We wish to express our gratitude to our colleagues Elizabeth Traugott, Arie Verhagen, and Bert Cornillie for their cooperation in preparing this paper. The paper is based on data collected by Hiroyuki Miyashita within a Ph.D. dissertation project.
1A primary goal of linguistics is to describe languages and to explain why they are structured the way they are. To this end, this paper attempts to account for a set of constructions of Modern High German all involving the verb drohen “to threaten”. The four different constructions associated with this item are illustrated in (1). We will refer to them as drohen‑1 through drohen‑4.
(1)
|
German
|
|
a.
|
Karl
|
droht
|
seinem Chef,
|
ihn
|
zu verklagen.
|
drohen‑1
|
|
|
Karl
|
threatens
|
to his boss
|
him
|
to sue
|
|
|
|
‘Karl threatens his boss to take him to court.’
|
|
b.
|
Uns
|
droht
|
nun
|
eine Katastrophe.
|
drohen‑2
|
|
|
to.us
|
threatens
|
now
|
a disaster
|
|
|
|
‘We are now threatened by a disaster.’
|
|
c.
|
Das Hochwasser
|
droht
|
die Altstadt
|
zu überschwemmen.
|
drohen‑3
|
|
|
the flood
|
threatens
|
the old town
|
to flood
|
|
|
|
‘The flood risks flooding the old town.’
|
|
d.
|
Mein Mann
|
droht
|
krank
|
zu werden.
|
drohen‑4
|
|
|
my husband
|
threatens
|
sick
|
to become
|
|
|
|
‘My husband risks falling ill/threatens to fall ill.’
|
- 1 One may wonder why we distinguish between (1c) and (1d) considering the fact that in Table 1 they (...)
2In syntactic terms, drohen in (1a) is described as a lexical main verb, while in (1c) it tends to be described as a “semi-modal” raising verb, more specifically as a subject-to-subject raising verb. We are not aware of any coherent description of the remaining two types of construction.1 The main properties distinguishing these constructions are summarized in Table 1.
Construction
|
Subject referent is human
|
drohen takes a subject argument
|
drohen expresses
a speech act
|
Meaning of drohen
|
drohen‑1
|
+
|
+
|
+
|
Lexical
|
drohen‑2
|
-
|
+
|
-
|
Functional
|
drohen‑3
|
-
|
-
|
-
|
Functional
|
drohen‑4
|
+
|
-
|
-
|
Functional
|
Table 1. Distinguishing properties of the four German drohen constructions.
- 2 In the French literature corresponding uses of the verbs menacer “to threaten“ and promettre “to p (...)
3Semantically, there is a twofold distinction, characterized in (2). According to this distinction there is on the one hand drohen‑1, which is generally described as a lexical verb, more specifically an activity (or manner) verb, a commissive verb, and a control verb taking an agent participant as its subject able to control the action described by drohen, and having the meaning sketched in (2a), roughly paraphrasable as “to intend to inflict something negative on someone” (Traugott 1993: 349).2 In the relevant literature, drohen‑1, or its equivalents in other Western European languages, is referred to as descriptive, performative, objective, or lexical drohen; we will henceforth call it the lexical drohen. On the other hand, there is what we will call functional drohen, paraphrased in (2b), which characterizes all remaining constructions, that is, drohen‑2, -3, and -4. It is referred to variously as epistemic, subjective, modal, semi-modal, evidential, or temporal-aspectual drohen.
(2)
|
A paraphrase of the meaning of German drohen‑1 (2a) vs. drohen‑2-4 (2b):
|
|
a.
|
“Someone points out that s/he intends to do something that is undesirable to someone else.”
|
|
b.
|
“Something undesirable is about to happen.”
|
- 3 German functionaldrohen, or its equivalent in languages such as Dutch (dreigen), English (threaten(...)
4It is widely held that the primary function of drohen‑2 in (2b) is to express epistemic modality.3 However, this is not a comprehensive characterization. There are the following components of meaning in addition: (i) the lexical concept “something undesirable happens”, (ii) aspectuality, in that almost invariably its use involves a change of state, and (iii) deictic time, in that the event expressed by the infinitival verb is conceived of as taking place later than at reference time – that is, a concept that in a number of languages is expressed by a distinct functional category, usually referred to as the near or immediate future tense (cf. Diewald 2004).
5The distinctions outlined above raise a number of questions, in particular the ones listed in (3).
(3)
|
Questions on German drohen
|
|
a.
|
Why are there four different constructions involving one and the same item drohen – one that behaves morphologically like a verb?
|
|
b.
|
Assuming that these different uses of drohen are not coincidental, how to account for this?
|
|
c.
|
Why are there two different meanings expressed by drohen, a lexical and a functional one?
|
|
d.
|
Whereas drohen‑1 exhibits a rich morphosyntactic potential which is characteristic of lexical verbs, all other constructions have only a severely limited range of morphosyntactic options – why is this so?
|
|
e.
|
Why does (1b) behave syntactically but not semantically like a lexical verb – exhibiting meaning (2b), that is, that of a functional marker, like (1c) and (1d)?
|
|
f.
|
Why is (1d) frequently ambiguous, in that it can be interpreted either with reference to lexical or to functional drohen? The example in (1d) is suggestive of functional drohen (2b), but in specific contexts it can also be understood as expressing lexical drohen: For example, when (1d) is placed in a context of the kind presented in (4), where the subject referent is portrayed as being a controlling agent, there more likely is an interpretation with reference to the lexical drohen of (2a).
|
|
g.
|
Why does drohen exhibit two contrasting forms of syntactic behavior, in that it controls the subject in the case of lexical drohen but is a subject-to-subject raising verb that is controlled by the infinitival complement in the case of functional drohen‑3, as the passive sentence in (5b) shows)?
|
(4)
|
Mein Mann
|
droht
|
krank
|
zu werden,
|
falls
|
ihm
|
|
my husband
|
threatens
|
sick
|
to become
|
if
|
to him
|
|
der Urlaub
|
verweigert
|
wird.
|
|
the leave
|
refused
|
is.
|
|
‘My husband threatens to fall ill in case he doesn’t get leave.’
|
(5)
|
a.
|
Das Hochwasser
|
droht
|
die Altstadt
|
zu überschwemmen.
|
|
|
the flood
|
threatens
|
the old town
|
to flood
|
|
|
The flood risks flooding the old town.’
|
|
b.
|
Die Altstadt
|
droht
|
vom Hochwasser
|
überschwemmt
|
zu werden.
|
|
|
the old town
|
threatens
|
by the flood
|
flooded
|
to become
|
|
|
‘The old town risks being flooded by the flood.’
|
6In the present paper we aim at looking for answers to these questions, using grammaticalization theory as an explanatory tool. To this end, we need to discuss a number of variables in the following paragraphs and will return to these questions in section 10. In the following section 4 we provide an outline of this theory – at least as far as it affects the questions posed in this section.
