Viewpointed morphology: A unified account of Spanish verb-complement compounds as fictive interaction structures

Abstract

Spanish verb-complement (VC) compounds, one of the most common compound types in Spanish, raise interesting questions, because they are inflected, prototypically containing a verb in the third-person singular of the present indicative. This complexity seems paradoxical, given the strong restrictions of Romance languages on word compounding.

Based on a self-compiled corpus of over 1,400 VC compounds, we show that the compound’s verb may display different persons and illocutionary forces. We claim that all Spanish VC compounds can be parsimoniously accounted for as involving a grammaticalized perspective-indexing structure, setting up a non-actual enunciation. We identify three subtypes of nominal VC compounds according to whether they refer to: (i) the fictive addresser of the non-actual enunciation it is composed of (e.g. metomentodo [I+put+myself+into+everything], ‘meddler’), (ii) the fictive addressee (e.g. tentetieso [hold+yourself+upright], ‘tilting doll’), or (iii) the fictive conversational topic (e.g. pintalabios [paints+lips], ‘lipstick’). We further argue that, despite undeniable morphological constraints, Spanish VC compounds involve a similarly complex semantic and morphological structure as English multi-word compounds like ‘wanna-be(s)’, ‘forget-me-not(s)’, or ‘bring-and-buy sale’. This reveals that intersubjectivity can be central to word formation.

1. Introduction

Nominal compounds can underlie complex semantic and morphological processes far beyond the establishment of a straightforward signifier-signified relation. Semantically opaque compounds like ‘bellbird’, ‘hot dog’, or ‘coffee headache’ pose a challenge for compositional accounts of meaning. Particularly complex are compounds with a multi-word specifier, such as ‘once-in-a-lifetime opportunity’ and ‘spur-of-the-moment decision’, which share formal and functional characteristics with phrases. Even more complex in form and meaning are compounds such as ‘wanna-be(s)’, ‘forget-me-not(s)’, and ‘bring-and-buy sale’, which display viewpoint information (i.e. person, tense, and/or mood), thus resembling (elliptic) clauses or sentences (e.g. Pascual 2006, 2014; Pascual, Królak & Janssen 2013). Such complex nominal compounds are productive and relatively frequent in Germanic languages, compound formation in general constituting ‘without doubt the most productive morphological process’ in languages like Dutch (Don 2009: 583).

In contrast, viewpointed nominal compounds are much more marginal in Romance languages (Pascual & Królak 2018), which have strong restrictions on word compounding (Val Álvaro 1999: 4759; Marqueta 2017). Some scholars even claim that Romance languages lack structurally complex compounds (Bisetto 2015), so-called phrasal compounds, that is, compounds with phrases in the non-head position (Wiese 1996: 185). This notwithstanding, Romance languages have verb-complement (VC) compounds. ² Spanish examples are abrecartas ([open(s)+letters], ‘letter opener’); hazmerreír ([make+me+laugh], ‘laughingstock’); and mandamás ([rule(s)+more], ‘boss’). This seems paradoxical given the constraints on compounding, because VC compounds, including its most prototypical form, that is, verb-noun (VN) compounds, are semantically and grammatically particularly complex. They are not formed by an infinitive or a bare verbal stem but by an inflected verb form (see Section 4). Thus, these compounds carry perspective information of person, tense, and/or mood, just as clauses and sentences do. Far from being rare, this in fact constitutes one of the most productive patterns of compound formation in most Romance languages (e.g. Bisetto & Scalise 1999: 75).

How a poor compounding language can have as one of its most productive compound types structures of such complexity remains an unresolved issue. The vast literature on Spanish VN compounds treats particularly complex instances as anomalies, for example, those containing pronouns (e.g. sabelotodo ([knows+it-all], ‘know-it-all’) or determinants (e.g. vivalavida [may-live+the+life], ‘overly laid-back a person’) (Val Álvaro 1999; Moyna 2011), with those displaying complex inner structures like vocatives (pasagonzalo [pass+Gonzalo], ‘punch’) or coordinates (e.g. correverás [run(+and)+you’ll-see, ‘moving toy’) being largely ignored by scholars.

Contrary to the general view, we claim that these more striking and rare subtypes of VC compounds in fact reveal what we argue is the covert structure underlying ordinary and frequently occurring VN compounds like limpiabotas ([clean(s)+boots], ‘boot polisher’) or saltamontes ([hop(s)+hills], ‘grasshopper’). We thus provide a unified account of these and ‘regular’ Spanish VN compounds as constituting different subtypes of the same skeletal viewpoint schema, sharing the same basic formal and semantic features. We attempt to show that, despite undeniable morphological constraints, all Spanish VC compounds carry as much perspective information as complex English multi-word compounds like ‘wanna-be(s)’, ‘forget-me-not(s)’, and ‘bring-and-buy sale’. This approach is consistent with the broad definition of compounds by authors such as Plag (2003: 135): ‘A compound is a word that consists of two elements, the first of which is either a root, a word or a phrase, the second of which is either a root or a word’. In our definition, however, the first element may be an inflected verb and the second one may constitute a pronoun. Our database provides ample evidence of the complexity and diversity of Spanish VC compounds that has so far been largely ignored and thus unaccounted for by Hispanists.

2. Database

This study is based on a self-compiled database of 1,417 VC compounds (i.e. 981 conventional and 436 creative ones), extended from Marqueta (2019b). These are mostly from Peninsular Spanish but also include instances from Equatoguinean Spanish and all main varieties of Latin-American Spanish. ³ Most examples are from contemporary Spanish, with approximately 100 instances from Late Medieval to Modern Spanish. The oldest examples in our database are from the twelfth century, a few no longer in use. We did not search for older sources. The most recent conventionalized examples are neologisms for new phenomena, such as salvapantallas ([save(s)+screens], ‘screensaver’), pescaclics ([fish(es)+clicks], ‘clickbait’), and cazaautógrafos ([hunt(s)+autographs], ‘autograph hunter’). The most recent creative compounds are from 2015 to 2020. Most entries are from standard Spanish, with a large percentage from informal language use. A few instances are from marked registers, such as professional jargons (e.g. nautics, the law, and the army) and sociolects (e.g. the speech of the youth, Casado Velarde & Loureda Lamas 2012; Sanmartín 2017).

