Available online at www.sciencedirect.

com

Lingua 120 (2010) 329–353

On the relation between V2 and the second position cliticization§

Abstract
Following the standardly adopted hypothesis of a uniform diachronic source of V2 in Germanic and second position cliticization
(2P), the paper investigates both operations in order to verify common claims concerning the motivation for the V2 and 2P
movement. It shows that neither of them can be analyzed as a PF-driven phenomenon, nor is it possible to attribute them to a uniform
syntactic trigger, such as Tense or Illocutionary Force feature checking. Special attention is given to the 2P effect in Slavic, whose
properties show that it is necessary to draw a distinction between the ‘‘generalized’’ 2P, which is a syntactic process but seems
unrelated to any feature checking mechanism, and Force-related 2P, which occurs for Force/Operator feature checking. The findings
concerning 2P are applied to draw conclusions about the trigger of V2 in Germanic. It is argued that V2 covers a number of unrelated
cases of movement or base generation, whose only common property is the position of the verb after the first constituent. Therefore,
it seems incorrect to attribute all cases of V2 to a uniform syntactic trigger, in particular to Force marking.
# 2009 Elsevier B.V. All rights reserved.

Keywords: V2; Clitics; Illocutionary force; Germanic; Slavic

This paper investigates possible links between the V2 requirement in Germanic and the second position cliticization
(2P) in Slavic. The hypothesis that the two phenomena are related has a long tradition in linguistics, and dates back to
Wackernagel’s (1892) study of word order in early Indo European languages. Wackernagel observed that the word order
was rather free in these languages, but a number of unrelated categories, such as pronouns, auxiliaries, adverbs and
discourse particles, always appeared after the first word in the sentence. These elements behaved like clitics, as they were
unaccented and occurred together forming clusters. Wackernagel declared that this was the basic word order rule in early
Indo-European languages and tentatively suggested that it is reflected in contemporary languages, either in the form of
second position cliticization (Wackernagel clitics), or the Verb Second rule in Germanic.
Wackernagel’s tentative idea has been adopted in a number of analyses, including Anderson (1993), Franks (2000),
and Bošković (2001). Thus, Franks (2000:17–21) conjectures that all languages are V2 at some level of representation,
so any account of V2 should also be valid for Wackernagel clitic placement. Obviously, V2 differs from second
position cliticization in not being related to any prosodic requirements in most cases and in being restricted to finite
verbs. However, Franks argues that prosodic differences aside, the actual crosslinguistic variation is limited to the
position where the finite verb is spelt out. Some other researchers have postulated that the V2 rule may be operative
also in the languages in which the verb comes second only in limited environments. For example, Rizzi (1996) states

§
This research was funded by the Netherlands Organization for Scientific Research (NWO) under the Rubicon grant 446-05-008.
E-mail address: krzysz75@yahoo.com.

0024-3841/$ – see front matter # 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.lingua.2008.05.005
330 K. Migdalski / Lingua 120 (2010) 329–353

that the subject-verb inversion in English is ‘‘residual V2’’, a special subcase of Verb Second that is restricted in
English to operator contexts.
Admittedly, not all linguists support the idea that V2 in contemporary Germanic is directly linked to the
cliticization pattern in early Indo-European. For instance, Kiparsky (1995:159–161) points out that the assumption is
based on the misguided premise that all unaccented words are clitics. There is no reason to assume that all verbs in the
second position in the languages studied by Wackernagel were clitics, because lexical categories such as main verbs
are never clitics. It is only auxiliaries and other functional elements that are attested as clitics. Moreover, modern
Germanic languages also have V1 constructions (e.g. in yes–no questions), and it is unclear how they could be related
to second position cliticization.
Even if some of Kiparsky’s objections are justified, this paper will provide arguments that there is indeed a syntactic
correspondence between the two processes. Moreover, it has recently been argued that V2 in Germanic and second
position cliticization in Slavic exhibit similar stages in their diachronic evolution (cf. Migdalski, 2007).1 If this is
indeed the case, it might be useful to investigate properties of Wackernagel clitic placement in order to verify common
assumptions concerning V2, especially with respect to the trigger of the V2 movement and the elements preceding the
verb. This paper is a step in this direction and is organized as follows. Section 1 looks into the possibility of analyzing
V2 as a phonologically conditioned operation, on a par with the clitic placement in early Indo-European languages.
Section 2 investigates syntactic accounts of V2, focusing on those that relate V2 to Force marking. It postulates that
there is no uniform trigger for V2, therefore it is necessary to distinguish Force-driven V2 from other types of
movement to the second position. This postulate receives more support from Slavic clitic data presented later in the
paper. Section 3 examines patterns of Wackernagel cliticization in contemporary Slavic languages. It demonstrates
that although clitics are phonologically deficient, they behave like syntactic units, as they are sensitive to syntactic
locality constraints. The section also argues that the ‘‘genuine’’ V2/2P cases2 described by Wackernagel cannot be
captured via a uniform syntactic feature checking mechanism. Neither can they be equated with other instances of verb
movement that result in the same V2 word order, but occur for clausal typing, Force or Tense feature checking. Section
4 analyzes Force-licensing in 2P and non-2P Slavic languages and shows that it is entirely independent of second
position cliticization. This fact is taken to indicate that V2 is not related to Force-marking, either.

1. Relation between V2 and phonological requirements

An intriguing property of V2 is the fact that finite verbs move to the position in which they can be preceded by
categorially unrelated elements. This makes the movement rather different from other syntactic operations, which
occur in order to establish a relation with a specific category resulting in feature checking. What is relevant in V2
contexts is that the finite verb follows the initial constituent, with seemingly few restrictions as to what this constituent
might be. This may give an impression that the movement does not occur in syntax, especially since it has been
claimed to have its origin in a prosodic requirement. Therefore, in order to determine if V2 is a syntactic operation, it is
necessary to investigate whether it can be influenced by PF restrictions in contemporary Germanic languages.
As was mentioned in the introduction, Wackernagel’s original insight was that the elements occurring in the second
position in main clauses in early Indo-European languages were unstressed. This insight raises two questions. First, it
is necessary to determine whether the prosodic deficiency of the elements occupying the second position affected the
syntax of these languages. Second, it is crucial to examine the nature of the material preceding the clitics. If it can be

1
Generalized V2 found in most contemporary Germanic languages is an innovation. There was no generalized V2 pattern in Old Germanic; at
first the rule was limited to Force-related contexts. This is what has been observed by Eythórsson (1995) and Fuss (2003), who notice that the oldest
Germanic texts (such as the Gothic translations of the Bible from the 4th century) show regular deviations from the original Greek texts, which
correspond to the ‘‘residual’’ V2 in Modern English (i.e. V-to-C movement in wh-questions, yes–no questions, neg-preposing, and imperatives). It
seems that second position cliticization in Slavic developed through the same pattern: in Old Church Slavonic it was restricted to clitics specifying
the Illocutionary Force of a clause, and in some languages (e.g. Serbo-Croatian, Czech, Slovak, Slovene) it was generalized to all clitics, but only at a
later stage (see Radanović-Kocić, 1988; Migdalski, 2007). There are also striking synchronic similarities between V2 and second position
cliticization, which seem difficult to account for, but indicate that these processes are related. For instance, in German the conjunctions und ‘and’
and aber ‘but’ do not trigger V2, whereas entweder ‘either’ is compatible with V2. I am not aware of any analysis explaining the relation between a
type of conjunction and the possibility of V2, but Ćavar and Wilder (1999) notice that the distribution of second position clitics with the related
conjunctions in Serbo-Croatian strikingly resembles the German V2 pattern.
2
By ‘‘genuine’’ V2/2P I understand the cases of verb or clitic movement to the second position that have no effect on the sentences interpretation
and which therefore cannot involve movement related to Focus, Topicalization or Polarity marking.
 K. Migdalski / Lingua 120 (2010) 329–353 331

shown that the elements occurring in front of the clitics are syntactic constituents (which can be verified by testing
whether they are able to move independently in the absence of the clitics), then it is clear that the lack of stress on the
element in the second position had no syntactic consequences. Such a scenario is in fact to be expected in the current
model of narrow syntax, which assumes that phonological requirements have no direct influence on syntactic
movement.
Dewey’s (2006) recent study of verb placement in Old Saxon, Old Icelandic, and Gothic confirms that the
unstressed categories comprised both auxiliary and lexical verbs. She also observes that the elements preceding the
verb in the second position do not form a natural syntactic class. They could be made up of unrelated categories, and
sometimes they seem to be non-constituents, as in (1a), where the verb follows the demonstrative/article, and (1b),
where the verb intervenes between the quantifier and the NP. Given this Dewey suggests that the placement of the verb
in the Early Germanic languages was determined prosodically.

However, it is far from obvious that the verbs in (1) indeed cause discontinuous constituency. It does not have to be
so if Early Germanic languages allowed Left Branch Extraction, and thus calculated constituency differently than
Modern Germanic languages.3 An important typological variation related to Left Branch Extraction concerns the
presence of the DP layer in a language. It has been known since Corver (1992) that languages without articles (hence,
without the DP projection) permit Left Branch Extraction. Tonya Dewey (p.c.) informs me that the definite article in
the languages she discusses was homophonous with the demonstrative. Therefore, it is not certain whether they had
articles at this stage at all.4 Observe that both of the initial elements in (1) show nominal morphology. This makes the
article in (1a) morphologically alike to demonstratives in Slavic languages, in which they show adjectival morphology
(they inflect for case, gender, and number), and in which they also have the same distribution as adjectives. It is
difficult to provide sufficient syntactic evidence that the preverbal material in (1) is adjectival without additional
corpus investigation, given the limited data that are available. However, there is compelling typological evidence
indicating that demonstratives as well as possessives, that is the elements commonly argued to occupy the DP layer in
the languages with articles, are adjectival in the article-less languages. For instance, they may appear in the predicative
position in an article-less language such as Serbo-Croatian. This is not possible in English, a language with articles.

