2.1. Formal V2

Due to its V2 characteristics, Old Spanish is typically viewed as differing from modern Spanish in terms of the position of its inflectional licensing site. In the modern language, this site is analyzed as T⁰, whereas for medieval Spanish, a higher, C-related position is often proposed (Benincà Reference Benincà, Zanuttini, Campos, Herburger and Portner2006; Wolfe Reference Wolfe2015, Reference Wolfe2016a; Poole Reference Poole, Bailey and Sheehan2017). I assume this position to be Fin⁰, following Haegeman’s (Reference Haegeman1996) proposal that Fin⁰ in V2 languages has INFL-like properties. In this framework, the V2 constituent necessarily merges into Spec-FinP to satisfy an EPP condition on Fin⁰.Footnote ⁴ Like the classical EPP on T⁰, this condition is primarily formal and can be satisfied by an expletive if necessary (Roberts Reference Roberts and Luigi2004). Holmberg (Reference Holmberg, Rebecca and Sam2020) further demonstrates that the condition can be satisfied through either IM or EM, the latter correlating in Old Spanish with enclisis (see Section 2.3 for detailed motivation of this claim).

Thus the proposed analysis adopts Haegeman’s fundamental insight (Reference Haegeman1996: 144) that Fin⁰ in V2 languages is an Agree head, a C-related analogue of T⁰ in English-type languages. It therefore probes φ-features rather than Criterial ones, and hence cannot license topic, focus or quantificational readings, which are associated with higher Criterial projections in the left periphery. However, XPs externally merged into Spec-FinP are not semantically vacuous; while they do not convey discourse-related interpretations, their surface position can still contribute to clause-level meaning. In this respect, Fin resembles Rizzi’s Mod projection, which hosts constituents that are neither topics nor foci but are nonetheless ‘prominent’ – that is, interpreted as satisfying an EPP-related prominence requirement (Rizzi Reference Rizzi and Adriana2004: 242). On this view, merge into Spec-FinP plays a dual role: it satisfies a formal EPP requirement and simultaneously enables a constrained interpretive effect – pragmatically inert, but semantically contentful. Section 5 examines this interaction in the context of adverbial licensing in Old Spanish, though the same mechanism plausibly extends to all externally merged XP categories.

In languages lacking an EPP on Fin⁰ (e.g. Modern Spanish), such XPs are likely instead to be merged into higher left-peripheral projections like FrameP (Benincà and Poletto Reference Benincà, Poletto and Luigi2004), given that V2 effects such as subject–verb inversion are not enforced in such contexts. The contrast with Old Spanish confirms, therefore, that FinP can function as a default host for available, independently derived XPs – especially those not selected by the verb – whenever the EPP on Fin⁰ is active. This allows these constituents to participate in clause structure without altering its propositional core.

Under these assumptions, the earlier example (4) can be seen as being derived as in (6); that is, without IM of the V2 element:Footnote ^5, Footnote ⁶

As implied by the above discussion (see also Section 5 below), the V2 constituent desi in the above example is unmarked in scope-discourse terms, reflecting a purely formal V2 structure characteristic of the medieval language. Batllori and Sitaridou (Reference Batllori and Sitaridou2020) argue that Old Spanish lacks formal movement into the C-system altogether, interpreting this as evidence against a V2 grammar. In contrast, the present analysis concludes that formal V2 in Old Spanish crucially involves external merge (EM) into Spec-FinP. Accordingly, given the proposed correlation between EM and enclisis, the typology of formal V2 constituents can be reconstructed from the distribution of enclisis. Table 1 presents this typology, along with the proportion of enclisis associated with each constituent type. This enclisis-based typology is exemplified, quantified and mapped to further diagnostics in Section 4. The data for Table 1 derive from an exhaustive analysis of finite enclisis in the thirteenth-century historical work General estoria I. Footnote ⁷ As Sitaridou (Reference Sitaridou2011) observes, this text is well-suited to reconstructing thirteenth-century Spanish syntax. Its copy date (1272) falls within the work’s composition window (1272–1284), suggesting that it faithfully reflects the original language. Moreover, its direct and relatively informal prose style enhances the likelihood that it captures authentic spoken usage.