Conventional examples were mostly obtained through native-speaker introspection, from dictionaries and grammars, and from academic publications. The oldest examples are mainly from Bustos (1986), Herrero Ingelmo (2001), and Moyna (2011) and are mostly from lexicographical sources as well as ancient novels and theater scripts by classical Spanish writers, such as Diego Sánchez de Badajoz (1479–1549), Miguel de Cervantes (1547–1616), and Pedro Calderón de la Barca (1600–1681). Creative examples, from oral as well as written Spanish, come from a wide array of different sources and genres, ranging from one-time occurrences in private blogs, social media posts, or spontaneous conversations to highly scripted language use, as in poems, newspaper articles, cartoons, or movies. These creative instances include, among others, nicknames for real or fictional characters (e.g. Matacuras [kill(s)priests]), the actual nickname for a man who killed five priests during the Spanish civil war), and new inventions, games, or products (e.g. Pintalenguas [paint(s)+tongues], ‘tongue painter’, i.e. a brand of candy that colors the consumer’s tongue).

The etymology of all compounds was checked to ensure their proper categorization. Ambiguous cases were not included. For instance, the toponym Matalascañas, where las cañas means ‘the reeds’, could either be a noun-noun (NN) or a verb-noun (VN) compound, because mata equally corresponds to the noun mata (‘bush’) and to the imperative and the third-person singular forms of the verb matar (‘to kill’). In all examples, italics (marking inflectional structures) and underlining (for noteworthy parts) are ours. Unless otherwise specified, examples in the text and the database are all found in dictionaries or directly retrievable from the internet.

This paper first presents the structural differences between Spanish and English compounds, arguing that such differences do not pose an obstacle for Spanish to encode viewpoint in compounding just as English does (Section 3). We then discuss evidence for our analysis of Spanish VC compounds as viewpointed structures (Section 4). Section 5 introduces the phenomenon of fictive interaction (Pascual 2006, 2014), which we believe can account for the presence of perspective information in a nominal structure. In Section 6, we lay out how this becomes manifest in Spanish VC compounds, in different semantic and formal types.

3. Spanish compounds: Formal restrictions

As discussed in the introduction, Spanish lacks compounds with phrases, like the English ‘once-in-a-lifetime opportunity’, or clauses, like ‘bring-and-buy sale’ (see Buenafuentes 2021 for an overview of the semantic, syntactic, and morphological properties of Spanish compounds). This fact is often accounted for by the small productivity of the Spanish compounding system, characterized by structures with simple syntax, which should consequently not display viewpoint information. Snyder (2001) attempted to account for the contrast in the compounding restrictions of Germanic vs. Romance languages through his ‘compounding parameter’. This argues that languages allow complex predicate constructions like verb particles, resultatives, and double objects if and only if they can productively form noun-noun compounds. Indeed, none of these complex predicative constructions appears in Spanish, whose noun+noun word-formation structure is also less productive than that of Germanic languages. However, noun+noun compounds are still productive in Spanish. Hence, Snyder’s (2001) account is not entirely satisfactorily (see Liceras, Díaz & Salomaa-Robertson 2002: 209) or cannot in itself account for the differences we encounter.

In this section, we show that the lack of such compounds can be accounted for by restrictions on only one compound pattern, that is, the modifier-head structure, which cannot be generalized over all Spanish compound types, specifically not to VC compounds. These are restrictions on complex modifiers (Section 3.1) and on indexicals and other functional categories (Section 3.2).

3.1 The complex modifier restriction

The first restriction on Spanish compounds composed of two nouns, a productive and frequently occurring structure (see Marqueta 2019a for an overview), is that the modifier needs to be simple. Compare, for instance, the following two Spanish compounds with their English counterparts:

Indeed, the Spanish equivalents of English nominal compounds with a complex first element are generally phrases with prepositional phrase (PP) complements, including a PP within a PP (2b), or even with a pronoun introducing a relative clause, as in (2c): ⁴

Thus, whereas in English a complex structure may be directly adjoined to the compound’s head, Spanish requires a second level of subordination in which the complex structure appears as a prepositional complement or a clause. Viewpointed compounds, such as ‘Will you marry me ring’ and ‘the dog ate my homework excuse’, have undeniably complex modifiers, as they involve an entire sentence or occasionally even a piece of discourse or dialogue (e.g. ‘“How-are-you-fine-thank-you-and-you-fine-thank-you” syndrome’, Pascual 2014: 63). As predicted by the complex modifier restriction, translating these examples literally into Spanish compounds results in ill-formed structures, as shown in (3). For these too, Spanish generally uses phrases headed by a noun with a PP complement or a clause modifying the head noun:

Indeed, most viewpointed English nominal compounds are translated into Spanish as phrases or even clauses (Pascual & Królak 2018). Consider the English compounds below, and their official Spanish translations as a prepositional phrase (4a) and a clause (4b), respectively (Pascual & Królak 2018: 409, 417):

Less frequently, such complex viewpointed structures appear in Spanish as noun appositions, following a pause, as in the attested advertisement line ‘Plan me quedo todo en uno’ (‘Plan I’ll take all in one’, Pascual 2010: 85). These are not to be considered proper compounds, however, and are most likely the result of preposition ellipsis, a fairly frequent phenomenon in Spanish noun appositions (e.g. plaza España, lit. ‘square Spain’, from ‘plaza de España’, lit. ‘square of Spain’). ⁶ Since these are not compounds, such inflected noun appositions are not counterarguments to the view that ‘Romance languages seem to lack phrasal compounds of the kind present in some Germanic languages’ (Bisetto 2015: 395).

3.2 The restriction on indexicals and other functional categories

It is often assumed that compounds lack phrasal functional categories, such as determiners and pronouns (Rainer & Varela 1992), and that compound elements are not accessible to syntactic processes, such as agreement and anaphoric relations (Ackema & Neeleman 2004: 341). These assumptions have been challenged theoretically, by frameworks like Distributed Morphology (Halle & Marantz 1993) as well as empirically (Lieber 1992; Weiskopf 2007). Our database also contains a few counterexamples, which include determiners (5) and, most importantly, pronouns (6):

Due to their unsystematic morphological properties, compounds with overtly indexical elements like the ones in (6), have been treated as oddities in the compound system (Val Álvaro 1999; Moyna 2011). We regard these compounds as belonging to the VC category, if as non-prototypical members.

As for the indexicality restriction, English and Dutch allow indexicals with anaphoric access to the compound, the clearest case in fact being viewpointed compounds with complex modifiers (Janssen 2007; Pascual, Królak & Janssen 2013). Spanish VC compounds may on occasion also show coindexation of an element with a pronoun outside the compound, as in (7):

These examples show that the internal semantics of Spanish VC compounds is transparent, because the compound complement in the three of them is accessible enough to be referred to through a pronoun later in the sentence. Interestingly, pronoun coindexing can occur even in a VC compound emerging from an idiom without number agreement, as in (7b).