Moreover, these elements exhibit a relatively free word order, on a par with adjectives (cf. 3), but in contrast with
determiners and possessives in English (cf. 4).

If the typological generalization is correct, and if the languages exemplified in (1) indeed lack the DP layer, the
‘‘constituent split’’ can be readily accounted for by, for example, appealing to Bošković’s (2005) proposal concerning

3
This is what also happens in contemporary Slavic 2P languages, which, as will be shown later in this paper, permit the same patterns of 2P clitic
placement as the finite verb in (1).
4
Giusti (1995) observes that the emergence of the definite article in Germanic languages coincided with the loss of morphological case. If her
observation is correct, it means that the languages exemplified in (1) have no articles, as all the nouns show rich case distinctions.
332 K. Migdalski / Lingua 120 (2010) 329–353

the structure of APs and NPs/DPs. Bošković assumes that in languages with articles A takes NP as its complement, as
in (5a), whereas article-less languages host APs in the Specifier of NP, so that NP dominates AP, as in (5b).

This gives an easy way to handle Left Branch Extraction: it is prohibited in DP languages, because AP is not a
constituent to the exclusion of NP, so Left Branch Extraction involves extraction of a non-constituent. AP is a separate
constituent in DP-less languages, and therefore it may be extracted. If this proposal is on the right track, the initial
elements in (1) end up preceding the finite verb, because originally they were located in Spec, NP; and hence, were
extractable. This option does not exist in the contemporary Germanic languages, because their NPs acquired the DP
layer, which makes their AP structure immobile.
In most contemporary Germanic languages V2 is not conditioned by phonological requirements. However, there
are some exceptions. Rice and Svenonius (1998) observe that V2 placement in the Tromsø dialect of Northern
Norwegian seems to be sensitive to prosody in wh-questions: the finite verb in V2 contexts must be preceded by at least
one foot (i.e. two syllables); otherwise, it occurs in the third position. Thus, the wh- words korfor, korsen, and katti
(‘why’, ‘how’, and ‘when’) always require V2 (cf. 6). By contrast, the finite verb is not required to appear in the second
position after the monosyllabic question words ka, kem, and kor (‘what’, ‘who’, and ‘where’), as shown in (7).

Whether V2 in the Tromsø dialect is indeed phonologically conditioned is a matter of debate. Westergaard and
Vangsnes (2005) suggest that information structure requirements decide about the V2/V3 variation. They observe that
V3 is preferred when the subject is definite or understood as familiar from the previous context, as well as when it is a
pronoun or an expletive. There seems to be even a definiteness restriction on the pre-verbal position in V3 contexts,
which is not observed in the V2 patterns.

Westergaard & Vangsnes (W&V) try to explain the variation by assuming Rizzi’s (1997) split CP for Norwegian. In
their view, the V2 versus V3 pattern is due to the availability of phrasal or head movement of a wh-word to a functional
projection within the complex CP.

They suggest that in declarative main clauses V2 in all Norwegian dialects is triggered by a Topic feature, and the
verb moves to the Topic head (cf. 11). In wh-questions, the derivation proceeds differently depending on the language
variant. In standard Norwegian V2 is the result of verb movement to Int, while the wh-word moves to Spec, IntP. In
 K. Migdalski / Lingua 120 (2010) 329–353 333

Tromsø Norwegian the longer wh-words also land in Spec, IntP and the verb moves to Int. As for the short wh-words,
W&V propose following Taraldsen (1986) that they are head-like elements in Tromsø Norwegian, so they do not
project. When they land in Spec, IntP, they c-command the Int head, and as a result this projection does not need to be
filled by a verb (cf. 12).

W&V do not present a complete derivation for the wh-movement in Tromsø Norwegian, but it is clearly
problematic, as it violates the Minimal Link Condition. The verb is closer to the Int-head than the wh-word at the initial
stage of the derivation. Therefore, if both the finite verb and the wh-word are able to check the same feature, the
wh-word should never be able to raise across the finite verb, even if one were to assume that the wh-word has clitic-like
properties, thus it raises via XP-movement and lands in a head position.
In order to account for the relation between the indefiniteness of subjects and the V2 patterns in wh-questions,
W&V propose that new information subjects represent a type of Focus and they target a position that is higher than
Spec, TP; namely Spec, AgrsP. Without providing much justification, they argue that the finite verb must raise to
Foc-head to license the Focus feature. If given subjects are present, there is no trigger for V-to-Foc movement, hence
the verb remains in the third position.
W&V propose a fairly complex structure to account for the observed word order variation. With the CP layer that is
so rich, it is difficult to test or falsify their claims, as well as to restrict or exclude the derivations that are not found in
Norwegian, but which might be permitted by the proposed structure. For instance, if the CP layer in Norwegian is so
elaborate, a question that may arise is why Norwegian disallows multiple wh-fronting. Moreover, it is difficult to relate
the preference for the V3 order with pronouns, expletives and definite subjects to the suggested lack of V-to-Foc
movement. The most basic distinction in the V2/V3 pattern is related to the phonological make-up of the wh-element.
Without taking recourse to building up a highly elaborate CP structure, the most intuitive explanation might be to
assume that the short wh-words cliticize onto short subjects, pronouns, and expletives, which are also clitic-like.
Clitics always refer to definite objects and represent old information. In this way, the definiteness effect in the V3
context can be captured straightforwardly.
Apart from Northern Norwegian, it has been suggested that the V2 requirement might be prosodically motivated
also in other Germanic languages. For instance, Boeckx (1998) claims that the V2 order is due to a PF parameter that
forces the finite verb to appear in the second position of its intonation phrase. An example of this PF parameter at work
is given in (13a) for German, where the finite verb heirate is the third syntactic constituent. The sentence is
grammatical as long as the verb is the second prosodic word following a pause.
334 K. Migdalski / Lingua 120 (2010) 329–353

Bošković (2001:173) adopts Boeckx’s idea and proposes that all V2 constructions in Germanic are subject to the
following set of PF requirements:

(14a) states that there must be a phonologically overt element preceding the verb; (14b) specifies that the verb must
be right-adjacent to an intonation boundary. However, Bošković admits that these PF constraints are too general. For
example, V1 constructions in German, such as yes–no questions, are problematic and are incorrectly excluded by (14).
In order for these requirements to work, it would be necessary to stipulate that they are ‘‘somehow suppressed’’ in the
syntactic environments requiring V1. This idea is obviously problematic, because it suggests that phonological rules
are able to detect a syntactic context.
Furthermore, there are some constructions in German in which the finite verb occurs as the third constituent, both
syntactically and prosodically, not because of some phonological constraints, as in Northern Norwegian, but rather due
to requirements related to sentence interpretation and parsing. Thus, Meinunger (2006:149–150), who quotes data
from Pittner (2003), points out that V3 is possible in free relatives if there is a case conflict between a pronoun and a
wh-word.

The examples in (15) present the ‘‘matching effect’’ related to case identity: a verb must appear there in the third
position (and the d-pronoun cannot be dropped) in case the w-constituent in the free relative and the d-pronoun are
marked for different morphological cases. Meinunger argues that in this instance V3 is not a purely morphosyntactic
operation. The placement of the verb is determined by the need to disambiguate readings and facilitate interpretation.
Thus, V2/V3 seems to have a pragmatic motivation in this context.
To summarize, this section has shown that even though V2 is different from other syntactic operations, it would be
difficult to maintain that it is triggered by prosodic considerations. Even though there seem to be a few sporadic cases
in which V2 placement is sensitive to phonology, there are also exceptional examples in which V2 is determined by
pragmatic requirements related to reference disambiguation. Therefore, it would be inappropriate to create a general
rule on the basis of a few exceptions.

2. Syntactic accounts of V2

It has been more common in the literature to claim that V2 is a purely syntactic phenomenon, not related to or
dependent on prosody. The earliest accounts, such as Den Besten’s (1977) analysis of Dutch and Thiersch’s (1978)
analysis of German, postulated that the finite verb targets C in non-embedded clauses in V2 languages. Den Besten
assumed that both C and the finite verb express Tense in these languages; therefore the verb may target (and replace)
the complementizer position. In this way he was able to capture the fact that in all Germanic languages with the
exception of Icelandic and Yiddish, V2 is disallowed in embedded clauses introduced by a complementizer.

Den Besten’s hypothesis sparked a whole range of proposals related not only to the explanation of the V2 effect,
but also aiming at pinpointing a parametric difference between V2 and non-V2 languages. For example, Koopman
(1984) argued that in V2 languages nominative case is assigned under government, rather than in a Spec-head
 K. Migdalski / Lingua 120 (2010) 329–353 335

configuration, therefore the verb must move to C to govern the subject. Platzack (1986) claimed that finiteness is
located higher than in I in V2 languages, so that they require that the verb target C rather than I. Müller (2004), on the
other hand, rejected the traditional head raising analysis of V2 and argued for movement of remnant vP to Spec, CP. A
critical review of different approaches can be found in Vikner (1995) and Brandner (2004). The present paper will
focus on the common idea of analyzing V2 as a means of specifying the Illocutionary Force of a clause. This
assumption will be later challenged by a comparison of second position cliticization and Force-marking effects in
Slavic.

