Jump to ContentJump to Main Navigation
Clause Structure and Word Order in the History of German$

Agnes Jäger, Gisella Ferraresi, and Helmut Weiß

Print publication date: 2018

Print ISBN-13: 9780198813545

Published to Oxford Scholarship Online: April 2018

DOI: 10.1093/oso/9780198813545.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2020. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use.  Subscriber: null; date: 28 March 2020

The prefield after the Old High German period

The prefield after the Old High German period

Chapter:
(p.64) 5 The prefield after the Old High German period
Source:
Clause Structure and Word Order in the History of German
Author(s):

Augustin Speyer

Helmut Weiß

Publisher:
Oxford University Press
DOI:10.1093/oso/9780198813545.003.0005

Abstract and Keywords

The filling of the prefield in Modern German is determined by information-structural constraints such as scene-setting, contrastiveness, and topichood. While OHG does not yet show competition between these constraints, competition arises from MHG onward. This has to do with the generalization of the V2 constraint (i.e. the one-constituent property of the prefield) for declarative clauses, in which context the information-structural constraints are loosened. The syntactic change whose result eventually was the loss of multiple XP fronting comprised a change of the feature endowment of C because the fronting of expletive thô (roughly in the OHG of the ninth century) led to the reanalysis of XP fronting as a semantically vacuous movement whose only function is to check the EPP feature of C. Data from doubly filled prefields in ENHG and post-initial connectives indicate that an articulated split CP-structure, as proposed within the cartographic approach, is also at play in German.

Keywords:   cartographic approach, split CP, one-constituent property, scene-setting, topic, expletive, contrastive, connectives

5.1 The prefield in Modern German

5.1.1 Syntactic constraints

The prefield is defined in the traditional topological or field model (see Chapter 1) as the zone immediately preceding the left sentence bracket, i.e. the position of finite verb forms in main clauses (e.g. Dürscheid 1989; Wöllstein 2010). In generative terms, it is identical with SpecCP (see e.g. Thiersch 1978; den Besten 1983; Sabel 2000). We deliberately use the pre-Rizzian cover term for the left periphery, the reason for which will become clearer when we turn to the possible analyses of the prefield.

A property which has been described for the prefield in classical topological literature as well as generative literature is that it usually contains only one constituent (in the following referred to as the ‘one-constituent property’). Whereas in the field model this is simply a descriptive generalization, in generative grammar—at least at the stage of the ‘Barriers’ model (Chomsky 1986)—this property directly follows from the fact that it is the specifier of a single projection, hence only one phrase can occur in this position.

The advent of a more fine-grained analysis of the left periphery, in the wake of Rizzi (1997), removed the explanatory power of the SpecCP scheme for the observable one-constituent property of the prefield, as now more than one specifier position is available in the left periphery. As a consequence, the one-constituent property has to be derived by an additional mechanism. Several proposals concerning this mechanism have been made. Grewendorf (2002), for instance, assumes that the prefield constituent has to pass through the lowest projection in the split C architecture, SpecFinP, before moving to an information-structurally more suitable projection in the split CP architecture such as SpecTopP or SpecFocP. As the constituent leaves a trace in SpecFinP, further movement of other constituents to the left periphery is excluded. On the other hand, Frey (2004a, 2006) suggests that only phrases bearing a contrastive feature can be moved directly to the left periphery via operator movement (p.65) (SpecKontrP in this case, which is probably identical to Rizzi’s SpecFocP, although Frey does not commit himself to this). In other cases, the left-most constituent of the middle field (that is, everything below Fin in Frey’s conception) is moved to SpecFinP by the operation of formal movement (the idea essentially goes back to Fanselow 2002, see also Chapters 3 and 4 of this volume on stylistic/formal movement).

By looking more closely at the data, one can observe that the generalization that the prefield always contains only one constituent does not hold in all cases. There are several cases reported in which the prefield contains more than one constituent (see Müller 2005 among others). A violation of the one-constituent property occurs only rarely, but obviously it does occur. Several attempts have been made to explain such exceptions from the one-constituent property against the background of a syntactic analysis that allows for only one phrase in the prefield. One approach is to say that cases of multiple constituents in the prefield are in fact instances of ‘remnant movement’, i.e. movement of a partially emptied verbal phrase to the prefield. Such a partially emptied VP can contain more than one phrase in complement or adjunct position, and these phrases can therefore be fronted to the prefield together (Müller 2005); consider [CP [VP Zum zweiten Mal die Weltmeisterschaft t1]2 errang1 [IP Clark t2 t1]] ‘for the second time the world championship won Clark’ (adapted from Müller 2005). Other researchers analyse multiple constituents in the prefield as exactly that and assume that more than one phrasal position in the C architecture is filled (e.g. Jacobs 1986; Speyer 2008a; Chapter 3.3.3 of this volume). Later on in this chapter, we will return to this question.

