2.1 The empirical base

The empirical base of the present investigation are the verum strategies in Spanish, English and German illustrated in (1). The three languages also possess other strategies that can result in an interpretation similar to verum focus for instance relying on adverbs such as Englisch really, German wirklich or Spanish de verdad. For Spanish, furthermore a left peripheral use of bien (que) ‘well (that)’ (Hernanz 2007) and a syntactic fronting strategy by which not the fronted expression but the whole proposition is contrastively focused (Escandell-Vidal & Leonetti 2009a), have been described as giving rise to emphatic polarity or verum. Although a comparison of the precise meaning would certainly be interesting, these alternative strategies are not part of the empirical base of the present article. But see Kocher (2023) for a comparison of different strategies in Spanish. The reason I focus on the ones presented in (1) is that in these strategies verum focus can be shown to result from focus on sentence mood. In the other cases, the verum interpretation might arise for different reasons.

In German declaratives, verum focus is marked through prosodic stress on the finite verb (Höhle 1992; Lohnstein 2016) (cf. (9-a) and (9-b)).

Superficially, the German strategy is ambiguous: In addition to verum focus, it can also be interpreted as focus on the lexical verb, compare (9-a) to (10-a), or as focus on TAM, compare (10-b) to (9-b).

In English declaratives, verum focus is expressed by stressing an auxiliary verb. In analytic tenses like in (11-b), the corresponding auxiliary is stressed. In synthetic tenses, the auxiliary do is inserted and stressed (11).

In English, only TAM focus is syncretic with verum focus. Focus on the lexical verb is expressed through stress on the verb itself (cf. the translations of (10)).

In Spanish, irrespective of whether the tense is synthetic (12-a) or analytic (12-b), the particle sí can be used to express verum focus (Batllori & Hernanz 2008; Escandell-Vidal & Leonetti 2009a, 2009b; Escandell-Vidal 2011; Kocher 2017; 2019; 2022; 2023; Villa-García & Rodríguez 2020a; b). The particle can be followed by the complementizer que. The pragmatic function of que is described in detail in Kocher (2022). Again, in this article I am only interested in the Spanish strategy that is associated with a verum meaning. A recent analysis of sí as a marker of emphatic affirmation can be found in Villa-García & Rodríguez (2020b).

In Spanish, neither focus on the lexical verb (13-a) nor TAM focus (13-b) is syncretic with verum focus. Both are realized as stress directly on the lexical or auxiliary verb.

For the present proposal, the behavior of verum focus in non-declarative clause types is of central interest. Verum focus in polar question is illustrated in (14). In English, just as in declaratives, there is stress on an auxiliary (14-a). Similarly, in German in (14-b), stress falls on the finite verb. The same happens in wh-questions: English and German, just as in a declarative, respectively stress an auxiliary (15-a) or (15-b) the finite verb.

The case is more complex in Spanish: While sí, but not sí que, is attested in polar questions to express verum (15-c), the same is not true for Spanish wh-questions. In these, verum is either left unmarked or a lexical strategy with the adverbial de verdad is used (14-c).

The fact that sí in polar questions cannot be combined with que can be explained straightforwardly from the pragmatic function of que Kocher (2022): It is used to attribute a commitment to the proposition to the hearer. The effect of que in polar questions is that the speaker is biased to expect an affirmative answer from the hearer. In verum focused polar questions, however, there is no speaker bias. On the contrary, the speaker does not express their expectation of one answer over the other. Rather, they are confronted with incompatible information and press the interlocutors to give the true answer and possibly to support their commitment to p with further evidence.

In Spanish wh-questions, verum is not expressed by sí. It will be argued that this is due to the fact that the position sí would be merged in, is occupied by the wh-operator (see section 4.3). There are, however, cases where sí follows the wh-expression in wh-questions. These are not verum focused wh-questions. They are rather cases where a contrast is established between alternative expressions. It usually appears in a context where the same wh-expression followed by the negative particle no is given.