2.1. V2 as Force-marking

One of the most common explanations of the V2 effect is to suggest that the movement is a formal way of
specifying the Illocutionary Force of the clause. On these approaches V2 is argued to license a clause as an assertion, a
question, a command, etc. To my knowledge, the earliest proposal of this kind is due to Wechsler (1991), who remarks,
following earlier observations by Andersson (1975), that in Swedish V2 is possible in embedded clauses when they are
introduced by complementizers of a certain semantic class. Namely, clauses that are embedded under the
complementizers fast(än) ‘although’, eftersom ‘since’, medan ‘while’ are optionally V2. These complementizers
indicate that the propositional content of the clause is asserted by the speaker, therefore they are termed ‘‘assertion-
type’’ complementizers.

V2 is also optional in embedded ‘‘indirect assertions’’, that is in clauses in which the speaker is not the assertor, but
rather the person who conveys somebody else’s assertions. These clauses are introduced by verbs påstod ‘claimed’,
gissade ‘guessed’, and trodde ‘believed’.

By contrast, the clausal complements that are introduced by the conditional-type complementizers, such as om ‘if’,
ifall ‘in case,’ cannot be V2.

Likewise, most Swedish speakears disallow V2 in the embedded clauses when the main clause is headed by factive
verbs such as ‘regret’. These verbs denote propositions whose truth is only presupposed but it is not asserted.

Wechsler concludes that V2 may apply in embedded clauses only if the clauses are direct assertions. He proposes to
capture this in formal terms by suggesting that there are two syntactic features FIN and C, which both constitute an
Illocutionary Force Indicator (in the sense of Searle (1969:30), who uses the term to refer to the devices such as word
order, stress, the mood of the verb, and others, which specify what illocutionary act is performed by the speaker in the
utterance of a sentence). The IRC has different illocutionary potentials, such as a polarity question, conditional, or an
(in)direct assertion. These potentials are assigned by the Illocutionary Rule, which applies across Germanic with
minor parametric differences.
336 K. Migdalski / Lingua 120 (2010) 329–353

It is difficult to render Wechsler’s account in a more recent theoretical framework, as it dispenses with construction-
specific rules. However, his general idea was recently revived by Bentzen et al. (2007), who postulate The Assertion
Hypothesis capturing Wechsler’s empirical observations.

They propose that the strength of assertion is reflected in the size of the structure of the embedded clauses selected
by the main verb. Those that express a higher degree of assertion contain more structure, including ForceP, which is
responsible for the verb movement.
According to some other proposals, all instances of Illocutionary Force marking are rendered by V2, not only those
that express a high degree of assertion. For example, Brandner (2004) argues that V2 licenses the Force value of a
clause in all cases, even in declarative clauses. She follows Cheng’s (1997) Clausal Typing Hypothesis, which states
that every clause needs to be typed. This means that every clause must contain formal means to express whether it is
declarative, a question, or some other type.
The most direct way to type a Force value is to insert a relevant lexical item, such as a Force-marking particle in
languages like Korean (clause-finally) or Persian (clause-initially). Languages that do not have any dedicated elements
for Force marking at their disposal must employ specific syntactic configurations to render Force marking. Examples
of this strategy include wh-movement and V2. Brander assumes that the following structural conditions must be met
for Force marking to be encoded. Force must take a complete clause as its complement, as it may only operate on
already formed propositions. However, to act as a proposition, an event must have a Tense value. Therefore, Tense and
Force values must be legible by having scope over all verbal projections and being part of the main projection line.
Brandner (2004:109) proposes that the following conditions must be met for a (Force) feature to have scope over the
verb.

In this way the Force value can either be read off directly from an inserted particle (as stated in (23a) or via V2
movement, by which the verb leaves the VP and receives its Force-value via an XP in its specifier that bears this feature
lexically (e.g. a wh-phrase). The latter strategy satisfies (23b).
A question that arises is how Force becomes licensed in the languages that have neither V2 nor Force-related
particles, on the assumption that Force is always encoded. Bradner argues that these languages only mark the deviation
from declarative. V2 languages, though, explicitly specify Force in all contexts, including declarative clauses. Thus,
V2 is motivated by Force, but the specific Force value is defined only in the second step, by the constituent that moves
to the Spec above the verb.
Just like Brandner, Koster (2003) assumes that V2 expresses clausal typing, but his view on the way this happens is
different. The easiest way to type a clause is via a complementizer, which indicates whether it is declarative, a
question, or any other type. Given that in languages like Dutch or German complementizers are in complementary
distribution with finite verbs, complementizers also code Tense (or finiteness), as already suggested by Den Besten
(1977). Since complementizers specify clause typing, Koster argues that complementizers are combined Type/Tense
markers, marking both the type of the sentence and the scope of its Tense operator. Given the close relationship
between complementizers and Tense, tensed verbs also type clauses in this way in V2 languages. Koster extends this
idea crosslinguistically and claims that all languages have Tense/Type marker in the second position, proposing the
following constraint.

He states that Slavic languages give support to this constraint, as some of them have tensed auxiliary clitics in the
second position. Section 3.1 will show that this is not always the case, because it is possible to have just pronominal
clitics in the second position, without any tensed elements adjacent to them.
 K. Migdalski / Lingua 120 (2010) 329–353 337

2.2. Problems with V2 as a uniform Force marker

So far all the analyses that have been presented account for V2 effects in a largely uniform way, in the sense that they
argue that the trigger for V2 is always the same, related to Tense or Force. However, there is a problem with this idea,
because some elements located in the prefield, that is in the Specifier of the projection headed by V2, give rise to special
semantic or pragmatic effects. Travis (1984) observes that the nominative subjects always have a neutral interpretation in
the prefield. Objects may appear there only when they are interpreted as topics or focused. For instance (25) could not be
used as an answer to ‘‘what has happened?’’, but only an answer to ‘‘who has Peter invited’’, or ‘‘who has invited Hans?’’.

For Travis this means that the subject is in Spec, TP in German, and that V2 is not always in C. However, the
interpretational differences triggered by the element located in the prefield are more fine-grained and cannot be
reduced to the subject versus non-subject dichotomy. Lenerz (1977) and Fanselow (2002) observe that passive
constructions receive the unmarked interpretation when the dative indirect object, rather than the subject, precedes the
finite verb. This means that in these constructions the indirect object does not move to Spec, CP, which is an operator
position, as it does not receive a topicalized or focused reading.

Likewise, Fanselow (2003:30) (cf. also Haider (1993), and Mohr (2005)) points out that when the prefield is filled
by sentential or temporal adverbs, the sentence also has a pragmatically neutral interpretation.

These data indicate that V2 occurs in many different contexts triggering various interpretations. This fact suggests
that there is no uniform trigger of V2 and definitely not all cases of V2 are related to Force-marking. Therefore, it is
crucial not to analyze all occurrences of the finite verb in the second position as a V2 effect. There might be a number
of syntactic operations which result in V2 placement, but not all of them necessarily exemplify the ‘‘genuine’’ V2 rule,
as described by Wackernagel. Frey (2006) suggests that there are three ways of satisfying the V2 requirement in
German. Frey’s division is related to the type of elements that fill in the prefield.
First, prefield may be filled as a result of ‘‘Formal Movement’’, which consists in raising of the highest constituent
in the middle field to the sentence-initial position preceding the verb. This operation is ‘‘contextually neutral’’, as it
preserves the semantic/pragmatic properties the constituent would have in the middle field. It also does not add any
new properties by raising, nor does it change any stress patterns. An example provided by Frey (2006:241) is given in
(28). It demonstrates scrambling of an instrumental PP, which is perceived as equally marked whether it is the highest
element in the middle field (28b) or in the prefield (28c).

Second, the prefield may also be filled via ‘‘Base Generation’’. The elements that are base-generated there include
expletives or sentential adverbs. They do not reach this position by movement, because they are not in any licensing
338 K. Migdalski / Lingua 120 (2010) 329–353

relation with the verb or any verbal functional head. They also may never be moved from the position in which they
were merged.

Finally, prefield may be filled by A‘-movement, which corresponds to a contrastive interpretation of the
moved element. For instance, it applies to long distance topics, which receive contrastive interpretation in
German.

Max is old information in both clauses in (30), but it does not have to receive a pitch accent in (30a). It receives pitch
accent in (30b). There are also interpretational differences. (30a) is a statement about Max without any reference to
other people. In (30b) Max is understood as discussed in relation to other people who have already been mentioned.
Since this is the only movement that according to Frey gives rise to interpretational differences, I will assume that this
is the type that corresponds to Force-marking. The other two movements bear no effect on interpretation, so I will take
them to be instances of the ‘‘genuine’’, generalized V2 effect. More evidence for this division will be provided in
subsequent sections on the basis of cliticization patterns in Slavic.
To summarize, the data presented in this section indicate that there is no uniform motivation for V2. In other words,
it seems that there are a number of unrelated operations termed V2, but their only common property is the position of
the verb with respect to the clause-initial constituent. Therefore, it is incorrect to attribute V2 to a single syntactic
trigger, such as the requirement to check Tense or a Force feature.

