Nonetheless, both analyses were suggested by generativist researchers. Hohle (1982) favors the first alternative, [[[anruf.sub.N]+[beantwort.sub.V]]].sub.V]+er, thus claiming the existence of bound verbal compounds like *anrufbeantwort, whereas Olsen (1986) and Leser (1990) opt for the second analysis, [anruf.sub.N]+[[[beantwort.sub.V]+er].sub.N], suggesting a bound -er-nominal *Beantworter. On the other hand, an alternative approach, namely the nominalization of word groups or phrases, has been employed by several representatives of traditional word formation theory, for example, Fleischer and Barz (1995 ), Muller (1993), and Doring and Eichler (1996). Since we have limited ourselves to the notion of structural complexity as defined above (cf. Section 3.2), we classify Anrufbeantworter and similar formations as multilexemic-based -er-nominals independent of their genesis.
There are only few examples, for example, those presented in (5), where the -er-suffix neither effects a transposition of the word category nor of the global concept (PERSON, OBJECT, EVENT) in our corpus. The indication of male sex by adding -er is limited to the case of Witwer 'widower' (5a) in our data.
(5) a. masculine marking: Witwer 'widower (widow+er)' b. pleonastic marking: Jesuiter 'jesuit (jesuit+er)', Offizierer 'officer (officer+er)'
In our corpus, pleonastic marking of PERSON nouns with the -er-suffix as in (5b) is mainly found in the 17th century. It is attested up until the 19th century (Paul 1920; Fleischer and Barz 1995 ); in our corpus, it can be spotted in isolated and rather opaque cases until 1950. It is not evident why pleonastic marking emerged initially. We assume that there was a need to mark borrowed or unknown PERSON nouns explicitly by means of the well-established PERSON-denoting -er-suffix, and that in certain contexts, this kind of marking may still be productive. Fuhrhop (1998) reports on nonstandard -er-nominals that people created when asked to name the inhabitants of an unfamiliar geographic area (e.g. Laoter, Senegaleser instead of Laote 'Laotian,' Senegalese 'Senegalese'). Further research should not only focus on the causes of pleonastic marking, but additionally try to determine which properties of the -er-suffix encouraged its use as pleonastic PERSON-marker.
We will concentrate on the discussion of unambiguous deverbal and denominal -er-nominals in the presentation of our corpus data in Section 4.2, thus neglecting masculine (5a) and pleonastic marking (5b), as well as nominals with other than nominal or verbal bases (3e-3g), ambiguous -er-nominals (4a), and nominals with bound stems (4c). Accordingly, the total number of types diminishes to 1,844, and the total number of tokens to 7,791.
In the project concerned with language acquisition, an experiment containing production and comprehension tasks with 56 preschool children ranging from 3;0 to 6;5 was carried out. The group was subdivided into seven age groups, each comprising a time span of six months, with eight children each. The starting point of the experiment was the outcome of a diary study of a German monolingual child from 2;0 to 4;11 (Meibauer 1995a). Since the analysis of available corpora of spontaneous German child language, for example the CHILDES database, did not yield a sufficient amount of data concerning the development of the -er-nominalization, an experimental approach to data collection was necessary. The design of the experiment basically follows and improves on related studies (cf. Berko 1958; Clark and Hecht 1982; Derwing 1976; Schaner-Wolles and Dressier 1985; Windsor 1994).
The entire experiment comprises 37 test items varying with respect to their base forms and semantic concepts. The experiment is subdivided into four parts. Task I and II aim at assessing the comprehension skins of -er-nominals. Task I is a multiple choice exercise. For each test item, the children are shown four different pictures; then they are instructed to point at the corresponding ones, for example the pictures showed a person writing, a book, a pencil, and two people standing. The children are asked, "Can you show me the Schreiber (writer)?" The intended reaction was to point at the person writing, or the pencil, or both. After they have pointed at one or more of the pictures, they are asked, "Can you see another Schreiber?" If they deny this, the next set of pictures is presented to them. In Task II, the children have to explain given items, which is more demanding but gives good insight into what the children already know about the different meanings of -er-nominals. Task III focuses on the evaluation of production skills. The children are provided with nonsense bases and asked to name corresponding illustrations of fantasy creatures and objects. The question for creatures was "Der tut immer.... Wie heisst der denn dann?"--"This one is always.... What would you call him then?" and for objects, "Damit kann man.... Kannst Du Dir denken, wie das dann heisst?"--"This can be used for.... Do you know what it is called then?" Task IV tests both comprehension and production skills. In the context of some activities with toys and other objects, the children are asked to produce intended word forms or perform simple actions as a reaction to given test items. For example, they are given some balls and are asked, "Bist Du schon ein guter Ballzahler?"--"Are you already a good ball-counter?" (8)
The experiment is repeated with the aim of capturing the development of these children.
4.2. Diachronic change
In this section, we will present data from our newspaper corpus (Scherer 2002) in order to evaluate the three tendencies of -er-nominalization. All data are based on unambiguous denominal and deverbal types only, -er-nominals with other than verbal or nominal bases are excluded.
In order to confirm potential correlations statistically, a regression analysis was carried out. The following variables have been tested with regard to their correlation with time: the total number of -er-nominals (types and tokens), the number of hapax legomena, the number of deverbal and denominal -er-nominals (types), the number of monolexemic and multilexemic -er-nominals (types), the number of PERSON and OBJECT -er-nominals (types), and finally, the number of deverbal respectively denominal PERSON -er-nominals (types) and the number of deverbal respectively denominal OBJECT -er-nominals (types). As a result, all of our data proved to be correlated with time to a highly significant degree.
Because our nine subcorpora vary considerably in size, the data presented in Figures 1-3 and 11-13 were calculated on the basis of 100,000 word forms each (see Appendix 2 for an overview of the historical data).
[FIGURES 1-3, 11-13 OMITTED]
4.2.1. Word category of the base. The first tendency relates to the shift from a high proportion of nominal bases to an increasing number of verbal bases (cf. Section 3.1). The numbers of deverbal and denominal -er-nominals found in German newspapers are illustrated in Figure 1. The numbers of both deverbal and denominal -er-nominals are significantly related to time (deverbal: p = 0.000/[R.sup.2] = 0.971, denominal: p = 0.004/ [R.sup.2] = 0.846).
The general tendency confirms our expectations: given the considerable general rise of -er-nominals from 1609 to 2000, the increase of -er-nominals with verbal bases is much more important than the increase of -er-nominals with nominal bases. While denominal nominals (Fleischer-type) figure prominently from 1609 to 1700 (roughly 60%); there has been a reverse trend since 1750. With the exception of 1800, between 65.2% and 73.6% of all types have a verbal base (Schneider-type). Hence, the first tendency in the development of -er-nominals is validated historically. On closer inspection, however, the proportion of denominal -er-nominals is very much dependent on the fact that names of origin like Hollander 'Dutchman,' which figure prominently in the corpus (e.g. 32.2% in 1609 and still 8.1% in 2000), are included in our count.
To be sure, even if names of origin were subtracted from the corpus, the general decrease of the nominal base remains to be observed, but its decline would start at a remarkably smaller quota of 35.5% denominal nominals and fall to 22.2% in 2000. (10) It seems that the name of origin type of -er-nominals is often grossly underestimated, while our newspaper corpus visibly shows the significance of this type and the necessity of an adequate analysis.