The alternatives that positively satisfy the proposition are not given in the previous context but only follow once the question has been asked. This is different from the verum-context in (15), where alternative answers are already given and the speaker presses the interlocutors to give the true answer, in other words the speaker presses the interlocutors to commit to one of the given alternatives and potentially provide support for it.

Sí (que) can also be found in echoic questions like (17) (cf. also Villa-García & Rodríguez 2020a; b).

Finally, verum focus in imperatives is illustrated in (18). Spanish does not permit sí (que) (18-c) while German and English stress the imperative verb.

This brief presentation of the empirical base shows that German and English make use of roughly the same strategy irrespective of the clause type. The case is different in Spanish: The sí (que)-strategy is only grammatical in declaratives and polar questions and ungrammatical in wh-questions and imperatives. In section 4, I review these facts again and show how they follow from the analyses I develop in this article.

4.1 Parametic differences in verb movement

It has been observed in the literature that there are parametric differences in the position a verb reaches in German, English and Romance declaratives (cf. Emonds 1976; Belletti 1990; Williams 1994; den Besten 1983; Suñer 1994; Benedicto 1998; Ordóñez 1998; Toribio 2000; Roberts 2001, Goodall 2002; Zagona 2002; Truckenbrodt 2006; Schifano 2018). The examples in (33)–(34) illustrate this by comparing the word order of finite verbs and functional adverbs in the three languages. The diagnostic employed here goes back to Emonds (1976) and Pollock (1989) and is based on the assumption that the position of the adverbs in the periphery of the VP is fixed cross-linguistically (for a proposal of a universal hierarchy of modifiers see Cinque 1999). The contrast between (33-a) and (33-b) shows that in English, the finite verb follows the VP adverb and cannot precede it. This has been taken to suggest that the verb remains in V⁰.

In Spanish, the opposite pattern emerges. The only grammatical order is the one where the finite verb precedes the VP adverb in (34-b). The reverse order is ungrammatical (34-a). This indicates that the verb moves to T⁰ in Spanish.

German superficially shows the same pattern as Spanish: The order where the VP adverb precedes the verb in (35-a) is ungrammatical and the order where the VP adverb follows the verb in (35-b) is grammatical. The example in (35-c) illustrates that German is a V2 language. The word order restrictions show that in a declarative the finite verb necessarily occupies the second position of the sentence. It is assumed for German that the finite verbs moves even higher than TP to a position in the CP. In line with Lohnstein (2016), I identify this projection with Mood⁰.

The core argument of this section is that the projection hosting the mood feature is occupied by different material in the various sentence moods and languages. The variation we observe is the result of the fact that focus on mood, giving rise to verum, is expressed on different things. This section presents a formalization of the empirical data from Spanish, German and English relying on existing accounts on verbal movement and feature inheritance. The aim is not to develop a new theoretical motivation for verbal movement, rather the more modest goal is to systematize the empirical data and find a formal way to describe them. I am fully aware that further research is pending to motivate certain aspects of the analysis.

There are various options to model head movement and its triggering factors. A comprehensive overview of the main conceptual alternatives and their merits can be found in Dékány (2018). For the present analyses, I adopt the reprojection approach developed in Biberauer & Roberts (2010) and Roberts (2010). In this account complex heads are externally merged in their inflected forms. Certain formal features on these complex heads require to be checked in a structurally higher position. This provokes movement of the head out of its position and internal merge of it as the sister of the previous phrase. The moved head projects the label of the new syntactic phrase. Biberauer & Roberts (2010) propose that in Romance varieties that have rich tense inflection such as Spanish, the finite verb carries a tense feature that provokes reprojection of a TP. In English, the verb lacks rich tense inflection and its unvalued tense feature cannot project. Head movement from V-to-T does not take place. The tense feature on the verb is valued through agree, which in their accounts is defined as a mechanism that copies feature values from a probe, T in our case, to a c-commanded goal, V, with a matching unvalued feature. Biberauer & Roberts (2010) argue that there is agreement between T and V even in cases when T is filled by an auxiliary. This is based on the data in (36) which show that the verb in V is non-finite.