Table 1. Enclisis in (non-imperative) finite main clauses (data from General estoria I, 1272; 535,516 words)

As in Germanic V2 (Holmberg Reference Holmberg, Kiss and Alexiadou2015: 353), a small minority (5.6%) of the enclitic clauses captured in Table 1 are V1. In these cases, EPP on Fin⁰ can be assumed to be satisfied covertly. For example, Fontana (Reference Fontana1993) and Wolfe (Reference Wolfe2015) analyze such cases as being instances of Narrative Inversion, which Holmberg (ibid.) suggests could involve a ‘covert temporal adverbial particle’ in initial position.Footnote ⁸

It is also striking that by far the most frequent V2 constituent in Table 1 is the coordinating conjunction, typically represented in the manuscripts by the Tironian et (here shown as the ampersand ‘&’). This differs from modern Germanic V2 languages, where a coordinating conjunction such as und ‘and’ cannot serve as the V2 constituent (Holmberg Reference Holmberg, Kiss and Alexiadou2015: 375). However, as Poletto (Reference Poletto, Batllori, Hernanz, Picallo and Roca2005: 227–234) shows for Old Italian e, clause-initial conjunctions in early Romance often functioned not as true coordinators but as expletive elements merged in the CP domain to satisfy a formal requirement. Diagnostics include their appearance clause-initially even when no coordination is present, their ability to introduce embedded infinitival and gerundial clauses, their frequent use after a full stop and the fact that they disappear in modern translations of the same texts. The distribution of Old Spanish ‘&’ conforms to this pattern and, within the Fin-based approach adopted here, it can be analyzed as an externally merged filler in Spec-FinP that satisfies the EPP on Fin⁰, rather than as a genuine coordinator. This view is consistent with Wolfe (Reference Wolfe2015: 135), who argues against the traditional treatment of Old Spanish et-initial clauses as V1 clauses.

Finally, it is worth noting that the preverbal elements listed in Table 1 can co-occur within the preverbal field, as illustrated in (7). In this example, a dislocated ‘subject’ (in bold) precedes an adverbial clause (in italics).Footnote ⁹ By definition, a dislocated ‘subject’ is not the syntactic subject but a DP coreferential with the null subject (see Villa García 2019, Barbosa Reference Barbosa2009). This arrangement closely resembles clitic left dislocation with a null clitic.

The adverbial clause in (7) is the V2 constituent and should be analyzed as undergoing EM into Spec-FinP, as per the paradigm in (6). The dislocated ‘subject’ occupies a higher, externally merged Spec, potentially inside the Frame field of Benincà & Poletto (Reference Benincà, Poletto and Luigi2004).Footnote ¹⁰ If such a dislocated ‘subject’ were the only preverbal element, however, it would have to merge into Spec-FinP to satisfy the EPP on Fin⁰. In that case, as shown in (31) below, enclisis arises, indicating that the XP does not undergo further movement. This confirms the observation made earlier in this subsection that both Fin and Frame can host externally merged constituents not implicated in the core propositional content of the clause. In non-V2 languages Spec-FinP may be bypassed altogether, but in V2 languages the EPP forces its activation.

2.2 Scope/discourse-related V2

In addition to the purely formal paradigm based on EM, as illustrated in (6), Old Spanish also exhibits structures where the V2 constituent undergoes IM into the C-system. This movement consistently correlates with a scope- or discourse-related interpretive effect. Many of these structures persist in the modern language as residual V2, retaining their interpretative function (see Leonetti & Escandell Vidal Reference Leonetti, Vidal, Dufter and Jacob2009, Jiménez Fernández Reference Jiménez Fernández2015). Examples (8) and (9) – the first from Old Spanish and the second from Modern Spanish – illustrate a well-documented instance of this phenomenon. Quer (Reference Quer, Beyssade, Bok-Bennema, Drijkoningen and Monachesi2002) refers to this as Quantificational QP-fronting and argues that the fronting operation imbues the QP with a ‘focus-affected’ reading in the sense of Herburger (Reference Herburger1997, Reference Herburger2000). In both examples, the fronted QPs are shown in bold.