5.1.2 Information-structural constraints

Another property of the prefield or SpecCP is that syntax does not determine what phrase is moved to that position. As the movement to SpecCP is not A-movement, constraints on A-movements such as locality do not hold. In other words, any phrase—as long as it is a phrase—can be moved to the prefield (see also Chapters 1 and 3).

The selection of which phrase is moved to the prefield in the end is mostly a matter of information structure. It can be observed that only phrases with certain information-structural properties appear in the prefield (see e.g. Hoberg 1997; Jacobs 2001; Speyer 2008b, 2009a):

  • phrases that determine the temporal or spatial situation in which the proposition is evaluated with respect to its truth value (‘scene-setting’ in Speyer 2008b; see (1a); definition following Jacobs 2001);

  • phrases that refer to an entity which stands in a partially ordered set relation to at least some other element in the discourse (‘contrast’ in Speyer 2008b; definition following Prince 1999. The clauses in (1b) illustrate this; the set consists of potential windows in the building);

  • phrases that refer to an entity that the rest of the sentence is about (‘topic’ in Speyer 2008b; see (1c); definition following Reinhart 1982). In order to be a candidate for the prefield, the topic must either be discourse-old (preferably at the same time being a centre, see Speyer 2007), or it must be of some macro-structural relevance, that is, it must function as a topic for at least one more sentence (see Speyer 2009a). (p.66)

(1)

The prefield after the Old High German period

A clause that contains no phrase of any of the information-structural types outlined above is likely to fill the prefield with an expletive, that is, an element base generated in the prefield that has no semantic content and whose sole purpose seems to be to turn the clause into a verb-second (V2) clause, as can be seen in (2a). Note that this type of expletive (the Vorfeld-es/prefield-es, see also Chapter 3) cannot appear in any other position, as illustrated in (2b). This indicates the properties identified above, namely that it is an element base-generated in the prefield (as the prefield, or SpecCP, is part of the highest projection in the tree, its content cannot be moved further upwards in the tree), and, on a more descriptive level, that its raison d’être is strongly connected to the prefield position as such.

(2)

The prefield after the Old High German period

A question that presents itself is what happens if a clause contains more than one phrase which would conform to the information-structural demands for movement into the prefield. In principle, there is competition as to which one is moved to the prefield in the end. A survey of German newspaper texts and texts of a similar stylistic level reveal that there is a ranking involved (Speyer 2008b); if one of the competitors is a scene-setting element, the prefield will most likely (in roughly 90% of (p.67) cases) be filled by the scene-setting element. If no scene-setting element is involved, but there is competition between a topic and a contrastive element, there is a slight tendency for the contrastive element to win out. This ranking is demonstrated in (3). There are two sentences, each of which contains a scene-setting element (marked with {…}), a contrastive element (marked with /…\), and a topic (marked by underlining). The set opened up by the contrastive elements is different publications of the author Hans Küng; the topic of the whole passage is the author himself and his activities leading to the institution of his foundation. In both clauses, the prefield is filled with the scene-setting element.

(3)

The prefield after the Old High German period

So, on the whole, there seems to be a ranking of prefield fillers of the following form in operation in Modern German if a clause contains more than one phrase that could be moved to the prefield:

(4)

The prefield after the Old High German period

Note that the ranking of scene-setting elements over the other two classes of elements is pronounced, whereas the ranking of contrastive elements over topics is much less clear-cut.

