1

2

Pathways are no One-Way Streets: German Indefinite Article(s)

3

4

5

In this article, it will be shown that the German indefinite article has lost admissible contexts since Middle High German times (≈ 1050–1350).

6

This data provides clear counterevidence against the claim by Roberts

7

& Roussou (2003) that grammatical change with interface effects always

8

involves syntactic upward movement. More generally, it provides evidence

9

against any theory assuming that grammatical change equates grammatical-

10

isation, that is, a shift to more grammatical meaning. I will further try to

11

show, in an attempt to specify the concept of semantic bleaching, that the

12

data involves semantic strengthening, and thus, that grammatical change

13

does not always boil down on its semantic side to meaning loss.

1

14

1 Introduction

15

It is an uncontroversial fact that there is a diachronic process that has come to be known

16

as ‘grammaticalisation’, that is, “[. . . ] the increase of the range of a morpheme advanc-

17

ing from a lexical to a grammatical or from a less grammatical to a more grammatical

18

status [. . . ] ”.1 However, the interpretation of grammaticalisation theory of this empiri-

19

cal phenomenon, namely that grammatical change is unidirectional and irreversible, is

20

highly controversial. The aim of the volume is to put grammaticalisation theory to the

21

test. I will focus on the unidirectionality claim/hypothesis.

22

It is difficult to assess the possible scope of this specific hypothesis in a general way,

23

since grammaticalisation theory in itself is a somewhat moving target. There are several

24

flavours and strengths associated with individual proponents of that theory, and even

25

individual articles sometimes do not seem (at least to me) coherent with respect to the

26

assumed scope of grammaticalisation and grammaticalisation theory.2

27

Some scholars, e.g. Roberts & Roussou (2003), go for a maximalist position and argue

28

that all instances of grammatical change affecting meaning and sounds (and not ‘simple’

29

reshuffling of SOV to SVO) are unidirectional. This position is open to testing, since

30

unidirectionality is in such a case an empirical claim. On the extreme opposite of the

31

spectrum, unidirectionality can be taken simply as a definitional property of grammat-

32

icalisation (as suggested in Campbell (2001)). In this case, it makes no sense at all to

33

argue against unidirectionality: a phenomenon that fails to show it, by definition, is not

34

grammaticalisation.

35

Still others however, e.g., Heine (2003: 575) describe grammaticalisation theory to

36

“hypothesize” that the change of grammatical forms is “essentially unidirectional ”, while

37

admitting at least in principle counterexamples. Such a position can be interpreted in

38

at least two ways: first of all, there are certain configurations that favour reversal of the 1. Cited from (Kuryłowicz, 1965: 69). For a survey of other definitions, cf. Lehmann (2002: 10ff.). 2. The survey article of Heine (2003) is such an example, or also Haspelmath (1999).

2

39

direction and override grammaticalisation.3 If this interpretation should be correct, as

40

far as I know, nobody has proposed clear criteria to determine what these overriding

41

configurations are; it would be simply a blind spot for grammaticalisation theory. A sec-

42

ond interpretation would be to suppose that the unidirectionality of grammaticalisation

43

is merely the statistically dominant type of grammatical change. Such a hypothesis is

44

explicitely evoked in Heine (2003: 599), however without giving an estimate to what

45

degree unidirectionality holds.4 While in principle both interpretations of this position

46

are empirically testable, such a move seriously damages the claims to the explanatory

47

power of grammaticalisation theory. It would be by no means any longer the study of

48

“the way grammatical form arise and develop through space and time”, as claimed by

49

Heine (2003: 576). Furthermore, if the statistical information should prevail, disproving

50

it would mean large-scale investigation on attested grammaticalisation phenoma, type

51

by type, in order to determine to what degree unidirectionality prevails. In this paper,

52

I cannot provide any large-scale investigation of such a type.

53

Instead, I will be more modest and argue that grammaticalisation does not equate

54

grammatical change, against Roberts & Roussou (2003). I will study a case of gram-

55

matical change which goes against a generally assumed path-way of grammaticalisation,

56

based on the study of the diachronic evolution of the German indefinite article ein from

57

Middle High German (around 1050–1350 CE) to contemporary German. In order to make

58

my point more clearly, I will try to establish clear criteria on how to detect “semantic

59

bleaching”, which is standardly seen as the semantic correlate of grammaticalisation.

60

This is necessary because at least in written standard German, no morpho-phonological

61

consequence of the grammatical change can be detected,5 and the change is purely 3. A less generous interpretation would be that such a position is a pure immunisation strategy against criticism, since such an assertion is extremely difficult to refute. 4. This, of course, would be crucial for evaluating claims on historical reconstructions based on the hypothesis that change is “essentially unidirectional”. Janda & Joseph (2003) have proposed to mark any reconstruction with indication of certainty. 5. In spoken standard German, this is not quite true: there is version of the indefinite article that is very much reduced, cf. ne in (1) instead of full eine:

3

62

semantico-pragmatic in nature.

63

The case study makes also clear the danger in using grammaticalisation theory in

64

reconstructing older stages purely based on synchronic variation. In the absence of di-

65

achronic data, a reconstruction of earlier stages of the German indefinite article based

66

on standard assumptions of grammaticalisation theory would go completely astray.

67

2 The data

68

I will present the data concerning the indefinite article ein. Following a general trend

69

in diachronic evolution, the German indefinite article derives from the numeral one. In

70

contemporary standard German (and also in earlier varieties), ein corresponds as well

71

to English a — the indefinite article — as well as to English one — the numeral. I will

72

start with the contemporary varieties, and then move on to the Middle High German

73

data (mainly with respect to the Nibelungenlied ), before discussing Martin Luther’s

74

ideolect. The rationale of this rather unconventional order is to point out that there are

75

circumstances where generally recognised semantico-pragmatic and morpho-phonological

76

criteria for grammaticalisation point unanimously into the same direction: the false one.