3. Cliticization in Slavic

Given that second position cliticization in Early Indo-European languages has been argued to be the source of V2, it
will be useful to compare V2 effects with Wackernagel clitic effects. Slavic languages are particularly helpful in this
regard, because many of them have retained Wackernagel clitics to various degrees. The languages with second
position clitics include Serbo-Croatian, Slovene, Czech, Slovak, and, in some contexts, Old Church Slavonic.5 Two
South Slavic languages, Bulgarian and Macedonian, have verb-adjacent clitics, largely on a par with contemporary
Romance languages: Polish has auxiliary clitics which are gradually becoming reanalyzed as affixes on the participle
in compound tenses and weak pronouns instead of pronominal clitics. East Slavic languages, such as Russian and
Ukrainian, do not have any pronominal or auxiliary clitics.
This part of the paper has the following organization. Section 3.1 will discuss the distribution of Wackernagel clitics
in Serbo-Croatian. Section 3.2 will show that although the placement of clitics is sensitive to prosodic requirements, it
is also subject to syntactic constraints. However, it is very difficult to provide a purely syntactic account of clitic

5
Old Church Slavonic is the language in which the oldest Slavic manuscripts were written. In descriptive literature it is regarded as a 2P language
(see Lunt (2001)). However, detailed corpus findings show that the distribution of Wackernagel clitics was not entirely uniform, and that it was
constrained to operator clitics that mark the Illocutionary Force of a clause, see Radanović-Kocić (1988), Migdalski (2007), and footnote (1).
 K. Migdalski / Lingua 120 (2010) 329–353 339

distribution, and virtually impossible to pinpoint a feature checking mechanism that is responsible for their location in
the second position.

3.1. Properties of second position clitics in Slavic

The ordering of clitics with respect to each other is the same in all Slavic languages, whether they have 2P clitics or
verb-adjacent clitics. It is presented in an abridged version in (31). The cluster opens with the particle li, which is
often termed the ‘‘interrogative complementizer’’. It occurs in questions and/or focus constructions; thus, it encodes
Force. It will be demonstrated in section 4 that li may have a different distribution from the other clitics. Li can be
followed by a modal clitic. The dative clitic precedes the accusative clitic, while the auxiliary clitics show an
intriguing split concerning the positions of the 3rd person singular form, which always appears as the last one in the
cluster.

Clitics cluster and appear in the second position in both matrix and embedded clauses. Unlike V2 in most Germanic
languages, clitics may co-occur with the complementizer.

Placement of the clitics in any other position than the second, as well as splitting them from each other results in
ungrammaticality.

As in the case of V2 in Germanic, some categories preceding second position clitics give rise to special discourse
effects. In the most neutral scenario, the clitics are preceded by the subject or the participle.

However, sentences with high adverbials, such as manner or sentential adverbs in the first position are also
perceived as discourse-neutral.
340 K. Migdalski / Lingua 120 (2010) 329–353

Other categories, such as the direct object in (36), are interpreted as contrastively focused when they appear in front
of clitics.

In contrast to non-second position clitics in Romance languages or Bulgarian and Macedonian, which are always
adjacent to the verb, Wackernagel clitics do not have to be adjacent to elements of any specific grammatical category.
For instance, they do not have to be adjacent to the participle (cf. 37a) or the finite verb (cf. 37b and c). The only
requirement that they must observe is that they appear in the second position.

The examples in (37b and c) are instructive for at least two reasons. First, they challenge Koster’s (2003) ‘‘Tense
Second Constraint’’, because the elements that occur in the second position in (37b and c) are non-tensed pronominal
clitics. Second, these data show that the clitics do not target a uniform syntactic position. In (37a) they follow the
subject, in (37b) they follow an adjunct, while in (37c) they are sandwiched between two wh-words. This fact also
suggests that it is very difficult to specify a formal feature that triggers the movement of clitic to the second position, as
they may target different projections in the clause structure.
An anonymous reviewer points out that the data in (47) are inconclusive about the clitic placement in the structure,
because even if all the clause-initial elements occupy different projections, it is possible that in those cases the clitics
are hosted by the same head below all of these elements. However, there are additional empirical facts which give more
support for the conclusion based on the examples in (47), and which show not only that there is no designated structural
position for Wackernagel clitics, but also that the auxiliary and pronominal clitics are hosted in different projections in
the tree structure, even though they seem to cluster. This is demonstrated by the interpretation of certain adverbs in the
presence of the two types of clitics.

The adverb pravilno ‘correctly’ is ambiguous and may have a sentential or a manner reading. According to
Bošković (1995), the sentence in (38a), which contains only the auxiliary clitic su, is acceptable under both
interpretations of the adverb. However, the string in (38b), which contains the auxiliary su followed by the dative clitic
joj, permits only the manner-oriented reading of the adverb. Given the standard assumption that sentential adverbs
reside higher in the structure than manner adverbs, this means that the auxiliary clitic su moves higher when it occurs
on its own, as in (38a), than when it is accompanied by a pronominal clitic, as in (38b).

3.2. Placement of clitics: syntax or phonology?

The data presented so far may give an impression that the placement of clitics is phonologically conditioned,
and that what matters is that they are preceded by a prosodic word, especially since they appear to be able
 K. Migdalski / Lingua 120 (2010) 329–353 341

to split syntactic constituent, as in (39a), where the auxiliary clitic is placed between a demonstrative and
a noun.6

This is the approach taken by Halpern (1992), who assumes that in (39b) the clitic appears at the beginning of the
clause in syntax. Since this position is prohibited by phonological constraints, it is moved after the first stressed word
through the operation of ‘‘prosodic inversion’’, which preposes taj in front of je at PF.
However, Wilder and Ćavar (1994) and Progovac (1996) show that there is a general option of separating
demonstratives from nouns in Serbo-Croatian by any overt material. Demonstratives in the Slavic languages have
adjectival morphology, and it is most likely that the clitic in (39b) is positioned after a phrase that has been extracted
from a larger constituent. This proposal gains support from the fact that the noun and its modifier can be split by non-
clitic material, such as the object Milena and the verb voli in (40).

Thus, it is evident that the positions targeted by clitics may also be filled up by non-clitic elements. However,
Progovac (1996) observes that this property should be viewed from the opposite angle, and as a one-way correlation,
because clitics may occur only in those positions that are accessible to non-clitic material. In other words, clitics may
not follow the first word unless this word is a constituent that is independently able to undergo syntactic movement.
This predicts that there will be instances in which the prosodic word can receive stress, but it will not be able to host a
clitic, because it cannot be extracted. This is indeed the case, and is exemplified by coordinate structures, in which the
clitics may not appear after the first conjunct, because in general it is not extractable.

Similarly, the contrast between the two sentences in (42) indicates that clitics may only follow complex NP, but not
the first word, on a par with non-clitic finite verbs.

The contrasts in (41) and (42) are unexpected if the clitic placement is motivated prosodically, because the nouns
sestra and roditelji are stress-bearers. However, from a syntactic point of view there is nothing exceptional about the
ill-formedness of these examples, given that it is impossible to extract the first conjunct, nor can the noun roditelji
undergo syntactic movement by itself.
Thus, it has been demonstrated that the placement of clitics is not phonologically conditioned, even though they are
prosodically deficient and need to be supported. It is possible to go one step further and demonstrate that cliticization in
Serbo-Croatian is in fact subject to the same constraints as syntactic movement. Suitable data showing this property is
provided by Progovac (1993), who investigates clitic movement across different types of verbs. By applying a number

6
Note that a very similar pattern is to be found in the Old Norse sentence exemplifying V2 in (1).
342 K. Migdalski / Lingua 120 (2010) 329–353

of different syntactic criteria, Progovac draws a distinction between verbs that take ‘‘subjunctive-like’’ complements,
and verbs that take ‘‘indicative-like’’ complements. For example, she shows that topics can raise from ‘‘subjunctive-
like’’ complements (cf. 43b), but not from ‘‘indicative-like’’ complements (cf. 43a). Likewise, wh-movement is
prohibited out of ‘‘indicative-like’’ complements (cf. 44a), but it freely occurs out of ‘‘subjunctive-like’’ complements
(cf. 44b). Similarly, negative polarity items may extend their domain only in ‘‘subjunctive-like’’ complements
(cf. 45b).

Importantly, Progovac argues that clitic placement is sensitive to the same dichotomy between the two types of
verbs. The examples in (46b) and (47b) demonstrate that clitics may climb out of ‘‘subjunctive-like’’ complements, but
not out of ‘‘indicative-like’’ complements. The movement occurs for focus reasons.

The data in (46) and (47) have been readdressed by Stjepanović (1999), who observes that if an embedded
subjunctive clause contains two pronominal clitics, only the higher one may climb to the main clause. This is shown in
(48), where only the dative clitic may raise.