5.2 The prefield in older stages of German

If one adopts a diachronic viewpoint, a question that suggests itself is whether the syntactic property of hosting exactly one constituent as a rule, the information-structural properties and the prefield ranking described above were in operation in earlier stages of German as well, or whether a change occurred. The question is the more urgent as the prefield in itself (as the whole V2-property of Germanic languages) is a relatively recent development, in that it does not seem to have been part of PIE syntax, and it is doubtful whether, within the Germanic branch, it is a very old innovation either (see e.g. Speyer 2009b). In fact, the prefield in OHG shows some properties that differ strongly from the modern setup of the prefield, as discussed in Chapter 3. Yet we see that something like the one-constituent property is developing, and it can be shown that, on the whole, if the prefield is filled with only one constituent, it follows a ranking not unlike the one determined for Modern German, with the difference, however, that contrastive elements are not preferred as prefield fillers. Scene-setting elements are highest in ranking, but topics can beat them in the competition more easily than in Modern German (see Speyer 2015b). So it looks as if (p.68) the ‘old’ property of the prefield as classical topic position seems to be still in operation (see Hinterhölzl and Petrova 2010; also Chapter 4 of this volume).

In this chapter, we will concentrate on the prefield after the OHG period. By MHG times, the one-constituent property seems to have been fairly established, although exceptions seem to be more numerous in MHG and ENHG (Early New High German) than in Modern German (see Section 5.2.2 on ENHG). We will concentrate on the information-structural content and the prefield ranking.

5.2.1 Middle High German

In order to gain an approximative view on the state of the prefield in MHG, we analysed 100 declarative clauses from a sermon by Berthold von Regensburg (sermon Von den siben planêten, in Pfeiffer’s edition Vol. I: 48–53). The numbers are given in Table 5.1 and Table 5.2. Table 5.1 is to be read as follows: the clauses that conform to the ranking identified for Modern German are in columns 2 to 4 and are printed in bold face. The clauses that violate this ranking are given in columns 5 to 7. The percentages are aligned such that each competition case occupies a row by itself. In the tables, S stands for scene-setting elements, C for contrastive elements, T for topic. In Table 5.2, the clauses that conform to the expected ranking are in column 2 and in bold face; the other possibilities are in columns 3 and 4. Only the number of tokens is given in this table.

Table 5.1 Clauses with two competitors for the prefield in MHG (Bavarian)

S > C

S > T

C > T

C > S

T > S

T > C

number of tokens

1

6

13

0

0

13

% C : T

50

50

% S : T

100

0

% S : C

100

0

Table 5.2 Clauses with three competitors for the prefield in MHG (Bavarian)

S > C, T

C > S, T

T > S, C

number of tokens

9

1

0

Examples for some of the competition cases are given in (5). Here we relate cases with three competitors: (5a) represents the ‘normal’ case in which the scene-setting element is in the prefield (note that is clearly temporal-referential here and not simply a cohesive device as e.g. also in (6)), (5b) represents the ‘special’ case in which the contrastive element is in the prefield. The information-structural properties are marked as in example (3). (p.69)

(5)

The prefield after the Old High German period

It can be noticed immediately that no ranking between contrastive elements and topics is established, not even the slight preference for contrastive elements visible in Modern German, whereas scene-setting elements are already as much preferred for the prefield as in Modern German. The total number of sentences showing competition captured by the prefield ranking is fair: out of 100 sentences, forty-three show competition of the relevant kind. It should be mentioned that in quite a number of sentences the prefield is occupied by an element not included in the ranking, most often an adverbial element that seems to serve mostly as a cohesive device, such as alsô ‘in this way’ or ‘in the same way’ or ‘consequently’, cf. (6). As a matter of fact, if there was competition between a cohesive device and some other prefield filler, which was the case in thirteen clauses, in twelve cases the cohesive device won out.

(6)

The prefield after the Old High German period

Note that this cohesive adverbial may oust ‘better suited’ prefield candidates; in (6), the topic (got ‘God’) and an element that can be analysed either as scene-setting or as contrastive element at the same time (alle naht und tac ‘all days and nights’) appears in the middle field.1

A similar study was made on a MHG sermon from the East Central German area (Mitteldeutsche Predigten). The numbers point in the same direction, although the tendencies are less pronounced, as can be seen from Table 5.3 and Table 5.4.

Table 5.3 Clauses with two competitors for the prefield in MHG (East Central German)

S > C

S > T

C > T

C > S

T > S

T > C

number of tokens

1

14

12

1

2

4

% C : T

75

25

% S : T

87.5

12.5

% S : C

50

50

Table 5.4 Clauses with three competitors for the prefield in MHG (East Central German)

S > C, T

C > S, T

T > S, C

number of tokens

4

0

1

Note that clauses in which competition occurs at all are infrequent (mostly there is only a topic as potential prefield filler), so even larger portions of text yield only a few examples. Therefore the calculations should be treated with some caution, but as they point in the same direction, one feels justified in drawing conclusions. Competition between scene-setting and contrastive elements was rare, so these numbers are not meaningful. With the other competition cases, it should be noted that—on the whole—there are clear tendencies in favour of scene-setting elements over topics (p.70) (although the percentage is lower than in the Bavarian text) and of contrastive elements over topics (the tendency is even clearer here as compared to Bavarian). In cases with three competitors, scene-setting elements are in general similarly successful as in Bavarian.