77

2.1 The Grammaticalisation Pathway according to Heine (1997)

78

In this section I will briefly review the pathway of indefinite articles as exposed in Heine

79

(1997: 66ff.). First of all, Heine presents several properties that hold for grammaticalised

80

indefinite articles. Among others, indefinite articles are stressless, and they tend to be

81

confined to determining the singular of count nouns (which is the case in both English (1)

Ich hab ne Katze gesehen. I have a cat seen. ‘I have seen a cat.’ However, ne is an allomorph of the full eine, as long as the latter is not emphasized; ne is never obligatory.

4

82

and contemporary standard German). He establishes the following implicational scale

83

for the distribution of indefinite articles:

84

(2)

85

This scale is couched in a more general, five-stage model of the evolution of the indefinite

86

article (Heine, 1997: 71ff.). I will be extremely sketchy with respect to the model, since

87

the only stages of interest to us in this article are Heine’s stages IV and V. Stage I

88

corresponds to the stage where the form is used only as a numeral. According to Heine,

89

stage IV corresponds to the indefinite article in languages like contemporary English

90

and German (see section 2.2 for a brief description of this stage). In stage V, we would

91

have a completely unrestricted, “generalized ” indefinite article, which can occur with all

92

types of nouns (i.e., also plural, and eventually mass). At stage V, all number-specific

93

behaviour has been bleached out.

Mass Nouns > Plural Nouns > Singular Count Nouns6

94

As has been noticed by Givón (1981: 51), as well as Heine, the development of the

95

indefinite article can be seen as a paradigm case of “semantic bleaching”, that is, the loss

96

of semantic substance commonly associated with a process of grammaticalisation.

97

Now, let us have a brief look at what a stage four indefinite article looks like.

98

2.2 Contemporary Standard German

99

As already mentioned, in the current standard variety of German, the indefinite article

100

exhibits a distribution that is familiar from other European languages. Indefinite articles

101

combine felicitously with singular count nouns. However, with unmodified 7 mass nouns

102

or plural count nouns, the indefinite article is either ungrammatical (cf. (3a)) or induces

103

a strong coercion effect from mass to count (cf. (3b)). This is to be explained, according 6. This scale does not appear as such in Heine (1997: 68f.), but it is a consequence of his properties 4–6. 7. For some reason, the presence of any kind of modification makes a mass noun much more amenable to indefinite determination.

5

104

to Heine, as an effect of the retained numeral meaning at this stage.

105

(3)

106

a.

Wir haben eine Katze | *ein Geld | *eine Katzen. We have a cat |a money | a cats.

b.

Ich hab ein Wasser getrunken. I have a water drunken. ‘I have drunken (some salient quantity of/some sort of) water.’

107

108

On a discourse level, the indefinite article is generally taken to introduce new discourse

109

referents (cf. Kamp & Reyle, 1993). Anaphora referring back to already mentioned dis-

110

course participants generally require definite articles or demonstratives, as shown in (4):

111

(4)

112

a.

Es war einmal ein König.113 It was once a king. ‘Once there was a king.’ 114

b.

Dieser König hatte eine Tochter. That king had a daughter. ‘The king had a daughter.’

115

As the indefinite article of contemporary standard German is virtually identical to the

116

English indefinite, I will not further pursue the presentation here, and turn instead to the

117

more interesting states of affairs that are attested in some Southern German varieties.8

118

2.3 The indefinite article in contemporary Southern German varieties

119

120

Contrary to the contemporary standard German, in the Alemannic dialect investigated

121

here, indefinite articles combine felicitously with singular count nouns and with unmod-

122

ified mass nouns, but not with plural count nouns:

123

(5)

mir hond { an öpfl | a wassr | *a öpfl } im kaschta. we have { an apple | a water | a apples } in the cupboard.

8. The dialect used here is a Low Alemannic dialect spoken in Vorarlberg (Austria). Various persons have pointed out that not all indefinite articles in all Alemannic dialects behave the same, but as far as I am aware, the indefinite article in the dialects of Vorarlberg are consistent with the judgements given for the specific dialect. Furthermore, Bavarian dialects in Austria also display at least some patterns that make them look like the studied dialect.

6

124

It is important to notice that the acceptability of the indefinite with a mass noun is

125

not a case of coercion of a mass-denoting expression into a count noun: this dialect has

126

separate forms for the definite article and the cardinality expression 1, just like English

127

a vs. one). The cardinality expression uj is infelicitous in contexts where the indefinite

128

is acceptable:

129

(6)

130

Küntsch mr { a gealt | *uj gealt } gea? Could you me { a money | one money } give? ‘Could you give me (some | one) money?’

131

Money is notoriously difficult to coerce into a count noun. As one should expect, the

132

cardinality expression 1, which requires coercion into a count expression (either as a

133

subkind, or a salient quantity), is infelicitous. However, the indefinite article does not

134

only remain fully acceptable in such a context,9 indeed, (7) with the indefinite article is

135

the standard way of asking somebody for money.

136

However, one can not conclude that the indefinite is a purely morphological reflex,

137

and has no semantic meaning attached to it at all. Mass nouns in this variety do not

138

always accept the indefinite article. For instance, they are unacceptable with generic

139

statements:

140

(8)

a. *an schnee isch wiiss10 a snow is white.

b. *a wassr isch duursichtig a water is transparent.

141

142

Intuitively, the indefinite article seems to entail some notion of quantity, and being

143

white or being transparent is not a property of some quantity of snow or water, but of 9. Such sentences are attested also for Bavarian varieties in Austria (web example, attributed to an agent of an employment centre in Vienna): (7)

sans froh dass wir so an sozialschmarotza wie ihnan übahaupt a göd gem be glad that we such a social welfare freeloader like you at all a money give ‘Be glad that we give money to social welfare freeloaders like you at all.’

10. A note on the transcription for the dialect: I write a double vowel in order to indicate a long vowel. Thus, “ii” corresponds to [i:].