The contrast in (46) and (47) as well as the examples in (48) lead to an important conclusion concerning the position
of the clitics in the clause structure. They indicate that although clitics are phonologically deficient, they behave like
syntactic units. Their placement adheres to syntactic locality conditions, which cannot be given a phonological or
 K. Migdalski / Lingua 120 (2010) 329–353 343

morphological explanation, because it is subject to the same constraints as the uncontroversially syntactic operations
like wh-movement and topicalization.7 Moreover, the clitics are located hierarchically with respect to each other, with
the dative clitic positioned higher than the accusative clitic, as indicated by (48c), where the lower (accusative) clitic
cannot be preposed across the higher (dative) clitic due to locality restrictions.
A number of approaches have been proposed to capture the second position requirement on clitics. One of them,
which appeals to the mechanism of prosodic inversion due to Halpern (1992; see also Radanović-Kocić (1988)), has
been briefly mentioned earlier in this section. Currently, an influential hypothesis is the ‘‘scattered deletion’’ approach
due to Franks (1998).8 Assuming Chomsky’s (1993) copy and deletion theory of movement, it proposes that it is not
always the case that the head of a chain is pronounced at PF. If the head of a chain happens to be an illegitimate PF
object, PF may act as a filter and force pronunciation of a lower copy, which does not violate PF requirements. An
example of an illegitimate PF element is a second position clitic that does not receive prosodic support. Franks
assumes that since the second position clitics are enclitic, they must adjoin to a prosodic word to their left in PF to be
appropriately supported. Franks also suggests that clitics must raise as high as possible, to the top-most head position.
If there is no overt material filling the Specifier of the highest head, the clitics may not be pronounced there, because
they are not phonologically supported. Therefore, the lower, second copy of the clitics is pronounced, while the highest
one gets deleted.
This approach crucially presupposes that the second position clitics move to the first, clause-initial position in
syntax. There are some serious problems with this idea. First of all, movement is a last resort operation in the
minimalist framework. It may occur only if it results in checking of an uninterpretable feature. Second position clitics
do not share any morphological or categorial features. What they have in common is just their prosodic deficiency.
Therefore, it seems impossible to identify a uniform feature that drives their movement to the second position, as it is
unlikely that a number of categorially unrelated elements could all check the same feature. Moreover, Bošković
(2001:63) points out that since the examples in (37) and (38) show that the clitics do not target a uniform position, it is
difficult to determine in a principled way what feature could be checked via their movement, given that they seem to
raise to different projections in different sentences.
Second, with the exception of auxiliary clitics, there is no strong empirical evidence to support the idea that clitics
ever move to the first position in syntax. All clitics in Serbo-Croatian have non-clitic, strong counterparts. For instance,
the past tense auxiliaries have non-clitic, pluperfect auxiliaries, which are prosodically independent and may occur
clause-initially. In fact, this is their preferred position: they may be preceded by the participle only when a ‘‘non-
neutral’’ (focused or topicalized) interpretation of a sentence is required.

Given that non-clitic auxiliaries move to the clause-initial position in (49b’), it might be empirically justified that
the clitic auxiliaries raise to the first position, but due to their phonological weakness, their second lower copy is
pronounced. However, in contrast to the auxiliaries, there is no evidence that pronominal elements or non-subject NPs
ever need to move to the clause-initial position. In fact, recall from (36), repeated below as (50), that clause initial
objects are interpreted as contrastively focused.

7
An anonymous reviewer points out that whereas the data in (46) and (47) indicate that the operation of clitic climbing is subject to syntactic
locality conditions, they do not show that the configuration of 2P in the main clause is subject to those conditions. It could be that in (47b) the
syntactic cross-clausal operation of clitic climbing is followed by a phonological movement. However, it seems to me that the previous examples
quoted in this section have demonstrated that a phonological movement is a very unlikely scenario. For instance, a phonological movement should
be insensitive to syntactic constituency, whereas examples (40) and (41) clearly show the opposite.
8
This approach is also adopted in a different version by Stjepanović (1999) and Bošković (2001).
344 K. Migdalski / Lingua 120 (2010) 329–353

On the assumption that pronominal clitics move to the first position, but are pronounced in the second position due
to a PF filter, they should still be interpreted at LF as occurring clause initially. Since objects in the first position are
interpreted as contrastively focused, all pronominal clitics in Serbo-Croatian are in this scenario expected to have
contrastive focus interpretation, contrary to fact.
Another problem with the scattered deletion approach concerns the interaction between clitic placement and
multiple wh-fronting. As pointed out by Penn (1999) and Lenertová (2001), in Serbo-Croatian all D-linked wh-phrases
move to the first position and appear adjacent to each other, whereas the clitics follow all wh-words.

Non-D-linked wh-words do not need to be extracted to the first position, but if they do, the clitics must follow the
first wh-element.

If the placement of clitics is determined by a PF filter, it is unclear how the filter is able to decide about the placement of
clitics with respect to the D-linked or non-D-linked status of the wh-words. D-linking is related to the semantics of the
wh-elements in question, while a PF mechanism should have no access to semantic properties of syntactic units.9
Furthermore, Lenertová (2001) observes that the scattered deletion approach faces empirical problems when
applied to Czech. Just as Serbo-Croatian, Czech has Wackernagel enclitics, but in certain very marked environments
(usually colloquial speech) they can procliticize on the element that follows them in the absence of a lexical element in
the first position, as illustrated in (53b). The direction of cliticization is marked with an arrow.

To account for the grammaticality of (53b), Franks (1998) claims that the first word is interpreted as ‘‘understood’’
from the context and therefore becomes deleted at PF. However, Lenertová (2001) points out that there are examples in
which there is no ‘‘understood’’ element that could have been deleted. For instance, in (54B) the clitics end up in the
highest position in syntax, but they are not hosted by any element preceding them. The sentence has a neutral
interpretation and can be used as an answer to the question ‘‘what happened?’’.

9
An anonymous reviewer points out that since D-linked wh-phrases are often analyzed as topics (Hornstein, 2001; Richards, 2001), and there may
exist multiple Topic projections (Rizzi, 1997), it is likely that there are substantial differences between the syntax of D-linked and non-D-linked
multiple wh-phrases. If this is so, it might well be the case that these differences surface at the PF level as well.
 K. Migdalski / Lingua 120 (2010) 329–353 345

The scattered deletion approach predicts that in standard Czech it is necessary to pronounce the lower copy of the
clitics, as in (55), in which the clitics are preceded by the temporal adverb. It also predicts that (54B) and (55) have the
same interpretation, because the outputs received by LF are exactly the same. However, this is contrary to fact, because
the adverb in (55) is a topic.

On a par with (54B), the sentence in (56) is topic-less and has a neutral interpretation. Here the clitics are hosted by
the participle, which has moved from a lower position across the adverb včera (compare a similar structure in (34b)
from Serbo-Croatian). Moreover, the clitics precede the adverb as in (55), but the sentence has a different reading,
which indicates that they are in a different syntactic position. As far as I can tell, the scattered deletion approach is
unable to capture this contrast.

In sum, these data suggest that the scattered deletion approach fails to predict the interpretations of clauses with
elements preceding the clitics correctly. In a colloquial register of Czech, the clitics may occur in the first position,
which does not result in a PF crash, because they procliticize on the second element. Thus, whenever they move to the
first position, they stay there. The sentences in which the clitics are in the second position do not arise due to a PF filter
which forces their lower copy to be pronounced. They are created via a syntactic movement of different elements to the
position preceding the clitics. Depending on the element that is moved, the sentences acquire different interpretations.
To summarize, it seems we are left with a serious dilemma concerning the mechanism of second position
cliticization in Slavic. The mechanism has clear properties of a syntactically constrained operation, as it is sensitive to
the same locality restrictions as wh-movement or scrambling. However, it is difficult to analyze second position
cliticization on a par with other syntactic movements, because it is not triggered by any identifiable formal feature, nor
does it target a uniform syntactic projection. It also applies to a number of categorially unrelated elements, which do
not have any morphosyntactic features in common. The only property that unifies them is their prosodic deficiency. As
far the syntactic structure of clitics is concerned, the data in (48) clearly indicate that pronominal clitics do not cluster
syntactically in Serbo-Croatian, so it is unlikely that they are adjoined to a single head. Therefore, I will assume that
the dative clitic is hosted in Spec, AgrIOP, whereas the accusative clitics is hosted in Spec, AgrOP, and that they move
to these positions to check case (see Bošković (2001) and Stjepanović (1999) for more arguments showing that the
pronominal clitics do not cluster and are not adjoined to a single head in Serbo-Croatian). The auxiliary clitic targets
T0, but it may also move higher, to a position above the wh-word and sentential adverbs, as indicated by (37c) and
(38b), respectively.
In many ways second position cliticization resembles V2. V2 does not seem to have a uniform trigger, and in a
number of cases it is difficult to pinpoint the landing site targeted by the verb, because it can be preceded by so many
unrelated categories. Thus, just like second position cliticization, V2 does not seem to be a ‘‘standard’’ syntactic
operation.

4. Force and second position cliticization

One of the most widespread explanations of the V2 effect in Germanic is to relate the movement to Force-marking.
The subsequent sections will analyze the relation between Force marking and second position cliticization in Slavic in
order to verify whether Force-licensing can be a uniform trigger for V2 and 2P. Section 4.1 will show that the
cliticization in Illocutionary Force contexts has different syntactic properties than ‘‘generalized’’ second position
cliticization, even though in both instances clitics follow the first constituent. It will also demonstrate that movement
for Force checking is the only context in which Wackernagel’s law is violated in Serbo-Croatian. Section 4.2 will
examine data illustrating movement of clitics to the second position in the languages that lack Wackernagel clitics
otherwise. It will show that the exceptional 2P placement of clitics in these languages is related to Force marking.
346 K. Migdalski / Lingua 120 (2010) 329–353

4.1. Force and cliticization in second position clitic languages

We have seen that all clitics in Serbo-Croatian must appear in the second position. However, there is an exception to
this requirement in the context of negation. Negation obligatorily attracts verbs in Serbo-Croatian (cf. Rivero, 1991;
Leko, 1996; Progovac, 2005), and in fact in many other Slavic languages as well. As a result, no overt material may
intervene between the negative marker ne and the verb in (57).10

Negation attracts lexical verbs in Serbo-Croatian also in the clauses with pronominal clitics. In this context the verb
is left-adjacent to negation preceding the clitic, which end up in the third position.