4.2.2. Complexity of the base. The corpus data given in Figure 2 provide the information about the complexity of -er-nominals in our corpus that is necessary to verify the second tendency of -er-nominalization, namely the trend toward more multilexemic bases (cf. Section 3.2). Again, the numbers of monolexemic- and multilexemic-based -er-nominals, respectively, correlate significantly with time (monolexemic: p = 0.003/ [R.sup.2] = 0.851; multilexemic: p = 0.000/[R.sup.2] = 0.975).
The historical data coincide with the results predicted by tendency II. Even though the total number of -er-nominals grows continuously, we can perceive a relatively faster increase of the proportion of multilexemic-based -er-nominals than of the proportion of monolexemic-based -er-nominals. In the 17th century, only one out of three bases possessed a multilexemic structure (Handballer-type), but the share of multilexemic-based -er-nominals rises constantly until 2000, when it finally reaches 69.5%, that is, almost twice as much as in 1609 (37.0%). Therefore, the data of our newspaper data confirm the hypothesis of an increasing complexity as stated in Section 3.2.
4.2.3. Concepts of the -er-nominal. To complete our analysis of the historical data, we finally turn to the concepts realized by -er-nominals. Tendency III suggests a trend towards more OBJECT concepts (cf. Section 3.3). To evaluate this tendency, Figure 3 illustrates the data for PERSON and OBJECT concepts of our newspaper corpus. The regression analysis again shows significant results for both concepts (PERSON: p = 0.000/[R.sup.2] = 0.952; OBJECT: p = 0.004/[R.sup.2] = 0.844).
It becomes evident that PERSON concepts constitute the large majority of types with a maximum share of 99.3% in 1609, and a minimum share of 87.0% in 1900. Despite this overwhelming dominance, an increase of the OBJECT concept can be observed. There is a rather small but continuous increase of OBJECT nominals from 1609 (0.7%) until 1850 (6.3%), followed by a sudden jump up to 13.0% in 1900, and 12.1% in 1950. Finally, we again observe a smaller proportion of OBJECT-denoting -er-nominals (7.1%) in 2000. The reason for the considerable expansion of OBJECTS in 1900 and 1950 is partly due to the high percentage of advertisements in our newspapers (roughly 37% in 1900, and 10% in 1950), and advertisements--trying to sell things--contain more OBJECT -er-nominals. This might have been caused by the industrial development at that time.
A closer look at our data shows that, between 1900 and 2000, we find ten to forty times as many OBJECT types as in the 17th century, whereas the number of PERSON types only doubled to quintupled during the same period (cf. Table 5). Even though OBJECTS only represent a small group in our corpus, the tendency towards more OBJECTS can therefore be confirmed. (12)
4.3. Ontogenetic change
4.3.1. Word category of the base. If one compares the proportions of nominal and verbal bases in the historical corpora of Muller (1993), Brendel et al. (1997), and the Mainz Newspaper Corpus (Scherer 2002) in Figure 1 with the proportions in a corpus of spontaneous coinages of a diary study (Meibauer 1995a), some parallels can be observed as can be seen in Figure 4.
[FIGURE 4 OMITTED]
Over the years, the number of denominal -er-nominals has been decreasing and the number of deverbal -er-nominals has been increasing; this holds true for monolexemic bases as well as for multilexemic ones.
This tendency is also exemplified by a subpart of Task IV of our experiment. The children were asked what they would call themselves at the moment when they were eating some cookies. Of the six children who produced an -er-nominal, only two children aged 4;3 and 5;10 produced the nominal based Kekser 'cookie+er,' whereas five children produced a multilexemic verbal-based -er-nominal Keksesser/Keksfresser 'cookie eat+er' (3;3, 4;7, 5;10, 6;1) and Sussigkeitenesser 'sweets eat+er' (6;0). (13) The test item Fussball+er/Fussball+spiel+er 'football+er/football+play+ er' (both forms are commonly used in German) generated more relevant data. The test instructor made a doll play football, told the children that this doll loved playing football, and then asked them what they thought this doll would like to become when it grew up. 28 children produced one of the intended word forms, 14 of them the noun-based Fussballer, and 14 of them the verbal-based Fussballspieler. Figure 5 shows the distribution of these word forms across the age groups. (14)
[FIGURE 5 OMITTED]
The chi-square test proves that there is a significant correlation between age group and multilexemic verb-based word form (p = 0.05; Cramer-V = 0.37), but not between age group and noun-based word form (p = 0.11). That means that the children in the oldest age group used verb-based derivations significantly more often than the younger children, whereas there is no significant development for noun-based word forms.
Of course, the results for the test items Kekser/Keksesser and Fussballer/ Fussballspieler are equally important for our tendency II that looks at the complexity of the base.
4.3.2. Complexity of the base. In the case of language acquisition, the development towards more multilexemic bases has to be seen in connection with the composition of the lexicon as a whole (Kauschke 2000). Thus, the child varies the categories of the first elements of multilexemic bases during the fourth year (e.g. [Streich.sub.N]+feg+er 'paint cleaner,' [Nett.sub.A] +richt+er 'nice judge') and acquires the combination of suffixes during the fifth year (e.g. diminutives like Schwer+horer+chen 'hard+hearer (dim.)' and Schlips+bind+er+in 'tie tier (fem.)').
A comparison of the diachronic data in Figure 2 with the diary data in Meibauer (1995a) reveals that N+V+er-nominals increase in the course of ontogenetic development, too. This is also true if other multilexemic bases (e.g. N+N+er, A+V+er) are added.
Central to this issue is Clark's principle of simplicity of form: "Speakers find it easier to interpret and coin a new word the simpler it is in form--that is, the less its root changes in its construction" (Clark 1993: 120). This implies that the simple is easier to acquire than the complex, which becomes evident in our experiment. When the children were asked to explain three multilexemic-based -er-nominals in Task II, (16) only 18 children ventured an answer. Figure 7 displays for each age group the mean value of the number of cases for which the children explained the whole test item or just the second part of the word form, respectively. That means, for example, that in the youngest age group none of the children gave an explanation for any of the three test items with multilexemic base. In the third age group, however, the children explained on average 0.75 times (out of a maximum of 3) the second part of the multilexemic based test items, but none explained the whole nominal. The pattern reverses from the fourth age group on, so that in the oldest age group, the children explain on average 0.88 times the whole word forms and only 0.25 times just the second part of them.
[FIGURE 7 OMITTED]
In accordance with the principle of simplicity, the younger children tended to explain, if anything at all, only the second part of the word form, (17) whereas the older ones showed a growing awareness of the first part of the word form and included it in their explanation. Remarkably, none of the children was inclined to explain just the first part of the word form.
By means of statistical tests, the observations concerning the explanation of the whole word form can be validated. Using the Mann-Whitney test, the results of the single age groups were compared. In agreement with Figure 7, the two oldest age groups behaved significantly differently (p < 0.05) from the younger groups (except age group 4). A regression analysis shows a highly significant correlation (p = 0.00/[R.sup.2] = 0.31) between the age of the children in months and how often they explained the whole word form, that is, the older they were, the more often they explained the whole word form.
4.3.3. Concepts of the -er-nominal. The corpora overview in Section 3.3 and the Mainz newspaper corpus (cf. Figure 3, Table 5) showed that the OBJECT-denoting -er-nominals increased to a much larger degree than the PERSON -er-nominals. With respect to child language, a similar tendency can be observed.
There is an early preference for PERSON-denoting nouns; however, OBJECT-denoting -er-nominals increase. Nevertheless, there apparently are children like Rafael, reported on in Neugebauer-Kostenblut (1914), who show a proper balanc, e between PERSON vs. OBJECT -ernominals; in the Rafael corpus, we find 18 PERSON nominals and 17 OBJECT -er-nominals from 1;8 to 2;10.
