2.1 Lexical-Realizational Functional Grammar

LFG is a constraint-based grammar framework that seeks to model the linguistic knowledge of the speaker of a given language as well as the way in which this knowledge is put to use in production and perception (Kaplan & Bresnan 1982). The architecture of LFG posits separate levels of linguistic representation which are connected via explicitly formulated mapping relations (Dalrymple & Lowe & Mycock 2019, § 7.2). Levels included in current versions of LFG are not only constituent structure (c-structure) and functional structure (f-structure), but also prosody, semantics (s-structure), information structure and argument structure, as illustrated in Figure 1.

Figure 1

Correspondence architecture of Lexical-Realizational Functional Grammar. Visualization by the author, following Asudeh & Bögel & Siddiqi (2023).

In the present article, relevant modules are, on the syntactic side, constituent structure and functional structure, on the phonological side, prosodic structure and the phonological string, and vocabulary structure as an interface mediating the syntactic and the phonological sides. Relations between modules are described as functions from (parts of) representations in one module to (parts of) representations in another module. In principle, all modules operate in parallel. Accordingly, the arrows in Figure 1 do not entail relations of succession or precedence. Rather, representations in different modules constrain each other in systematic, yet selective ways, where the exact nature of the constraints governing the relationship between the levels continues to be a matter of ongoing empirical research.

As to morphology or word-internal structure and its interaction with syntax, only few explicit proposals have been made in the past decades (for a review see Dalrymple 2015; Sadler & Nordlinger 2018; Asudeh & Siddiqi 2023). A recent development is Lexical-Realizational Functional Grammar (LRFG), combining the formal explicitness of Lexical-Functional Grammar (LFG) with the realizational, morpheme-based approach of Distributed Morphology (Melchin & Asudeh & Siddiqi 2020; Everdell et al. 2021; Asudeh et al. 2023); cf. Asudeh & Siddiqi (2023) for a brief introduction. Similarly to DM, LRFG assumes that the terminal elements of constituent structure (c-structure) are morphosyntactically relevant features rather than words. c-Structure is mapped onto phonological form and onto functional structure (f-structure) and semantic structure (s-structure), respectively. Key aspects of LRFG for the analysis presented here include the distinction between constituency and grammatical function in the representation of negation, and the capacity to model the competition between various negation expressions, such as Spanish no and negative prefixes like des- (‘dis-’) or in- (‘in-, un-’), through constraints akin to DM’s subset principle (Halle 1997; Embick & Noyer 2007: 298), termed in LRFG MostInformative (Asudeh & Siddiqi 2023: 890) and MostSpecific (Asudeh et al. 2023: 32); see Section 7.3.

In the approach adopted here, verbalization of a piece of meaning and structure, i.e., lexical insertion from a procedural point of view, is conceived of as a mapping or function, termed the ν-function: An exponendum, or something that is to be verbalized, is mapped onto an exponent, i.e., a structure by means of which the exponendum is verbalized (Asudeh et al. 20231; Asudeh & Siddiqi 2023). The ν-function is of the general form illustrated in (5a) (Asudeh et al. 2023: 23). A partial description of the ν-function for Spanish no, discussed in more detail in Section 7, is given in (5b).

The exponendum is thus a pair consisting of a list which specifies the syntactic distribution of the item in question and of a set of descriptions of its function and meaning. Simplifying matters somewhat, one might say that the complex representation in (5), which comprises information about distribution, meaning and phonological form, is what is denoted a lexeme in certain other approaches.

2.2 Spanish prefixes

The present study considers Spanish no in contexts where it competes with negative affixes such as Spanish des- ‘dis-’ or in- ‘in-/un-’. In the descriptive literature no has likewise been classified as a prefix in some of its uses. The terms affix and prefix shall therefore briefly be discussed here. By affix, I refer to an element that is functional rather than lexical, and that is bound rather than free. Following Haspelmath (2021), a third criterion could be added, that of being restricted to a specific class of roots. Let us briefly consider the first two criteria, ignoring the third. With respect to the dichotomy between lexical and functional morphemes (Lyons 1968: 435–438), negation in general and Spanish no more particularly appear as a prototypical instance of a functional element. On the semantic side, no has a grammatical function rather than a denotation; it is an operator that (at least in some cases) reverses the truth value of its complement. On the formal side, in most, if not all, languages negation is expressed by only a small and closed set of negative exponents (Zeijlstra 2007).

With respect to the second criterion, matters are more complex. A bound element can be defined as an element that is in some way prosodically integrated into the phonological domain of the host to which it is attached. Relevant phonological domains for the definition of boundedness are, e.g., domains of phonological processes such as stress assignment or featural harmony processes (Hildebrandt 2014). A (language-specifically defined) domain onto which various such processes converge is the phonological word (Dixon & Aikhenvald 2020). In Spanish, the phonological word is the domain of stress assignment (Harris 1983), and thus in Spanish the bound nature of a formative is dependent upon whether it is part of this domain – in which case it is bound – or not. In contrast to an affix, a free form is one which is not prosodically integrated into some prosodic host; it forms a separate phonological word. Between affix and free forms there is a third category of clitic. Clitics are forms which are phonologically dependent on a host, e.g. with respect to segmental processes such as vowel deletion or consonant insertion, without being part of the host’s domain of stress assignment.