In German, a V2 language that also lacks rich tense inflection, the finite verb surfaces in a C-position. Just as in English, German’s poor tense inflection means that V does not carry a tense feature. Reprojection of TP and V-to-T movement do not take place. Instead, according to Biberauer & Roberts (2010) the German verb does not pass through T but moves directly to a C-position. This contrasts with English and Spanish declaratives, where the verb does not reach a C-position. It reprojects T in Spanish, and remains in V in English.⁸

A second central ingredient to analyze the parametric differences in verb positions is Ouali’s system of feature inheritance. He distinguishes between three logical options of how features can be inherited by a head: keep (the head retains the features), share (the head copies features to another head) and donate (the head passes features to another head without keeping a copy). Biberauer & Roberts 2010 propose C donates φ- and tense features to T in Romance and English and keeps them in German.

In order to formalize my ideas within this system, I have to add some additional assumptions. I am of course aware that some of these assumptions are in need of more empirical proof. As I stated at the beginning of this section, as a first step, the goal is to find a formal description relying on existing tools from the theoretical literature. More elaboration and proof should follow in the future.

With this caveat in place, I make the following assumptions: I assume that there is a third important feature involved, namely mood. I also assume that the C-head in Biberauer & Roberts (2010) and Ouali (2008) can be identified with Mood. I furthermore assume that it is possible to donate only a subset of the features and keep the rest. This possibility has also been explored in Biberauer & Roberts (2010: 291–293) for Welsh. Finally, I assume that feature inherence can differ between clause types. What I propose for German, English and Spanish is presented in Table 3, leaving out φ features for simplicity.

In the present approach, every clause contains a mood feature (in MoodP). This picks up a traditional idea maintained in the literature (cf. for instance Baker 1970; Rizzi 1990; Cheng 1991; Rivero & Terzi 1995, Han 1998). The alternative view, that clauses are not typed by an abstract feature but through a combination of grammatical properties is maintained for instance by Altmann (1987),Brandt et al. (1992); Lohnstein (2000); Truckenbrodt (2006); Sode & Truckenbrodt (2018). It is also a prominent position in construction grammar (cf. Michaelis & Lambrecht 1996; Kathol 2000; Ginzburg & Sag 2000).

There are differences across languages and across clause types in terms of whether the mood feature is kept in Mood or donated to T. In all German clause types, Mood keeps the mood and tense feature, triggering movement of the finite verb to this position. In English declaratives and imperatives, the sentence mood features along with the tense features are donated to T. In declaratives, T agrees with V and values its unvalued features. In imperatives, the finite verbs moves to T. In English interrogatives, in contrast, Mood keeps the mood and tense feature and the finite verb moves there. In Spanish, Mood keeps the sentence mood feature in all clause types. In imperatives the tense feature is also kept by Mood resulting in head movement of the finite verb to this position. In Spanish declaratives and interrogatives, however, the tense feature is donated to T where the finite verb moves and is valued with its tense feature. Altogether, these patterns support the argument presented in Biberauer & Roberts (2010) that head movement is triggered by tense rather than φ (or mood) features. Building on the typology in Table 3, in the following sections I show how verum focused declaratives (4.2), imperative and interrogatives (4.3) are derived.

The mood feature is different from the tense feature, whose behavior and inheritance, according to Biberauer & Roberts (2010) has to do the richness and poverty of the tense morphology of a given language. None of the three languages have a distinct morphological form for the different sentence moods declaratives, interrogatives and imperatives. Only imperatives are distinguished morphologically. The mood feature impacts the structural properties that are used to encode the sentence moods. Therefore, rightly, English declaratives and interrogatives have a different configuration since their structural properties are in fact very different.