Regarding our experiment, the results of the comprehension task provide further data on this issue. The task made use of ten test items with four pictures each. For each test item, a corresponding picture for the concept PERSON and the concept OBJECT existed, so that the maximal result for each of the concepts would be a value of 10. Figure 9 illustrates the results of the calculation of the mean values of how often the children pointed at the pictures depicting OBJECTS or PERSONS, respectively.
[FIGURE 9 OMITTED]
Figure 9 shows that the youngest age group prefered OBJECTS to PERSONS, even in cases where a PERSON interpretation would have been more like the target. Following the development across the older age groups, both concepts balance out. Moreover, the average mount of answers increased for both concepts with the age of the children.
An ANOVA of these data reveals that there is a significant difference (p = 0.00) between the age groups for the concept PERSON, but not for the concept OBJECT (p = 0.57). Furthermore, post-hoe tests verify that the behavior of the three-year-olds (age group 1 and 2) differs significantly from the rest of the children, that is, the four- to six-year-old children pointed significantly more often at the PERSON interpretation of the test items than the three-year-olds. A regression analysis, done on the basis of the age of the children in months and their behavior, supports these findings. For the PERSON concept, there is a significant correlation between age and behavior (p = 0.00/[R.sup.2] = 0.24), but not for the OBJECT concept (p = 0.12/[R.sup.2] = 0.04).
In Clark (1993: 180; cf. Appendix 3) it is argued that children typically start to assign the agentive reading to the English -er-suffix, and only later extend the -er-suffix to both agentive and instrument readings, (18) If so, the principle of contrast (i.e. speakers take every difference in form to mark a difference in meaning, thus excluding the assumption that there are synonyms or homonyms) would be confirmed. However, in contrast to Clark, our data show that there is a preference for OBJECTS for the youngest age group, and that both concepts can be related to the -er-suffix from early on. An explanation for this could proceed from observations on early conceptual development and the composition of the small lexicon. There seems to be a predominance of nouns in most small children's lexicons and most nouns that are not proper names denote OBJECTS, which, therefore, make good candidates for word formation processes (cf. Smiley and Huttenlocher 1995; Bloom 2000; Kauschke 2000).
To be sure, there is a great deal of variation to be observed among the individual children, as can be seen from Figure 10. In this diagram, the age groups on the x-axis were replaced by the individual cases sorted by age (months) with the purpose of illustrating the considerable range of interindividual variation. The values on the y-axis represent the actual number of times the children pointed at either the PERSON or the OBJECT pictures.
[FIGURE 10 OMITTED]
Figure 10 shows that some children do have a more or less strong preference toward the concepts of PERSON or OBJECT respectively, whereas others treat both concepts more or less equally.
4.4. Comparing diachronic and ontogenetic change
Although studies in language acquisition and research in diachronic development both focus on phenomena of language change, the circumstances of their research are fairly different. Whereas language acquisition studies examine several months or years of language development, diachronic research looks at decades or centuries. Besides that, acquisition studies usually focus on one child or a relatively small group of children, while diachronic analyses typically investigate the language use of whole populations. Furthermore, first language acquisition data generally consist of recordings of spontaneous speech or elicitation tasks as opposed to corpora of written speech used in diachronic studies.
Therefore, when comparing our diachronic data with the ontogenetic data, it has to be kept in mind that both differ not only in origin and quantity of data but also in methodological aspects. The corpus of Meibauer (1995a) concentrates on new coinages in spontaneous child language, whereas the Mainz newspaper corpus of Scherer (2002) includes any morphological transparent -er-nominal independent of its degree of lexicalization. Moreover, the experimental data are quite a different matter because the test items were artificially selected for specific purposes, namely, to check the children's receptive and productive abilities as to the different word formation patterns and concepts connected to the--er-suffix. Taking these factors into account, we will compare our data with respect to tendencies that can be found in all sets of our data in spite of their differences.
Regarding the word category of the base, a comparison of verbal and nominal bases in the Mainz newspaper corpus in Figure 1 to the child's new coinages as presented in Figure 4 confirms that both historical and ontogenetic development veer towards an increase of deverbal -ernominals. This becomes even more evident when the proportions given in the appendix are taken into account. Whereas the child shows a constant growth of deverbal nominals, the historical development of deverbal -ernominals is--even though progressing as well--more oscillating (cf. Appendix 1B; 2D,II).
As far as the complexity of the base is concerned, we expect the -er-suffix to attach to simple, that is, monolexemic bases from early on and to spread to muitilexemic bases afterwards. This assumption is verified by both the historical newspaper corpus in Figure 2 and the language acquisition data as given in Figures 6 and 7. We can conclude that in language change as well as in language acquisition, there is a preference for monolexemic bases in the first place; only later in the development of the suffix do multilexemic bases come into play.
[FIGURE 6 OMITTED]
A look at the PERSON and OBJECT concepts in the Mainz newspaper corpus in Figure 3 and the diary study of Meibauer (1995a) in Figure 8 reveals an overwhelming dominance of PERSON-denoting -er-nominals throughout the whole period explored. As a result, we find three to four times as many -er-nominals within the last third of the period explored than in the first for both the diachronic and the ontogenetic studies (cf. Appendix 1A; 2B). The experimental data in Figure 9 illustrate that our children up to 3;11 showed a preference for the OBJECT interpretation of-er-nominals, whereas the older children in our experiment preferred PERSON reading to OBJECT reading. Yet, a closer look at the range of interindividual variation among the children as seen in Figure 10, and the data of Meibauer's (1995a) and Neugebauer-Kostenblut's (1914) diary studies, reveals that early preferences for either the OBJECT or the PERSON reading might be very much dependent on individual factors and that only later in their development do the children adapt their preference to the more productive PERSON reading of the -er-suffix.
[FIGURE 8 OMITTED]
5. Productivity in diachronic change and ontogenetic change
An attempt at explaining the observed similarities between language change and language acquisition in the realm of word formation must proceed from the morphological notion of productivity. This notion has recently been discussed by a number of authors from different points of view (e.g. Clark 1993; Plag 1999; Hay 2000; Bauer 2001; Baayen 2001), and there is no doubt that it is at the heart of word structure theory. As emphasised by van Made and Koefoed (2000), productivity is to be seen as a language-internal property of word types or word formation processes that is ultimately constituted by speaker's choices in language use and, hence, submitted to processes of change. Productivity is the result of many forces, some of them structural, some of them social, and some of them cognitive in nature.
For every word type or formation process, there are certain productivity restrictions (Rainer 2000). Since productivity is a gradual phenomenon (a word type or word formation process may be more or less productive), it is very important to be able to measure productivity in a given corpus. Following the work of Baayen and collaborators, we may differentiate between the following aspects of a given corpus which bear on the issue of productivity (Baayen and Renouf 1996; Plag 1999): first, the number of forms with a given affix ("extent of use"); second, the degree of exhaustion ("pragmatic potentiality"); third, the probability of encountering new formations ("productivity in the narrow sense"); and fourth, the number of new formations ("global productivity"). In Section 5.2, we will show the relevance of these notions of productivity for the analysis of our findings.
In our view, it is productivity that constrains the dynamics of-er-nominals. Adults prefer certain patterns of word formation at a given period, and children strive for adaptation to the adult patterns of word formation to which they are exposed. In fact, this is the only link between the history of-er-nominals and the acquisition of-er-noimnals that can be reasonably assumed.