7

144

snow or water in general. This is why the indefinite is infelicitous in (8).

145

Let us recapitulate the data. The Southern German indefinite article shows less re-

146

strictions than its standard German correspondent (thus, one might suppose that it

147

has undergone semantic bleaching). Furthermore, the Southern German indefinite is re-

148

duced to a schwa with case endings, whereas the standard German form retains more

149

phonological substance (thus, one might suppose that the southern German version has

150

undergone phonological erosion). So, whereas the contemporary standard German indef-

151

inite article is at stage IV, the southern German indefinite article clearly is beyond that

152

stage, and situated at least at halfway to stage V.

153

assumptions from grammaticalisation theory, one may be tempted to reconstruct the fol-

154

lowing scenario of the diachronic development: the Southern German indefinite article

155

has undergone grammaticalisation from a base that looked roughly like contemporary

156

standard German.

157

11

Based on these facts and standard

However, in applying grammaticalisation theory in this way, one would fail the target,

158

as the Middle High German data shows.

159

2.4 The indefinite article in Middle High German

160

The data I present here has been known for quite some time in the literature on Middle

161

High German (cf. Paul et al. (1982)). However, as far as I know, it has never entered

162

the discussion on grammaticalisation, or been discussed in a theoretical framework. In

163

any case, it has not informed the work on indefinites in Heine (1997).

164

The Middle High German indefinite article combines felicitously with singular count

165

nouns (cf. (9a)), mass nouns (cf. (9b)) and at least some plural nouns (namely pluralia

166

tantums, cf. (9c)):

11. Actually, the distribution of the southern German indefinite article goes against the implicational scale in (2), so things are not quite as neat as they may seem.

8

167

(9)

a.

iu hât erworben iuwer hant || ein schœne wîp unde ein lant.12 youDat has earned your hand a beautiful women and a country. ‘Your hand has earned you a beautiful women and a country.’

b.

ouch enwas her Îwein niht verzaget: || im hete diu minne einen also not-was lord I. not despaired: him had the love a muot || gegeben. . . 13 courage given. . . ‘Lord Iwein was not despaired: love had given him courage.’

c.

Daz was in einen zîten, dô vrou Helche erstarp 14 That was in aPl timesPl , when lady H. died. ‘It succeeded at the time when lady Helche had died.’

168

169

170

171

172

173

174

Thus, the Middle-High German indefinite article seems to be a full-fledged stage V

175

article.

176

There are more surprising facts about this article than its distribution with nonsingular

177

count nouns. We see that indefinites also appear in generic comparisons with mass nouns:

178

(10)

180

Sol aber mir iemer mê || geliuhten dur die naht || noch wîzer danne Shall but me ever more light through the night still whiter than ein snê || ir lîp vil wol geslaht? 15 a snow her body much well noble. ‘Shall her very beautiful body, whiter than snow, ever light again through the

181

night for me?’

179

182

‘Being white’ is not a property of a quantity of snow, but of snow in general. Such a

183

construction is infelicitous in contemporary Southern German (cf. (11a.)) and also in

184

contemporary standard German (cf. (11b)).

185

(11)

a. *sia isch so wiiss as wia an schnee. she is as white as like a snow.

12. Hartmann von Aue: Iwein: v. 2781-82. The double bar “||” indicates the end of a verse. 13. Hartmann von Aue, Iwein, v. 1419-21. 14. Nibelungenlied, 20, 1143. If not mentionned otherwise, all references to the Nibelungenlied follow the edition by Karl Bartsch and Helmut de Boor, published by Reclam, Stuttgart. 15. Heinrich von Morungen, Minnesangs Frühling, 30:1,2-5.

9

b. *Sie ist so weiß wie ein Schnee. she is so white like a snow.

186

187

Generics are often associated with bare nouns (cf. Carlson & Pelletier, 1995). The fact

188

of an article combining in a generic setting is often taken to be the end of a grammati-

189

calisation process.

190

Finally, Middle High German indefinite articles display also rather uncommon dis-

191

course effects. They appear with discourse-old referents (cf. Paul et al. (1982)) which is

192

impossible in any contemporary variety of German I know.

193

(12)

a.

“daz tuon ich”, sô sprach Gunther, ein riter küen’ unde balt.16 that do I, so spoke G., a knight brave and bold ’“I shall do that”, said Gunther, the brave and bold knight.’

b.

des antwrte Gernot ein riter chüen vnt gemeit17 a knight bold and proud on that responded G. ‘To that responded Gernot, a bold and proud knight.’

194

195

196

197

The sentences in (12) as such are both acceptable in contemporary German, as long as

198

Gunther or Gernot are introduced with the respective verse. However, this is not the case

199

here. In the Nibelungenlied, both persons are mentioned extensively several times and

200

immediately before the occurrences of these sentences. In modern German, one would

201

have to use a definite or a demonstrative in such a context.

202

The indefinite article is not obligatory in Middle High German for such contexts,

203

either. Like in modern German, one finds as well the definite article:

204

(13)

205

206

Úrloup dô némen wolde Sîvrit der hélt guot.18 Farewell then take wanted S. the hero good. ‘Then, Siegfried, the good hero, wanted to take farewell.’

A special case of this pattern of having very unrestricted specific indefinites can be seen

16. Nibelungenlied, 7, 391. Gunther is introduced at the very beginning of the poem. 17. NibelungenB, 1474, http://public.univie.ac.at/fileadmin/user_upload/inst_germanistik/ NibHsBReichert.txt. Gernot is mentionned several times immediately before this verse. 18. Nibelungenlied, 5, 320.

10

207

in the fact that the indefinite article may also appear in vocatives (cf. Paul et al. (1982);

208

Hildebrand & Wunderlich (1984)):

209

(14)

a.