Spanish VC compounds may also display person features. For instance, metomentodo ([I-put+myself+into+everything], ‘meddler’) contains the Spanish first-person verbal ending (i.e. –o) and sabelotodo ([knows+it-all], ‘know-it-all’) shows an unequivocal third-person verbal form. One may wonder why such indexical features are in the verb element of VC compounds, since they are not required for agreement purposes. It is well known that the verb does not show regular inflection of tense, mood, person, or number with elements outside the compound. For example, the verb in the Spanish VN compound for birthday, cumpleaños ([turn(s)+years] ‘birthday’), remains unchanged when referring to the twins’ last birthday, its corresponding noun *cumplieronaños ([turn-pst.3.pl+years]), with tense, person, and number agreement, being ungrammatical (Rainer & Varela 1992; Jiménez Ríos 2001). ⁸

In the next section, we discuss the formal and semantic evidence for our claim that all Spanish VC compounds are viewpointed, despite the formal constraints on compounding outlined in this section.

4. Spanish VC compounds as viewpointed

The most controversial aspect of Spanish VN compounds concerns their verbal inflectional features of tense (or mood) and person. These are overly clear in compounds involving a verb whose imperative and third-person indicative forms are phonetically different from the corresponding infinitive or the verbal stem. Examples are VN compounds with verbs of the third conjugation (-ir), which show a theme vowel /e/ instead of the /i/ of the unconjugated infinitive form (Lang 1990; Val Álvaro 1999). Consider the examples in (8a, b), from the verbs abrir (‘to open’) and cubrir (‘to cover’): ⁹

The compound cubrepán ([cover(s)+bread], ‘bread cover’), first attested in 1196, is the oldest VN compound in Moyna’s (2011) diachronic dataset, showing that morphological inflection in Spanish VN compounds is not a recent phenomenon. Neologisms composed of a third conjugation verb, such as abrecaminos ([open(s)+paths], i.e. a ritual to improve one’s life) or cumpledías ([celebrate(s)+days], i.e. the day-by-day celebration of life), both from the Uruguayan writer Mario Benedetti (1920–2009), illustrate that inflection of the compound verb is still productive today. Another piece of evidence for an inflectional reading is provided by VN compounds composed of verbs undergoing diphthongization of a stressed /o/ or /e/ in the verbal stem into /we/ or /ie/, respectively, in the inflected form (Bermúdez Otero 2013; Marqueta 2019b). This is illustrated with a few VN compounds with the verbs contar (‘to count’) and reventar (‘to blow up’): ¹⁰

The phonological pattern of verb inflection in VN compounds in (8)–(9) is systematic and entirely productive, as shown in its appearance in one-time creations, like tropiezapiedras ([stumble(s)+stones], ‘clumsy person’) from the verb tropezar, or neologisms like cierrabares ([close(s)+bars], ‘partygoer’, Casado Velarde & Loureda Lamas 2012), from the verb cerrar. The verb element in the two novel compounds duerme-bebés ([sleep(s)+babies], ‘baby sleeper’) and duermemonas ([sleep(s)+monkeys], lit., ‘sleeps-it-off-er’, from the idiom ‘dormir la mona’, i.e. ‘to sleep it off’), show both the diphthongization of /o/ to /we/ that characterize inflected verbs and the phonetic change from the third conjugation ending /i/ to /e/ (compare with *dormi-bebés and *dormimonas respectively).

The phonological evidence of verb inflection is thus indisputable. ¹¹ It does, however, support both an imperative and a third-person indicative reading, since both forms share the same diphthongization and ending in the great majority of varieties of present-day Spanish. There are indeed supporters of both analyses. Romanist studies have long interpreted the verb element in VC compounds as a singular imperative form (see Lloyd 1968 and Floricic 2008 for an overview). This hypothesis is based on formal evidence that various verbal forms in Italian compounds unequivocally show an imperative rather than a third-person indicative form. The argument is that Romance compounds must have evolved from a common Proto-Romance language morphological schema, originating in the precursor of Latin from ancient Greek (Bader 1962). Critically, however, VN compounds, which are extraordinarily productive in Spanish, were in fact almost inexistent in Latin (Moyna 2011), even though they may have coexisted with the predominant Object-Verb (OV) pattern in vulgar Latin (Bork 1990). Moreover, the verbal systems of Romance languages differ considerably, with Spanish and Catalan having three verb types or conjugations, while Italian and French have four. Hence, the structure of the Italian verbal element in VN compounds does not seem a good candidate for inferring the corresponding verbal structure in Spanish.

Other scholars argue that Spanish VC compounds contain a verb in the third-person singular of the present indicative (Menéndez Pidal 1940; Val Álvaro, 1999). In the handful of Spanish verbs that show different stems in the imperative and the third-person present form, Spanish speakers select the present form for compounding (Val Álvaro 1999: 36). This is shown in Val Álvaro’s (1999: 4789) own neologism compound, entretieneniños ([entertains+kids], *entrete(n)niños), for someone who amuses children for a living, from the verb entretener (‘to entertain’), whose singular imperative form is entretén. Further evidence is provided by the conventional compound detienebuey ([stops+ox], ‘herbaceous plant’), from the verb detener, whose imperative form is detén, and from compounds with the verb poner (‘to put/lay/assign’), whose third-person indicative form (i.e. pone ) differs from the imperative form (i.e. pon). Examples of the latter are ponemedias ([puts+socks], ‘shoehorn’) – instead of *ponmedias – and ‘gallina pone huevos’ (lit. ‘hen lays eggs’, ‘fertile hen’) – rather than *gallina pon huevos (lit. ‘hen lay eggs’). ¹² The fact that these are not individual cases but part of a productive morphological pattern is evidenced by novel compounds with the indicative pone instead of the imperative pon, such as the following creative compounds:

This pattern can also be observed in VC compounds with other irregular verbs, such as tentar (‘to tempt’) and hacer (‘to do’), which also show third-person indicative forms (i.e. ‘tienta’ and ‘hace ’, respectively) that differ from their imperative counterparts (tenta vs. haz , respectively). Consider the following compounds composed of these verbs, which display the corresponding non-ambiguous third-person form, constituting conventionalized instances in (11) and creative ones in (12):