This position does not imply that a certain correlation observable in language change and language acquisition can be ultimately derived from the same source. While such an explanation cannot be totally excluded (e.g. historically as well as acquisitionally "new" types of word formation start with monolexemic instead of multilexemic bases when the respective types principally allow for both), we will see that there are a number of different sources as well. This is predictable, however, because the conditions under which adults influence the productivity of-er-nominals in, say, 1650 are in some respect different from those under which children acquire the -er-suffix in 2000.
In Section 5.3 we will sketch some possible explanations. More sophisticated analyses are, of course, dependent on advances in other areas of research, for example, research on early conceptual development or incorporation processes in ENHG. It goes without saying that gaining empirical data about other word formation devices and processes is a precondition for further progress in the discussion.
5.2. Productivity in diachronic change
Above, we calculated the extent of use of the -er-word formation pattern, that is, we made use of the first notion of productivity (cf. Figures 1-3). This was done on the basis of types only. Now compare Figure 11.
The figure shows that the number of tokens as well as the number of types rises from 1609 to 2000. This development proves to be statistically significant (types: p = 0.000/[R.sup.2] = 0.974; tokens: p = 0.001/[R.sup.2] = 0.889). However, the rate of types increases relatively faster than the rate of tokens. Altogether, the number of types more than tripled between the first and the last measuring point, whereas the number of tokens only doubled (cf. Table 6), thus giving rise to a larger number of hapax legomena or rare types in our corpus. How can this be explained?
We assume that the decrease of the token/type ratio reflects the rising productivity of the -er-suffix, as more and more new types are used. Because productivity is closely linked with transparency, we consider the increase of productivity as an indicator of a tendency toward more transparency of the -er-word formation pattern.
The second notion of productivity, pragmatic potentiality, plays a role when the reasons for the change in the PERSON/OBJECT parameter are considered. The late occurrence of OBJECT-denoting -er-nominals is pragmatically motivated, as this development correlates with changes in socioeconomic conditions such as the rise of newspapers and industrial development.
Figure 12, which shows the number of denominal and deverbal PERSON types only, is consistent with the development of the general extent of use as represented in Figure 11. A fact that is not astonishing, given that 90% of all types in the corpus represent PERSON concepts. For both denominal and deverbal PERSON nominals, the member of types is significantly increasing over time (deverbal PERSON: p = 0.000/[R.sup.2] = 0.943; denominal PERSON: p = 0.006/[R.sup.2] = 0.820).
A comparison with the bases of the OBJECT nominals in Figure 13 shows roughly the same development until 1900: a constant rise of denominal and deverbal OBJECT types. In this case, too, the regression analysis shows a significant correlation (deverbal OBJECT: p = 0.015/ [R.sup.2] = 0.754; denominal OBJECT: p = 0.007/[R.sup.2] = 0.810). In the 20th century, the proportion of denominal OBJECTS (which reaches its maximum in 1900) finally drops remarkably in 1950 again, whereas the same pattern can be observed for deverbal OBJECTS in 1950 and 2000. Thus, the extent of use of OBJECT nominals seems to decrease again. However, two major points have to be considered. First, the data in Figure 13 are based on a rather small sample (less than 200 types in the corpus as a whole, with a range of 1 to 45 types per period), which might lead to corpus-based artefacts. Second, it is not yet obvious how the trend will continue. The decline of deverbal OBJECTS from 1950 to 2000 may just be an intermezzo before a further rise--a tendency that might be anticipated in the development of denominal OBJECT nominals.
The third and fourth notion of productivity deal with new formations. The rationale behind these notions is that new formations show speaker preferences in the application of word formation patterns. Many new formations are an indicator of high productivity. A calculation of productivity in the narrow sense for the Mainz newspaper corpus, via the quotient of hapax legomena and the total number of -er-nominals, can be seen in Figure 14.
[FIGURE 14 OMITTED]
Between 1609 and 2000 we find a constant growth in productivity showing a kind of cycle characterized by a period of rather small growth (1609-1650, 1800-1850), followed by a period of a slight decline (16501750, 1850-1900), and completed by a period of a strong rise in productivity (1750-1850, 1950-2000). Altogether, in 2000, there are roughly five times as many hapax legomena as in the 17th and 18th century, and about twice as many as in 1900 and 1950 (cf. Table 6). The number of hapax legomena is correlated significantly to the respective periods (p = 0.000/[R.sup.2] = 0.975). The growing number of hapax legomena is in good accordance with the decrease of the token/type ratio as discussed above. Both hint in the same direction, namely, an increase in the productivity of the -er-word formation pattern over the last four centuries.
5.3. Productivity in ontogenetic change
It may very well be the case that a certain word formation pattern has already been acquired by a child, but that there is no particular need to make use of his or her competence. For example, we would suspect that a child who is more interested in things than in persons would coin more OBJECT derivations, even if the child knew the pattern for expressing PERSON derivations quite well (Meibauer 1995a). Therefore, the second notion of productivity, pragmatic potentiality, is of great importance for language acquisition. When young children coin new -er-nominals, they want to fill lexical gaps, that is, they compensate for their small lexicon. Word play seems to be another reason for coining new words, especially for older children. The observation of spontaneously coined new word formations is methodologically very important (especially for children under 3;0 who are too young to participate in experiments), since only these coinages show that the young child has not learned a complex word as a whole. The ability to produce a new formation shows that the child has analyzed a certain pattern (Bates and Rankin 1979; Clark 1993).
In the production and comprehension tests described above, the four notions of productivity also play a role. The first notion, the extent of use, is relevant in so far as the child has alternative possibilities to express a certain concept of PERSON or OBJECT, for example, syntactic constructions or, more interesting from the point of view of word formation, precursor elements like -mann ('man') or -maschine ('machine'). This option, however, was rarely used by our group of children in our Task III (word formations of nine given nonsense bases), and if so, only by four of the older ones (5;4, 6;1, 6;2, 6;5). Only one child (6;2) used it three times including both PERSON and OBJECT concepts. (19) On the other hand, one third (30.4%) of all the other answers in Task III consist of-er-nominals. (20)
The second notion of productivity, pragmatic potentiality, is affected to the extent that the children must make a decision as to PERSON vs. OBJECT readings (Task I) where both readings are appropriate. A preference of the child indicates a preference for the exploitation of the word formation pattern. The fourth notion of productivity dealing with new formations plays a prominent role in the experiments, because they focus on the ability of children to coin new -er-nominals (Tasks III and IV).
5.4. Explaining similarities
The observed similarities in the development of-er-nominals can be traced back to a number of factors, some being language-internal, some being language-external. In the following section we point out some of these factors without aiming at a comprehensive theory.
5.4.1. Base parameter. Why are verbal bases in -er-nominals increasing? The overall diachronic tendency may be interpreted as adaptation to the typical category changing function of German suffixes. When combined with nominal bases, -er acts like a small class of suffixes that do not change the category of their base as, for example, diminutive suffixes like -chen and -lein. Therefore, the rising proportion of verbal bases shows that the historical development of German favors category-changing -erformations.
According to most researchers, more nouns than verbs can be found in the small lexicon of children (cf. Kausclake 2000 for an extensive discussion). Therefore, nouns seem to be good candidates for the base of-ernominals. However, children probably hear more deverbal -er-nominals than denominal -er-nominals in the adult input, and they adapt to the pattern of productivity they find in adult language. Maybe verbal bases are better suited for the expression of PERSON concepts anyhow, because verbs typically describe actions of persons. Thus, if children are more interested in persons, there is an additional motive to opt for verbal bases.