7Grammaticalization theory is concerned with describing and explaining the genesis and structure of functional categories on the basis of a restricted set of parameters. The main motivation underlying grammaticalization is to communicate successfully. One salient human strategy consists in using linguistic forms for meanings that are concrete, easily accessible, and/or clearly delineated to also express less concrete, less easily accessible and less clearly delineated meaning contents. To this end, lexical or less grammaticalized linguistic expressions are pressed into service for the expression of more grammaticalized functions. Accordingly, grammaticalization is a process whereby expressions for concrete (e.g., lexical) meanings are used in specific contexts for encoding grammatical meanings.
- 4 Some students of this paradigm of linguistics argue that obligatorification, whereby the use of li (...)
8A wide range of parameters has been proposed (see e.g. Heine, Claudi & Hünnemeyer 1991; Hopper & Traugott [1993]2003; Bybee, Perkins & Pagliuca 1994; Lehmann [1982] 1995). In our model, it is the four parameters listed in (5) (see Heine & Kuteva 2002). Alternative parameters that have been proposed, such as syntacticization, morphologization, obligatorification,4 etc., can be accounted for as being derivative of these parameters.
(6)
|
Parameters of grammaticalization:
|
|
a.
|
extension (or context generalization): use in new contexts suggests new meanings,
|
|
b.
|
desemanticization (or “semantic bleaching”), i.e. loss in meaning content,
|
|
c.
|
decategorialization, i.e. loss in morphosyntactic properties characteristic of lexical or other less grammaticalized forms, and
|
|
d.
|
erosion (or “phonetic reduction”), i.e. loss in phonetic substance.
|
9Some students of grammaticalization do not distinguish desemanticization as a distinct parameter and rather prefer to describe semantic development in terms of notions such as invited inferences, pragmatic strengthening, or subjectification (e.g. Hopper & Traugott [1993]2003). It would seem, however, that the parameter desemanticization allows for more generalizations than any of these alternative notions do. There are many grammaticalization processes that involve desemanticization but no inferencing or strengthening. For example, the process from case-gender-number suffixes in Latin to plural suffixes in Romance languages such as Italian (-i) or Spanish (-s) involved the desemanticization of case and gender but not, as far as we are aware, inferencing, strengthening, or subjectification. In a similar fashion, when the English adjective full was grammaticalized to a derivational suffix –ful, none of these factors appears to have played any significant role. Rather, the adjective was simply decategorialized to a derivational category, losing its status as an independent word.
10While three of these parameters involve loss in properties, there are also gains. In the same way as linguistic items undergoing grammaticalization lose in semantic, morphosyntactic, and phonetic substance, they also gain in properties characteristic of their extension to new contexts. Grammaticalization requires specific contexts to take place and it can be, and has been, described as a product of context-induced reinterpretation. Accordingly, context extension is a crucial factor in shaping the structure of grammatical forms – to the extent that these forms may come to express meanings that cannot immediately be derived from their respective source forms.
11The examples presented in (1) can be interpreted as reflecting a canonical process of grammaticalization leading from lexical/verbal to functional/auxiliary structures. This process has been well described (e.g., Heine 1993; Bybee, Perkins & Pagliuca 1994; Kuteva 2001). Summarily, it can be described as leading from (7a) to (7b).
- 5 For an alternative account of auxiliation within the Chomskyan Minimalist framework, see van Gelde (...)
(7)
|
A morphosyntactic model of auxiliation5:
|
|
a.
|
main verb (V1) - (non-finite verb) complement (C)
|
|
b.
|
auxiliary (A) - main verb (V2)
|
12German drohen can be viewed as a canonical instance of (7). In terms of the constituents distinguished in (7), we predict that (7a) and (7b) differ from one another in the properties listed in (8).
(8)
|
Predictions on the structure of auxiliation on the basis of (7):
|
|
a.
|
Compared to the subject referent of V1, A lacks salient semantic properties, in particular the property of referring to human (or animate) referents (desemanticization).
|
|
b.
|
Whereas V1 denotes the semantics of a lexical verb, A expresses some schematic meaning, most appropriately described as a grammatical function (desemanticization). In the present case of auxiliation, such a function is most likely to relate to aspect, tense, and modality.
|
|
c.
|
While the meaning of A contrasts with that of V1, it can be derived from that of A in a principled way.
|
|
d.
|
Compared to the rich lexical potential of V1, A disposes only of a severely limited range of morphosyntactic options (decategorialization).
|
|
e.
|
The verbal valency is determined by V1 in (6a) but by V2 in (6b).
|
13The predictions in (8) are based on synchronic generalizations. While grammaticalization theory aims at accounting both for synchronic and diachronic phenomena, unlike most other approaches within the paradigm of cognitive linguistics it has a diachronic foundation, in that it rests on generalizations on language use through time and space. This means that its hypotheses can be verified or falsified by means of diachronic evidence. Accordingly, it hypothesizes that the structure of (7a) and the properties associated with it are temporally prior to those of (7b). With reference to German drohen, this leads to the following hypotheses.
(9)
|
Hypotheses on the history of German drohen:
|
|
a.
|
Lexical drohen preceded functional drohen in time, and the latter is historically derived from the former.
|
|
b.
|
The use of lexical drohen was extended to new contexts where the subject referent of drohen was inanimate, hence where its semantic property of standing for an agent, able to control an action, was ruled out.
|
|
c.
|
Desemanticization had the effect that the lexical component of drohen was largely bleached out; what remained was the schematic function paraphrased in (2b).
|
|
d.
|
In these new contexts, drohen lost most of its morphosyntactic properties, especially that of controlling the subject of the clause (decategorialization) – a function that naturally shifted from drohen to the infinitival verb, the result being what has been called subject-to-subject raising.
|
|
e.
|
Since lexical drohen has the function of a main verb, drohen‑2, which also functions as a main verb, must have preceded drohen‑3 and drohen‑4 in time. Hence, the most likely scenario of historical development is the one sketched in (10).
|
(10)
|
The historical development hypothesized for German drohen
|
|
drohen‑1 > drohen‑2 > drohen‑3/4
|
14There remains a problem where grammaticalization theory does not provide a conclusive hypothesis, namely the question of what accounts for the difference between drohen‑3 and drohen‑4. We will return to this problem later.
- 6 The data on which this diachronic analysis rests is taken mainly from Der digitale Grimm (2004). A (...)