It is generally assumed that negation is a Force marker, as it licenses negative assertion acting as an operator. If this
is correct, the Serbo-Croatian data indicate that second position cliticization is unrelated to Force, because the only
context in which it is violated instantiates a type of Force-licensing.11
Moreover, like most Slavic languages, Serbo-Croatian has the designated clitic li to mark the Illocutionary Force of
a clause. It is used to in yes-no questions and focus constructions, as in (59), where the object knjige ‘books’ and the
wh-word receives additional emphasis.

10
Leko presents data from Bosnian, but the Serbo-Croatian facts are exactly the same.
11
An anonymous reviewer remarks that it is not necessary to analyze ne as an independent word; rather, it might be that the sequence ‘‘ne+verb’’ is
a single complex head. The same pattern is observed in many other Slavic languages, such as Czech and Polish, where negation forms a prosodic
word with the verb it precedes, but not with other constituents, such as nouns (see Ozga, 1976:137, and Błaszczak, 2001:120). This is a plausible
option, and it would indicate that the verb adjoins to the negation in syntax, while the second position requirement is taken care of only at PF, where
the ‘‘ne+verb’’ sequence is treated as a single word. In Bulgarian, however, negation adjoins not only to the verb, but if pronominal clitics are
present, it attracts the verb together with the clitics (cf. (ia)). In the presence of li, it attracts the highest pronominal clitic, changing the usual ordering
of clitics (cf. (ib) and (ic)).

Moreover, in Bulgarian and Serbo-Croatian negation may attract only those verbs that undergo head movement (e.g. the finite verb in (iic)), but
not the participle, which I argue in Migdalski (2006) undergoes remnant XP movement (cf. (iia) and (iib)).

This all indicates that negation is selective about both the categorial and syntactic status of the elements it attracts, therefore the postulate that the
complexes of ne + other constituents come from the lexicon does not seem to be on the right track.
 K. Migdalski / Lingua 120 (2010) 329–353 347

Along with the other clitics in Serbo-Croatian, li is always placed in the second position. However, unlike all the other
clitics, it must appear after the first word (cf. (60)), even if the other clitics may appear after the first phrase (cf. (61)).

The placement of li might give an impression as being prosody-driven. Bošković (2001:28–29) observes, however,
that li is not permitted after the words that are not syntactically mobile, such as prepositions (cf. 62a) or the first
conjunct in coordinated structures (cf. 62b).

Interestingly, native speakers inform me that li cannot be placed after the first syntactic constituents, either, as
illustrated in (63). In fact, it seems impossible to insert li anywhere in these sentences.

In all the examples with li the word that precedes it is the only one that is focused. Therefore (61a) has the
interpretation ‘‘Does Ana read expensive books?’’ rather than ‘‘Does Ana read expensive books?’’, with the focus
restricted to the adjective. Bošković (2001:31ff) takes this to mean the focal feature of li is checked via head
movement, and that li in Serbo-Croatian is defective in the sense of not being able to support a specifier. It is impossible
to check this feature via a Spec, head configuration, which is the reason why adjunction of XP elements to li gives
ungrammatical outputs in (63).
Some deviations from the second position cliticization in relation to Force marking are also found in Old Church
Slavonic. Old Church Slavonic is regarded as a Wackernagel clitic language in traditional grammar descriptions
(cf. Lunt, 2001:77), although more detailed studies question this assumption (cf. footnote (5)). Regardless of this
assumption, embedded clauses in Old Church Slavonic show an interesting split in the cliticization patterns depending
on the complementizer type. Willis (2000) observes that the conditional auxiliary clitic is always right adjacent to the
348 K. Migdalski / Lingua 120 (2010) 329–353

complementizer a (cf. 64), irrespectively of the ordering of the elements following the clitic (cf. 64b). Conversely, the
complementizer da does not have to be adjacent to the conditional clitic. This is demonstrated in the two different
translations of Luke 4,42’s gospel. In the first one in (65a) bi is separated from da by negation. In the second one in
(65b) the two constituents are adjacent, and negation occurs to the right of the conditional auxiliary bi.

The data indicate that certain types of complementizers can attract clitics depending on their semantics. The
movement is obligatory in the case of a, which introduces conditional sentences. However, the complementizer da, which
usually introduces indicative (declarative) clauses (cf. Bräuer, 1957), does not attract clitics, so they do not have to be
adjacent to it.
Summarizing, this section has described the properties of the Force-related clitics in the languages with generalized
second position clitics. It has shown that they may have a different distribution than the other clitics with respect to the
elements that can precede it: in Serbo-Croatian, li may only follow heads, even though all other clitics may follow
heads and phrases alike. Moreover, the section has demonstrated that certain clitics may target the second position
depending on the Force value of the clitic host, while in some other Force environments, the second position
requirement may be revoked. This indicates that Force-related 2P in Slavic cannot be equated with generalized 2P,
because it clearly involves a different syntactic operation.

4.2. Second position effects in languages without second position clitics

This section will analyze Force-related cliticization in those Slavic languages that have no other Wackernagel
clitics, such as Bulgarian, Macedonian, and Polish. To make the comparison with Serbo-Croatian more transparent, it
will focus on li. Even though li is found in many Slavic languages, its exact Force value varies crosslinguistically. For
instance, in both Bulgarian and Macedonian li licenses focus on the fronted element preceding it.

In addition to this, when the word in front of li is a verb in Bulgarian, the sentence is interpreted as a yes/no question
in Bulgarian (cf. Rudin (1986:64)).

A crucial property of li concerns its syntactic position. Even though Bulgarian and Macedonian have no second
position pronominal or auxiliary clitics (they all both verb adjacent, as in Romance), li is the Wackernagel clitic.
 K. Migdalski / Lingua 120 (2010) 329–353 349

It normally follows the clause initial word, and can be even separated from the other clitics occurring lower in the
sentence.

In Macedonian li in addition displays different properties concerning the direction of its cliticization from other
clitics. The examples in (69) show that the pronominal and auxiliary clitics procliticize on the verb, so they stay left-
adjacent to it. Li is an enclitic, and needs to be supported by some overt material in front of it, such as the main verb in
(70b).

Moving on to Bulgarian, even though this language does not have second position clitics, it requires the auxiliary
clitic to come second in wh-questions.

However, the requirement is not restricted only to clitics, because non-clitic pluperfect auxiliaries as well as finite
verbs must appear in the second position in wh-questions as well. This restriction is related to the obligatory rule of
subject-verb inversion in Bulgarian (cf. Rudin (1986), Izvorski (1995), Rivero (1994)).

Observe that English and many Romance languages display the same rule of the subject-verb inversion, albeit
restricted to auxiliary verbs. Rizzi (1996:63) suggests that this is ‘‘a special case of Verb Second’’ and employs the
term ‘‘residual V2’’ to describe these data. If V2 and second position cliticization are indeed the same phenomenon,
Rizzi’s generalization is incorrect, given that the inversion applies to clitics in the language which does not have any
Wackernagel clitics otherwise.
Finally, let us turn to Polish. In general, Modern Polish does not have any pronominal clitics, and by default, the
auxiliaries are affixed to the participle.12

12
This is a simplification, because the perfect and conditional auxiliaries in Polish are undergoing a diachronic change from a clitic to an affix on
the main verb. There are still environments in which the affixed auxiliary shows a clitic-like behaviour, and, for example, can be deleted in
coordinated constructions under identity. See Franks and Bański (1999) for details.
350 K. Migdalski / Lingua 120 (2010) 329–353

However, in Force-related contexts the auxiliaries behave like second position clitics in Polish (cf. Migdalski, 2006
ch. 5). For instance, Mikoś and Moravcsik (1986) and Borsley and Rivero (1994) observe that the conditional auxiliary
is obligatorily attracted by some clause-initial conjunctions and complementizers, such as gdy+by ‘if’, jak+by ‘as if’,
o+by ‘I wish’/‘may’, and że+by ‘so that’, and as a result, the auxiliaries appear in the second position in embedded
clauses. The complementizers express various types of modal meanings, such as condition and potentiality in (75) or
desire in (76). If the auxiliary is affixed on the participle instead of being adjoined to the complementizer, the result is
ungrammatical (cf. 75b and 76b).

The same requirement holds for the position of the auxiliary in subjunctive clauses introduced by volition verbs
such as prosić ‘to ask’. It must occur right-adjacent to the complementizer że and may not be affixed to the
participle.

If the verb in the matrix clause does not require a complement in the subjunctive mood, the auxiliary does not need
to be adjoined to the complementizer and can be affixed on the l-participle. However, only the indicative meaning is
then possible.