An extra piece of evidence for the third-person singular indicative theory can be found in entries from Medieval and Early Modern Spanish, which did not show equivalent forms in the imperative and the third-person indicative. While the third-person indicative showed the same form as in today’s Spanish, the second-person singular imperative was conjugated like the present-day’s Peninsular Spanish form for the second-person plural. That verbal form did not undergo a phonetic change from /i/ to /e/ in verbs with an –ir ending or undergo diphthongization in verb stems with an /o/ or /e/. Thus, the old Spanish imperative of suplir (‘to replace’) was suplid , instead of its modern form suple , which coincides with the third-person indicative. Similarly, while old verbs like contar (‘to count’) and venir (‘to come’) underwent diphthongization in the indicative, as in today’s Spanish (‘cuenta’ and ‘viene’), the old forms for their imperative counterpart did not (‘¡contad (vos)!’ and ‘¡venid (vos)!’). Early Spanish compounds from that period involving a stressed /o/ or /e/ vowel in the stem are thus not ambiguous regarding the verb’s inflected form being a third-person indicative. This can be illustrated by the following old compounds (from Moyna 2011), with the date when they were first attested:

Hence, the third-person indicative explanation seems more convincing than the imperative one. However, a few Spanish VN compounds are in fact unambiguously imperative instead of indicative forms, such as the conventional instances in (16) (two conventional and one creative), from the verbs tenerse (‘to hold oneself’), salir (‘to exit’), and ponerse (‘to become’):

The data discussed in this section show that some Spanish VC compounds undeniably involve a verb in the present indicative, whereas a few others unequivocally comprise the imperative mood. Neither of these groups involves loanwords and they thus both need to be accounted for. Therefore, we reject previous approaches which commit to one single form (but see Rainer 2001). Instead, we propose an umbrella account of Spanish VC compounds as all comprising an inflected verb and thus being viewpointed, like ordinary direct speech. We sustain that by assuming that they constitute different instantiations of the same grammatical pattern involving perspective information we can explain their formal diversity.

5. Fictive interaction

As outlined in the previous sections, our main tenet is that Spanish VC compounds are made out of inflected verb forms. This is not a disputed fact among scholars, despite the disagreement on whether they constitute imperative or declarative forms. However, no study to date has managed to account for the fact that Spanish VC compounds display information on tense or mood, and person. This is nontrivial, because viewpoint information is what we find in a sentence, a piece of text, or a conversation, with perspective constantly shifting between interlocutors in the latter. In this section, we discuss the notion of fictive interaction (Pascual 2006, 2014; Pascual & Sandler 2016), which we believe is critical in understanding viewpointed compounds and other grammatical constituents. Consider first the following extract from an interview with a renowned political analyst and linguist:

Here, an ordinary noun phrase, that is, ‘pure narcissism’, is paraphrased as the first-person pronoun ‘Me’, followed by a string of speech in the first person ascribed to the third-person referent at issue. The utterer shifts perspectives, taking the voice of the individual spoken about in order to demonstrate – rather than denote or describe – the kind of narcissism that he sustains characterizes that individual (cf. Clark & Gerrig 1990; Clark 2016; Ferrara & Hodge 2018). This does not constitute an ordinary free quote of a previously produced utterance by the referent. Instead, it is an entirely constructed piece of dialogue (cf. Tannen 1986, 2007), while not being fabricated or fictitious. The non-genuine enactment in (17) is entirely conceptual in nature, between the real and the imaginary, and thus ontologically fictive in the sense of Talmy (2000). It is non-genuine, but it does serve to express something actual about the world, or better, the speaker’s view of the world.

In a large number of unrelated languages of the world, such non-actual direct speech is in fact widespread and may appear at different grammatical levels (Pascual 2006, 2014). Take the attested English examples:

In (18), a fictive enunciation in the first-person singular appears in the position of a clause (18a), a phrase (18b), a nominal modifier (18c), and even a lexeme with a suffix (18d). While having received little attention from linguists and barely any attention from Hispanists, this viewpointed structure is as grammatically possible and frequently occurring in Spanish as in English, becoming manifest at all grammatical levels (Pascual 2010, 2014; Pascual & Królak 2018). Consider the following example from an opinion column:

Note the appearance of deictic pronouns (used as generics) and the second-person singular present indicative and subjunctive affixes in puedes and quieras as well as the diphthongization of a stressed /o/ and /e/ of the verbs poder (‘to be able to’) and querer (‘to want’), further indicating inflectional information. The constituent ‘tú-puedes-llegar-a-ser-lo-que-quieras’ has the syntax of a clause or sentence, while operating as a head noun preceded by an article, and the first-person pronoun yo is used as a lexeme, also following an article and having a suffix. Indeed, in the examples in (17)–(19), linguistic units appear in the syntactic slots of phrases, nouns, and even lexemes, while displaying first- and second-person pronouns and, in some cases, even verbs inflected for tense and/or mood, and number, showing agreement with those pronouns (e.g. ‘I-1.sg am-1.sg’ in (17) and ‘ tú -2.sg -puedes -2.sg …quieras -2.sg’ [you can…you-want] in (19)). These both serve to refer to some concept the way ordinary nouns and lexemes do and to fictively enact speech ascribed to some specific or generic individual or group. It should thus not be too surprising that Spanish nominal compounds may also contain a fictive kind of interaction, even one involving only one conversational turn.

We argue that Spanish nouns may display different types of fictive enunciations, depending on their semantic structure vis-à-vis the frame of the conversation. They may: (i) fictively speak for and serve to refer to or characterize the fictive enunciator (20), (ii) fictively speak to and refer to or characterize the fictive addressee (21), or (iii) fictively speak of and refer to or characterize the fictive topic (22). Instances of fully conventionalized viewpointed Spanish compounds that enact and serve to refer to the fictive speaker of the non-actual enunciation it is composed of are as follows:

Note that these compounds involve complex sentential structures like subordination (20a) and coordination (20b). Also, because the enunciation that characterizes and gives name to the referent is non-actual, a beggar for instance can be referred to as a ‘pordiosero’ even when not uttering the words ‘Por [el amor de] Dios’ (‘for [the love of] God’) when begging. Critically, such fictive enunciations can also be ascribed to non-living entities. Long and narrow shoes or boots are called ‘zapatos/botas de chúpame la punta’ (‘shoes/boots of lick my tip’, Pascual 2014: 107), as though the footwear were verbally demanding to be licked. Take now the following examples of fictive enunciations to refer to the fictive addressee (from two spontaneous conversations and a dictionary entry):

Examples like these, in which the referent is characterized by a non-actual enunciation fictively addressed at them, are common in colloquial Spanish. The superlative idiomatic expression ‘hasta decir basta’ (lit. ‘until saying “stop”’), in which the entity or individual categorized by it is presented as the addressee of the fictive command, is even fully conventionalized.

Instances of a viewpointed structure serving to categorize or refer to the topic of the fictive enunciation seem by far the most common type of fictive enunciation in Spanish noun phrases. Consider the following (Pascual 2010):

The use of a clause involving a communicative verb introducing an opinion on something or someone through what one may say about them is extremely common in informal Spanish (Pascual 2010, 2014: 107). As examples (22b)–(22c) show, this may become grammaticalized and lexicalized.