15The hypotheses in (9) will now be tested by looking at earlier written documents of German.6
16These documents suggest, in fact, that hypothesis (9a) can be verified: Neither in Old High German nor in Middle High German is there any significant trace of functional drohen while there is abundant evidence of lexical drohen, which was used as a full-fledged verb with human (in a minority of cases with non-human animate) subjects, with dative objects (experiencers), accusative objects (undergoers), finite complement clauses, direct and indirect clause complements, etc. Furthermore, the construction frequently takes instrumental participants introduced by mit “with”. (11) is an example from Middle High German; let us call this earliest use stage 1, reflecting drohen‑1.
(11)
|
Middle High German (Hartmann von Aue, ca. 1170-1210 AD) (Stage 1)
|
|
hêrre
|
waz
|
wil
|
der
|
leu
|
uns
|
dunket
|
daz
|
er
|
uns
|
dreu.
|
|
man
|
what
|
wants
|
the
|
lion
|
to us
|
seems
|
that
|
he
|
to us
|
threaten
|
|
‘Sir, what does the lion want? It seems to us that he is threatening us.’
|
17A new situation, stage 2, arises in Early New High German in the first half of the 16th century. Rather than standing for a human or animate concept, the construction begins to be extended productively to contexts with inanimate subject referents. This situation has all the properties of drohen‑2. (12) is a typical example of this situation. In Luther’s writings, drohen‑2 is not yet very common; it accounts for 13.2 per cent of all his uses of drohen, with drohen‑1 being clearly predominant (86.8 per cent). But subsequently its popularity rises dramatically, reaching its peak with authors born between 1650 and 1699, where it occurs with the same rate of frequency (48.0 per cent) as drohen‑1 (see Table 2).
(12)
|
Early New High German (Hans Sachs, 1494-1576 AD) (Stage 2)
|
|
dergleichen
|
auch
|
ohn-zahlbar
|
sorgen,
|
troen
|
im
|
abendt
|
|
Such
|
also
|
countless
|
sorrows
|
threaten
|
him
|
evening
|
|
und
|
den morgen.
|
|
and
|
the morning
|
|
‘Countless sorrows of this kind were threatening him evening and morning.’
|
18Another important innovation characterizing stage 2 in the 16th century can be seen in the fact that, occasionally, there is now an accusative complement consisting of a verbal noun denoting a dynamic situation, like fall“collapse” in (13).
(13)
|
Early New High German (proverb, Georgius Agricola, 1560 AD) (Stage 2)
|
|
an eine wand,
|
so
|
den fall
|
drohet,
|
pissen
|
die hunde.
|
|
At a wall
|
which
|
the collapse
|
threatens
|
pee
|
the dogs
|
|
‘A wall which threatens to collapse, the dogs pee at.’
|
19There are already some occurrences of drohen with inanimate subjects in Middle High German, but they are not suggestive of a pattern. Rather, they appear to be individual instances of metaphorical transfer, where human agents serve as metaphorical vehicles for certain natural phenomena (in the sense of “personification” of Verhagen 2000; see also Cornillie 2004a; 2005a; 2005b). Thus, in (14) the noun karker “dungeon” is treated like a human threatening agent.
(14)
|
Middle High German (Ulrich von dem Türlîn, Willehalm; second half of 13th century)
|
|
Es was
|
nû
|
wol
|
gelîch
|
gewegen,
|
swaz
|
im
|
der karker
|
|
It was
|
now
|
well
|
equally
|
balanced
|
whatever
|
him
|
the dungeon
|
|
het
|
gedröut,
|
daz
|
in
|
diu liebe
|
nû
|
mêre fröut,
|
|
had
|
threatened
|
as
|
him
|
the love
|
now
|
more delights
|
|
die
|
er
|
der
|
küniginnen
|
truoc.
|
|
that
|
he
|
to the
|
queen
|
carried
|
|
‘It was now all the same, whatever he might have been threatened by prison, as long as he now enjoyed the love that he felt for the queen.’
|
20The drohen‑2construction illustrated in (12) and (13) has inanimate subjects and expresses what we portrayed in (2b) as functional drohen. Still, while expressing a grammatical (modal) concept, the morphosyntax of lexico-functional drohen is that of a main verb. What characterizes this stage-2 construction is that the subject referent is no longer conceived of as controlling the event described by the verb.
21It took roughly another 200 years before functional drohen of stage 3 began to emerge as a new construction. First attestations of this category are found in the first half of the 18th century, and in the course of the 18th century it establishes itself as a distinct category, accounting for between 10 and 20 per cent of all uses of drohen. This construction, illustrated in (15), has in particular the following properties:
-
It occurs with inanimate subjects that cannot be understood to be agents controlling the event concerned,
-
it expresses the grammatical function paraphrased in (2b),
-
it is likely to refer to events located in the future, and – unlike stage 2 –
-
it takes infinitival verbs as complements but does not allow for any other complements.
(15)
|
New High German (Gottsched, 1738 AD) (Stage 3)
|
|
[…]
|
wasser
|
in
|
ein meer
|
zu giessen,
|
|
|
water
|
in
|
a sea
|
to pour
|
|
welches
|
ohnedem
|
überzulaufen
|
drohet.
|
|
which
|
anyway
|
to run over
|
threatens
|
|
‘[…] to pour water into a sea that risks running over anyway.’
|
22Finally, there is a fourth stage in the development of this item, leading to the rise of drohen‑4: Since the late 18th century, functional drohen is freed from the constraint of being used only with inanimate subjects. It is now generalized to contexts where the subject is human. Still, unlike stage 1, the meaning is that of epistemic modality. There are two phases in the development of what in Present-Day German constitutes drohen‑4 (see (1d)). In the second half of the 18th century there are first attestations where there are human subject referents but where there is ambiguity between lexical and functional (epistemic) drohen, that is, drohen can be understood alternatively either with reference to drohen‑1 or drohen‑4. Thus, in the text passage of (16), which is one of the first attestations of this kind, the subject referent can be understood either as expressing a physical threat (= drohen‑1) or as being about to do the flogging (= drohen‑4) – we will refer to this (intermediate) situation as drohen‑1/4.
(16)
|
New High German (Bräker, 1735-1798, 129)
|
|
[…]
|
eines
|
(preusz.)
|
offiziers,
|
der
|
mit furiosem gesicht
|
und
|
|
|
of a
|
Prussian
|
officer
|
who
|
with angry face
|
and
|
|
aufgehobnem
|
stock
|
vor uns
|
stund,
|
und
|
alle augenblick
|
|
raised
|
stick
|
in front of us
|
stood
|
and
|
every moment
|
|
wie
|
unter kabisköpfe
|
drein
|
zu hauen
|
drohte.
|
|
like
|
under cabbage heads
|
into
|
to hit
|
threatened
|
|
‘[…] of a (Prussian) officer who stood in front of us with a raised stick and any moment threatened/was about to flog [us] like heads of cabbage.’
|
23The second phase, beginning roughly around the turn of the 18th and 19th centuries, marks the rise of drohen‑4 as a distinct and unambiguous construction, where the subject referent can no longer be interpreted meaningfully as controlling the action, or as acting intentionally. Thus, in (17) the context makes it clear that the event expressed by fallen “to fall” is not intended. It is the classic German poets Friedrich Schiller and Johann Wolfgang von Goethe who appear to have been the first to use drohen‑4 productively and to propagate this construction. However, neither drohen‑1/4 nor drohen‑4 attained any popularity; they remain minor options accounting for less than one tenth of all occurrences of drohen.