Those -er-nominals whose base cannot be uniquely determined, for example Oler 'oiler' in child language or Fischer 'fisherman' in adult language, may function as an interface between denominal and deverbal -ernominals and allow "die Ubertragung der Suffixe von einer Wortart auf die andere" [the transfer of the suffixes from one part of speech to another], as Wilmanns (1986 : 14 f.) put it (Meibauer 1999).
5.4.2. Complexity parameter. Why are multilexemic bases increasing? In historical word formation, we found that there is an increase of multilexemic-based -er-nominals. There are also good reasons in the study of language acquisition to assume that young children prefer simple word structures (Clark 1993). Only later in their development are complex structures acquired. Thus, it is reasonable to assume that these similarities can be traced back to a general (cognitive) principle that the simple arises earlier in a process than the complex because it is easier to process and produce. Thus, when a suffix first comes into play, it should attach more productively to what children regard as simple bases instead of complex ones.
The notion of transparency is crucial for the identification of complex bases. Therefore, if a base is not transparent for the speaker, a suffix may be attached to a multitexemic base. For example, when the child in Meibauer (1995a) coined Festhalter 'handle (tight-holder)' with 2;0, the base festhalt- possibly was not transparent to him. What is transparent to the speaker also depends on the frequency of the base. Hay (2000) has recently shown that speakers tend to regard a derivation as complex if its base is frequent. Consequently, if the base of a multilexemic -er-nominal is low in frequency, it may be regarded as being not transparent, and therefore simple as well.
Another relevant factor is the competition with rival suffixes. A new suffix that competes with an established suffix may have a greater chance to promulgate if its base is monolexemic. Finally, the pool of possible bases for a given suffix should be taken into account. Thus, if there is a general development toward more motivated and transparent bases in a language, the chance for a given affix to attach to a multilexemic base increases additionally.
Historically, incorporation processes and the growing need to designate more and more specialized professionals and instruments were further reasons to enhance the number of -er-nominals with multilexemic bases (Pavlov 1983; Joeres 1995). Whether there are independent reasons for the children to use more and more multilexemic-based -er-nominals in their development besides adaptation to adult productivity, an increasing awareness of transparency, and the growing ability to compress information, for example, an acquisitional equivalence to incorporation, is still unclear.
5.4.3. Concept parameter. Why is there a preference for PERSON concepts? The access of the -er-sulfix to the relevant concepts PERSON, OBJECT, etc. can be analyzed in terms of conceptual shift (Booij 1986; Meibauer 1995b), a process that figures prominently in language change as well as in language acquisition. It seems that the PERSON concept is the starting point for conceptual shifts. In the history of language, the possibility to shift from PERSON to OBJECT concepts is exploited because of the growing need for words denoting artefacts in the ENHG and NHG periods (Muller 1993; Brendel et al. 1997; Doring and Eichler 1996).
In language acquisition, the issue is more complicated (Guttropf and Meibauer 2003). On the one hand, it can be argued that the concept of PERSON and its bootstrapping into the notion of an agent (e.g. as a thematic proto-role) is very important from early on (Budwig 1989; Smiley and Huttenlocher 1995); on the other hand, it can be noticed, contrary to the assumptions of Clark (1993), that there are young children who prefer OBJECT-denoting -er-nominals. Rafael, as reported by Neugebauer-Kostenblut (1914), seems to produce more OBJECT-denoting than PERSON-denoting -er-nominals during his third year. Our data from the comprehension task point in the same direction. Thus, there seems to be a sort of balancing out the saliency of the PERSON concept, the filling of lexical gaps, and last but not least, the proportions of -er-nominals representing PERSONS and OBJECTS in the adult input.
Appendix 1. Data from the corpus of spontaneous -er-coinages in child language (Meibauer 1995a) A Nominals total 2;0-2;11 3;0-3;11 4;0-4;11 Total 35 94 90 B Base 2;0-2;11 3;0-3;11 4;0-4;11 Nominal base 17 16 9 Verbal base 14 50 46 C Morph. structure 2;0-2;11 3;0-3;11 4;0-4;11 Monolexemic 30 49 48 base Multilexemic 5 45 42 base D Concepts 2;0-2;11 3;0-3;11 4;0-4;11 PERSON 32 77 58 OBJECT 3 16 20 A Nominals total Total B Base 2;0-2;11 (%) 3;0-3;11 (%) 4;0-4;11 (%) Nominal base 54.8 24.2 16.4 Verbal base 45.2 75.8 83.6 C Morph. structure 2;0-2;11 (%) 3;0-3;11 (%) 4;0-4;11 (%) Monolexemic 85.7 52.1 53.3 base Multilexemic 14.3 47.9 46.7 base D Concepts 2;0-2;11 (%) 3;0-3;11 (%) 4;0-4;11 (%) PERSON 91.4 82.8 74.4 OBJECT 8.6 17.2 25.6 Appendix 2. Data from the Mainz newspaper corpus (Scherer 2002) A. Corpus overview: -er-nominals 1609 1650 1700 1750 1800 Word forms (in 99 98 98 103 101 thousands) Types 146 87 145 198 207 Tokens 607 289 538 628 554 Hapax legomena 47 25 40 46 67 1850 1900 1950 2000 total (a) Word forms (in 109 150 137 137 1,031 thousands) Types 307 471 520 678 1,844 Tokens 757 1,388 1,396 1,634 7,791 Hapax legomena 107 190 217 362 1,101 (a.) The total number of types in the corpus does not equal the sum of types in all sub-corpora, as -er-nominals that occur in different subcorpora are only counted once. B. -er-nominals: extent of use (per 100,000 word forms) 1609 1650 1700 1750 1800 Types 147.6 88.5 148.1 192.6 204.8 Tokens 613.5 293.9 549.3 610.8 548.2 Hapax legomena 47.5 25.4 40.8 44.7 66.3 1850 1900 1950 2000 Types 282.6 315.0 380.9 493.6 Tokens 696.9 928.3 1,022.5 1,189.6 Hapax legomena 98.5 127.1 158.9 263.5 C. -er-nominals: productivity in the narrow sense 1609 1650 1700 1750 1800 Productivity (%) 7.7 8.7 7.4 7.3 12.1 1850 1900 1950 2000 Productivity (%) 14.1 13.7 15.5 22.2 D. -er-nominals: base (types) I. Extent of use (per 100,000 word forms) 1609 1650 1700 1750 1800 Nominal base 86.9 48.8 79.6 67.1 84.1 Thereof toponyms 47.5 26.4 35.7 22.4 43.5 Verbal base 60.6 39.7 68.4 125.5 