210

genâde, ein sælic wîp19211 merci, a praised woman. 212 ‘Merci, o praised woman!’

b.

trœste, ein süeZe minne, mich20 comfort, a sweet love, me. ‘Comfort me, o sweet love.’

213

Ein once again is not obligatory in vocatives: the NP may stand without determiner

214

(cf. (15b)). One finds variation in one and the same text, in (nearly) the same environ-

215

ments:

216

(15)

a.

sît willekomen, her Sîvrit, ein edel riter guot.21 Be welcome, Lord S., a noble knight good. ‘Welcome, ô Lord Sîvrit, you noble and good knight.’

b.

sît willekomen, her Sîvrit, riter lobelîch.22 Be welcome, Lord S., knight glorious. ‘Welcome, ô Lord Sîvrit, glorious knight.’

217

218

219

220

Vocatives with indefinites (as well as with definite articles) are completely impossible

221

in any modern variety of German I know:

222

(16)

223

a.

Hey, { (*d) | (*a) } Mari, kum ummea! Hey, { the | a } Mary, come over here!

b.

{ (*die) | (*eine) } Königin, erhöre mich! { the | a } queen, answer my prayers!

224

To sum up, we have seen so far that the Middle High German indefinite article disposes

225

of more admissible contexts of use than the indefinite article of contemporary German

226

varieties. It clearly should be classified within stage V of Heine’s development scale.

227

Within the contemporary varieties, the Southern dialect investigated disposes of more

228

admissible contexts than the standard variety, and should probably seen as a “stage 19. MS, 1,200. All examples in (14) taken from the Textarchiv des Mittelhochdeutschen Wörterbuchs. 20. MS. 1,198. 21. Nibelungenlied, 5, 292. 22. Nibelungenlied, 9, 552.

11

229

IV+”.

230

The data presented in this section shows clearly that the Southern German indefinites

231

can not be the fruit of a grammaticalisation process starting from a state of affairs similar

232

to that of contemporary standard German. On the contrary, we will see that there is a

233

process leading from the Middle High German state to the contemporary states. I will

234

examine in what follows the writings of Martin Luther.

235

2.5 The Indefinite Article in Luther’s German

236

Martin Luther’s ideolect is probably the best-documented in older German history. Fur-

237

thermore, he exemplifies a more northern variety than the Middle High German examples

238

and the Austrian Alemannic variety we have been dealing with so far. Last, but not least,

239

Luther’s German was a rather influential ideolect.

240

It is therefore interesting to notice that Luther’s German provides us with an in-

241

termediate position between the Middle High German examples and the uses of the

242

modern varieties. I will focus here on properties of the indefinite article that set apart

243

the different varieties.23

244

245

First of all, indefinites appear with mass nouns: (17)

a.

‘Torrentem’ vocat, non dicit calicem, Torrens heist ein wasser, das ‘Torrentem’ calls, not says ‘calicem’, Torrens is a water, which 24 da geschwind leufft, ein starck wasser. there fast runs, a strong water.

b.

[. . . ] denn spiritus heist ein wind, da mit got uns anblesit und [. . . ] because spiritus means a wind, with which god us on-blows and macht geistlich menschenn aus uns [. . . ]25 makes spiritual men of us [. . . ]

246

247

248

23. All examples have been cited following Luther (2002). 24. Schriften, 41. Band, [Predigten des Jahres 1535], Sermo 9. et ultimus dominica 3. (S. 234a) 25. Schriften, 8. Band, Deutsche Auslegung des 67. (68.) Psalmes. 1521, Der Ixvij Psalm von dem Ostertag Hymelfart unnd Pfingstag (p. 5).

12

c.

249

250

Aber das bleybt gleych wol, das ein gelt, das man auff eynen But it remains same yet, that a money, that one on a grund mag legen, [. . . ]26 ground may lay, [. . . ]

251

Notice that in (17a-b), the indefinite article is support for a translation, X means Y . In

252

modern German use, here one would most normally use a bar noun. Luther frequently

253

(but not always, cf. (19a)) uses ein wasser to in order to denote some body of flowing

254

water (cf. Latin torrens), where in modern German, one would probably use the collective

255

Gewässer ). So, this might be seen as a case of a “lexicalised”, derivational use of the

256

indefinite article. In (17c), however, we have a combination with the indefinite and

257

money, which cannot fall into that class. Note that in all cases in (17), we have a

258

subordinate clause or an adjectival modifier.

259

260

However, we also find cases where the mass noun is not modified: (18)

261

Ich thun ein wein in ein gulden gefeß und thun desselben weins I put a wine in a golden recipient and put of the same wine.GEN auch in ein huelczern gefeß, [. . . ]27 also in a wooden recipient, [. . . ]

262

(18) is particularly revealing in that it combines a construction indefinite plus mass noun

263

with a sort of genitive partitive construction (which has not survived in the modern

264

varieties of German under consideration here).

265

Then, the indefinite article appears in comparatives with mass, and in generic com-

266

parison (cf. (19b)):

267

(19)

268

a.

[. . . ] es ist ein leib und blut, der vol Gottes ist oder das [. . . ] it is a body and blood, which full of God is or which durchgoettert ist wie ein wasser, das durchzuckert ist [. . . ]28 through-godded is like a water

26. Schriften, 6. Band, (Großer) Sermon von dem Wucher. 1520, IHESVS. Eyn Sermon von dem Wucher Doctoris Martini Luther Augustiner zu Wittenbergk. (p. 54). 27. Schriften, 10. III Band, Predigten des Jahres 1522, Der fünfft Sermon martini Luthers uff den nechsten Suntag vor Simonis und Jude uff dem schlos gescheeen. 28. Schriften, 33. Band, Wochenpredigten über Johannes 6-8, Die 17. Predigt am 15. tag Aprilis, oder

13

b.

269

270

So ist auch das Manna weis gewesen gleich wie ein Taw, wie ein schnee So is also the manna white been as like a dew, like a snow 29 oder reiff. or ice.