(17)
|
New High German (von Goethe, 1749-1832, Hermann und Dorothea [40, 320]) (Stage 4)
|
|
es
|
knackte
|
der
|
fuss,
|
sie
|
drohte
|
zu
|
fallen, […].'
|
|
it
|
cracked
|
the
|
foot
|
she
|
threatened
|
to
|
fall
|
|
‘Her foot cracked, she was about to fall down […].’
|
- 7 The fact that drohen‑2 exhibits a comparatively high frequency of occurrence in our Present-Day Ge (...)
- 8 Askedal (1997: 15-7) gives the following percentages for drohen in Present-Day German: 45.5 % for d (...)
24As Table 2 shows, by the first half of the 19th century all four drohen constructions had established themselves and have remained largely unchanged for the next two centuries up to Present-Day German7 - that is, a grammaticalization process that started in the 16th century came to a standstill by the early 19th century. Thus, the presence of the four constructions distinguished in (1) is due to historical processes that took place within a time span of hardly more than two and a half centuries, and functional drohen has remained a weakly grammaticalized “auxiliary” up to now.8
Time
|
drohen‑1
|
drohen‑2
|
drohen‑3
|
drohen‑1/4
|
drohen‑4
|
Totals
|
Keisersberg
(1445-1510)
|
15
(100)
|
-
|
-
|
-
|
-
|
15
(100)
|
Martin Luther
(1483-1546)
|
33
(86.8)
|
5
(13.2)
|
-
|
-
|
-
|
38
(100)
|
From Luther to 1599
|
52
(76.5)
|
16
(23.5)
|
-
|
-
|
-
|
68
(100)
|
Authors born between
1600 and 1649
|
29
(61.7)
|
18
(38.3)
|
-
|
-
|
-
|
47
(100)
|
Authors born between
1650 and 1699
|
12
(48.0)
|
12
(48.0)
|
1
(4.0)
|
-
|
-
|
25
(100)
|
Authors from
1700 to Goethe
|
54
(40.0)
|
68
(46.6)
|
22
(15.1)
|
2
(1.3)
|
-
|
146
(100)
|
Authors during Goethe’s and Schiller’s time
|
19
(54.3)
|
9
(25.7)
|
4
(11.4)
|
4
(5.4)
|
-
|
75
(100)
|
Friedrich Schiller
(1759-1805)
|
17
(39.5)
|
12
(27.9)
|
9
(20.9)
|
3
(7.0)
|
2
(4.7)
|
43
(100)
|
J. W. von Goethe
(1749-1832)
|
25
(33.8)
|
16
(21.6)
|
27
(36.5)
|
4
(5.4)
|
2
(2.7)
|
74
(100)
|
Authors from
Schiller to 1799
|
37
(30.6)
|
50
(43.3)
|
24
(19.8)
|
6
(5.0)
|
4
(3.3)
|
121
(100)
|
Authors born between 1800 and 1900
|
30
(27.3)
|
39
(35.5)
|
30
(27.3)
|
4
(3.6)
|
7
(6.3)
|
110
(100)
|
Present-Day German
(MannheimerKorpus 1-2)
|
62
(32.1)
|
77
(39.9)
|
44
(22.8)
|
2
(1.0)
|
8
(4.2)
|
193
(100)
|
Table 2. Frequency of occurrence of the four drohen‑constructions from the 15th century to Present-Day German (percentages in parentheses; sources: Der digitale Grimm 2004, Mannheimer Korpus, 1-2).
25To conclude, at least from Old High German to roughly until 1500 AD, lexical drohen‑1 was essentially the only way in which German speakers could express themselves. There are sporadic uses of inanimate subjects in Middle High German but, as we noted above, they are suggestive of idiosyncratic metaphorical uses.
- 9 What Verhagen hypothesizes on the development of Dutch beloven“promise“ from lexical to functional (...)
26The 16th century saw a revolutionary innovation, in that lexical drohen‑1 came to provide a conventional metaphorical vehicle for use in contexts where certain inanimate forces encoded as clausal subjects could be treated like human beings. These subject referents expressed undesirable situations, frequently consisting of nominal referents denoting negatively valued emotions or sensations like diseases or psychological problems, natural forces like storm, lightning, or threatening celestial bodies. Thus, this innovation is suggestive of the rise of a metaphorical convention whereby inanimate concepts could be treated productively like human agents capable of instigating and controlling actions. But with the presence of such inanimate subject referents, which Verhagen (2000: 205) refers to with reference to the development of Dutch beloven “promise” as “metaphorical subjects”, the lexical meaning of drohen was backgrounded and the functional/modal meaning foregrounded.9 In Martin Luther’s work, this new convention is visible but does not appear to really have been established.
27It was a natural step to exploit the new construction, drohen‑2, by allowing it to take infinitival complements rather than nominal objects. Note that the extension from nominal object to non-finite verbal complements is a universally attested process, especially when certain cognitive, speech-act, or volitional/intentional verbs are involved (that is, languages allowing “I want X” are likely to also develop a use pattern of the kind “I want to do X”). At the time in the 16th century when drohen‑2 evolved, this item was still a transitive verb taking direct objects in the accusative case, as example (13) shows. The transition from drohen‑2 plus object nouns to drohen‑3 plusinfinitival verbs marked the beginning of auxiliation. Rather than being conceptual in nature, the motivation for this grammaticalization can be seen primarily in a morphosyntactic process of extension from nominal to non-finite verbal complements.
28This option does not yet appear to have been available to Martin Luther and his compatriots. It is only a century later that it crops up occasionally in writings, although only in exceptional uses (1.4 per cent of all uses of drohen in the 17th century). A dramatic change occurred in the second half of the 18th century, when drohen‑3 establishes itself as a functional category. The German poet Friedrich Schiller may be taken to be representative of the use patterns characterizing the late 18th century, and his compatriot Johann Wolfgang von Goethe made excessive use of drohen‑3, which accounts for the majority (36.5 per cent) of his uses of drohen.