In the next section we present viewpointed Spanish VC compounds as instances of fictive enunciations whose referent is one participant in the fictive conversation that the compound is composed of, that is, the fictive enunciator or the fictive addressee, or the topic of the fictive utterance itself.

6. Spanish VC compounds as involving fictive interaction

As discussed, VN compounds are the prototypical subtype of Spanish VC compounds. We argue that they share the same schematic structure as other VC compounds (verb-verb [VV], verb-adjective [VAdj], verb-adverb [VAdv], etc.) and that they have all grammaticalized from a fictive enunciation. The Spanish literature on compounding notes the existence of VC compounds that are far from the canonical VN pattern, but these have been excluded by all scholars who have written on them, pointing to their unproductive character and their formal irregularity. In contrast, our proposal aims to unify prototypical and non-prototypical VC compounds under an umbrella account.

Given the low frequency of non-prototypical VC compounds in Moyna’s (2011) historical corpus of Spanish compounds, she concludes that examples with determinants (e.g. rascalacría ([scrap(s)+the+offspring], ‘method against mites’), ‘are the result of folk etymology or of lexicalization of syntactic phrasal formulas, and thus not true exponents of the pattern’ (Moyna 2011: 200). She further determines that the compound structure with an adverb in complement position (e.g. cantaclaro [sing(s)+clear], ‘popular folk song composer’) ‘is such a small group that the label [V+N] can be used to refer to that type of compounds indistinctly’ (Moyna 2011: 201). However, our database shows that although these non-prototypical VC compounds are certainly rare in standard Spanish from conventional lexicographic sources, such nominal compounds with an adverb are productive and relatively frequent in social media and colloquial conversation. This may be the reason why they are scarce in Moyna’s (2011) corpus, which is mostly based on conventional, highly scripted written lexicographic sources. Val Álvaro (1999: 4804) agrees with Moyna’s (2011) analysis of VAdj and VAdv combinations, confirming that they are of less importance, due to their lack of both productivity and frequency. However, it is worth noting that the number of so-called ‘exceptional’ examples provided by Val Álvaro (1999) and Moyna (2011) is still substantial. For compounds like bienmesabe ([well+me+tastes-3.sg], ‘sweet’) and nomeolvides ([not+me+forget-2.sg], ‘forget-me-not’), which show a sentential structure, Val Álvaro (1999) creates a composition category that is different from that of productive schemes in that it presents peculiarities that endow it with a specific nature within the syntagmatic composition.

In our analysis, the viewpointed compound family includes different manifestations of a fictive enunciation at different stages of grammaticalization. As discussed in previous sections, Spanish grammar is more restrictive than that of English and other Germanic languages, in which fictive direct speech may be directly introduced in the compound modifier position. Spanish does allow, however, more grammaticalized word-size structures. In what follows, we discuss the three main subtypes of nominal VC compounds we identified, referring to (i) the fictive enunciator (Section 6.1), (ii) the fictive addressee (Section 6.2), and (iii) the topic of the fictive enunciation (Section 6.3).

6.1 Compounds whose referent is the fictive enunciator

This class is the most marked one, thus the least productive and most morphologically heterogeneous one. This type does not display the prototypical and syntactically simplest VN structure, instead including clause-like elements such as determiners and pronouns. These are also frequent in VC compounds whose referent is the fictive addressee, while being entirely absent in the compounds whose referent is the fictive topic. The schematic formal pattern of this class of VC compounds is thus [V_x + C_x (X)].

6.1.1 The compound’s referent

The referents in this class are typically humans, since they are the ones to whom the non-genuine enunciation in the compound is ascribed. Hence, the resulting compound can refer either to a masculine or feminine referent, depending on the natural sex of the fictive enunciator. Examples are:

The two synonymous compounds in (23a) literally depict the referent’s characteristic nonchalant attitude by a jovial expression that epitomizes that attitude, namely ‘¡Viva la virgen!’ and ‘¡Viva la vida!’. The syntactic structure of these compounds is thus directly imported from the fictive message, as a demonstrative verbal formula serving to denote the one to whom that non-genuine quotation is ascribed. A similar example is (23b), in which a linguistic unit that is typically used to start an expression of remorse is used to refer to the remorseful. In all examples in (23), the fictive message that is presented as ascribed to the referent as best characterizing that referent is entirely fictive ontologically, as it represents a person’s demeanor or attitude through the gist of what they might say to express it verbally, as opposed to an utterance they actually produced.

VC compounds with a fictive enunciator may also refer to inanimate entities, such as objects and plants, metaphorically construed as the anthropomorphized fictive speakers of the non-genuine enunciation or representing what we may communicate through them:

Thus, in this class, the compound constitutes the fictive enunciation ascribed to an animate or inanimate fictive enunciator that most clearly defines it.

6.1.2 The compound’s morphosyntax

Contrary to the other two types, compounds whose referent is the fictive enunciator allow for verbs conjugated in any person, tense, or mood. Consider:

In Spanish VC compounds, the complement serving to refer to the fictive enunciator may be any element in regular Spanish phrasal and clausal complements. Possible compound elements in this category are:

As the least common and least grammaticalized, this is the most heterogeneous category regarding form, with barely any instances showing the prototypical VN compound structure. This contrasts with the other two categories of VC compounds, discussed in Section 6.2, and especially those in Section 6.3, in which the complement is predominantly a bare noun.

6.2 Compounds whose referent is the fictive addressee

Spanish VC compounds referring to the addressee of the fictive enunciation are highly productive. Their verb form can plausibly be analyzed as an imperative and in many cases it in fact non-ambiguously shows an imperative form. Take the following:

In these cases, the individual or entity referred to is not the one presented as characterized by uttering the message in the compound but by being addressed with that fictive message (i.e. the one being ‘told’ to make others laugh, to stay upright, or to go tell gossip to others). Note that the examples in (27) comprise verbs unequivocally in the imperative, their third-person present indicative equivalents resulting in ungrammaticality (*hacemerreír, *tienesetieso, and *correvaidícele). The basic structure of the fictive addressee pattern is thus [V-2.sg.imp+N-pl]. The existence of such a class shows that VC compounds do not display only one possible form, as discussed in Section 4.