271

As already observed in (11), sentences like (19b) are not possible in modern varieties of

272

German, whereas they were attested in Middle High German.

273

Finally, let me come to phenomena I did not find in Luther’s German: discourse-

274

old infinites, vocatives with infinitives and indefinites with pluralia tantums. As such

275

elements are difficult to check by an automatized research, I cannot guarantee that

276

there are none; I will however indicate for each case how I searched (so that future

277

research may improve on my tentatives).

278

With respect to the discourse-behaviour of indefinites, I have checked the Genesis of

279

Luther’s translation of the Bible. The rationale behind this choice was the following: odd

280

discourse behaviour of indefinites appears most clearly in longer narrative passages, and

281

the Genesis provides such narrative passages. In Luther’s Genesis, the indefinite article

282

has the same discourse properties as the modern German varieties. The only suspicious

283

case I found was the following:

284

(20)

286

Vnd nennet jn, den heimlichen Rat, Vnd gab jm ein weib Asnath And called him, the secret counsellor, and gave him a woman Asnath die tochter Potiphera des Priesters zu On.30 the daughter of Potiphera the priest at On. ‘And he named him minister, and gave him Asnath, the daughter of Potiphera,

287

the priest of On, to be his wife.’

285

288

However, (20) is an example of a ‘lexical’ use (maybe an idiomatic expression) of the

289

indefinite article, namely the “to give X somebody a N” construction, which is frequently den Sonnabend nach Quasi modo geniti. 29. Schriften 16. Band, Predigten über das 2. Buch Mose 1524-27, Auslegung D. Mar: Luthers, uber etliche Capitel des andern Buchs Mosi, Geprediget zu Wittemberg, Anno 1524. 1525. und 1526, (p. 307c). 30. Genesis, XLI, 45.

14

290

attested in Middle High German (cf. (21)), but which exists no longer in contemporary

291

standard and southern German varieties. Its meaning can be roughly given as “X is

292

given somebody as a N”. Today, one would use in such a context “zur Frau geben” (lit.

293

‘to give to the woman’).

294

(21)

296

gîstu mir dîne swester, sô wil ich ez tuon || die scœnen Kriemhilde, ein give-you me your sister, so want I it do the beautiful K., a 31 küneginne hêr. noble. queen ‘If you give me your sister, the beautiful Kriemhild, as the noble queen, I shall

297

do it.’

295

298

I have not found indefinites with pluralia tantum, nor any case of a vocative with an

299

indefinite. For both cases, I checked what I assumed to be the ‘usual suspects’.

300

the vocative examples, I rely additionally on the authority of Grimm’s dictionary, which

301

states that “there exist only some Middle High German examples for the use of the

302

indefinite article before a vocative”.33 Notice, however, that even if there were vocatives,

303

pluralia tantums and discourse-old indefinite articles in Luther’s writing, this would not

304

disconfirm the following finding: Luther’s indefinite article is less restricted than the ones

305

in the two contemporary varieties under discussion.

306

2.6 Summary of Findings

307

At this point, it will be profitable to summarize the data we have seen so far in a table,

308

in order to contrast the different varieties:

32

For

31. Nibelungenlied, 6, 333. 32. That is, for the pluralia tantum, Leute (‘people’), Eltern ‘parents’, Geschwister (‘siblings’), Kosten (‘costs’), Einkünfte (‘revenues’). For the vocative, I checked Christ, Jungfrau (‘virgin’), König (‘king’), Königin (‘queen’), and Fürst (‘prince’). 33. Cf. Hildebrand & Wunderlich (1984), entry ‘ein’ §D, 10: für den unbestimmten artikel vor dem vocativ des bloszen substantivums stehen nur einige mhd. beispiele zu gebot.

15

309

(22)

MHG

Luther

SG

CSG

unmodified mass

OK

OK

OK

*

plurals

OK

*

*

generic statements mass

OK

*

*

discourse-old

OK

*

*

vocatives

OK

*

*

OK

310

MHG: Middle High German; SG: Contemporary Southern German (Sulzberg);

311

CSG: Contemporary Standard German

312

An ‘OK’ marks a context as possible for an indefinite article; a ‘*’ marks a context as

313

ungrammatical with an indefinite article. In case of Luther, three fields remain white.

314

On one side, this is due to the non-exhaustive nature of the inquiry; on the other hand,

315

negative data — such as grammaticality judgements — are impossible to extract from

316

a corpus. We cannot know what Luther could have said; we only have a (rather large)

317

sample of things he has said or written.

318

If we look at the admissible contexts, one sees that the admissible contexts of use

319

of the indefinite article in each variety form a subset of those situated to its right;

320

i.e., according to the table in (22), the admissible contexts of contemporary varieties

321

form a proper subset of the admissible contexts of earlier varieties.34 The interesting

322

question is now the following: what (if anything) does this data tell us with respect to

323

the phenomenon of grammaticalisation or the theory of grammaticalisation?

34. There are contexts in which in Middle High German the indefinite article was optional where it is obligatory in modern varieties, cf. (23) (Nibelungen, 3, 50). However, I conjecture that this is a consequence of the different status of negation in Middle High German, and not a consequence of a less grammaticalised indefinite article. (23)

a. b.

nie keiser wart sô rîche, der wolde haben wîp35 never emperor become so rich, who wanted have woman Es war nie *(ein) Kaiser so reich, . . . it was never a emperor so rich, . . .

16

324

The reason why I ask this is because for French, where similar (though not quite as

325

extreme) losses of admissible contexts for the indefinite article have been found,36 Carlier

326

(2001) concludes that this does not constitute any evidence against grammaticalisation

327

theory.

328

3 What is Grammaticalisation?

329

As an empirical phenomenon, grammaticalisation has been defined as the change of a

330

lexical item into a grammatical item, or of a less grammatical item into a more gram-

331

matical one. Now, there are several imaginable criteria on how to recognize when an

332

item has become grammaticalised (semantic bleaching, erosion, obligatorification, etc.)