4.2 Analysis of verum focused declaratives

For verum focus in declaratives I propose the following analyses. In German, Mood keeps the tense and the mood feature and attracts the verb to this position. Focus on mood is expressed through prosodic stress on the finite verb. A simplified scheme of the derivation is found in (37).

In synthetic tenses the finite verb moves from V to Mood (cf. Biberauer & Roberts 2010), in analytic tenses the auxiliary moves from T to Mood. The mood and tense feature are valued in Mood. As explained at length in the previous sections, I take verum to result from focus on the sentence mood feature. Thus, if the feature is focused it is realized as prosodic stress on whatever checks the mood feature, which is a lexical verb in (38-a) and an auxiliary in (38-b).⁹

In English, Mood donates tense and mood features to T. The finite lexical verb remains in the VP. Its unvalued features are valued via agree between T and V.

In analytic tenses, an auxiliary is merged in TP where focus on mood is expressed through stress on this auxiliary. In synthetic tenses, ‘do’ is merged in T triggered by the focus feature on mood. Biberauer & Roberts (2010: 272), despite not using the term verum, suggests that English overt do is triggered by an additional feature they call [+ Affect] that results in a special discourse effect. I propose what they call [+ Affect] is in fact a focused mood feature.

In Spanish declaratives, tense (and φ) features are donated to T, but mood features are kept in Mood. The Spanish finite verb moves from V to T in synthetic tenses and is merged directly in T in analytic tenses. Crucially it does not move beyond this position. The particle sí is merged directly in the specifier of MoodP.

There are reasons to believe that Spanish developed the particle-based strategy because of a restriction on the verb movement, offering further support for the underlying relation between the Spanish particle based and German and English verb-based verum strategies. The empirical point in favor is the fact that the first attestation of sí (que) coincides with the loss of V2 in Old Spanish (cf. Fontana 1993 on Old Spanish V2, and Rodríguez Molina 2014 and Kocher 2017 on the history of sí que). In other words, the need to merge sí in MoodP to check the focus feature only arose after verb movement to the left periphery was no longer available. A link between the emergence of the particle and the loss of V2 also extends to other Romance varieties: Ledgeway (2008) proposes that Old Neapolitan sí was merged in the CP to satisfy a V2 requirement. In the context of the present argumentation, the loss of V2 in Old Spanish might constitute the parametric shift between a system akin to the one we find in modern German where tense features triggering head movement are kept by Mood and the system of modern Spanish where these tense features are donated to T.¹⁰

To summarize, the analyses I propose for verum focused declaratives boil down to the following: German Mood keeps tense and mood features, which triggers head movement of the verb to this position. The focus feature, which gives rise to the verum interpretation, is expressed on the element that carries the mood feature, i.e. the finite verb. In English declaratives tense and mood features are donated to T. Focus is realized as prosodic stress and needs lexical material to be expressed on. This is why, in the absence of an auxiliary required by the tense, ‘do’ is inserted in T. In Spanish declaratives there is no movement of the finite verb to Mood either. While the tense feature is donated to T, the mood feature remains in Mood in this clause type. The focus feature on mood in MoodP triggers the merger of sí in the specifier of MoodP.

4.3 Analysis of verum focused interrogatives and imperatives

In the present section, I look at the strategies the three languages use to express verum focus in non-declarative clause types and show how I account for them. This section provides crucial evidence for the link between verum focus and sentence mood. The interesting empirical observation is that German and English use a similar strategy as in declaractives, while in Spanish the sí (que) strategy is not available in imperatives and wh-interrogatives.

Verum focus in polar questions, as illustrated in section (18), is a device to press the interlocutor to provide a true answer out of the two salient alternatives. Similarly, verum focus in wh-questions is used as a means to press the interlocutor for a true answer out of n salient alternatives. German employs the same strategy in wh- (43-a) and polar questions (43-b) as in declaratives: It stresses the finite verb.