Some VN compounds referring to the fictive addressee remain ambiguous between the third-person declarative and the imperative forms but only from a morphological point of view. The referent of these compounds is presented as called upon through the use of a derogatory nickname related to their profession (28a) or overall attitude or behavior (28b), regarded in a negative light:

These compounds thus originate in vocatives, calling upon the addressee, and are used mostly as insults (see examples in Herrero Ingelmo 2014). Vocatives overtly imply the interlocutor of the fictive interaction, who also constitutes the compounds’ referent. ¹⁴

6.2.1 The compound’s referent

Being the addressee of the fictive enunciation expressed in the compound, the referent of this compound type can be either masculine or feminine, depending on the referent’s natural sex or grammatical gender in the case of inanimate referents like tentetieso (‘tilting doll’).

6.2.2 The morphosyntax

As discussed previously, this compound subtype referring to the fictive addressee favors the imperative mood. Its prototypical pattern also displays the noun complement in the plural, just as in the case of fictive enunciators. This is related to the use of the schema, as demonstrated by the following types of compounds, which cannot be accounted for through compositionality:

1. i) Compounds from an idiomatic phrase with the noun element in the singular, for example, aguafiestas ([spoil+parties], ‘party blower’), from aguar la fiesta (‘to spoil the party’), and tiratoallas ([throw+towels], ‘halfhearted person’), from ‘tirar la toalla’ (‘to throw the towel’)

2. ii) Compounds with a complement noun referring to one single individual or entity, for example, golpeaesposas ([beat+wives], ‘wife beater’) and cazadotes ([hunt(s)+dowries], ‘man attempting to marry a rich woman for her wealth’)

3. iii) Compounds with a complement mass noun, for example, quemasangres ([burn+bloods], exasperating person’) and atrapasuertes ([grab+lucks], ‘serendipitous person’)

Apart from nouns, VC compounds referring to the fictive addressee may include non-prototypical complements, such as pronouns (29a), adjectives (29b), and adverbs or verbal phrases (29c):

Lastly, in the least prototypical examples, the complement may not be a direct object. Consider the following alternatives:

As was the case with VC compounds referring to the fictive enunciator, note that the fictive addressees may also constitute personified inanimate entities, such as plants (30b, d), as well as things or events (30c).

6.3 Compounds whose referent is the fictive topic

The third class of Spanish VC compounds is the one whose referent is the topic of the non-actual enunciation itself. Thus, the compound verb in this class rarely is in the first or the second person, favoring instead the third person (see Section 4 for unambiguous examples). This is unquestionably the most productive and prototypical pattern of one-word, single-stressed compounds in Spanish. Its basic structure is: [V3-sg+N-pl].

We follow the much-supported tenet that this class of VN compounds results from synchronic grammaticalization, in particular an operation of clausal reduction of free relatives, that is, descriptive relative clauses invariably containing an inflected verb in the third-person present indicative (for details, see Contreras 1985; Di Sciullo 1991; and Franco 2015). ¹⁵ The source and target forms of this grammaticalization process are exemplified in the following:

The kind of referent is only relevant in order to explain the gender of the resulting compound. Its default grammatical gender is the masculine when referring to an entity (afilacuchillos [sharpens+knives], ‘knife sharpener’). Only a few examples receive feminine grammatical gender (e.g. tragaperras [swallows+coins], ‘slot machine’, from máquina tragaperras [machine swallows+coins], máquina being feminine). When referring to a person, the VN compound’s gender corresponds to the referent’s natural gender (e.g. un/a cazatendencias [hunts+trends] for a male or female coolhunter). The predicate introducing the topic, however, is not affected by the referent, which is why this kind of VN compounds can serve to refer to a person and/or an object. For instance, guardajoyas ([keeps+jewels]) may equally refer to the officer in charge of keeping royal jewels and the container where jewels are kept. We did not find any such cases of ambiguity in the referent as a person or a (personified) object in either of the other two classes of compounds.

Other typical referents of VN compounds in this category are animals described by their most salient habits (e.g. saltamontes [jumps+hills], ‘grasshopper’, and picamaderos [bites+trunks], ‘woodpecker’), plants or drinks described by their effects when ingested (e.g. matacán [kills+dog], ‘poison’, and quitapenas [removes+sorrows], ‘liquor’), and entities or events described by their effects on people (e.g. trabacuentas [tangles+sums], ‘mistake’, and comecocos [eats+heads], ‘problem’).

This compound class is characterized by offering a non-encyclopedic description of individuals or entities. In fact, it is very frequent for compounds in this class to act as modifiers of an already existing referent, displaying a much more distinct descriptive function, specifying the function or characteristics of a given entity:

VN compounds involving (elliptic) prepositional complements are also clearly descriptive. In this case, the PP does not describe a person or object but rather an event (e.g. tocateja [touches+tile], ‘method of payment’, and pasatoro [passes+bull], ‘a bullfighters’ technique’).

Another less prototypical class within this compound subtype includes examples in which the topic of the fictive conversation expressed in the compound already involves an enunciation. In these cases, a fictive utterance (quoting what is being said, recited, or sung) appears embedded within a fictive utterance (saying what someone does, i.e. say, recite, or sing something):

Regarding morphosyntactic variants, most complements in this VN compound class are in the plural. Just like VN compounds whose referent is the fictive addresser or addressee, we find more instances of complements in the singular in VN compounds that name an inanimate referent than in those refereeing to an animate one. Such compounds may display a singular complement if the noun element is (i) a mass noun (34a), (ii) a noun referring to a single entity (34b), or (iii) a singular monosyllabic noun ending in a consonant (34c):

Complements of VC compounds referring to the fictive topic may be adjectives or adverbs or even adverbial phrases, as in the following:

As for the complement’s structure, they not only are direct objects but can also be locative adjuncts (with elliptic preposition and article), as in (36a); directional (36b); or even sources (36c). We also find coordinated clauses in the description of events (36d):

Overall, there is less variety in the type of complements that VC compounds referring to the fictive topic can take than in those referring to the fictive enunciator or addressee. Critically, we found no VC compound referring to the fictive topic containing a first- or second-person pronoun, except for bienmesabe ([well+me+it+tastes], ‘sweet’) and nomelopongas ([don’t+me+it-acc+put-2.sg], ‘don’t-serve-it-to-me’, ‘canceled coffee order’). This is significant, because it shows that it is the topic that is highlighted, instead of the fictive interlocutors, a fact that we believe constitutes a powerful argument to distinguish between these three subtypes of VC compounds.