333

In order to assess to validity of a theory claiming like grammaticalisation theory stating

334

that grammatical change works in one direction only, and that this change is irreversible

335

(see Haspelmath, 1999), it is crucial that these criteria be as reliable and generally agreed

336

upon as possible.

337

So, what are the criteria that allow to deduce that grammaticalisation has taken place?

338

I will start the review with Roberts & Roussou (2003), which are very clear how to

339

detect grammaticalisation; namely, by checking for upward movement. Then, I will go

340

on to the more problematic notion of “semantic bleaching”, and try to make it as precise

341

as possible.

342

3.1 Upward movement

343

Roberts & Roussou (2003), who use the minimalist framework, see upward movement as

344

the syntactic counterpart (or definition) of grammaticalisation. Their argument is based

345

on the assumption that there is a universal hierarchy (i.e., stratification) of functional

36. The French indefinite article could be combined at some point with plural nouns, cf. Carlier (2001).

17

346

categories (cf. Cinque (1999)), and elements becoming grammaticalised get access to

347

higher positions in this hierarchy.

348

According to Roberts & Roussou (2003: 208), every grammatical change which has an

349

impact on the category of the item involved and/or is associated with semantic bleaching

350

and phonological reduction entails upward movement in the syntactic structure: before

351

the movement, the item was semantically and phonologically richer.

352

The advantage of this approach is that it gives very clear predictions and is therefore

353

relatively easy to test. Its major inconvenient is that it is very dependent on the particular

354

syntactic theory they use. For instance, one is committed to the existence of a hardcoded

355

hierarchy (in Universal Grammar) of functional categories, which is far from being a

356

universally shared assumption, even in the realm of minimalist theorising.

357

That being said, let us test if our example of indefinite articles in German involve an

358

instance of downward movement, and thus a counterexample to the theory of Roberts

359

& Roussou. In order to prove this, one would have to show that Middle High German

360

indefinite articles had access to positions which were higher in the functional hierarchy

361

than those accessible to them in contemporary varieties. Such proof can indeed be given.

362

In Middle High German, the indefinite article could be placed in some circumstances

363

above the definite article, and co-occur with that latter:

364

(24)

366

er truoc in sîme sinne ein minneclîche meit, || und ouch in ein diu he wore in his sense a lovely girl, and also him aNom theNom frouwe die er noch nie gesach . . . 37 lady whom he still never seen . . . ‘He was in love with a lovely girl, and the lady, whom he had never seen,

367

was in love as well.’

a.

365

368

369

b.

scham ist ein diu hœhste tugent38 sense of honour is a the highest virtue ‘The sense of honour is the highest virtue.’

37. Nibelungenlied, 3, 132. 38. MS, 2, 175.

18

370

Notice that in both cases in (24), we do not have an instance of indefinite article +

371

genitive (cf. ‘one of the’): the definite diu can only be nominative case, as is the indefinite.

372

In current varieties of German, the indefinite is in complementary distribution with the

373

definite article (as shown in (25a-b) for the Southern German variety). Notice that the

374

numeral can co-occur with the definite article, but only in a lower position (cf. (25c-d)).

375

(25)

376

377

378

a. *schpaara isch a d hökschte tugent. to save is a the highest virtue. b. *schpaara isch d a hökschte tugent. to save is the a highest virtue. c.

der ujne maa hot mr gseit dass . . . the one man has me said that . . .

d. *ujn dr maa hot mr gseit dass . . . one the man has me told that . . .

379

So, (24) and (25) show that the Middle High German indefinite article was higher in

380

the functional hierarchy than the contemporary ones. Therefore, the former is more

381

grammaticalised than the latter, and diachronically, the German indefinite article has

382

undergone downward movement. This is thus a clear counterexample to the theory of

383

Roberts & Roussou (2003).

384

But what about more traditional accounts of grammaticalisation? In this case, we will

385

have to look more closely at the notion of “semantic bleaching”.

386

3.2 Semantic Bleaching

387

The notion of semantic bleaching seems to be a non-controversial ingredient of gram-

388

maticalisation theory; all versions I am aware appeal to it in some way or another. The

389

base idea behind semantic bleaching is that an element, as it grammaticalises, loses

390

its restrictions with respect to non-grammatical (or non-logical) elements of its original

391

meaning (cf. von Fintel, 1995).

19

392

The main advantage of semantic bleaching is that it is an a priori theory-neutral

393

approach to the meaning-part of grammaticalisation. Furthermore, the basic idea is very

394

intuitive. There are, however, also some inconvenients attached to it, both empirical and

395

theoretical. Empirically, it is rather difficult to test, especially when there is no native-

396

speaker intuition available — which unfortunately is always the case when working

397

on diachrony. In diachrony, practically, the semantics vs. pragmatics distinction breaks

398

down: while this distinction has considerable heuristic value when working in synchrony,

399

and while there is a battery of tests to distinguish whether some meaning component

400

is semantic or just contextually induced, in diachrony, these levels can normally not

401

be reliably distinguished.39 On a theoretical level, it is hard to come up with a precise

402

definition of what is to count as ‘semantic bleaching’, and it is also not clear that semantic

403

bleaching is a sufficient condition for grammaticalisation; as we will see, the proposed

404

definition is not restricted to grammaticalisation-cases of semantic bleaching.

405

Why has semantic bleaching been associated with grammaticalisation in the first

406

place? The basic reasoning is probably the following: meaning components can be seen

407

as tests on an input: the more meaning there is, the less possible contexts are compatible

408

with it (in other words: the bigger the intension, the smaller the extension). If an item

409

shows a less restricted distribution at one moment of history that at an earlier one, one

410

can assume that there is less meaning associated with that item at the latter time.40

411

412

So, I propose the following definition of semantic bleaching: (26)

a.