Mackenzie & Wurff (Reference Mackenzie and van der Wurff2012) analyze Old Spanish examples similar to (8) in terms of a so-called Quant-Criterion – analogous to the traditional wh-Criterion – whereby a [+Quant] feature on C attracts a quantified object into its Spec. Reinterpreting their analysis within the framework of the Spec-head or Criterial Theory of V2 (Poletto Reference Poletto2000: 89; Samo Reference Samo2019), example (8) appears at first glance to involve the configuration shown in (10), where the [+Quant] feature is licensed in Rizzi’s Focus head, as per the licensing typology of Rizzi (Reference Rizzi and Adriana2004: 243).

However, this analysis is more appropriately applied to the residual V2 structure observed in (9), as the Spec–head paradigm cannot be fully generalized in Old Spanish. This limitation arises because the medieval language permits multiple A-bar movements, as demonstrated in (11), which displays a characteristic Spec–Spec–head linear sequence.Footnote ¹¹ The two Specs are highlighted in bold and italics, respectively.

Here, the fronted object este ruego & esta razon is implicitly discourse-given and is thus naturally analyzed as occupying Spec-TopP (cf. Wolfe Reference Wolfe2015: 139), while the raised negative adverb nunca is appropriately placed in Spec-FocP, following the licensing typology proposed by Rizzi (Reference Rizzi and Adriana2004: 243). Under a Spec-head analysis, therefore, the structure would be as in (12):

The difficulty with this analysis is that the final position of the finite verb seems arbitrary. If it can raise to Foc⁰, then in principle it could just as well raise to Top⁰. One possible solution to this problem arises from the proposal in Samo (Reference Samo2022), according to which there is parametric variation as regards whether a given Criterial head triggers verb movement, but finite V always raises to the highest verb-movement–triggering head. Applied to (12), however, this would entail that in Old Spanish Top⁰ never triggers verb movement, while Foc⁰ consistently does – an asymmetry that is difficult to reconcile with the fact that both projections invariably show surface adjacency (ignoring clitics) between their specifier and finite V when they occur in isolation (see (8) above and (20) below). A more data-consistent assumption, therefore, is that Criterial heads in Old Spanish are uniformly inactive for verb movement: they attract phrasal constituents but do not serve as landing sites for the verb. On this view, the finite verb remains in Fin⁰, its inflectional licensing position, and its observed adjacency with the Criterial specifier reflects surface linearity rather than a Spec–head configuration.Footnote ¹² Under this view, the QP-fronting in (8) can be analyzed as in (13):

This representation also takes into account the assumed merge requirement on Spec-FinP. Thus, while the Criterial head Foc⁰ attracts the QP directly from v*P, the EPP feature on the Agree head Fin⁰ (see Section 2.1) also attracts the same XP into Spec-FinP, where it receives a null spellout. In other words, in scope- or discourse-related V2 the initial XP participates in two separate chains: a Criterial chain and an Agree chain, much as who in English subject questions is simultaneously part of both an A-chain (to Spec-TP) and an A-bar chain (to Spec-CP) – see Chomsky (Reference Chomsky, Freidin, Otero and Zubizarreta2008: 149–150).Footnote ¹³ This dual-chain structure contrasts with purely formal V2, where the initial XP participates only in a single-membered Agree chain.

For the reasons that are advanced in Section 2.3, the dual chain structure triggers proclisis rather than enclisis, if a weak pronoun is present. This is exemplified in (14) below, which features a quantificationally fronted QP (in bold) and a proclitic pronoun (in italics).