As to Spanish, many suffixes are clearly bound, as they interact with stress assignment. They can be the locus of primary word stress, (6b), or have a prestressing effect, (6c).

Spanish prefixes, in contrast, do not tend to have any effect on stress assignment to their base. Based on this observation and on observations about the syllabification of prefixed words, it has been argued that at least some prefixes in Spanish (Elordieta 2014) and other Romance languages (Peperkamp 1995; Nespor & Vogel 2007: 125) lie outside of the phonological domain of the root they precede. Yet it is also clear that with respect to phonological integration not all prefixes pattern alike. First, some Spanish prefixes are phonologically integrated with their bases (RAE-ASALE 2009a: §10.3c). This is the case for the Spanish negative affix in- ‘un-, in-’, which surfaces in different phonological forms, reflected in different orthographic representations, depending on the onset segment of the base, (7).

Other Spanish prefixes, in contrast, are less phonologically coherent. To give an example, the prefix sub- ‘under’ forms its own syllabification domain (RAE-ASALE 2009a: §1.7a), as in (8), blocking resyllabification, and more precisely onset maximization, across the boundary between prefix and root.

Spanish no, however, shows no phonological integration with a following element. In descriptions of Spanish word structure, a third category of “separable prefixes” has been postulated, and Spanish no is sometimes included in this class (Rainer 1993: 350; Montero Curiel 1999; Varela Ortega & Martín García 1999; RAE-ASALE 2009b: § 48.2p; Gibert-Sotelo 2021). Under the term “separable prefix”, the Nueva gramática de la lengua española (RAE-ASALE 2009a: § 10.2f) mentions forms like ex ‘ex’, pro ‘pro’, sin ‘without’, medio ‘half’ or cuasi/casi ‘almost’ (cf. also Felíu Arquiola 2018; 2019), (9).

As evident from (9), separable prefixes can be orthographically separated from their base by a space; occasionally they are hyphenated, or written together with the base (RAE-ASALE 2009a: § 10.3i). no, in contrast, is always separated by a space. Some separable prefixes are formally identical to adverbs (casi/cuasi ‘almost’) or prepositions (sin ‘without’), others occur exclusively to the left of a lexical head (ex- ‘ex’, pro ‘pro’). In this study, I use the term modifier of X°, and more particularly modifier of N°, rather than “separable prefix”, see Section 7.2. Note that in contrast to adverbs, preverbal separable prefixes, i.e., modifiers of V°, as in (9b), occur closer to a verbal root than preverbal clitic pronouns. Since here I focus on no preceding nouns (and adjectives), this criterion is not applicable.

While the Nueva gramática de la lengua española is not explicit with respect to the precise distributional criteria relevant for the distinction between separable prefixes and other prefixes, the possibility of being stranded in ellipsis in coordination structures appears to be one criterion; for a discussion of this and other relevant criteria cf. Felíu Arquiola (2018; 2019). As shown by Felíu Arquiola (2018; 2019), Spanish separable prefixes can occur as stranded elements in constructions with ellipsis under coordination, in which case the prefixal element bears contrastive stress, (10). Other Spanish prefixes, in contrast, cannot surface as remnants in ellipsis under coordination, (11).

I will not expand further on this, but it should have become clear that if the definition of affixes as bound elements is to be maintained, Spanish “separable prefixes” cannot be affixes. Furthermore, certain other prefixes, such as sub- ‘under-’, (8), or post- ‘post’, (10a), are likewise not affixes in this sense, as they are not prosodically bound (cf. Fábregas & Krämer 2020 for a critical discussion of the concept of prefix in prosodic and structural respects). In Spanish, at least two types of ‘prefixal’ formatives need to be distinguished, contrasting with respect to the phonological integration with their base. As shown below, Spanish no preceding nouns patterns with certain “separable prefixes” like sin ‘without’ or ex ‘ex’: Like these, it functions as a modifier of a word, i.e., of an N°. Clearly, the precise phonological and morphological properties of different classes of Spanish prefix-like elements remain to be studied more in detail, which goes beyond the scope of this article.

2.3 The typology of negation constructions

As shown in (1) to (4) above, negative markers can differ in their scope as well as in their form. The element over which a negative marker has scope can be smaller or larger, stretching from lexical roots, (2), to finite projections, (1). As to form, different types of formatives are attested in the languages of the world: In Spanish, besides the free form no, (1), one finds affixes such as des-, (2). In languages like Turkish, negative affixes attach to the verb and have scope over the entire sentence, (12).