In English verum focused interrogatives, just as in declaratives, there is stress on do in synthetic tenses and on the auxiliary in analytic tenses (cf. (44) repeated from (23) and (45-a) repeated from (25)).

Notably, do is also present in subject questions that do not require the auxiliary in a neutral, non-verum focused context (Who lives in Seville?) (cf. (45-b)). In my proposal, this is expected because, just as in declaratives, do is triggered by a focus feature on mood.

In Spanish polar questions, verum can be expressed with the particle sí but not with sí que (cf. (46-a)). As has been explained in section 2.1, this is because que is used in polar questions to express a speaker bias, which is not compatible with the effect of verum in polar questions. In the contexts that elicit a verum interpretation of a wh-question, speakers resort to lexical strategies, employing for instance the adverbial de verdad ‘really’ (46-b).¹¹

Imperatives express the speaker’s demand to make the propositional content of the imperative a fact. Verum focus in imperatives is licensed when the speaker insists on their demand that the addressee should do whatever the proposition expresses and rejects an alternative behavior, defined as any salient behavior that differs from the one requested by the speaker. In contexts like these, German once again stresses the finite verb (47).

In the parallel context in English, there is also stress on the finite imperative verb ((48) repeated from (27)). So in English verum focused imperatives, different from what we find in declaratives or interrogatives, do is not inserted.

There is another version of an English imperative featuring do. It is however not licensed in a context such as that illustrated in (48). It typically appears in contexts where the addressee is hesitant about which behavior is required of him or when the speaker wants to signal politeness. The contexts also differ from those that elicit verum focus in that the proposition does not have to be given.

In Spanish imperatives, just as in wh-questions, sí (que) is not an option (50).¹²

The empirical facts outlined above are compatible with the following analyses of verum focus in non-declarative clause types. In German, it appears that in all unembedded sentences irrespective of their clause type, the finite verb moves to Mood where the focus feature is expressed. In my account, this means that Mood keeps all relevant features in all clause types in German, so the structure in (38) also holds true for interrogatives and imperative.

In English interrogatives there is movement of the auxiliary from T to Mood. This means that the verb moves to a higher position than in a declarative. This high structural position of the auxiliary in English questions finds support in the fact that (neutral) polar questions are verb initial (51-a)¹³ and wh-questions (51-b) do not permit an intervening subject between the wh-pronoun and the verb. Neither would be expected if the auxiliary remained in T.

In my account, Mood in interrogatives is different from Mood in declaratives in English: It keeps its features. The focus feature is realized as prosodic stress on the auxiliary that occupies Mood (cf. (52)).

In imperatives, no auxiliary is merged. There is evidence that the lexical verb also moves higher than in declaratives in this clause type. This is illustrated by the word order contrasts in (53). In the declarative in (53-a), the adverb often precedes the verb and in the imperative in (53-b), it follows the verb.

I propose that the lexical verb moves to T in imperatives (see also Han 2000) and does not remain in V as in declaratives. The sentence mood feature and the focus feature, if present, are donated to T. This explains why do-insertion does not take place in imperatives.

Motivation for locating the finite verb in T rather than Mood stems from examples like (54-a) and (54-b) contrasting with (54-c). They show that if a subject is realized in English imperatives, it precedes the verb (cf. (55)).

In Spanish, the strategy employed in declaratives is only grammatical in polar questions, but not in wh-questions nor imperatives. I suggest that this is the case because Mood is not empty but occupied in Spanish imperatives and wh-interrogatives, which makes the merger of the particle impossible. For wh-interrogatives, I propose that Mood is occupied by the wh-operator (cf. also for instance Rizzi 2001 and Prieto & Rigau 2007 for an operator account of interrogatives).