In summary, Old Spanish V2 represents a composite of two distinct mechanisms: purely formal V2 and scope/discourse-related V2. Purely formal V2 targets Spec-FinP (an A-position in all but name) and exhibits default enclisis. In contrast, scope/discourse-related V2 targets higher, Criterial projections, such as Spec-FocP, and the cliticization pattern is proclisis rather than enclisis. This empirical duality mirrors the theoretical divide between the formal/FinP approach to V2 (Haegeman Reference Haegeman1996; Roberts Reference Roberts and Luigi2004) and the Spec-head/Criterial approach (Poletto Reference Poletto2000; Samo Reference Samo2019). The underlying data pattern is presented in greater detail in Sections 3, 4 and 5.

2.3. Enclisis as an EM diagnostic

A central claim of this paper is that the enclisis–proclisis alternation in Old Spanish finite main clauses reliably distinguishes between external merge (EM) and internal merge (IM) of the preverbal constituent. This correlation is not stipulated but derives from Shlonsky’s (Reference Shlonsky and Luigi2004) account of clitic placement, which in turn provides a theoretical foundation for Benincà’s (Reference Benincà, Zanuttini, Campos, Herburger and Portner2006) observation that proclisis in Old Romance reflects A-bar movement into a Criterial position. While Benincà identifies this position as Spec–FocP, the present analysis does not hinge on that specific label. Crucially, the proposed diagnostic is empirically robust: familiar A-bar constructions such as wh-movement and focus fronting consistently co-occur with proclisis and are categorically incompatible with enclisis, as confirmed by corpus data (see Section 2.3.2). This empirical asymmetry supports the conclusion that enclisis signals the absence of a Criterial chain and thereby serves as a diagnostic for EM into the preverbal field.

2.3.2. Empirical support

The theoretical link between IM and proclisis is strongly corroborated by quantitative corpus evidence. In a 3.5-million-word sample from the Old Spanish Textual Archive (OSTA; Gago Jover and Pueyo Mena Reference Gago Jover and Mena2020),Footnote ¹⁴ automated searches identified 24,846 tokens of finite pronominal enclisis – but none occur in clauses containing familiar A-bar movement structures such as wh-movement or focus fronting.Footnote ¹⁵ Wh-movement, in particular, systematically co-occurs with proclisis, as illustrated by the bold and italicized elements in examples (15) and (16), with no enclitic counterparts attested.Footnote ¹⁶

A similar pattern is observed with focus fronting, a structure widely assumed to involve A-bar movement rather than EM into the left periphery.Footnote ¹⁷ Example (17) illustrates proclisis in this context in Old Spanish, with the focused constituent in bold and the relevant clitic in italics. The (contrastive) focal reading is reinforced by the negative coda & no en encubierto ‘and not in secret.’

Establishing that enclisis is incompatible with focus fronting is methodologically difficult, because focus is an interpretive feature that cannot easily be targeted by automated searches. However, no instances of finite enclisis were found in constructions where a cliticized finite verb is followed by a negative coda, as in (17), supporting the hypothesis that enclisis is systematically excluded in such contexts. This is consistent with Sitaridou’s (Reference Sitaridou2011: 173) observation that the focus–proclisis linkage in Old Spanish ‘enjoys global consensus’ and Benincà’s (Reference Benincà, Zanuttini, Campos, Herburger and Portner2006) view that enclisis in Old Romance is the reflex of an empty FocP.

Additional evidence linking A-bar movement to proclisis comes from cases of long-distance movement out of CP, a traditional hallmark of A-bar movement (Cinque Reference Cinque1990). Old Spanish manuscripts provide numerous examples where long movement co-occurs with proclisis on matrix verbs, as shown in (18) and (19), with the long-moved element in bold and the relevant clitic in italics.Footnote ¹⁸ By contrast, no instances of long movement co-occurring with enclisis on matrix verbs were found in OSTA, suggesting a fundamental incompatibility between long-distance movement and enclisis in Old Spanish.

In sum, the extensive textual evidence strongly supports the conclusion that A-bar movement chains are incompatible with finite pronominal enclisis in Old Spanish. This suggests that whenever enclisis appears in finite clauses, any specifiers in the C-system must be externally merged, reinforcing the validity of enclisis as a reliable diagnostic for identifying EM-driven structures.