In the East Central German text (Mitteldeutsche Predigten), a similar phenomenon occurs as in Bavarian, namely that the prefield is relatively often occupied by a cohesive device (in this case mostly , which indicates the discourse relation of narration, and , which corresponds more or less to ‘yet’), most notably in cases in which a prefield filler would be available. In twenty-four out of twenty-six cases, a cohesive device wins over other types of prefield fillers. This phenomenon can be interpreted as a strategy to cope with the V2-constraint which might be either particular to the two dialect areas from which the texts come or might be the general state of the MHG language in this respect, and which is different from the strategy that Modern German is following. Whereas in Modern German the prefield is filled by actual movement of some phrasal element that is part of the clause proper into SpecCP, in the MHG texts there is a tendency to base-generate elements in SpecCP rather than to fill this position by movement. The fact that verb-first clauses are not infrequent in OHG might indicate that at that time, the V2-constraint was not yet generalized; MHG might reflect a stage of the evolution of the V2-constraint in which the link between V2 and declarative sentence mood was already firmly established, but the movement operations filling the prefield were not yet as freely available as they are in Modern German. In some ways, this might be a continuation of the hypothetical pre-OHG state-of-affairs where particles indicating sentence mood (and probably discourse particles) regularly, if not compulsorily, appeared in the left periphery (Axel 2007: 35ff and references therein; see also Chapters 3 and 4 of this volume).

(p.71) 5.2.2 Early New High German

A similar, but broader, investigation was conducted with texts from the fifteenth century, that is, the early part of the ENHG period. Texts of two comparable dialect areas, viz. Upper Saxon (again an East Central German dialect) and Central Bavarian, were selected. In contrast to the MHG investigation, texts of three text types were chosen from each dialect area, namely chronicle, narrative, and sermon. The texts constituting the corpus are given in Table 5.5. For each text type in each region, 100 clauses were analysed, giving a total of 600 clauses. If a text turned out to be too short, a second text of the same text type was added.

Table 5.5 The ENHG corpus

Region

Text type

Text

Date of composition

Number of analysed clauses

Upper Saxon

chronicle

Urkunden des Oberlausitzer Hussitenkrieges und der gleichzeitigen die Sechslande angehenden Fehden.

1423–5

29

Pilgerfahrt des Wilhelm von Thüringen

not before 1461

71

narrative

Ritter, Bürger, und Bauern

first half fifteenth century

65

Das Leben der heiligen Dorothea von Johannes Marienwerder

between 1402 and 1417

35

sermon

Predigt (Anonymous, probably from Leipzig)

first half fourteenth century

100

Central Bavarian

chronicle

Jörg Kazmair’s Denkschrift über die Unruhen zu München

mostly 1397–1402

100

narrative

Die Denkwürdigkeiten der Helene Kottanerin

not before 1445; probably c.1450

100

sermon

Das ,Beichtbüchlein‘ des Thomas Peuntner

first half fifteenth century

100

In the sample, there was no significant difference between the dialect regions, so the numbers are combined.

The first observation is that it is the same three information-structural types of elements that are found in the prefield as in Modern German. So here the modern state-of-affairs is already present. An innovation compared to MHG is that cohesive devices do not play such a central role, although one Bavarian text (Die Denkwürdigkeiten der Helene Kottanerin) tends heavily towards filling the prefield with discourse connectives, mostly , which indicates the discourse relation of narration. In forty-nine cases, was used in the prefield although other potential prefield fillers were present. In the Upper Saxon texts combined, it was only eight cases. As (p.72) occurs only in the prefield in the texts, it could be regarded as a prefield expletive.2 So the strategy identified at the end of the last section—to fill the prefield by base-generating cohesive devices rather than moving something out of the clause proper to SpecCP—seems to linger on in Bavarian longer than in East Central German.

Turning to the competition cases, we see a slight difference from Modern German. The numbers of competition cases are reported in Table 5.6 and Table 5.7, which are to be read as explanations for Tables 5.1 and 5.2 respectively.