120.7 II. In % of -er-nominals per measuring point Nominal base (%) 58.9 55.2 53.8 34.8 41.1 Thereof toponyms (%) 32.2 29.9 24.1 11.6 21.3 Verbal base (%) 41.1 44.8 46.2 65.2 58.9 I. Extent of use (per 100,000 word forms) 1850 1900 1950 2000 Nominal base 77.3 92.3 119.4 141.2 Thereof toponyms 26.7 26.8 37.4 40.0 Verbal base 205.3 222.7 261.5 352.4 II. In % of -er-nominals per measuring point Nominal base (%) 27.4 29.3 31.3 28.6 Thereof toponyms (%) 9.4 8.5 9.8 8.1 Verbal base (%) 72.6 70.7 68.7 71.4 E. -er-nominals: morphological structure (types) I. Extent of use (per 100,000 word forms) 1609 1650 1700 1750 1800 Monolex. base 93.0 55.9 90.9 99.2 111.8 Multilex. base 54.6 32.5 57.2 93.4 93.0 II. In % of -er-nominals per measuring point Monolex. base (%) 63.0 63.2 61.4 51.5 54.6 Multilex. base (%) 37.0 36.8 38.6 48.5 45.4 I. Extent of use (per 100,000 word forms) 1850 1900 1950 2000 Monolex. base 114.2 125.1 148.7 150.7 Multilex. base 168.5 189.9 232.2 342.9 II. In % of -er-nominals per measuring point Monolex. base (%) 40.4 39.7 39.0 30.5 Multilex. base (%) 59.6 60.3 61.0 69.5 F. -er-nominals: concepts (types) (b) I. Extent of use (per 100,000 word 1609 1650 1700 1750 1800 PERSON 146.5 85.4 142.9 179.0 188.0 OBJECT 1.0 3.1 4.1 9.7 11.9 II. In % of -er-nominals per measuring point PERSON (%) 99.3 96.6 97.2 94.8 94.1 OBJECT (%) 0.7 3.4 2.8 5.2 5.9 I. Extent of use (per 100,000 word 1850 1900 1950 2000 PERSON 260.5 268.2 325.9 445.5 OBJECT 17.5 40.1 45.4 34.2 II. In % of -er-nominals per measuring point PERSON (%) 93.7 87.0 87.8 92.9 OBJECT (%) 6.3 13.0 12.2 7.1 (b.) EVENT nominals, which are otherwise included in our counts, have been excluded from the calculation of the total number of concepts resulting in slightly lower totals. G. PERSON -er-nominals: base (types) I. Extent of use (per 100,000 word forms) 1609 1650 1700 1750 1800 Nominal base 86.9 47.8 77.6 60.3 75.2 Verbal base 59.6 37.6 65.3 118.7 112.8 II. In % of -er-nominals per measuring point Nominal base (%) 59.3 56.0 54.3 33.7 40.0 Verbal base (%) 40.7 44.0 45.7 66.3 60.0 I. Extent of use (per 100,000 word forms) 1850 1900 1950 2000 Nominal base 69.0 78.3 113.5 134.7 Verbal base 191.5 189.9 212.4 310.9 II. In % of -er-nominals per measuring point Nominal base (%) 26.5 29.2 34.8 30.2 Verbal base (%) 73.5 70.8 65.2 69.8 H. OBJECT -er-nominals: base (types) 1. Extent of use (peword forms) 1609 1650 1700 1750 1800 Nominal base 0.0 1.0 2.0 6.8 8.9 Verbal base 1.0 2.0 2.0 2.9 3.0 II. In % of -er-nominals per measuring point Nominal base (%) 0.0 33.3 50.5 70.0 75.0 Verbal base (%) 1.0 66.7 50.5 30.0 25.0 I. Extent of use (per 100,000 1850 1900 1950 2000 Nominal base 8.3 14.0 5.1 6.6 Verbal base 9.2 26.1 40.3 27.7 II. In % of -er-nominals per measuring point Nominal base (%) 47.4 35.0 11.3 19.1 Verbal base (%) 52.6 65.0 88.7 80.9 Appendix 3. Consistency in producing agent and instrument nouns (Clark and Hecht 1982; Clark 1993) (c) Age group Mean age: 3;10 Mean age: 4;5 Consistent Consistent only Consistent only on neither on agent on instrument 1 (3;0-3;8) 6 4 -- 2 (3;9-4;5) -- 4 2 3 (4;6-5;2) 2 3 -- 4 (5;3-6;0) 1 3 1 Total 9 14 3 Age group Mean age: 4;9 Consistent on both 1 (3;0-3;8) 2 2 (3;9-4;5) 6 3 (4;6-5;2) 7 4 (5;3-6;0) 7 Total 22 (c.) The numbers in the four columns are numbers of children. Table 1. Classification of -er-nominals in present-day German PERSON OBJECT EVENT Verbal base Lehr+er Steck+er Seufz+er 'teach+er' 'plug (stick+er)' 'sigh (sigh+er)' Nominal base Fleisch+er Benzin+er 'butcher 'car that runs (meat+er)' on petrol (petrol+er)' Table 2. Bases of -er-nominals in OHG and present-day German Nominal Verbal Types base (%) base (%) (n) a. OHG 21.7 78.3 263 (Weinreich 1971) b. present-day 15.4 84.6 [1.242.sup.3] (Wellmann 1975) German Table 3. Structure of -er-nominals in OHG and ENHG Monolexemic Multilexemic base Types(n) base (%) (%) a. OHG 69.2 X+base+er 30.8 263 (Weinreich 1971) b. ENHG 83.5 N+V+er 17.5 510 (Brendel et al. 1997) c. ENHG 68.5 N+V+er 31.5 92 (Muller 1993) Table 4. Concepts of -er-nominals in present-day German PERSON OBJECT EVENT Types (%) (%) (%) (n) Present-day 79.1 13.3 3.5 [1,190.sup.6] (Wellmann 1975) German Table 5. -er-nominals: concepts (types) per 100,000 word forms 1609 1650 1700 1750 1800 PERSON 146.5 85.4 142.9 179.0 188.0 OBJECT 1.0 3.1 4.1 9.7 11.9 1850 1900 1950 2000 PERSON 260.5 268.2 325.9 445.5 OBJECT 17.5 40.1 45.4 34.2 Table 6. -er-nominals: types, tokens, and hapax legomena per 100,000 word forms 1609 1650 1700 1750 1800 Types 147.6 88.5 148.1 192.6 204.8 Tokens 613.5 293.9 549.3 610.8 548.2 Hapax leg. 47.5 25.4 40.8 44.7 66.3 1850 1900 1950 2000 Types 282.6 315.0 380.9 493.6 Tokens 696.9 928.3 1,022.5 1,189.6 Hapax leg. 98.5 127.1 158.9 263.5
* Parts of this paper were presented at Tubingen ('Workshop on Nominalization', April 2001) and Stockholm. We thank the audiences for stimulating questions and helpful comments. We also thank our student assistants Claudia Faust, Edda Gerhart, Stefanie Nartschik, and Stephan Scheffe, who did most of the data input. We finally thank Ingo Plag (Siegen), Markus Steinhach (Mainz), and two anonymous reviewers for critical remarks on a previous draft of this paper. Our research was made possible through a grant of the Deutsche Forschungsgemeinschaft. Correspondence address: Jorg Meibauer, Deutsches Institut, Universitat Mainz, Welderweg 18 (Philosophicum), 55099 Mainz, Germany. E-mail: email@example.com.
(1.) Phonological reflexes of the ancient Germanic suffix *-warja are still present in the NHG word Bajuware 'Bavarian.'
(2.) Actually, the -er-suffix is used to form masculine persons and animals out of feminine denotating nouns as well (Witwe 'widow' > Witwer 'widow+er'). As we focus on those cases where -er effects a transposition of the word category or at least of the global concept, we only note that the indication of [+male] via -er or other suffixes is limited to a very small range of nouns in German (of. Doleschal 1992).
(3.) To allow exact comparison, 28 -er-nominals with bases other than nominal or verbal were subtracted from Wellmann's corpus.
(4.) We thank one of our reviewers for proposing the term "multilexemic base," to which we oppose the term "monolexemic base." The dichotomy of monolexemic and multilexemic avoids possible misunderstandings, as the terms "complex" and "simple" are not used consistently throughout the linguistic community.