4.1. Externally merged V2 constituents and their rates of occurrence

As outlined in Section 2.1, the typology of externally merged V2 constituents in Old Spanish is inferred from the distribution of enclisis. Table 1 in Section 2.1 defines this typology, identifying six categories of preverbal constituents, which are exemplified below (in bold) with enclitic pronouns (in italics). Table 2 provides estimates for the occurrence rates of these categories in the V2 position of main clauses in thirteenth-century Spanish. These estimates reflect the proportion of enclitic tokens for each category (as detailed in Table 1) and are adjusted using the 75%-to-25% enclisis-to-proclisis ratio reported by Bouzouita (Reference Bouzouita, Cooper and Kempson2008: 238, N = 2026) for main clauses of that period. This ratio serves as a proxy to distinguish formal V2 constructions (involving EM) from scope- or discourse-related V2 constructions (involving IM) in main clauses. Separate estimates are provided for ‘all main clauses’ and for ‘referential V2 clauses’, i.e. those in which the V2 position is occupied by an overt referential element (excluding both the covert adverb associated with Narrative Inversion and the conjunction marker).

Table 2. Externally merged V2 constituents: estimated occurrence rates in thirteenth-century Spanish

Coordinating conjunction:

Adverbial clause:

Dislocated argument:

Adverb:

Non-argumental PP:

Vocative expression:

4.2. Compatibility with recomplementation

As discussed in Section 2, enclisis serves as the primary diagnostic for external merge (EM) into the V2 position. Additional evidence comes from the correlation, first identified by Fontana (Reference Fontana1993), between a constituent’s ability to co-occur with pronominal enclisis and its capacity to appear in the recomplementation position, i.e. sandwiched between two iterations of the complementizer que at the beginning of an embedded clause. According to Shlonsky (Reference Shlonsky and Luigi2004: 344), the first que realizes the Force head, while the second realizes the Fin head. Villa García (Reference Villa García2019) further demonstrates that the second que imposes locality restrictions, blocking movement and preventing elements from crossing it. As a result, the intervening constituent must be externally merged, as internal merge (IM) would violate these locality constraints. Fontana (Reference Fontana1993: 163–169) had already observed this phenomenon, proposing – prior to Rizzi’s split CP hypothesis – that the element between the two instances of que is externally merged as the sister of CP.

The following examples illustrate recomplementation for each of the constituent types listed in Section 4.1 (with the relevant item shown in bold), excluding the coordinating conjunction and vocatives. The coordinating conjunction is excluded due to the requirement for the intervening element in the recomplementation structure to carry descriptive content, inherently ruling out expletive elements and Boolean operators. Vocatives, meanwhile, are confined to main clauses, whereas recomplementation is, by definition, an embedded structure.

Adverbial clause:

Dislocated argument:

Adverb:

Non-argumental PP:

A related phenomenon is the overt realization of a single complementizer in main clauses. As with the lower que in recomplementation, elements that other diagnostics suggest are externally merged into the left periphery appear before this overt complementizer, whereas items that undergo IM are spelled out to its right. Desi, the most frequent enclisis-triggering adverb (see Table 4 in Section 5), exemplifies the first case in (39), while the argumental PP de njnguno (‘of nobody’) in (40) illustrates the second.

Like the lower instance of que in recomplementation, the single que in (39) and (40) appears to impose a locality constraint, preventing IM across it. As a result, EM is the only mechanism that allows an item to be positioned in the preceding structural space.