29The late 18th century saw a new form of context extension, when drohen‑3, until then reserved for inanimate subjects, was extended to human subject referents, thereby giving rise to drohen‑1/4 and eventually to drohen‑4. This new construction inherited the function of drohen‑3 – with the effect that in this construction there were human subjects but they were no longer agents able to control the event described by the infinitival new main verb. With the emergence of drohen‑4 there was now a situation of potential ambiguity, in that drohen‑4 shared with drohen‑1 the property of having human subjects. It is only in a contextual frame where the subject referent could be understood to be incapable of controlling the event that unambiguous instances of drohen‑4 can be found; Goethe’s example in (17) illustrates this contextual frame. This new construction appears to have gained in popularity in the 19th century, but its frequency of occurrence never came close to that of the other three drohen‑constructions, as Table 3 shows (Sources: Der digitale Grimm 2004, Mannheimer Korpus, 1-2).
Authors born between
|
drohen‑1
|
drohen‑2
|
drohen‑3
|
drohen‑1/4
|
drohen‑4
|
Totals
|
1445-1599
|
83.2
|
16.8
|
-
|
-
|
-
|
100 (125)
|
1600-1699
|
56.9
|
41.7
|
1.4
|
-
|
-
|
100 (72)
|
1700-1799
|
36.3
|
37.0
|
20.5
|
4.3
|
1.9
|
100 (419)
|
1800-1900
|
27.3
|
35.5
|
27.3
|
3.6
|
6.3
|
100 (110)
|
After 1900
|
32.1
|
39.9
|
22.8
|
1.0
|
4.1
|
100 (193)
|
Table 3. Relative frequency of occurrence of the four drohen‑constructions in five phases of development (percentages of total uses).
30As Table 4 shows, the morphosyntactic development of drohen can be described in terms of three different stages. Since the 15th century, drohen‑1 was, as it still is today, characterized by a rich syntactic configuration structure. In addition to the subject, constructed in the nominative case (S), it could take direct objects in the accusative case (A), indirect objects in the dative case (D), prepositional adjuncts headed by mit “with” or other prepositions (P), finite complement clauses (C), direct and indirect speech sentences (DS), and/or infinitival complements (I). Table 4 shows that with the rise of drohen‑2 there was massive decategorialization, leading to a shrinking of morphosyntactic options: Of the seven different argument types of drohen‑1, only four survive in drohen‑2. The only complements that can be used were A and D. Further decategorialization occurs with the transition to drohen‑3 – there is now only one syntactic configuration that remains: The only complement now tolerated is I, and this is the situation also characterizing drohen‑4.
drohen‑1
|
drohen‑2
|
drohen‑3
|
drohen‑4
|
S (D) P
|
S (P)
|
|
|
S (D) A
|
S (D) A
|
S I
|
S I
|
S (D) I
|
|
|
|
S (D) C
|
|
|
|
S (D) DS
|
|
|
|
S (D)
|
S (D)
|
|
|
Table 4. The development of participant marking associated with the four drohen‑constructions since the 15th century (S = subject, D = dative object, A = accusative object, P = prepositional phrase, I = infinitival complement, C = finite complement clause; DS = direct speech proposition. Source: Der digitale Grimm 2004).
31Considering that infinitival complements (I) were already present at the stage of drohen‑1, the question arises of how it was possible that what was present at stage 1 disappeared at stage 2 and re-surfaced at stages 3 and 4? As the analysis of the written evidence suggests, this is not exactly what happened: The infinitival complements of stages 3 and 4 can not be traced back to stage 1. Rather, they are the result of a development from nominal accusative complement (A) at stage 2 to infinitival complement (I) at stages 3 and 4.
32Diachronic processes of grammaticalization tend to surface in the synchronic structure of a language in the form of structural variation, where the different stages of the process co-exist side-by-side as distinct but conceptually related categories. This is also the situation of German drohen, which has inherited all the four different constructions illustrated in (1). However, there is one exception. With the transition from drohen‑2 to drohen‑3, accusative complements disappeared from the language in the 20th century, that is, none of the four Present-Day German constructions allows for accusative objects. In other words, synchronic structure reflects what happened in the past, but it does not do so in every detail. We will return to this issue.
33The historical sketch that we presented in section 4 confirms all five hypotheses formulated in (9) within the framework of grammaticalization theory on the basis of the predictions on auxiliation proposed in (8). And it also confirms the sequence of development proposed in (10), according to which drohen‑2 preceded drohen‑3 and -4 in time.
34But this historical sketch also shows that there is one fact that is not covered by grammaticalization theory, namely that drohen‑3 is historically earlier than drohen‑4. Crosslinguistic findings on this issue are inconsistent. What this suggests is that, while grammaticalization theory provides a skeleton of functional evolution, it does not offer a complete match of historical processes, and hence it cannot replace historical reconstruction.
35The limits of this theory can also be demonstrated with another issue that we mentioned earlier, namely the loss of accusative objects. In the case of drohen‑2, loss is exhaustively accounted for by the fact that nominal accusative objects were replaced by infinitival verbs in the transition from drohen‑2 to drohen‑3, involving an intermediate stage where the accusative object consisted of a verbal noun denoting a dynamic situation, as we saw in example (13). But there is no explanation of why accusative objects also disappeared from drohen‑1. They were part of the valency of drohen throughout the recorded history of German; example (18) shows that drohen was still a fully transitive verb at Luther’s time in the 16th century. As Table 5 suggests, accusative objects were never really popular, but in the course of the 20th century they disappeared entirely from the language, the function of drohen as a transitive verb having been taken over by the derived form androhen.
(18)
|
Early New High German (Martin Luther, Jerem. 11, 17)
|
|
[…]
|
denn
|
der
|
herr
|
Zebaoth
|
der
|
dich
|
gepflanzt
|
hat,
|
|
|
for
|
the
|
Lord
|
Zebaoth
|
who
|
you
|
planted
|
has
|
|
hat
|
dir
|
ein
|
unglück
|
gedrewet.
|
|
has
|
to you
|
a
|
misfortune
|
threatened.
|
|
‘[…] because the Lord Zebaoth, who has planted you, threatens you with a misfortune.’
|
Time
|
Number of text occurrences of drohen
|
Number of accusative objects with drohen
|
Percentage
|
Authors born between 1445 and 1599
|
127
|
24
|
18.9
|
Authors born after 1600
|
67
|
11
|
16.4
|
Authors born after 1700
|
417
|
56
|
13.4
|
Authors born after 1800
|
110
|
6
|
5.5
|
Authors born after 1900
|
193
|
0
|
0
|
Table 5. Occurrences of drohen with accusative objects since 1445 (Sources: DerdigitaleGrimm 2004, MannheimerKorpus, 1-2).