An item φ undergoes semantic bleaching between two moments t and t0

413

iff the set of grammatically admissible contexts (noted Γ[φ]) for φ at t is

414

a proper subset of the grammatically admissible contexts for φ at a later 39. Notice, however, that a learner acquiring his native language is in a position similar to the one of the diachronic linguist with respect to semantics vs. pragmatics: he has to learn from what has been called a “text” (cf. Niyogi, 2006: for an overview of learning theory). 40. This reasoning presupposes that at the earlier moment, the more restricted distribution echoes a semantic fact, and not just a purely accidental non-use in contexts which would have been grammatically acceptable.

20

moment t0 ; or, written in set notation:

415

b.

416

{Γ[φ] : Γ[φ] ∈ L at t} ⊂ {Γ[φ] : Γ[φ] ∈ L0 at t0 }

417

where t ≺ t0 , L0 is a descendent of L, x ∈ L means that sentence x is

418

grammatical in language L41

419

Let me elaborate a bit on this definition, starting from its end. First, in order to be

420

maximally comparable, I have to assume that there are not different types of (grammat-

421

ical) context arising and disappearing arbitrarily. Then, we have to make sure there is

422

some sort of language-continuum L − L0 .42 between the moments t and t0 . Let us now

423

turn to the main definition. We cannot consider individual lexical or grammatical items

424

on its own; they come embedded in sentences (which come in discourses). So we will

425

compare sentences in which φ arise. Then, (27) says that we all contexts in which φ was

426

acceptable at t remain acceptable for φ at t0 , and additionally φ becomes acceptable at

427

t0 in contexts where it was inacceptable at t.

428

This definition (which will probably be seen by some to state the obvious, and in

429

failing that) is not without (potential) problems. First of all, there is sometimes no

430

subset relation between the less grammaticalised and the more grammaticalised item,

431

but rather overlap: some contexts are acquired, but others lost. This is not handled

432

within (27), and I think that it should be like that: in such cases, we should assume 41. This definition assumes that the item φ remains in some sort identical in the period ranging from t to t0 . If one would not like to commit to this assumption, one could go for the following definition of semantic bleaching: (27)

a.

b.

φ is a semantically bleached version of ψ iff at a moment t, the set of admissible grammatical contexts Γ for ψ (noted Γ[ψ]) is a proper subset of the admissible grammatical contexts Γ for φ (noted Γ[φ]) at a moment t0 , or φ is a semantically bleached version of ψ iff {Γ[ψ] : Γ[ψ] ∈ L at t} ⊂ {Γ[φ] : Γ[φ] ∈ L0 at t0 } where t ≺ t0 , L0 is a descendent of L, x ∈ L means that sentence x is grammatical in language L, φ is a diachronic continuation of ψ, and under the assumption that the global set of contexts is stable.

42. The notion of language-continuum is just a way of not commiting to a position on the issue whether two diachronic varieties which are distinct at least with respect to the semantics of φ are different languages or not.

21

433

two different processes: one of meaning extension (i.e., bleaching), and another of a

434

competing item entering the arena, eventually eliminating the original occupier from

435

parts of its original distribution.

436

Second, (27) makes clear that semantic bleaching is not restricted to grammatical

437

contexts. Under the given definition, ‘lexical’ bleaching will also fall under the category.

438

This, in my mind, is not a shortcoming either: lexical elements should be able to lose

439

meaning, while still remaining lexical elements.43

440

Third, if one wants to equate the semantic side of grammaticalisation with semantic

441

bleaching, we will use all track of the open-class to closed-class criterion, or the obliga-

442

toriness of grammatical items. I do not think that this is a real problem either: what

443

is obligatory seems to me to be a syntactic problem rather than a semantic one. Put

444

in other words: I doubt it that one could find purely semantic criteria for telling that

445

such and such an information must be encoded as a grammatical category.44 Then, the

446

question is whether being obligatory in a given context is a good indicator for the gram-

447

matical status of a particular item. I suspect that this is just a byproduct of alternative

448

items that might be used in the same context.45

449

However, my main aim is to find a definition that can be applied in the case of

450

the diachrony of the German indefinite articles. (27) gives us a testable prediction: a

451

semantically bleached version should be compatible with more contexts than the original

452

version. Should we observe an augmentation in restrictions (which seem to be of a

453

‘grammatical’ sort), this should constitute evidence for a full stop and a reversal of the

454

grammaticalisation process, independent of a particular definition or proposed pathway.

455

In this sense, there clearly is a reversal of the grammaticalisation process, as shown 43. The evolution of English ‘dog’ is a well-known and illustrative case: it was initially restricted to a particular sub-kind of dogs, but bleached out and is now used for the whole dog-kind. 44. Von Fintel (1995) has made such a tentative, but I think that this article misses the gradual nature of grammaticalisation, in opposing strictly lexical vs. grammatical categories. 45. For instance, Spanish has what is often described as two forms for the imperfect subjunctive, hubiera and hubiese, which are largely (but not entirely) substituable one for the other. This does not prevent any of these items from being grammatical.

22

456

in (22), and evidence for semantic strengthening. Therefore, I conclude that we have a

457

counterexample to grammaticalisation that any theory must acknowledge. Notice that

458

the definition of semantic bleaching does not pay any attention to why a particular item

459

undergoes loss or strengthening of its meaning. The only thing that matters is whether

460

there is strengthening or loss. Therefore, an argument like the one by Carlier (2001),

461

according to which the loss of admissible contexts of the French indefinite article un

462

does not constitute a counter-example to grammaticalisation theory because that loss

463

was induced by phonological reasons, is to be rejected on this base.