The word order of Spanish interrogatives differs from declaratives: There is subject-verb inversion which is obligatory in wh-questions and optional in polar questions. This could be taken to suggest that the verb itself occupies Mood. In spite of this, my analysis does not assume movement of the verb to the left periphery in interrogatives (contra Torrego 1984 and Rizzi 1996). In line with Suñer (1994); Belletti (2004) and Cardinaletti (2007) I assume that the verb remains in T (cf. (57)). One of the arguments against T-to-Mood movement put forward in the literature is related to TP peripheral adverbs in interrogatives. Just as in declaratives, adverbs like todavía ‘still’ tend to precede the finite verb, suggesting that the verb remains in TP in polar (56-b) and wh-questions (56-a).¹⁴

Notably, in English, only the order in which the verb precedes the adverb is grammatical (cf. the translation of the examples in (56)). This fact again supports the analysis of English do in Mood.

The properties of Spanish (verum focused) interrogatives are captured by the analyses in (57). The finite verb moves from V to T and the sentence mood feature is checked by the wh-operator in Mood in the wh-question and expressed on sí in the polar question.

The properties of Spanish imperatives support an analysis that assumes T-to-Mood movement of the imperative verb. The verb precedes high adverbs like todavía. The reverse order is ungrammatical (58-a) vs. (58-b). As stated above, this contrasts with the orders we find in unmarked declaratives (58-c) and interrogatives (58-d) where todavía precedes the verb.¹⁵

Another strong argument in favor of the T-to-Mood movement of imperative verbs is the postverbal position of clitic pronouns, illustrated in (59-a) and (59-b).

In interrogatives (60-a) (vs. (60-b)), just as in declaratives, clitics are preverbal. This once again supports the idea that in these clause types finite verbs remain in T.

The analysis I assume is given in (61). The finite verb moves from V to T to Mood.

Table 4 summarizes positions the finite verb occupies in the three languages in the different clause types. Based on these distributional contrasts, the present section has provided further evidence in favor of the sentence mood theory of verum focus. I showed that it offers an adequate analysis for verum focus in non-declarative clause types in the three languages. The underlying principle is the same in all contexts: A focused mood feature gives rise to a verum interpretation. The superficial differences are explained by independent properties that condition verb movement.

4.4 Verum in embedded sentences

The position of MoodP within the functional field in (62) repeated from (30) is adopted from Lohnstein (2016) and is independently supported by the proposal made in Haegeman (2006).

Additional evidence for the position of MoodP within the left periphery will be reported in this section. It comes from the behavior of verum focus in embedded sentences. The relevant empirical point is that Spanish and English use the same strategy to express verum focus in embedded declaratives as in their unembedded counterparts, whereas German does not. One distributional property that has to do with the V2 nature of the language is that the finite verb in German in embedded clauses is not in the second but in the final position. What is interesting for the present discussion is that in embedded sentences, the prosodic stress marking of verum focus does not fall on the sentence final verb, but on the complementizer (63-a) or the relative pronoun (63-b). These facts support that verum results from stress on mood and that mood is hosted by a structural position in the left periphery.

I do not assume that relative pronouns, being phrases rather than heads, occupy Mood⁰. They are more adequately analyzed in the specifier of MoodP. The position of the finite verb, the complementizer and the relative expression are therefore not the same, but crucially, the positions are inside the same projection. Note that, the specifier of MoodP is also the position I propose for Spanish sí. The data in (64) illustrate how verum is realized in a relative with a PP as the relative expression in (64-a) and a dependent interrogative in (64-b). They were pointed out by an anonymous reviewer. They show that the verum focus stress is inside the whole relative or wh-phrase.

In English embedded clauses, the same strategy is employed as in unembedded declaratives: An (inserted) auxiliary is stressed (65).

Verum in Spanish embedded clauses is also expressed through the same means as in unembedded declaratives. Notably, in these cases, sí (que) is preceded by another instance of the complementizer (66).

I propose the same analyses for embedded declaratives as for their unembedded counterparts. This means that the locus of the sentence-mood features is again MoodP, and focus on these features gives rise to the verum interpretation. In English, tense and focused mood are donated from Mood to T where the auxiliary is merged. The complementizer is merged in a higher position at the upper edge of the left periphery in SubP (67).