(5.) An anonymous reviewer points to the possibility that the need for derivation (with a multilexemic base) correlates to the complexity of a situation to be named. While such a motivation cannot be excluded altogether, it cannot easily explain the substitution of korber by korbmacher, because the complexity of the situtation to be named remains the same. We assume that analogy to other professional names ending in -macher is at work here.
(6.) Another 80 nominals that could not be classified as belonging to one or only one of these concepts were not taken into account.
(7.) According to Kluge (1999 ) Gegner is a loan translation of Latin adversarius, the German preposition gegen translating the Latin adversus which could serve as a preposition, an adverb, or an adjective, respectively.
(8.) The whole experiment was audiotaped and transcribed. The results of Task I and parts of Task IV were entered in preset forms. For the coverage and evaluation of the data, we used SPSS 10.0.
(9.) Note that the proportion of each base is not represented by the line, but by the space occupied by the corresponding color.
(10.) These calculations are based on the total number of types (e.g. 146 in 1609) reduced by the number of deonomastic types (i.e. 47 in 1609, thus leaving a total of 99 types for 1609). Therefore, they differ from the result of a simple subtraction of the proportion of names of origin from the proportion of denominal -er-nominals.
(11.) In this calculation, the concept OBJECT comprises plants and animals as well. The EVENT concept, which constitutes 13 out of 1,844 types in the corpus, is not included in this analysis.
(12.) However, the rather small quantity of OBJECT nominals may not be representative for the German language in general, as newspapers seem to focus more on persons and incidents than on objects. Note that Baayen (1994) convincingly shows that the productivity of a suffix is dependent on text type. Comparing our data with similar historical corpora, focussing on different text forms or topics would allow a more general interpretation of our results.
(13.) The other children either did not reply to the question at all, gave a different answer like hungrig 'hungry,' or used a suppletive like Krumelmonster (the character "Cookie Monster" from the German Sesame Street).
(14.) In this case, the group of children was divided into just four age groups: 3;0-3;11, 4;0-4;11, 5;0-5;11, and 6;0-6;5. This allows for a better analysis of the data.
(15.) Children who produced Fussballer. 3;10, 4;1, 4;10, 4;3, 4;7, 5;0, 5;2, 5;5, 5;7, 5;10 (3x), 6;3, 6;4, and Fussballspieler: 3;6 (2x), 4;10, 4;4, 4;5, 4;8, 5;4 (2x), 5;11, 6;0, 6;2, 6;4, 6;5 (2x).
(16.) The test items are Bauehplatsch-er 'belly landing into the water (belly-splash-er),' Abendgahn-er 'evening-yawn-er,' Grossstudt-er 'person coming from a large city (big-city-er).'
(17.) For example, they explained Abendgahner just by "somebody who yawns," or Grossstadter by "person coming from a city."
(18.) Clark's agentive reading equals approximately our PERSON reading, whereas her instrument reading is best compared with our OBJECT reading.
(19.) The answers were: Klitzeding (5;4); Ripfmaschine (6;1); Klitzmaschine, Tulkmaschine, Dengmann (6;2); Schloffmann (6;5).
(20.) The remaining answers contain suppletives, bases, repetition of the infinitive, or no reply.
(21.) An anonymous reviewer proposed that there might be a parallel between developing cultural complexity through the ages and the child mastering increasingly complex skills and knowledge.
Aronoff, Mark (1976). Word Formation in Generative Grammar. Cambridge, MA: MIT Press.
Baayen, R. Harald (1994). Derivational productivity and text typology. Journal of Quantitative Linguistics 1, 16-34.
--(2001). Word Frequency Distributions. Dordrecht: Kluwer.
--; and Renouf, Antoinette (1996). Chronicling the times: productive lexical innovations in an English newspaper. Language 72, 69-96.
Baron, Naomi S. (1977). Language Acquisition and Historical Change. Amsterdam: North-Holland.
Bates, Elizabeth; and Rankin, Jane (1979). Morphological development in Italian: connotation and denotation. Journal of Child Language 6, 29-52.
Bauer, Laurie (2001). Morphological Productivity. Cambridge: Cambridge University Press.
Berko, Jean (1958). The child's learning of English morphology. Word 14, 150-177.
Bloom, Paul (2000). How Children Learn the Meanings of Words. Cambridge, MA: MIT Press.
Booij, Geert E. (1986). Form and meaning in morphology: the case of Dutch "agent nouns." Linguistics 24, 503-517.
Bowerman, Melissa (1988). The "no negative evidence" problem: how do children avoid constructing an overly general grammar? In Explaining Language Universals, John A. Hawkins (ed.), 73-101. Oxford: Blackwell.
Brendel, Bettina; Frisch, Regina; Moser, Stephan; and Wolf, Norbert Richard (1997). Wortund Begriffsbildung in fruhneuhochdeutscher Wissensliteratur: Substantivische Affixbildung. Wiesbaden: Reichert.
Budwig, Nancy (1989). The linguistic marking of agentivity and control in child language. Journal of Child Language 16, 263-284.
Clark, Eve V. (1993). The Lexicon in Acquisition. Cambridge: Cambridge University Press.
--; and Hecht, Barbara F. (1982). Learning to coin agent and instrument nouns. Cognition 12, 1-24.
Clark, Robin; and Roberts, Ian (1993). A computational model of language learnability and language change. Linguistic Inquiry 24, 299-345.
Croft, William (2000). Explaining Language Change: An Evolutionary Approach. London: Roufledge.
Derwing, Bruce L. (1976). Morpheme recognition and the learning of rules for derivational morphology. The Canadian Journal of Linguistics 21, 38-66.
Doleschal, Ursula (1992). Movierung im Deutschen: Eine Darstellung der Bildung und Verwendung weiblicher Personenbezeichnungen. Unterschieissheim: Lincom Europa.
Doring, Brigitte; and Eichler, Birgit (1996). Sprache und Begriffsbildung in Fachtexten des 16. Jahrhunderts. Wiesbaden: Reichert.
Erben, Johannes (1993 ). Einfuhrung in die deutsche Wortbildungslehre, 3rd ed. Berlin: Schmidt.
Fleischer, Wolfgang; and Barz, Irmhild (1995 ). Wortbildung der deutschen Gegenwartsspraehe, 2nd ed. Tubingen: Niemeyer.
Fuhrhop, Nanna (1998). Grenzfdlle morphologischer Einheiten. Tubingen: Stauffenburg.
Guillaume, Gustave (1973 ). Le developpement des elements formels clans le langage de l'enfant. Journal de Psychologie Normale et Pathologique 24, 203-229. [English translation by Eve Clark: The development of formal dements in the child's speech. In Studies of Child Language Development, C. A. Ferguson and D. I. Slobin (eds.), 240-251. New York: Holt, Rinehart, and Winston.]
Guttropf, Anja; and Meibauer, Jorg (2003). Konzeptuelle Entwicklung und Wortbildungserwerb. Eine empirische Studie zum PERSON- mad OBJEKT-Konzept. In Spracherwerb und Konzeptualisierung, Stefanie Haberzett1 and Heide Wegener (eds.), 139-159. Frartkfurt/M.: Lang.
Haeckel, Ernst (1866). Generelle Morphologie der Organismen. Berlin: Reimer.
Halle, Morris (1962). Phonology in generative grammar. Word 18, 54-72.
Haspelmath, Martin (1999). Optimality and diachronic adaptation. Zeitschrift far Sprachwissenschaft 18, 180-205.