Table 5.6 Clauses with two competitors for the prefield in ENHG

S > C

S > T

C > T

C > S

T > S

T > C

number of tokens

53

70

34

1

2

16

% C : T

68

32

% S : T

97

3

% S : C

98

2

Table 5.7 Clauses with three competitors for the prefield in ENHG

S > C, T

C > S, T

T > S, C

number of tokens

8

0

0

Examples for some of the rankings are given in (7). In (7a), a scene-setting element is positioned in the prefield outranking a contrastive element, in (7b), a scene-setting element stands in the prefield, while the topic is in the middle field, and in (7c) a contrastive element is in the prefield, outranking a topic.

(7)

The prefield after the Old High German period

(p.73) So, the ENHG sources indicate that indeed there was a continuous development between MHG and Modern German, namely a gradual establishment of the modern V2-syntax and as a consequence a change in the ranking of prefield fillers. The modern version of the V2-constraint (obligatory prefield filling by movement in declarative clauses) gained ground; whereas in MHG, V2 clauses in which the prefield is not filled by movement of some phrase out of the VP-domain into the left periphery are common, in ENHG they are only exceptions. In the cases in which the prefield is filled by movement, the modern ranking (scene-setting element > contrastive element > topic) is on the way to being established in ENHG; whereas in MHG, the OHG state-of-affairs—where the competition of topics and scene-setting elements often ends up with the topic as the winner, and contrastive elements do not play a role at all—is still traceable. In information-structural terms, one might interpret this as a development of the prefield from a position specialized for topics to a less specified position that can host any element that has a context-linking quality (to which scene-setting elements and contrastive elements belong). In structural terms, we see that the association between the filling of SpecCP and the sentence mood ‘declarative clause’ is already more or less compulsory by MHG times. So we might assume that in MHG at the latest, the declarative sentence mood was implemented as a feature that could be satisfied either by base-generation of elements (as it presumably was in pre-OHG times) or by movement of some phrase to the left periphery (as is the normal case today). In the course of the development, the movement option became more and more favourable. It is not clear why this should be the case; one could assume that base-generation is a costly operation (which in optimality-theoretic (OT) terms could be implemented as a violation of a highly ranked faithfulness constraint DEP-IO(DS-PF-pair)3) whereas movement is less costly (which, in OT terms could be implemented as a violation of another lower-ranked faithfulness constraint IDENT-IO(DS-PF-pair)4). This would hinge on base-generation being more costly than movement, however, which is evidently not true. At a stage of the language in which sentence mood was indicated almost compulsorily by particles, there would be no violation of DEP-IO(DS-PF-pair), as the mood particle is ‘meaningful’ and therefore not pure expletive.

(p.74) 5.3 On the doubly filled prefield in Early New High German

5.3.1 The data

In Chapter 3, we saw that the one-constituent property holds in most OHG declarative clauses. In the preceding sections of this chapter, we saw the one-constituent property at work in MHG and ENHG. In fact, V2 declarative clauses—clauses in which the prefield is filled by one constituent—are the rule. Yet the implementation of the one-constituent property is not absolutely categorical; whereas in Modern German, most if not all counterexamples are only apparent exceptions (in the sense that the prefield is filled via remnant movement of a partially emptied VP, see Müller 2005), in ENHG clear violations of the one-constituent property appear every now and then. A case is counted as a clear violation if the order of the constituents in the prefield does not correspond to the base-generated order that the constituents would have if they were in the middle field and no scrambling had applied. Examples are given under (8); they are cited from the Bonn ENHG corpus. (8a) is an example in which a PP functioning as local adverbial (information-structurally more precise: scene-setting element) occurs together with an NP functioning as subject and representing the topic of the sentence in the prefield; in (8b) an adverbial connector (see Chapter 6) and a personal pronoun subject share the prefield. The constituents are indicated by square brackets.