Lastly, we briefly discuss Spanish compounds representing the fictive message (or part of it) expressed in the compound itself. While considering that these involve fictive interaction, we do not find them revealing regarding word formation in general or the prototypical structure of Spanish VC compounds in particular. These compounds are mere grammaticalized enactments of a verbatim quote (or part of one). Thus, the atypical morphosyntax of examples like besalamano ([kisses+the+hand], ‘short note’) and sepancuantos ([know-3.pl+how-many-pl], ‘punishment’) directly reflects the syntactic structure of the original verbal formula in the short note and legal warning that are metonymically referred to through these messages. The same goes for onomatopoeic compounds. For instance, the different nicknames for the great kiskadee bird all depict the bird’s high-tone call. It is called comechile ([eat(s)+chili]) in Peru; bichofué ([beast+it+was]) or cristofué ([Christ+it+was]) in Colombia, Venezuela, and Honduras; and diostedé [God+you-ACC+may-give]) or bienteveo ([well+you-ACC+I-see]) in Ecuador, Venezuela, and other parts of Latin America. The different internal structures do not arise from dialectal differences in compounding; they simply reflect what the bird seems to be saying with its call, manipulated as a unit. Their apparent paradoxical formal variety can parsimoniously be explained under our umbrella account of all Spanish VC compounds instantiating an inflected meta-schema, standing for a non-actual enunciation enacting or describing the compound’s referent.

6.4 Statistical distribution

A quantitative analysis of our database confirms that the most frequently occurring Spanish VC compound is type 3 (1,023/1,417; 72%), in which the compound’s referent is the fictive topic, followed by type 2 (366/1,417; 26%), in which the referent is the fictive addressee, with type 1, referring to the fictive enunciator, being the least frequent one (28/1,417; 2%). This is the case for both conventional compounds (981 instances in our database) and creative ones (436). We did, however, find robust differences between conventional and creative compounds in the exact percentages. As Tables 1 and 2 show, VC compound type 2, serving to refer to the fictive addressee, constitutes a much lower percentage of the total of conventional compounds (18.6%) than of creative ones (42%). A higher relative percentage of type 2 in new creations with respect to lexicographic is to be expected, because spontaneous one-time creative insults, as in ¡Eres un arruinapueblos! (‘You are a town-ruiner!’), which are frequent in this type, do not usually end up represented in lexicographic sources.

Table 1 Conventional Spanish VC compounds.

Table 2 Creative Spanish VC compounds.

As for structures that are commonly regarded as not belonging to or not prototypical of nominal compounding (those containing personal pronouns, determiners, etc.), our database confirms that these are indeed the least frequently occurring ones. Contrary to what the available literature assumes, however, the total percentage of peripheral forms (177+32/1417 = 14.7%) is high enough to deserve the attention of linguists. Tables 1 and 2 further show that the percentage of non-prototypical structures is much lower among creative than among conventionalized compounds (7.3% vs. 18%). This may be due to the fact that novel compounds have not undergone grammaticalization. It is striking, however, that both conventional and creative VC compounds of type 1, referring to the fictive addresser, display many more non-prototypical forms than prototypical ones. There indeed seems to be no systematic generic structure for this type, although we did find a predominance of structures with pronouns or nominal phrases used as subjects or vocatives as well as coordinate and subordinate clauses. Type 2 VC compounds, referring to the fictive addressee, show a larger variety of non-prototypicality, but with better defined groups, namely structures with reduplicated verbs, V+PP combinations, and constructions with pronouns. Type 3, referring to the fictive topic, also contains numerous non-prototypical elements and structures, despite being the unmarked and thus most prototypical category of Spanish VC compounds. These are verb structures with an adjective or adverb, transitivized verbs, occurrences with a noun interpretable as a prepositional phrase, parasynthetics and locutions, and structures with quantifiers and complements with determinants. Regarding compounds involving adverbs in complement position, they are rare when referring to the topic of the fictive conversation (approximately 10%) but represent nearly 20% of those referring to the fictive addressee and more than 50% of those referring to the fictive enunciator.

7. Discussion

On the basis of our self-compiled database, we conclude that Spanish VC compounds constitute a grammaticalized schema, especially in its most productive subtypes (i.e. when referring to the fictive topic or the fictive addressee of the imagined conversation expressed in the compound itself). That is, we regard Spanish VC compounds as lexically stored form-meaning pairs in the sense of Construction Grammar models (see Pascual 2014: 115–140 and Sandler & Pascual 2019 for a similar analysis of other perspective-indexing structures), and particularly the recent proposal by Jackendoff & Audring (2020).

As our database shows, VC compounds in general and VN ones in particular are maximally productive in contemporary Spanish, giving rise to numerous new words that are often yet to be included in lexicographic sources. According to Jackendoff & Audring (2020), there is much morphological regularity in the lexicon (e.g. number in idioms, as in ‘raining cats and dogs’), which must be stored and thus cannot be accounted for by ordinary and exceptionless syntactic principles.

Constructionist models account for such regularities by postulating for grammatical schemas instead of rules. Schemas allow for different degrees of specificity and grammaticalness, ranging from word formation schemas (the most grammaticalized lexical units) to individual instances. This approach also allows for intermediate realizations and numerous variables, which are lexically interconnected with each other. This helps explain why even the most peripheral compounds do not behave idiosyncratically. Hence, we disagree with their categorization as random lexical innovations in the new grammar of the Spanish language (Real Academia Española & Asociación de Academias de la Lengua Española 2009: 779–782) and Val Álvaro’s (1999: 4837) claim that ‘there is no general structure that constitutes a common denominator of constructions, such as besalamano ([kisses+the+hand], ‘short note’), bienmesabe ([well+me+it+tastes], ‘sweet’), and nomeolvides ([not+me+forget-2.SG], ‘forget-me-not’)’ (our translation and glosses). Under our account, these – as well as prototypical VN compounds such as pintalabios ([paints+lips], ‘lipstick’) – are all grammaticalized fictive enunciations, that is, they share an inflected skeletal meta-schema structure, involving perspective information as in ordinary reported speech. The schematic formal pattern of the ‘skeletal meta-schema’ is [Vx + Cx (X)], its prototypical form being [V3-sg+N-pl]. Our database shows evidence that non-prototypical VC compounds involving determiners or pronouns are in fact very common in the subtype referring to the fictive enunciator (the least entrenched pattern in Spanish compounding), not rare in VC compounds referring to the fictive addressee or even in those referring to the topic of the fictive conversation (i.e. the most abstract, productive, and grammaticalized subtype of VN compounds).