Hay, Jennifer Bohun (2000). Causes and consequences of word structure. Unpublished doctoral dissertation, Northwestern University.
Henzen, Walter (1965 ). Deutsche Wortbildung, 3rd ed. Tubingen: Niemeyer.
Hirt, Hermann (1932). Handbueh des Urgermanisehen: Tei111: Stammbildungs- und Flexionslehre. Heidelberg: Winter.
Hohle, Tilman (1982). Uber Komposition und Derivation: Zur Konstituentenstruktur von Wortbildungsprodukten im Dentschen. Zeitsehrift fur Sprachwissenschaft 1, 76-112.
Joeres, Rolf (1995). Wortbildungen mit -macher im Althochdeutschen, Mittelhochdeutschen und Neuhochdeutschen. Heidelberg: Winter.
Kauschke, Christina (2000). Der Erwerb des fruhkindlichen Lexikons: Fine empirische Studie zur Entwieklung des Wortschatzes im Deutschen. Tubingen: Narr.
Kerswill, Paul (1996). Children, adolescents, and language change. Language Variation and Change 8, 177-202.
Kiparsky, Paul (1965). Phonological change. Unpublished doctoral dissertation, MIT.
Kluge, Friedrich (1999 ). Etymologisches Worterbuch der deutschen Sprache, 23rd ed. Berlin: Walter de Gruyter.
Leser, Martin (1990). Dos Problem der 'Zusammenbildungen': Eine lexikalistische Studie. Trier: Wissenschaftlicher Vedag.
Lightfoot, David (1979). Principles of Diachronic Syntax. Cambridge: Cambridge University Press.
--(1991). How to Set Parameters." Arguments from Language Change. Cambridge, MA, MIT Press.
--(1999). The Development of Language: Acquisition. Change, and Evolution. Oxford: Blackwell.
Marle, Jaap van; and Koefoed, CJeert (2000). Productivity. In Morphologie/Morphology: Ein internationales Handbuch zur Flexion und Wortbildung/An International Handbook on Inflection and Word-Formation. vol. 1, Handbucher zur Sprach- und Kommunikationswissenschaft 17, Geert Booij et al. (eds.), 1303-1311. Berlin: Walter de Gruyter.
Meibauer, Jorg (1995a). Neugebildete -er-Derivate im Spracherwerb. Ergebnisse einer Langzeitstudie. Sprache und Kognition 14, 138-160.
--(1995b). Wortbildung und Kognition: Uberlegungen zum deutschen -er-Suffix. Deutsche Sprache 23, 97-123.
--(1998). Kunst vertrucker und kolengreber: Zum Wortbildungswandel der N+V+er-Bildungen im Fruhneuhochdeutschen. In Zwischen Grammatik und Lexikon. Irmhild Barz and Gunther Ohlschlager (eds.), 81-102. Tubingen: Niemeyer.
--(1999). Uber Nomen-Verb-Beziehungen im fruhen Wortbildungserwerb. In Dos Lexikon im Spracherwerb, Jorg Meibauer and Monika Rothweiler (eds.), 184-207. Tubingen: Francke.
Muller, Peter O. (1993). Substantiv-Derivation in den Scbriften Albrecht Darers. Berlin: Walter de Gruyter.
Neugebauer-Kostenblut, Hanna (1914). Sprachliche Eigenbildungen meines Sohnes. Zeitschrift fur Kinderforschung 19, 174-181; 242-246; 362-370.
Olsen, Susan (1986). Wortbildung im Deutschen: Eine Einfuhrung in die Theorie der Wortstruktur. Stuttgart: Kroner.
Pavlov, Vladimir M. (1983). Zur Ausbildung der Norm der deutscben Literatursprache im Bereich der Wortbildung (1470-1730): Von der Wortgruppe zur substantivischen Zusammensetzung. Berlin: Akademie.
Paul, Hermann 0975 ). Prinzipien der Sprachgeschicbte, 9th ed. Tubingen: Niemeyer.
--(1920). Deutsche Grammatik: 5. Teil 4: Wortbildung. Halle a.d.S.: Niemeyer.
Plag, Ingo (1999). Morphological Productivity: Structural Constraints in English Derivations. Berlin: Mouton de Gruyter.
Rainer, Franz (2000). Produktivitatsbeschrankungen. In Morphologie/Morphology: Ein internationales Handbuch zur Flexion und Wortbildung/An International Handbook on Inflection and Word-Formation. vol. 1, Handbucher zur Sprach- und Kommunikationswissenschaft 17, Geert Booij et al. (eds.), 877-885. Berlin: Walter de Gruyter.
Ravid, Dorit D. (1995). Language Change in Child and Adult Hebrew. New York: Oxford University Press.
Romaine, Suzanne (1989). The role of children in linguistic change. In Language Change: Contributions to the Study of Its Causes, L. E. Breivik and E. H. Jahr (eds.), 199-225. Berlin: Mouton de Gruyter.
Schaner-Wolles, Chris; and Dressier, Wolfgang (1985). On the acquisition of agent/ instrument nouns and comparatives by normal children and children with Down's syndrome: a contribution to natural morphology. Acta Linguistica Academiae Scientiarum Hungaricae 35, 133-149.
Seherer, Carmen (2002). Wortbildungswandel und Produktivitat. Eine empirische Studie zur nominalen -er-Derivation im Deutschen. Unpublished doctoral dissertation, University of Mainz.
Schottelius, Justus Georg (1967 ). Ausfuhrliche Arbeit vonder Teutschen Haubt-Sprache: vol. 1. Braunschweig: Zilligern. Reprinted: Tubingen: Niemeyer.
Slobin, Dan I. (1977). Language change in childhood and history. In Language Learning and Thought, John MacNamara (ed.), 185-214. New York: Academic Press.
Smiley, Patricia; and Huttenlocher, Janellen (1995). Conceptual development and the child's early words for events, objects, and persons. In Beyond Names for Things. Young Childrens's Acquisition of Verbs, Michael Tomasello and William E. Merriman (eds.), 21-61. Hillsdale: Erlbaum.
Stricker, Stefanie (2000). Substantivbildung durch Suffixableitung um 1800: Untersucht an Personenbezeichnungen in der Sprache Goethes. Heidelberg: Winter.
Weinreich, Otto (1971). Die Suffixabldsung bei den Nomina agentis wahrend der althochdeutschen Periode. Berlin: Schmidt.
Wellmann, Hans (1975). Deutsche Wortbildung: Typen und Tendenzen in der Gegenwartssprache." Zweiter Hauptteil: Das Substantiv. Dusseldorf: Schwann.
Wilmanns, Wilhelm (1986 ). Deutsche Grammatik: Gotisch, Alt-, Mittel- und Neuhochdeutsch: Zweite Abteilung: Wortbildung, 2nd ed. Strassburg: Trubner.
Windsor, Jennifer (1994). Children's comprehension and production of derivational suffixes. Journal of Speech and Hearing Research 37, 408-417.
Received 5 February 2002 Revised version received 23 July 2002
|Printer friendly Cite/link Email Feedback|
|Author:||Meibauer, Jorg; Guttropf, Anja; Scherer, Carmen|
|Publication:||Linguistics: an interdisciplinary journal of the language sciences|
|Date:||Jan 1, 2004|
|Previous Article:||On a partial, strictly word-order based definition of grammatical relations in Liangshan Nuosu *.|
|Next Article:||Signs of a developing grammar: subject drop and inflection in early child Dutch *.|