36As may have become apparent from the preceding discussion, the development of drohen is the product of three main forces, namely of cognitive-conceptual, of pragmatic-contextual, and of morphosyntactic forces. The question that we want to look into now is: What is the relative contribution of each of these forces in this and the many other similar processes whereby prototypical discourse situations involving human agents and kinetic processes turn into situations where there are participants that are unable to instigate and control actions, and predications that express a grammatical concept rather than an action –with the result that, ultimately, there are functional categories rather than lexical discourse structures?
37The text evidence that is available on the emergence of German drohen as a functional category suggests that at the beginning of the process in the early 16th century, at stage 2, there was a creative activity whereby the use of this action verb was extended to non-human subject referents, in particular to natural phenomena like “lightning” or “thunderstorm”, and abstract concepts such as “death”, “disease”, “war”, etc., that were treated like human agents. Once this extension was established, it was only a minor step to extend the range of subject referents to other inanimate concepts.
38There is reason to call this process overall a metaphorical transfer, whereby inanimate concepts were treated like human agents (see below). However, this transfer did not happen abruptly but rather in incremental stages via contexts that were compatible with both the metaphorical vehicle and the metaphorical topic – it was made possible by a mechanism of gradual context extension.
39The second major step, from stage 2 to stage 3, was not primarily cognitively but rather morphosyntactically motivated: With the presence of inanimate subject referents and functional drohen, the ground was cleared for an extension from nominal to non-finite verbal complements. It can be described essentially as a morphosyntactic process induced by the pragmatic process of context extension.
40The final step, leading to stage 4, was once more essentially both cognitively and pragmatically motivated, in that a process that had happened earlier was, so to speak, undone. A construction typically reserved for human subject referents at stage 1, changing into a construction with inanimate referents at stages 2 and 3, was now extended and generalized to another construction allowing for both human and inanimate referents.
41To conclude, with reference to the three main variables that we are concerned with in this paper, the rise of a new functional category such as drohen‑3 in German can be sketched in an abstract manner as in Table 6. What this sketch suggests is:
-
that the only mechanism that is stable across the entire process of grammaticalization is pragmatics,
-
that cognitive and morphosyntax innovation are inextricably linked to pragmatics, and
-
that linguistic parameters are not crucially involved at the initial stage, that is, in the transition from lexical to functional categories.
Mechanism
|
Stage 2
|
Stage 3
|
Stage 4
|
Cognitive (conceptual innovation)
|
+
|
|
|
Pragmatic (context extension)
|
+
|
+
|
+
|
Morphosyntactic (new structure)
|
|
+
|
|
Table 6. Mechanisms primarily involved in the auxiliation of German drohen.
42Table 7 raises an issue that is at the heart of grammaticalization theory, namely the unidirectionality principle. Earlier studies within this paradigm suggested that there are no exceptions to this principle, but later work (especially Newmeyer 1998) showed that there are exceptions, even if they are rare. Table 7 might be interpreted as suggesting that grammaticalization is not directional. As a matter of fact, however, this impression is wrong.What actually happened, as Table 7 shows, is that the meaning of the construction became more general, in that at the final stage 4 there are no more semantic constraints, neither on the subject nor on the compatibility of the verb. In addition to inanimate subjects at stage 3, there are now also human ones. To conclude, the development of German drohen, like those of the other auxiliaries mentioned earlier, does not contradict the unidirectionality principle.
|
Stage 1
|
Stages 2, 3
|
Stage 4
|
Subject
|
Human
|
Inanimate
|
Unconstrained
|
Verb
|
Action
|
Modal
|
Modal
|
Table 7. Major semantic change in the auxiliation of German drohen.
43In (3) at the beginning of the paper, we posed a number of questions and subsequently we were searching for answers. On the basis of the findings presented above we are now in a position to deal with these questions in turn.
44In (3a) we were wondering why there are four different constructions involving drohen. The answer is in accordance with the predictions made by grammaticalization theory and confirmed by diachronic evidence: The four different constructions represent different stages of grammaticalization, and these stages surface in the synchronic structure of German in the form of distinct constructions. This also answers question (3b), in that the presence of these constructions is not coincidental but rather is motivated by principles of grammaticalization, which also predicts that the four constructions differ both in their relative age and in their relative degree of grammaticalization.
45In much the same way, question (3c) can be answered with reference to the grammaticalization parameter of desemanticization, whereby the lexical meaning of drohen is bleached out – with the effect that there surfaces the grammatical function of modality, more precisely, a more complex grammatical function as paraphrased in (2b).
46An answer to question (3d) is provided by the grammaticalization parameter of decategorialization, which predicts that the more a construction is grammaticalized, the more categorial properties it will lose. Accordingly, drohen‑1 has distinctly lexical properties while all other constructions exhibit only a small range of lexical properties.
47Question (3e) relates to what may be called a “hybrid” stage on the way from drohen‑1 to drohen‑3. What has changed in drohen‑2 vis-à-vis drohen‑1 is that the subject referent is no longer an agent capable of controlling the action concerned – with the effect that there is a schematic meaning that is also present in drohen‑3. But since there is no other verb in the construction, drohen still maintains the status of the main verb of the construction.
48That drohen‑4 is frequently ambiguous, in that it can be interpreted either with reference to lexical or to functional drohen, as we were asking in question (3f), can be explained with reference to the new contexts that were responsible for the rise of this construction. With the extension of drohen‑3 to human subject referents, the new construction entered into competition with the lexical drohen‑1 construction. It is only in a contextual frame where the subject must be conceived as being incapable of controlling the event that there are unambiguous instances of drohen‑4, as in the Goethe quote of (17).
49Finally, in (3g) we asked why drohen controls the subject in the case of lexical drohen but is a subject-to-subject raising verb that is controlled by the infinitival complement in the case of functional drohen. The answer is that, once more, we are dealing with a predictable property of grammaticalization. Once a verb is desemanticized, turning into an auxiliary, it loses most of its syntactic properties, such as taking arguments other than its verbal complement, which is the new (infinitival) main verb. One of these properties concerns subject selection, which automatically shifts from drohen to the new main verb – with the effect that rather than the actual subject referent being the “controlling” participant, it is now the speaker (or conceptualizer) that determines the modal content of the predication expressed by the new main verb.
50To conclude, grammaticalization theory allows us to account for salient properties of each of the four constructions distinguished in (1), even if it does not answer all the questions that one may have, in particular the question of why drohen‑4 is more grammaticalized than drohen‑3, or why drohen ceased to be a transitive verb in the course of the 20th century.