All the examples which require encliticization of by onto the complementizer in the embedded clauses express
some kind of Force-related meaning: condition or potentiality in (75), optative mood in (76), and command in (77).
Therefore, it can be assumed that the auxiliary is attracted by a Force-related feature located in a functional head in
the left periphery of the clause, possibly the highest one, as the auxiliary always ends up in the second position,
adjacent to the complementizer. Observe that this pattern is by no means specific to Polish. I mentioned in section
4.1 that in Old Church Slavonic the complementizer introducing conditional clauses obligatorily attracted the
auxiliary clitic (cf. (64) and (65)). Likewise, the V2 requirement in some Scandinavian dialects is also related to
the semantics of the complementizer (see section 2.1. and Wechsler (1991)). Since Polish does not have any
Wackernagel clitics in other contexts, this indicates again that V2 or second position cliticization should not be
attributed to Force.
 K. Migdalski / Lingua 120 (2010) 329–353 351

To summarize, the variation in the position of the clitics clearly indicates that apart from the ‘‘generalized’’ second
position rule, some of the clitics may end up in the second position because they are attracted by a Force feature
specifying the clause type. These two processes are completely independent of each other. This fact holds for
Wackernagel clitic languages as well. Serbo-Croatian data analyzed in section 3.2 show that although the clitic moves
to the second position in both instances, the operations display different restrictions concerning the size of the
constituent moved to the position in front of the clitic: operator clitics may be preceded only by heads, whereas other
types of clitics can be preceded by heads and phrases alike.
Let me suggest that the Force-related movement is triggered by a feature that is located in the functional head
S (following Laka (1994)), and that clitics always move to S whenever the sentence deviates from declarative in
order to mark the clause type. Thus, in (64) and (75) the clitic by has to raise to S because the sentence is marked as
conditional. Importantly, recall that V2 has a very similar distribution in some Scandinavian dialects mentioned in
section 2.1, and is related to the degree of assertion or presupposition. The fact that this operation may occur in
violation of the usual second position requirements in Old Church Slavonic or results in the exceptional clitic
placement in the second position in Polish suggests that movement for Force checking is unrelated to generalized
second position cliticization, even though both operations may give rise to the same linear position of the clitics in
the structure.

5. Conclusion

In conclusion, let us juxtapose the properties of Wackernagel clitics and V2. The major contrast concerns
phonological deficiency, which is not observed in V2 in Germanic, with a possible exception of some Northern
Norwegian dialects. Another difference is related to the categories occupying the second position. In Germanic,
the requirement is limited to finite verbs. In the Slavic languages, it holds not only for auxiliary forms, but also for
pronominal and modal clitics, which move to the Wackernagel position even when they are not accompanied by
any other finite verb. Since second position clitics comprise different grammatical categories, they do not form a
natural class with respect to any morphological features. For this reason, it is difficult to identify a single syntactic
feature that possibly drives their movement. All other differences between second position cliticization and V2
can be reduced, it seems, to independent variation in the clause structure in Germanic and Slavic. For instance,
clitics appear to be able to split syntactic constituents in Slavic. This property, however, is directly related to the
way constituency is defined in Slavic, and the possibility of Left Branch Extraction, which is not available in
Germanic.
Apart from the ‘‘generalized’’ second position, I have also identified clitic movement that is driven by Force. It
locates clitics in the linear second position in main or embedded clauses as well. It has been pointed out that in spite of
the ordering similarities, these two operations are unrelated. Force-related movement occurs crosslinguistically in
languages with and without the 2P requirement on clitics. It may also conspire against the default 2P clitic placement,
as in the case of negation in Serbo-Croatian.
The empirical facts analyzed in the paper suggest that Wackernagel’s insight is correct and both V2 and 2P are
instances of the same linguistic process in different guises. They had similar stages in their diachronic development,
and show related patterns with respect to the syntactic environments in which they occur: in some languages V2 and 2P
are limited to Force-related environments, whereas in other languages they are required to hold in all contexts. Given
this variation, which applies to both 2P and V2, it seems incorrect to postulate a uniform trigger for V2 in Germanic or
2P in Slavic. It is evident that the operations commonly described as V2 or 2P cover a number of unrelated cases of
movement or base generation, whose only common property is the position of the verb or the clitics after the initial
constituent.

Acknowledgements

I would like to thank Željko Bošković, Hans Broekhuis, Anne Breitbarth, Tonya Dewey, Jutta Hartmann, Nataša
Milićević, Henk van Riemsdijk, Vera Hegedüs, Jan-Wouter Zwart, and the audiences at the Models of Grammar Staff
Seminar at Tilburg University and the V1/V2 workshop at Leiden University for comments. I am also grateful to the
two anonymous reviewers for their very insightful remarks. All mistakes are my own. This research was funded by the
Netherlands Organization for Scientific Research (NWO) under the Rubicon grant 446-05-008.
352 K. Migdalski / Lingua 120 (2010) 329–353

References

Aguado, M., Dogil, G., 1989. Clitics in lexical phonology: alleged counterevidence? Linguistische Berichte 120, 99–116.
Andersson, L.-G., 1975. Form and Function of Subordinate Clauses. Monographs in Linguistic 1, Department of Linguistics, University of
Gothenburg.
Anderson, S., 1993. Wackernagel’s revenge: clitics, morphology, and the syntax of verb-second position. Language 69, 68–98.
Bański, P., 2000. Morphological and Prosodic Analysis of Auxiliary Clitics in Polish and English. Ph.D. dissertation, Warsaw University.
Bentzen, K., Hrafnbjargarson, G.H., Hróarsdóttir, Th., Wiklund, A.-L., 2007. The Force Behind V2. Paper presented at The 22nd Comparative
Germanic Syntax Workshop, Stuttgart, Germany.
Błaszczak, J., 2001. Investigation into the Interaction Between the Indefinites and Negation. Akademie Verlag, Berlin.
Boeckx, C., 1998. A Minimalist View on the Passive. UConn Working Papers Occasional Papers in Linguistics 2. Department of Linguistics,
University of Connecticut, Storrs.
Borsley, R., Rivero, M.-L., 1994. Clitic auxiliaries and incorporation in Polish. Natural Language and Linguistic Theory 12, 373–422.
Bošković, Ž., 1995. Participle movement and second position cliticization in Serbo-Croatian. Lingua 96, 245–266.
Bošković, Ž., 1997. The Syntax of Nonfinite Complementation: An Economy Approach. The MIT Press, Cambridge, MA.
Bošković, Ž., 2001. On the Nature of the Syntax-Phonology Interface. Cliticization and Related Phenomena. Elsevier, Amsterdam.
Bošković, Ž., 2005. Left branch extraction, structure of NP, and scrambling. Studia Linguistica 59, 1–45.
Brandner, E., 2004. Head-movement in Minimalism, and V2 as FORCE-marking. In: Lohnstein, H., Trissler, S. (Eds.), The Syntax and Semantics of
the Left Periphery. Mouton de Gruyter, Berlin, pp. 97–138.
Bräuer, H., 1957. Untersuchungen zum Konjunktiv im Altkirchenslavischen un im Altrussischen. Slavistische Veröffentlichungen II. Osteuropa-
Institut an der Freien Universität Berlin, Wiesbaden.
Caink, A., 2000. Full form auxiliaries in Serbian/Croatian/Bosnian. In: King, T.H., Sekerina, I. (Eds.), Formal Approaches to Slavic Linguistics 8:
The Philadelphia Meeting. Michigan Slavic Publications, Ann Arbor, pp. 61–77.
Ćavar, D., Wilder, C., 1999. ‘‘Clitic Third’’ in Croatian. In: Van Riemsdijk, H. (Ed.),Clitics in the Languages of Europe. Mouton de Gruyter, Berlin,
pp. 429–467.
Cheng, L., 1997. On the Typology of Wh-Questions. Garland, New York.
Chomsky, N., 1993. A minimalist program for linguistic theory. In: Hale, K., Keyser, S. (Eds.), The View from Building 20: Essays in Linguistics in
Honour of Sylvian Bromberger. The MIT Press, Cambridge, MA, pp. 1–52.
Corver, N., 1992. On deriving certain left branch extraction asymmetries: a case study in parametric syntax. In: Proceedings of the Northeast
Linguistic Society 22, University of Delaware, pp. 67–84.
Den Besten, H., 1977/1983. On the interaction of root transformations and lexical deletive rules. In: Abraham, W. (Ed.), On the Formal Syntax of the
Westgermania. John Benjamins, Amsterdam, pp. 47–131.
Dewey, T.K., 2006. The Origins and Development of Germanic V2. Ph.D. dissertation, University of California, Berkeley.
Dogil, G., 1987. Lexical phonology and floating inflection in Polish. In: Dressler, W.U. (Ed.), Phonologica 1984. Cambridge University Press,
Cambridge, pp. 39–47.
Embick, D., Izvorski, R., 1997. Participle-auxiliary orders in Slavic. In: Browne, W., Dornisch, E., Kondrashova, N., Zec, D. (Eds.), Formal
Approaches to Slavic Linguistics: The Cornell Meeting. Michigan Slavic Publications, Ann Arbor, pp. 210–239.
Eythórsson, Th., 1995. Verbal Syntax in the Early Germanic Languages. Ph.D. dissertation, Cornell University.
Fanselow, G., 2002. Quirky subjects and other specifiers. In: Kaufmann, I., Stiebels, B. (Eds.), More Than Words. Akademie Verlag, Berlin, pp.
227–250.
Fanselow, G., 2003. Surprising specifiers and cyclic spell-out. In: Bański, P., Przepiórkowski, A. (Eds.), Generative Linguistics in Poland:
Morphosyntactic Investigations. Instytut Podstaw Informatyki PAN, Warszawa, pp. 29–46.
Franks, S., 1998. Clitics in Slavic. Paper Presented at Comparative Slavic Morphosyntax Workshop, Bloomington.
Franks, S., 2000. Clitics at the interface: an introduction to clitic phenomena in European languages. In: Beukema, F., Den Dikken, M. (Eds.), Clitic
Phenomena in European Languages. John Benjamins, Amsterdam, pp. 1–46.
Franks, S., Bański, P., 1999. Approaches to ‘Schizophrenic’ Polish person agreement. In: Dziwirek, K., Coats, H., Vakareliyska, C.M. (Eds.),
Formal Approaches to Slavic Linguistics: The Seattle Meeting. Michigan Slavic Publications, Ann Arbor, pp. 123–143.
Franks, S., King, T.H., 2000. A Handbook of Slavic Clitics. Oxford University Press, New York.
Frey, W., 2006. Contrast and movement to the German prefield. In: Molnár, V., Winkler, S. (Eds.), The Architecture of Focus. Mouton de Gruyter,
Berlin, pp. 235–264.
Fuss, E., 2003. On the historical core of V2 in Germanic. Nordic Journal of Linguistics 26.2, 195–231.
Giusti, G., 1995. A unified structural representation of (abstract) case and article. Evidence from Germanic. In: Haider, H., Olsen, S., Vikner, S.
(Eds.), Studies in Comparative Germanic Syntax. Kluwer, Dordrecht, pp. 77–93.
Haider, H., 1993. Deutsche Syntax Generativ. Narr, Tübingen.
Halpern, A.L., 1992. Topics in the Placement and Morphology of Clitics. Ph.D. dissertation, Stanford University.
Hornstein, N., 2001. Move! A Minimalist Theory of Construal Blackwell, Oxford.
Izvorski, R., 1995. Wh-movement and focus movement in Bulgarian. In: Eckardt, R., Van Geenhoven, V. (Eds.), ConSole II Proceedings, Holland
Academic Graphics, The Hague, pp. 54–67.
Izvorski, R., King, T.H., Rudin, C., 1997. Against Li lowering in Bulgarian. Lingua 102, 187–194.
Kiparsky, P., 1995. Indoeuropean origins of Germanic syntax. In: Battye, A., Roberts, I. (Eds.), Clause Structure and Language Change. Oxford
University Press, Oxford, pp. 140–169.
Koopman, H., 1984. The Syntax of Verbs. Foris, Dordrecht.
 K. Migdalski / Lingua 120 (2010) 329–353 353