(8)

The prefield after the Old High German period

The problem of doubly filled prefields in ENHG was investigated at length in an earlier study (Speyer 2008a), on which this section is based. In this study, eight texts from four dialect areas (Bavarian, Alsatian, Ripuarian, Upper Saxon) were used that are available via the Bonn ENHG corpus. For each dialect area, two texts were chosen: one from the period between 1350 and 1450, and one from the period between 1450 and 1550. In order to assess the further development, an analysis of texts from the period between 1550 and 1650 was added in the present chapter. The latter texts are listed in Table 5.8. In Speyer (2008a), all declarative main clauses following the one-constituent property were identified from a text sample, likewise those in which more than one constituent occupies the prefield were gathered from the sample. So it was possible to assess the percentage of declarative main clauses with doubly filled prefield out of the total of declarative main clauses that have a filled prefield. These percentages are reported in Table 5.9. For the clause count we refer to Speyer (2008a); the separate clause count of the latest period is additionally given in (p.75) Table 5.8. If we estimate the rate of clauses with double prefield filling, we get different results for each dialect area (Table 5.9, Figure 5.1).

Table 5.8 Additional texts for the period 1550–1650

Author

Text

Clause count

With double prefield

Bavarian

Georg Scherer

Christliche Leichpredigt, Bey der Käyserlichen Besing- und Begängnuß, […] (1603) (only part I)

115

1

Alsatian

Johann Michael Moscherosch

Wunderliche und wahrhafftige Gesichte Philanders von Sittewalt (1642) (pp. 9, 1.7–28, 1.2)

141

7

Ripuarian

Matthias Quad

Johannes Gropper

Imago Flandriae. Erinnerung an den Leser über gegenwärtige Figur (1604)

Briefwechsel II (1547–59)

(pp. 25–45)

36

79

1

1

sum of both texts

115

2

Upper Saxon

Jacob Böhme

Morgenröte im Aufgangk (1619) (pp. 48–56)

123

0

Table 5.9 Percentage of double prefield filling in declarative main clauses between 1350 and 1650

Bavarian

Alsatian

Ripuarian

Upper Saxon

1350–1449

0.61

1.96

3.85

2.11

1450–1549

1.21

2.05

2.64

0.69

1550–1650

0.87

4.96

1.74

0

The prefield after the Old High German period

Figure 5.1 Development of double prefield filling between 1350 and 1650

(p.76) While it is impossible to do either a χ‎2-test for all data (some cells are 0) or a Fisher’s test (the numbers are partly too big), a χ‎2-test for Alsatian versus Bavarian showed weak significance (p = 0.06), and a Fisher’s test for Alsatian versus Upper Saxon showed significance (p = 0.02). It is obvious that some dialect areas are more permissive with respect to multiple prefield filling, whereas others are more restrictive to begin with. It is noteworthy that the rate decreases significantly in the more northern dialects which had a relatively high frequency in the first period, whereas in the Southwest it even seemed to increase over that period. Bavarian kept a relatively low frequency. Upper Saxon started with a relatively high frequency, but it went to zero in the observed timespan, being the only dialect that did away with double prefield filling completely. It is perhaps no coincidence that Modern German shows virtually no double prefield filling, since its main source is the dialect which is the most restrictive with respect to prefield filling.

In terms of syntactic change, we can say that some change must have happened in the Northern dialects and possibly Bavarian (although it is possible that the underlying structure remained unchanged, and it is only a change in frequency of possible outputs, which could be modelled in a (stochastic) OT framework as a change in the relative value of the constraints involved), a change whose outcome is a more restrictive version of the one-constituent property, whereas in Alsatian the one-constituent property is on the whole not obeyed as strictly as in other areas. The exceptions become even more numerous in the latest period.

5.3.2 Implications for the syntactic analysis

The mere presence of sentences with more than one phrase in the left periphery challenges the standard view of German clause structure that goes back to den Besten (1983) (see Chapters 1 and 3 of this volume). According to him, the prefield is equivalent to SpecCP, as shown in (9). This means that only one position is available for movement to that position.

(9)

The prefield after the Old High German period

Consequently, the few instances of multiple prefield filling in Modern German are often accounted for by remnant movement of a partly emptied VP as discussed above, thus salvaging den Besten’s model (Müller 2005). This analysis is possible as long as the prefield fillers occur in the linearization in which they would be base generated within their VP, as we have to assume that no scrambling of the VP-internal constituents has occurred prior to remnant movement. With Müller’s examples—if we choose the reading in which für die Geschworenen ‘for the jury’ is an immediate constituent of VP—this analysis works quite well, cf. (10). (p.77)

(10)

The prefield after the Old High German period

The analysis does not work equally well, however, for several ENHG examples such as (8a). Here the adverbial appears before the subject. If we assume that the subject is in some functional projection higher than VP, the order of the adverbial before the subject would be an instance of scrambling.5 But remnant movement of scrambled constituents is excluded.