A schema-based account of productive morphology can further shed light on the fact that a particular pattern may be productive in a language with an unproductive structure, and vice versa, as discussed in Section 2. VN compounds are the most productive compounds in Spanish, a language with very restrictive compounding. In contrast, their English counterparts, like pickpocket, killjoy, turnkey, and turncoat, are extremely rare (Kageyama 2009: 818), while English does show very productive word formation. The adopted approach can also account for the apparent heterogeneity of VC compounds structures among Romance languages. As discussed in Section 4, Italian and Spanish, but also French, differ historically regarding the source of the verbal stem. This difference is probably a result of the mere arbitrariness of how each language has evolved (Rainer 2001), leading to the one illocutionary force (imperative or declarative) becoming prototypical for VC compounds in one language and another illocutionary force being allowed but more peripheral and therefore less productive. We further hope to have shown that it is mostly semantic factors that can account for the formal variation in the verbal component.

Our data also support Jackendoff & Audring’s (2020) departing hypothesis concerning the need to eliminate clearcut boundaries between different grammatical components, that is, the claim for a continuum between morphology and syntax and even discourse. This is also the accepted view in Cognitive Linguistics (Langacker 1987: 18–19, 1991: 511–525), which leads to a treatment of grammatical embedding as gradual (cf. Matthiessen & Thompson 1988). Indeed, fictive interaction constructions emerge from discourse, or rather situated talk-in-interaction, but appear manifested at all grammatical levels, as outlined in Section 5.

The diversity within the VC compound pattern itself is also worth noting. The loss of formal contrasts relevant in other areas of grammar within the context of the family of fictive interaction compounds shows the hallmarks of grammaticalization. In order to fit a word-like template, Spanish VC compounds have undergone reduction in the phonological and functional structure of the enunciation we claim they emerge from. Phonological structure reduction in VC compounds involves (cf. Santana et al. 2013: 81): (i) reduction of the weak vowel in a diphthong (e.g. fregasuelos [freγa’swelos] from friegasuelos [frjeγa’swelos], ‘floor mop’], (ii) reduction of conjoint vowels in pronunciation and spelling (e.g. tapagujeros [tapaγu’xeros] from tapaagujeros [tapaaγu’xeros]), (iii) dropping of the unstressed vowel next to another vowel in pronunciation and spelling (e.g. abrojos [a’βroxos] from abreojos [aβre’oxos], ‘thistle’), or (iv) dropping of a consonant in a consonant cluster (e.g. guardapero [gwarða’pero] from guardaapero [gwarðaa’pero], ‘boy who brings basic supplies to reapers or mowers’). The existence of VC compounds without phonological reduction (e.g. cagaaceite [kaγaa’θejte], ‘missel thrush’), even as alternatives to phonologically reduced ones (e.g. cagaceite [kaγa’θejte]), further shows an ongoing process of grammaticalization and lexicalization from a discourse structure to a compound word.

Grammaticalization in the prototypical and most frequent types of Spanish VC compounds is also evident in ellipsis of the original article or even preposition in the phrase they derive from. Examples still displaying the original determiner, such as vivalavirgen and besalamano, are very infrequent, and they very closely reflect the enunciation they originate from. This shows the loss of functional material – the hallmark of grammaticalization – in Spanish compounds (cf. Buenafuentes 2007), due to the constraints on complex specifiers examined in Section 3. Another piece of evidence concerns the unstable character of phonological reduction resulting in ellipsis of a coordinator in pairs such as quitaipón vs. quitapón (from [quita+y+pon] [get out+and+put in], ‘removable’). The coordinator is lost in numerous compounds, such as arrancasiega ([starts+mow], ‘poor grain half mowed and half pulled up’); duermevela [sleep(s)+hold(s)-awake], ‘light sleep’; alzapón ([lift(s)+put], ‘front opening in pants or pants with such an opening’); and callacuece ([shut-up+cook], ‘hypocrite’). Lastly, VC compounds also display reduction of functional structure, for instance in idiomatic phrases (6.2.2), as in aguafiestas ([spoil+parties’], ‘spoilsport’) from ‘aguar la fiesta’ (‘to spoil the party’) and buscavida(s) ([‘looks-for+life’], ‘self-starter’) from ‘buscarse la vida’ (‘to fend for yourself’). Another case concerns compounds referring to the topic of the fictive conversation, resulting from the formal reduction of relative phrases (6.3), in which the relative pronoun is also lost. In English such processes are unnecessary, because Germanic languages lack a restriction on complex modifiers and can thus create fictive interaction compounds without grammaticalization.

8. Conclusions

In this paper, we argued that all Spanish VC compounds are viewpointed and can be parsimoniously accounted for as involving a grammaticalized fictive interaction construction (Pascual 2006, 2014). We claim that Spanish VN compounds have emerged from non-actual, ontologically conceptual enunciations in which the referents are most typically the fictive enunciator, the fictive addressee, or the conversational topic. We further sustain that these different semantics are what results in the seemingly paradoxical formal variety of Spanish VC compounds.

We argue for a fundamental role of fictive interaction in word formation, specifically in the creation of VC compounds as well as numerous other nominal compounds and structurally simple lexical items like ‘vosear’ ([you-inf], ‘address somebody with the second-person pronoun ‘vos’); ‘pordiosero’ ([for+God+er], ‘beggar’); and ‘recibí’ ([I-received], receipt with the message ‘recibí’ written on it). Hence, Spanish does seem to have productive viewpointed compounds, just like English and other Germanic languages. These languages only differ in how they are grammatically expressed. Moreover, this great morphological variation, both inter- and intralinguistically, supports the view that these formal structures are variants of a family of form-meaning pairs defined functionally, namely fictive interaction compounds.

We hope to have shown that the fundamentally interactional dimension of language is reflected in its very structure down to the lexical level. Hence, shared intentionality and intersubjectivity, the presumed common denominator underlying the human communication potential (Enfield & Levinson 2006; Enfield 2008), is also at the very core of language structure and use (cf. Verhagen 2005; Zlatev et al. 2008). This view has far-reaching theoretical implications. If grammatical embedding, as in VN compounds, is gradual, ultimately emerging from sequential turn-taking, then this suggests that the structure of grammar primarily reflects its mode of usage rather than some context-independent, sui generis linguistic pattern. Furthermore, our approach contradicts most current morphological and semantic theories today, which are largely monologic, adopting a referential view of word formation and its semantic processes as primarily emerging from arbitrary signifier-signified relations (see overview in Sandler 2016). While it is undeniable that such symbolic relations are fundamental in most probably all natural languages of the world and thus also part and parcel of word formation, we sustain that this process alone cannot account for all linguistic and morphologic phenomena. We claim that a dialogic view, in which language is not just accounted for by denotation and association but also by demonstration (cf. Clark & Gerrig 1990; Clark 2016; Ferrara & Hodge 2018), can help elucidate a large number of understudied phenomena and throw new light on old but poorly understood ones, like VC compounds with inflectional information. Indeed, intersubjectivity seems central to referential semantics, including morphological semantics.