51In a number of accounts of cognitive linguistics, metaphor has been evoked as a descriptive, sometimes also as an explanatory tool of accounting for grammaticalization. At the same time, metonymy has been proposed as an alternative trope, and the literature is rife with controversies on whether a given process can be attributed to one or the other (see Cornillie 2004b on Spanish amenazar “to threaten” and prometer “to promise”). As has been argued in Heine, Claudi and Hünnemeyer (1991), this is essentially a non-issue since grammaticalization processes tend to have simultaneously a metaphorical and a metonymic component. Taking a standard definition of metaphor, there is reason to support the claim made above that metaphor was involved in the development from drohen‑1 to drohen‑2. According to such a standard definition, metaphor has the following properties:
(19)
|
A definition of metaphor
|
|
a.
|
The source and the target concept are different referents.
|
|
b.
|
The transfer involves two different domains of experience.
|
|
c.
|
The transfer is not formally expressed.
|
|
d.
|
The predication expressed by a metaphor is, if taken literally, false.
|
52With reference to our example in (1), we may say that the construction surfacing in (1b) is the result of a metaphorical transfer from (1a), in that the subject referents are different (19a) and belong to different domains of human experience (19b), in that the referent Karl in (1a) belongs to the domain of human concepts while the referent Katastrophe “disaster” in (1b) belongs to the domain of inanimate concepts. The transfer from human to inanimate concepts is not formally expressed (19c) and, accordingly, the predication in (1b) is literally false (19d) since inanimate concepts such as Katastrophe in (1b) are incapable of the action expressed by lexical drohen. There is, in fact, historical evidence to suggest that metaphor was a sine qua non for the process to happen: It was responsible for the rise of drohen‑2 in the 16th century, as we saw in section 4. In the subsequent development to drohen‑3 and drohen‑4, however, metaphor does not appear to have played any noticeable role.
53The present treatment is not the first analysis of German drohen within a grammaticalization framework (see Askedal 1997; Heine & Miyashita 2004), but there is also a range of alternative theories that have been applied to understand its behavior, or that of its analogues in other European languages, such as Dutch dreigen, English threaten, French menacer, Italian minacciare, or Spanish amenazar.
- 10 Some of her examples (Traugott 1993: 350, (9), (10)), however, are more difficult to classify.
54In her treatment of English threaten and promise, Traugott (1993; 1997) presents a three-stage scenario of the historical development of threaten which is similar to the scenario proposed here for German drohen: Her stage 1 corresponds to our stage 1. It concerns lexical threaten as used from Old English on, taking nominal objects, and both finite and non-finite complements. Her stage 2 involves the development in the 16th century of an epistemic meaning “portend, presage” in a non-intentional and non-commissive use with nominal object complements. As the following example shows, this is suggestive of our stage 2, involving inanimate subject referents.10
(20)
|
English, bef. 1627 AD (Traugott 1993: 350, (7))
|
|
This fire was the more terrible, by reasone it was in a conspicuous place, and threatned danger unto many […].
|
55Her stage 3 corresponds to our stage 3, characterized by an infinitival complement. It involves the development in the 18th century from control verb to “raising” verb and non-intentional epistemic uses of threaten, as exemplified in (21).
(21)
|
English, 1780 AD (Traugott 1993: 350, (11))
|
|
I am sometimes frightened with the dangers that threaten to diminish it [my estate].
|
- 11 Traugott (1993: 351) observes: “The development of raising threaten […] cannot be reduced to a gen (...)
56Traugott does not distinguish stage 4, but there is evidence that this stage has also been reached in English, as the following example from Present-Day English suggests, where there is a human subject referent but functional threaten:11
(22)
|
English, 1992 (Traugott 1993: 351, (14))
|
|
[…] the hapless, aggrieved house-husband threatens to become as rigid and unexamined a comic invention as the grotesquely intrusive mother-in-law once was.
|
- 12 An issue that we will not pursue here but that requires further attention is whether the rise of s (...)
57As we saw in Table 4, the evolution of drohen is characterized by massive decategorialization: drohen takes a wide range of participants at stage 1, of which three survive at stage 2, and only two at stages 3 and 4. Traugott’s (1997: 191-7) analysis suggests that English threaten of stage 3 is characterized by roughly the same kinds of decategorialization as its German counterpart drohen. The semantic development from stage 1 to stage 3 is interpreted by her as a process of subjectification, “whereby epistemic meanings gradually shift from more ‘objective’ possibility based in general beliefs and attitudes to a more ‘subjective’ possibility based in the individual speaker’s belief or attitude” (Traugott 1997: 198-9).12
58Most discussions of the relationship between lexical and functional “threaten” in Western European languages have in fact been phrased in terms of the notion subjectification: cf. Langacker (1995) on English threaten, Verhagen (2000) on Dutch dreigen, and Cornillie (2004a; 2004b; 2005a; 2005b) on Spanish amenazar. A common denominator in many of the discussions of subjectification appears to be that with lexical “threaten” (stage 1), the speaker reports a factual event carried out by the subject referent and expressed by the “threaten”-verb (and its complement), whereas with functional “threaten” (specifically stages 3 and 4) the speaker indicates that there is evidence for a certain expectation that the event expressed by the infinitival verb will take place, and s/he volunteers a “negative” evaluation of this event (see especially Verhagen 2000).
59But “subjectification” is also used as a technical term of cognitive grammar in the Langackerian framework. In this tradition, subjectification concerns a shift from an “objective” construal of an event to a “subjective” construal of the relation between the concepts expressed by the “threaten”-verb and the infinitival verb, involving a change in profiling from the former to the latter and a realignment of the locus of potency from the subject referent of the “drohen”-verb to the speaker (or conceptualizer) (see especially Langacker 2000, chapter 10; Cornillie 2004b; 2005a). This shift crucially involves, or is caused by, attenuation of the nature subject referent, which relates in particular to loss of control exerted by an agentive subject (Langacker 2000: 297) – in other words, to the desemanticization of agents acting intentionally, yielding patients incapable of intentions.
- 13 In these discussions, no distinction is made between drohen‑1 and drohen‑2, or between drohen‑3 an (...)
- 14 That an analysis in terms of a Langackerian subjectification hypothesis is compatible with diachro (...)