Thus it appears that not all cases of double prefield filling in EHNG can be accounted for by Müller’s remnant movement hypothesis. The only way out is to concede that the prefield consists of more than one syntactic position, although it is most often the case that only one of these positions is the target of movement or otherwise overtly filled. Den Besten’s analysis with one structural prefield position, on the other hand, covers the majority pattern and modern grammaticality judgements about the left periphery very well, so one should be hesitant to forgo this analysis. But is there a way to reconcile the ENHG findings with den Besten’s analysis?

There are basically two ways to achieve this: one is to allow multiple specifiers in the C-architecture (cf. Fuß 2008). Historically, German would then have had multiple specifiers in C, just as for instance Bulgarian does today (see Bošković 2002), but some mechanism normally prevented more than one of these specifiers being filled. According to Fuß (2008), there was only a single CP in OHG, but the C-head could project ‘multiple specifiers hosting fronted XPs’ (Fuß 2008: 270). In the unmarked case in OHG (as well as in NHG), C was endowed with an EPP feature, so it could only attract the closest XP in the middle field which is then raised to SpecCP. In cases, however, where C was endowed with an additional feature (e.g. a topic or focus feature), more than one XP was raised to the prefield where the XPs occupied different SpecCP positions. The syntactic change whose eventual result was the loss of multiple XP fronting comprised a change of the feature endowment of C because the fronting of expletive thô, which emerged (roughly) in the OHG of the ninth century, led to the reanalysis of XP fronting as a semantically vacuous movement whose only function is to check the EPP feature of C.

The second option of analysing the findings is to follow a cartographic approach to the left periphery of the clause, such as outlined by Grewendorf (2002). A sketch is given in (11). (p.78)

(11)

The prefield after the Old High German period

Here, we have more than one C-projection (and thus more specifiers, i.e. phrasal landing positions). As indicated above, the way to restrict movement to more than one C-position outlined by Grewendorf (2002) is that movement of any phrase necessarily goes through SpecFinP. The phrase would move on to a suitable landing site, but the trace that it leaves would prevent further movement of some other phrase to SpecFinP and thus into the left periphery. An alternative was suggested by Speyer (2008b), where long movement leaving out FinP is in principle possible, but dispreferred; therefore, long movement occurs only very rarely. This dispreference was modelled using Stochastic Optimality Theory (Boersma and Hayes 2001) in that normally a high-ranked constraint 1-VF (‘Only one constituent is in the prefield’) preventing more than one constituent in the C-architecture does its job; only occasionally a paradox ranking occurs, in that the constraint 1-VF receives a lower value than other constraints that take care to move for example topical or contrastive material to the prefield. These are the instances of multiple prefield filling. It is perhaps noteworthy in that connection that the order of referential phrases in multiple filled prefields are consistent with the serialization forced by the cartographic structure in (11).

The syntactic change mentioned above would then be caused by the increase (in Upper Saxon and Ripuarian) or decrease (in Alsatian) of the median value of the constraint 1-VF in that model. Such changes in value can be thought of as a feedback effect: if for some reason (or just randomly) violations of a given constraint occur less frequently than expected, language learners may assume a higher value of this constraint, which then leads to even fewer violations in production etc. (similarly for violations occurring more frequently: language learners assign a lower value, which shows in even frequencies of violations in production). When the value of 1-VF reaches such a high number that violations are practically non-existent, the language learner would lose the evidence for more than one position in the prefield and reanalyse the sentence structure as a conflated CP-pattern with just one CP (alternatively, if we follow Grewendorf’s analysis, the reanalysis consists of a ban on long movement over a trace in FinP). The few instances of double prefield filling which such a language learner might occasionally hear would be analysed as remnant movement; this language learner would admit doubly filled prefields as instances of (p.79) remnant movement in his own grammar and occasionally even use them—but only those that are compatible with the remnant movement analysis. This would explain why modern instances of multiple prefield filling are usually compatible with a remnant movement analysis.