4.3. Compatibility with V3

An additional empirical signature of EM-driven V2 in Old Spanish is that the V2 constituent can also occur in V3 structures, merged to the left of the V2 position (cf. Fontana Reference Fontana1993: 114). This matches Holmberg’s (Reference Holmberg, Rebecca and Sam2020) findings for Swedish, where ‘as-for’ phrases can be externally merged either into Spec-FinP as the V2 constituent or into a more peripheral outer topic position, preceding an alternative V2 constituent. Analogous findings are reported by Vance et al. (Reference Vance, Donaldson, Steiner, Colina, Olarrea and Carvalho2010) – albeit within an ostensibly movement-based approach – who show that in Old French and Old Occitan, fronted subordinate clauses play a central role in distinguishing V2 from V3 orders. Example (41) illustrates the V3 pattern for Swedish (where EM of the position 1 element is assumed):

Examples (42)–(44) illustrate the same pattern in Old Spanish, involving an adverb, a non-argumental PP and an adverbial clause, respectively. Each of these elements can be assumed to merge into a high clausal layer, such as the Frame field. In these examples, the V2 element arises through internal merge: Quantificational QP fronting in (42) and (43) and Adverb fronting in (44). Consequently, the cliticization pattern is proclisis rather than enclisis. Externally merged items are highlighted in bold, while internally merged ones are italicized.

The placement of the expressions in bold outside the V2 position highlights their syntactic independence from the rest of the clause. This is consistent with the assumption that these constituents are directly merged into the left periphery, rather than being derived through movement from within TP.

4.4. Incompatibility with long movement

A final piece of evidence for the EM-based analysis of enclisis-triggering preverbal elements comes from long movement, i.e. extraction from an embedded CP. The availability of such movement has long been recognized as a hallmark of A-bar dependencies (Cinque Reference Cinque1990), distinguishing operations driven by Criterial features (wh-movement, focus fronting, topicalization) from Agree- or EPP-related processes such as passivization and subject raising. Although this distinction is usually framed in terms of A-bar versus A-movement, the exclusion of long movement follows equally where no movement takes place. This helps explain the contrast in (45): while the wh-adverb when can be extracted from the embedded clause (45a), the non-wh adverb yesterday cannot (45b) – unless corrective focus or topicalization is applied, which upgrades it to an A-bar dependency.

This contrast makes the broader point that only elements capable of A-bar movement can undergo long extraction. It follows that items which resist extraction from an embedded CP cannot, when surfacing to the left of the finite verb, have reached that position by internal merge: if internal merge were an option for them in the matrix, it should be available uniformly across clause boundaries. The fact that, as shown in Section 2.3.2, long movement in Old Spanish is compatible only with proclisis and never with enclisis therefore further confirms that the preverbal items that trigger enclisis (listed in Section 4.1) are externally merged in their surface position, rather than raised from a lower position within the clause.Footnote ¹⁹

4.5. Data summary

Table 3 summarizes the key distinctions between externally and internally merged V2 elements, structured around the five core variables central to this paper’s analysis (further discussion of the ‘scope–discourse interpretive correlate’ variable is provided in Section 5 below).

Table 3. EM-driven V2 versus IM-driven V2

The most straightforward explanation for the contrasting data matrices in the table is a derivational one, based on whether the V2 constituent is externally or internally merged. Holmberg’s (Reference Holmberg, Rebecca and Sam2020) findings for Swedish make this unsurprising for adverbial clauses, which he recognizes can be externally merged into Spec-FinP. What is particularly striking, however, is that non-clausal adverbials – including both adverbs and non-argumental PPs – follow the same pattern.Footnote ²⁰ In other words, they are genuine V2 constituents but consistently meet the diagnostics for EM. This outcome is not predicted by Holmberg’s approach, which, as he acknowledges (ibid. 60), is based on a priori reasoning rather than empirical evidence. In light of this, the next section considers non-clausal adverbial constituents in more detail.

Old Spanish Manuscripts Cited

Espéculo (BNE MSS/10123), Estoria de España I (Escorial Y-I-2), Estoria de España II (Escorial X-I-4), Fazienda de Ultramar (Salamanca: Biblioteca Universitaria 1997), General estoria I (BNE MSS/816), General estoria II (BNE MSS/10237), General estoria IV (Vaticana Urb. Lat. 539), General estoria V (Esc. I-I-2), Lapidario (Escorial h-I-15), Libro de las leyes (BL Add. 20787), Libros del saber de astronomía (Madrid: Biblioteca Universitaria Complutense, BH MSS 156), Poridat de las poridades (Escorial L-III-2).