60Discussions phrased in terms of subjectification13 of one kind or other have yielded a rich literature on the semantics of “threaten”-verbs, and they are of help in understanding certain properties of functional “threaten”-verbs. Nevertheless, we do not use subjectification as a central notion to account for the problem that this paper is concerned with, for the following reasons. First, there are differences in the interpretation of this term. Whereas Traugott (1997) uses it to account for regularities in semantic change (see also Cornillie 2004b on Spanish amenazar), hence as a phenomenon that can be falsified with reference to historical evidence, Langacker and associates use it within a framework that is essentially synchronic or, perhaps more appropriately, achronic in nature.14 While Langacker argues that his notions of subjectification and attenuation are in accordance with diachronic change, diachrony is not within the scope of his methodology:
I should also emphasize that this work has not been based on serious historical investigation. In presenting series of examples representing progressive degrees of attenuation, I have not intended to suggest that they necessarily correspond to the actual order of diachronic development, which clearly has to be established in its own terms. (Langacker 2000: 315)
61Second, as a term of diachronic semantics, it remains to be established that subjectification was at any stage of development of German drohen a concept or strategy that was central to the communicative intentions of the speakers concerned. We observed earlier that the primary motivation of speakers of German developing a stage-2 construction in the 16th century was to present “threatening” forces like human agents – in other words, to use the existing drohen‑1 construction metaphorically for novel purposes, and the historical evidence available suggests that the situation in English was not much different. As we saw above, in subsequent phases of development there were alternative pragmatic and morphosyntactic motivations; as our reconstruction suggests, none of them had a shift in general beliefs or attitudes towards more subjectivity as a noticeable motivation.
62As a term of a synchronic or achronic model of linguistic description there remains the question of how to find appropriate empirical evidence to show that speakers of Present-day English, Spanish, or German do in fact establish a systematic cognitive relationship between “threaten”-verbs of stages 1 and 3, and that they conceptualize this relationship in terms of notions such as subjectification or profiling shift. However appealing these notions are intuitively, we are not aware of any empirically backed morphosyntactic or psycholinguistic evidence that would support them for the kind of data examined in this paper. In other words, both the descriptive and explanatory basis of these terms, and the notions underlying them, remain to be established. Until such evidence has become available, we maintain that subjectification is a by-product of the grammaticalization parameters of extension and desemanticization.
63And the same applies to the term “attenuation”. On the surface, this term covers what desemanticization stands for. As a matter of fact, however, the two are different in the following respect: Whereas desemanticization is falsifiable by means of diachronic evidence, attenuation is used in a descriptive framework that is synchronic in nature, and it is not entirely clear how hypotheses on attenuation can be tested by means of empirical methods of research. Take the following instance of attenuation. (23b) is said to be attenuated vis-à-vis (23a) because the mailbox is static, occupying only a single position vis-à-vis the landmark (across the street): While being an intuitively plausible hypothesis, the question remains as to what exactly the empirical evidence is that allows us to verify or falsify that (23a) and (23b) differ in their relative degree of attenuation?
(23)
|
Langacker (2000: 299)
|
|
a.
|
The child hurried across the street.
|
|
b.
|
There is a mailbox right across the street.
|
64And fourth, as we saw in section 3, the question remains of what the scope of subjectification is within a framework of grammaticalization theory. In cases like the present one we will therefore assume that what has been taken to be suggestive of subjectification can be interpreted as an epiphenomenal outcome of extension and desemanticization. The processes that we discussed – semantic loss of drohen as a lexical speech-act verb and of its subject as an agent in specific contexts (= desemanticization), extension of nominal complements to infinitival verbal complements, with the latter assuming the role of the new main verb – predictably trigger an interpretation to the effect that, rather than the subject referent doing something, it is now the speaker (or conceptualizer) who suggests a certain expectation that the event expressed by the infinitival main verb is going to take place.
65To conclude, subjectification is an attractive notion that is of help in understanding certain semantic properties of “threaten”-or “promise”-verbs in Western European languages, but so far it has not been possible to establish satisfactorily that speakers at some stage in the past, or else in Present-day English, Dutch, Spanish, or German really used – or are using – subjectification as a strategy for structuring discourses. More research is required on this issue.
66The observations made in this paper substantiate in more detail what has been maintained in many other works on the structure and evolution of auxilation (see especially Heine 1993; Bybee, Perkins & Pagliuca 1994; Kuteva 2001). Grammaticalization theory has both a synchronic and a diachronic component, being based on findings on language structure and language use across time and space. Accordingly, it contributes to both these fields. In section 3 we formulated a set of questions on the synchronic structure of drohen, and in section 9 we provided answers to these questions. These answers are essentially outside the scope of alternative models of linguistics – hence, grammaticalization theory can contribute to a better understanding of language structure, even if there remain a number of issues that are not within the scope of this theory.
67With reference to diachronic linguistics, grammaticalization theory is able to predict a major outline of historical evolution – hence, it provides a tool for historical analysis. Thus, we saw that the hypotheses proposed in (9) by means of this theory are confirmed by historical data. At the same time, it also became clear that reconstruction work carried out within this theory cannot replace diachronic analysis: Certain historical facts, such as the loss of transitivity of drohen in the 20th century or the relative age of drohen‑3 and drohen‑4, are simply not accessible to this theory.
68As we observed in section 11, notions such as subjectification and attenuation have been employed in cognitive-linguistics approaches to account for conceptual differences between stage-1 and stage-3 uses of “threaten”-verbs in some Western European languages. But the processes that these terms allude to took place more than three hundred years ago. Now, what may or may not have been subjectification or attenuation for speakers of German three centuries ago does not necessarily correspond to what contemporary speakers of German have in mind when using drohen. The question therefore arises what the synchronic significance of these notions is: What is the empirical evidence to establish that they can be defined in terms of cognitively significant parameters? Furthermore, what is the empirical evidence to establish that contemporary speakers of German do in fact relate drohen‑1 and drohen‑3 in a systematic way to one another – that is, in a way that makes it possible to establish a descriptive or explanatory relationship between the two?The evidence that we are aware of is as yet not entirely satisfactory.
69There are a number of issues that we ignored in the present paper. One relates to functional drohen. In treating it as a functional category we are aware that such a treatment is not without problems. What is beyond reasonable doubt is that functional drohen is only weakly grammaticalized, as is suggested in particular by the fact that it has retained a considerable part of the meaning of its lexical source, as we saw in (2b), that it differs from epistemic modals such as können “can”, sollen “should”, etc., in that it requires the infinitive marker zu on the main verb, and that it exhibits a number of constraints on the tense-aspect behavior of the main verb, which is, e.g., barred from occurring in the perfect aspect. The question therefore remains where exactly to locate functional drohen on a scale extending from full lexical to full auxiliary structure. This is a question that is beyond the scope of this paper. At the same time, it is also one that is contingent upon the particular theory one may wish to adopt.
70Another issue concerns the fact that German drohen is not an isolated case. Rather, similar situations can be found in other Western European languages, e.g., in English (threaten; Traugott 1993; 1997), Dutch (dreigen; Verhagen 2000), Spanish (amenazar; Cornillie 2004a; 2004b; 2005a; 2005b), French (menacer) or Italian (minacciare). Second, all the authors just mentioned treat “threaten”-verbs jointly with “promise”-verbs, which show a strikingly similar structure and development. These two issues will be the subject of separate treatments.