5.3.3 Excursus on post-initial connectives

It is hard to find clear and independent empirical evidence for the existence of a split-CP in German. The examples discussed so far involve two phrases preceding the finite verb in Fin (or C) and this can be captured either with a remnant VP analysis or with assuming multiple specifiers (see above). There is, however, another kind of data which may provide better evidence for a split-CP in German:

(12)

The prefield after the Old High German period

In the sentences in (12) there are also two constituents which precede the finite verb, but in these cases the second one is an adverbial connective (see also Chapter 6). Adverbial connectives which can appear in this special position between an XP in the prefield and the finite verb in Fin/C are called nacherstfähig (‘capable of post-initial position’) in German descriptive literature (e.g. Breindl 2008), and sentences such as in (12) are often considered to be V3-clauses (see Breindl 2008: 32). As shown in Volodina and Weiß (2010), there are in principle two possibilities to analyse the syntax of connectives in post-initial position: (i) the XP and the connective form a single phrase which is built in the middle field and then raised to SpecCP; (ii) the adverbial connective is merged in Top and the preceding XP is moved from its base position to SpecTopP. The first analysis would be in accordance with the V2-property of German, the second one with a cartographic structure. Although it is hard to decide which analysis is the appropriate one only on the basis of simple sentences such as (12), more complex sentences (see (13)) and the semantics of the connectives—they mark topic change (Onea and Volodina 2009)—speak in favour of the second analysis (Weiß 2011):

(13)

The prefield after the Old High German period

(p.80) (13a) is a left dislocation structure where, according to the assumptions made by Grewendorf (2002), an XP is raised to SpecTopP via SpecFinP leaving behind a resumptive d-pronoun in that position, see (14a). A comparable analysis is possible for (13b): the complex DP [die Antwort, die Thales gibt], presumably forming a big DP (Grewendorf 2002), is raised first to SpecFinP, where the upper part of the big DP is extracted and raised to SpecTopP, see (14b).

(14)

The prefield after the Old High German period

According to this analysis, the left periphery of German sentences would be more complex than the common V2-analysis assumes.6

Interestingly, post-initial connectives already occurred in the earliest OHG texts (see also Chapter 6 for further details). The adverb/conjunction aber ‘but’ is probably the lexeme which is attested earliest in this position: Example (15a) is from the OHG Isidor, which, however, mirrors the word order of the Latin original (filius autem datus est nobis). This is mostly the case with OHG examples. However, at least since MHG times, we find examples from texts which are not translations from Latin (see (15b)), so we can conclude that the post-initial positioning of these adverbial connectives reflects a genuine option of German syntax (see also Chapter 6 on this point).

(15)

The prefield after the Old High German period

5.4 Summary

The one-constituent property of Modern German declarative main clauses was basically in operation already in MHG in that most declarative main clauses conformed to this property. The decision regarding which phrase is moved to the prefield is information-structurally governed in Modern German. In MHG and ENHG, roughly the same information-structural tendencies hold, namely that scene-setting elements are the preferred prefield fillers in cases of competition with other potential prefield fillers (contrastive elements, topics). In contrast to Modern (p.81) German, the competition is not so clear-cut in disfavour of topics in MHG. Another difference between MHG and ENHG is that in MHG, the prefield is more often filled not by movement, but by base-generation of elements indicating the logical connection to the preceding sentence; in ENHG the development towards the Modern German filling by movement is visible. The one-constituent property is not so established that it does not allow for exceptions; these exceptions can be captured by a cartographic analysis.

Notes:

(1) Elements in which the assignment to one of the three classes of preferred prefield-fillers is ambiguous were excluded from the database from which the ranking was inferred.

(2) In modern Bavarian, where the prefield -es is lacking, da still occurs in this function, see Weiß (1998: 102). On different properties of clause-initial versus clause-internal thô (OHG predecessor of ), see also Chapter 4.

(3) DEP-IO stands for ‘the output is dependent on the input’. DS-PF specifies what the input and the output is: the input is a Deep Structure, the output a Phonetic Form. The optimal candidate under this constraint is one in which all material in the output corresponds to material in the input (which does not mean that all material from the input must be represented in the output).

(4) IDENT-IO stands for ‘the output is identical to the input’. For the other abbreviations see preceding footnote. The optimal candidate under this constraint is one in which there is no difference between input and output (that is: the Phonetic Form directly spells out the Deep Structure).

(5) The analysis would still apply, however, if a VP-internal subject position were assumed which is preceded by the respective adverbial in VP-adjoined position.

(6) For more details of, and problems with, this structural proposal see Volodina and Weiß (2010); Weiß (2011); and Catasso (2015). For an alternative treatment of post-initial connectors such as aber see also Chapter 6.2.2.3.