Koster, J., 2003. All languages are tense-second. In: Koster, J., Van Riemsdijk, H. (Eds.), Germania et alia: a linguistic Webschrift for Hans den
Besten, available online at: http://odur.let.rug.nl/koster/DenBesten/contents.htm.
Laka, I., 1994. On the Syntax of Negation. Garland, New York.
Leko, N., 1996. Notes on Negation and Related Topics in Bosnian. GenGenP (Geneva Generative Papers) 4, pp. 15–34.
Lenertová, D., 2001. On clitic placement, topicalization and CP-structure in Czech. In: Zybatow, G., Junghanns, U., Mehlhorn, G., Szucsich, L.
(Eds.), Current Issues in Formal Slavic Linguistics. Vervuert Verlag, Franfurt/Main, pp. 294–305.
Lenerz, J., 1977. Zur Abfolge nominaler Satzglieder im Deutschen. Narr, Tübingen.
Lunt, H.G., 2001. Old Church Slavonic Grammar. Mouton, The Hague.
Meinunger, A., 2006. Interface restrictions on verb second. The Linguistic Review 23, 127–160.
Migdalski, K., 2006. The Syntax of Compound Tenses in Slavic. Doctoral dissertation, Tilburg University, LOT Publications, Utrecht. Down-
loadable at: http://www.lotpublications.nl/index3.html.
Migdalski, K., 2007. On the Emergence of Second Position Cliticization in Slavic. Paper presented at the Formal Description of Slavic Languages 7,
University of Leipzig, December 2007, To appear in the conference proceedings.
Mikoś, M.J., Moravcsik, E.A., 1986. Moving clitics in Polish and some crosslinguistic generalizations. Studia Slavica 32, 327–336.
Mohr, Sabine, 2005. Impersonal Constructions in the Germanic Languages. John Benjamins, Amsterdam.
Müller, G., 2004. Verb-second as vP-first. Journal of Comparative Germanic Linguistics 7, 179–234.
Ozga, J., 1976. Clitics in English and Polish. Papers and Studies in Contrastive Linguistics 4, 127–140.
Penn, G., 1999. Linearization and Wh-Extraction in HPSG: evidence from Serbo-Croatian. In: Borsley, R., Przepiórkowski, A. (Eds.), Slavic in
HPSG. CSLI Publications, Stanford, CA, pp. 149–182.
Pittner, K., 2003. Kasuskonflikte bei freien Relativsätzen. Eine Korpusstudie. Deutsche Sprache 31, 193–208.
Platzack, C., 1986. COMP, INFL and Germanic word order. In: Hellan, L., Koch Christensen, K. (Eds.), Topics in Scandinavian Syntax. Reidel,
Dordrecht, pp. 185–234.
Progovac, L.J., 1993. Locality and subjunctive-like complements in Serbo-Croatian. Journal of Slavic Linguistics 1, 116–144.
Progovac, L.J., 1996. Clitics in Serbian/Croatian: comp as the second position. In: Halpern, A., Zwicky, A. (Eds.), Approaching Second: Second
Position Clitics and Related Phenomena. CSLI Publications, Stanford, pp. 411–428.
Progovac, L.J., 2005. Syntax of Serbian: Clausal Architecture. Slavica, Bloomington.
Radanović-Kocić, V., 1988. The Grammar of Serbo-Croatian Clitics: A Synchronic and Diachronic Perspective. Ph.D. dissertation, University of
Illinois, Urbana.
Rice, C., Svenonius, P., 1998. Prosodic V2 in Northern Norwegian. Paper presented at the West Coast Conference on Formal Linguistics 17,
University of British Columbia, Vancouver.
Richards, N., 2001. Movement in Language: Interactions and Architectures. Oxford University Press, Oxford.
Rivero, M.-L., 1991. Long head movement and negation: Serbo-Croatian vs. Slovak and Czech. The Linguistic Review 8, 319–351.
Rivero, M.-L., 1994. Clause structure and V-movement in the languages of the Balkans. Natural Language and Linguistic Theory 12, 63–120.
Rizzi, L., 1996. Residual verb second and the Wh-Criterion. In: Belletti, A., Rizzi, L. (Eds.), Parameteres and Functional Heads. Essays in
Comparative Syntax. Oxford University Press, Oxford/New York, pp. 63–90.
Rizzi, L., 1997. The fine structure of the left periphery. In: Haegeman, L. (Ed.), Elements of Grammar: Handbook of Generative Syntax. Kluwer,
Dordrecht, pp. 281–337.
Rudin, C., 1986. Aspects of Bulgarian Syntax: Complementizers and Wh-Constructions. Slavica, Columbus.
Rudin, C., Kramer, C., Billings, L., Baerman, M., 1999. Macedonian and Bulgarian li questions: beyond syntax. Natural Language and Linguistic
Theory 17, 541–586.
Searle, J., 1969. Speech Acts. Cambridge University Press, Cambridge.
Stjepanović, S., 1998. On the placement of Serbo-Croatian clitics: evidence from VP ellipsis. Linguistic Inquiry 29, 527–537.
Stjepanović, S., 1999. What do Second Position Cliticization, Scrambling, and Multiple wh-fronting have in Common? Ph.D. dissertation,
Department of Linguistics, University of Connecticut at Storrs.
Taraldsen, K.T., 1986. On verb second and the functional content of syntactic categories. In: Haider, H., Prinzhorn, M. (Eds.), Verb Second
Phenomena in Germanic Languages. Foris, Dordrecht, pp. 7–25.
Thiersch, C., 1978. Topics in Germanic Syntax. Ph.D. dissertation, Department of Linguistics, MIT.
Toman, J., 1996. A note on clitics and prosody. In: Halpern, A., Zwicky, A. (Eds.), Approaching Second: Second Position Clitics and Related
Phenomena. CSLI Publications, Stanford, pp. 505–510.
Tomić, O., 1996. The Balkan Slavic clausal clitics. Natural Language and Linguistic Theory 14, 811–872.
Travis, L., 1984. Parameters and Effects of Word Order Variation. Ph.D. dissertation, MIT.
Vaillant, A., 1977. Grammaire comparée des langues slaves. Tome V: La syntaxe. Éditions Klincksieck, Paris.
Vikner, S., 1995. Verb Movement and Expletive Subjects in the Germanic Languages. Oxford University Press, Oxford.
Wackernagel, J., 1892. Über ein Gesetz der indogermanischen Wortstellung. Indogermanische Forschungen 1, 333–436.
Wechsler, S., 1991. Verb second and illocutionary force. In: Leffel, K., Bouchard, D. (Eds.), Views on Phrase Structure. Kluwer, Dordrecht, pp.
177–191.
Westergaard, M., Vangsnes, Ø., 2005. Wh-questions, V2, and the left periphery of three Norwegian dialects. Journal of Comparative Germanic
Linguistics 8, 117–158.
Wilder, C., Ćavar, D., 1994. Long head movement? Verb movement and cliticization in Croatian. Lingua 93, 1–58.
Willis, D., 2000. Verb movement in Slavonic conditionals. In: Warner, A. (Ed.), Diachronic Syntax: Models and Mechanisms. Oxford University
Press, Oxford, pp. 322–348.