In Spanish, the mood feature is kept by Mood, the focus feature triggers the merger of sí. The low complementizer is merged in Fin and moves to Mood (see Kocher 2022 for a detailed motivation and analysis). The high complementizer is merged in SubP (68).

In German, the verb remains in V but the complementizer is merged in Mood. The complementizer checks the sentence mood feature (cf. also Roussou 2000 and Bayer 2004 on arguments for complementizers as clause typers) and the focus feature on the sentence mood feature is expressed as prosodic stress on the complementizer (68). Although this is not a central concern and requires further investigation, for the purposes of this analysis, I assume that the finite verb remains in V and its tense feature is valued via agree.

As stated above, the data in (63) suggest that the complementizer in embedded declaratives in German occupies the same projection the finite verb occupies in unembedded sentences. In English and in Spanish, the same strategies are used as in unembedded declaratives. The embedding complementizer, therefore, occupies a higher position than MoodP. This suggests that the two languages have a richer left periphery than German. This conclusion is also supported by the data in (70).

In English (70-a) and Spanish (70-c) a topic below the complementizer is grammatical in embedded declaratives with and without verum focus. This supports my conclusion that, in English, the complementizer is not merged in Mood but in Sub. The data indicate that in English there is a least one TopP sandwiched between these two projections.

In Spanish, the space between the embedding complementizer and the verum particle permits even more intervening material. This is illustrated by the corpus data in (71). In (71-a) there is a clitic left dislocated topic that targets a TopP between embedding que and sí que. In (71-b), there is a fronted adverb, which is generally analyzed as a focus in FocP (Rizzi 2013) and a topicalized subject in TopP. In (71-c) an evidential modifier analyzed in ModP surfaces between the high complementizer and the verum particle.

German appears to be more restricted with respect to permitting topics below the complementizer in embedded declaratives with and without verum focus. The equivalent structures that are grammatical in Spanish and English, are ungrammatical in German (70-a). The issue, however, could also be be related to the subject pronoun: The version without verum focus in (70-b), i.e. the question posed by A, improves if er ‘he’ is made more prominent through focus stress in (72).

Both version with and without verum focus improve even more if the subject pronoun er is replaced by a prosodically heavier phrase (73).

Further research is necessary to shed light into this peculiar restriction and to determine how the left periphery of German is structured precisely. For the present discussion, it is important to highlight that these facts do not go against my claim that the German embedding complementizer is merged in MoodP. The fact that a topic can surface below the complementizer in contexts as (73), does not contradict the assumption of a relatively low merge position of the complementizer in German compared to Spanish and English. For one thing, in the cartographic structure I posit that is given in (62), there is a TopP below MoodP. The existence of such a position in German can furthermore be corroborated by the findings in Frey (2004). Based on data like (74-b), the author independently shows that German has a position to host topics that is lower than the projection that hosts the finite verb and the embedding complementizer.

In German, topics are often moved above the finite verb to a left peripheral called the prefield in the German linguistic tradition (74-a).

The interesting case in the present context is given in (74-b). Here, the sentence topic, namely the object of the verb, Paul, surfaces below and not above the finite verb. It still occupies a relatively high position, which is evident from the fact that it precedes a sentential adverb. According to Frey (2004) the neutral base position of German sentential adverbs is higher than any argument or adverbial position. Frey (2004) does not map the position of the intermediate topic in (74-b) in a cartographic structure like the one in (62). Still, it is worth noting that the structure does have a low TopP that it could correspond to.

To conclude, the behavior of verum focus in embedded declaratives provides support for the left peripheral position of Mood. This is particularly clear when considering the Spanish examples in (66). Furthermore, the data suggest again that there are discrepancies in the size of the left periphery in the three languages investigated and compared here (see also the discussion around the examples in (32) in section 4).