464

4 Conclusion and perspectives

465

In this paper, I have presented a grammatical item, namely the German indefinite ar-

466

ticle, losing admissible context through its history from the Middle High German stage

467

on. I have shown that it constitutes an instance of syntactic downward movement and an

468

instance of what one may call ‘semantic recolouring’, providing decisive counter-evidence

469

against Roberts & Roussou (2003). Grammatical change with meaning effects doesn’t al-

470

ways involve upward movement or semantic bleaching. Grammatical change is reversible,

471

or at least, does not always go in the predicted direction (against Haspelmath (1999),

472

and following Campbell (2001)): therefore, grammaticalisation theory, if claiming that

473

the evolution of a grammatical forms through space and time is always unidirectional,

474

must be false. The data provided here also shows that reconstructions based purely on

475

synchronic evidence are unreliable and should be avoided.

476

But where should we go to from now? I think that what must be investigated in detail

477

is under which circumstances one observes three things: (i) the stability of grammatical

478

forms; (ii) grammaticalisation; and (iii) the reversal of grammaticalisation. It also would

479

be important to know to what extent grammatical change goes into one sense or the

480

other. French and German indefinite articles have experienced a reduction of their admis-

23

481

sible contexts against a predicted pathway; the present perfect tenses of Portuguese and

482

some Spanish varieties of Latin America have experienced a reduction of their admissi-

483

ble contexts against a predicted pathway. In order to understand grammatical change

484

in a more general way, such phenomena have to be carefully described and analysed;

485

they should not be dismissed as unimportant exceptions to an overwhelming general

486

tendency. Our final aim should be to understand grammatical stability and change; and

487

not just some particular kind of it.

488

Let me finish with a brief remark. Grammaticalisation is often said to be a consequence

489

of language use, or the “harnessing of meaning by a grammar” (cf., e.g., Hyman, 1984).

490

But as we have seen, grammaticalisation theory seems to feel concerned only by the loss of

491

meaning. Somehow paradoxically, on the other hand, if we look at pragmatics — at least

492

in the Gricean tradition —, that is the discipline of linguistics concerned with (mainly

493

synchronic) language use and its impact on meaning, the general consensus is to study

494

exclusively the strengthening of meaning. Clearly, unilateral synchronic strengthening

495

cannot lead to unilateral diachronic bleaching. There is little doubt that this state of

496

affairs is the consequence of selective perception on both sides. But if the two perspectives

497

could be put together, there would be new and exciting discoveries to be made.

498

References

499

Lyle Campbell (2001). “What’s Wrong with Grammaticalization?” In: Language Sci-

500

501

502

503

504

ences 23 , pp. 113–161. Anne Carlier (2001). “La genèse de l’article un”. In: Langue française 130 , 1, pp. 65–88. Gregory N. Carlson, Francis J. Pelletier (eds.) (1995). The Generic Book . University of Chicago Press.

24

505

506

507

508

509

510

511

512

513

514

515

516

Guglielmo Cinque (1999). Adverbs and Functional Heads. A Cross-Linguistic Perspective. New York: Oxford University Press. Kai von Fintel (1995). “The Formal Semantics of Grammaticalization”. In: Jill N. Beckman (ed.), Proceedings of NELS 25 . Amherst: GSLA, vol. 2, pp. 175–189. Kurt Gärtner, Klaus Grubmüller, Karl Stackmann (eds.) (2006–).

Mittel-

hochdeutsches Wörterbuch. Stuttgart: S. Hirzel Verlag. Talmy Givón (1981). “On the development of the numeral ‘one’ as an indefinite marker”. In: Folia Linguistica Historica 2 , 1, pp. 35–53. Martin Haspelmath (1999). “Why is Grammaticalization Irreversible?” In: Linguistics 37 , 6, pp. 1043–1068. Bernd Heine (1997). Cognitive Foundations of Grammar . Oxford: Oxford University Press.

517

— (2003). “Grammaticalization”. In: Joseph & Janda (2003), pp. 575–601.

518

Rudolf Hildebrand, Hermann Wunderlich (eds.) (1984). Deutsches Wörterbuch von

519

Jacob und Wilhelm Grimm. München: dtv. URL http://germazope.uni-trier.de/

520

Projects/DWB.

521

522

Paul J. Hopper, Elizabeth Traugott (1993). Grammaticalization. Cambridge: Cambridge University Press.

523

Larry Hyman (1984). “Form and Substance in Language Universals”. In: Brian Butter-

524

worth, Bernard Comrie, Östen Dahl (eds.), Explanations for Language Universals.

525

Berlin: Mouton, pp. 67–85.

25

526

Richard D. Janda, Brian D. Joseph (2003). “On Language, Change, and Language

527

Change — Or, Of History, Linguistics and Historical Linguistics”. In: Joseph & Janda

528

(2003), pp. 3–180.

529

530

Brian D. Joseph, Richard D. Janda (eds.) (2003). The Handbook of Historical Linguistics. Oxford: Blackwell.

531

Hans Kamp, Uwe Reyle (1993). From Discourse to Logic. Introduction to Modelthe-

532

oretic Semantics of Natural Language, Formal Logic and Discourse Representation

533

Theory. Dordrecht: Kluwer.

534

535

Jerzy Kuryłowicz (1965). “The Evolution of Grammatical Categories”. In: Diogenes 51 , pp. 55–71.

536

Christian Lehmann (2002). Thoughts on Grammaticalization. Arbeitspapiere des Sem-

537

inars für Sprachwissenschaft der Universität Erfurt. Erfurt: Universität Erfurt, 2 edn.

538

Martin Luther (2002). “Luthers Werke auf CD-ROM”. 2 CD’s, ProQuest Information

539

540

541

542

543

544

545

and Learning Company, Third Release. Partha Niyogi (2006). The Computational Nature of Language Learning and Evolution. Cambridge, MA: MIT Press. Hermann Paul, Hugo Moser, Ingebord Schöbler, Siegfried Grosse (1982). Mittelhochdeutsche Grammatik . Tübingen: Niemeyer, 22 edn. Ian Roberts, Anna Roussou (2003). Syntactic Change. A Minimalist Approach to Grammaticalisation. Cambridge: Cambridge University Press.

26