In the research literature, types of negation have been described by different terminologies with varying denotations, including terms like “sentential negation” (Jackendoff 1969 and much subsequent work), “syntactic negation” (Dahl 1979), “lexical negation” (e.g., Andreou 2015). In the typological literature the terms “morphological negation” and “affixal negation” are often used to refer to morphologically dependent negative markers (i.e., affixes) that have scope over the finite verb (e.g., Dahl 2010), (12). Other researchers use “morphological” and “affixal” negation (e.g., Zimmer 1964; Gyurko 1971; Joshi 2020) as synonymous with “derivational negation” (Miestamo 2017: 427), exemplified by Spanish prefixal negation as in in (2) above.

Given the inconsistent use of the terms morphological negation and affixal negation in the research literature, this study tries to avoid them and employs the terms syntactic negation and lexical negation. Both will be used as mainly descriptive concepts, denoting two categories which are based on (partly overlapping) properties – to be studied here for Spanish no – and thus not strictly distinct. In what follows, the term syntactic negation is used to refer to instances of clausal or sentential negation (Miestamo 2017), in which the negative expression has scope over the sentence, as in (1) or (12) above, as well as to instances of constituent negation (Klima 1964), where negation has scope over a maximal projection such as an AP, as in (21a) below, or a DP or PP; following the terminology in Bond (2023). The term lexical negation, in contrast, is used to refer to expressions in which negation has scope over a word, such as no in (4b) above, or over a part of a word, cf. (2) above or (17) below, as in the case of Spanish prefixes like des- or in- or English un-.

The point of departure of the present study is the observation that Spanish no occurs in instances of syntactic negation as well as of lexical negation. Unlike Spanish, English does not allow this syncretism; cf. De Clercq (2020: 35) on English not vs. non-. In English, one exponent of syntactic negation is not, whereas one exponent of lexical negation is non-, separated from its base by a hyphen.

As shown by De Clercq (2020), who develops a systematic typology of the expression of negation that goes beyond the dichotomy between syntactic and lexical negation, languages differ as to whether they use the same or different negative markers to lexicalize instances of syntactic and lexical negation. De Clercq (2020) distinguishes four types of negating structures, which are all verbalized differently in English, i.e., by the negative markers n’t, not, non- and un-/in-/dis-, respectively (De Clercq 2020: 59). Adopting a typological approach, De Clercq shows different patterns of syncretism that languages present in the lexicalization of the four types of negation structures, but Spanish is not part of the sample considered by that study.

The analysis presented here is compatible with the fourfold typology of De Clerq (cf. De Clercq 2020, Chapter 3), who distinguishes between T^Neg-markers (English n’t), corresponding roughly to what is termed sentential negation here, Foc^Neg-markers (English not), corresponding roughly to what is considered here as (certain kinds of) constituent negation, Class^Neg-markers (English non-), taking scope over a classifier phrase, and Q^Neg-markers (English prefixes un- and dis-), taking scope over a projection that hosts quantifiers, where the latter two types correspond roughly to what is termed lexical negation here. I will not dive into de Clercq’s precise characterization of Class^Neg-markers and Q^Neg-markers, but rather focus on the descriptive question of whether Spanish no, besides expressing constituent negation with nouns and adjectives (thus functioning as a Foc^Neg-marker), can express lexical negation of the two types corresponding to, on the one hand, the English affix non-, a Class^Neg-marker in De Clercq’s typology, and, on the other hand, to Spanish affixes like des- and in-, which arguably correspond to Q^Neg-markers in that typology; cf. Fábregas (2023: 53) for a more nuanced view with respect to Spanish in-.

4.1 no with intersective adjectives

A common case of negation with no preceding adjectives is that of contradictory negation. This can be illustrated particularly clearly with adjectives that express class membership, i.e., adjectives that have an intersective interpretation, (19).

Here, the negated adjective restricts the denotation of the nominal predicate: A negated adjective predicated over a noun specifies the intersection of the sets of referents that fall under the nominal and that do not also fall under the adjectival terms.

Among the intersective adjectives that are attested with no in the corpus (esTenTen11, see Section 1) are many derived adjectives, (20). Denominal adjectives such as anual ‘yearly’ express set membership with respect to a set determined (in some way) by the meaning of the base noun año ‘year’, and their negation express non-membership in that set.

Deverbal adjectives with a passive interpretation are derived by the suffix -able/-ible, (21a), or from past participles, (21b). Since the interpretation of such derived adjectives is – in the aspects relevant here – identical to that of a finite verb, negation with no has, much like sentential negation taking scope over the finite verb, a contradictory interpretation, (21a’–21b’). The same holds for deverbal adjectives in -ivo with an active dispositional meaning, as in (21c–21c’).

4.2 no with qualifying adjectives

Most occurrences of no preceding attributive adjectives are with relational adjectives, as illustrated in the preceding section. With qualifying and scalar adjectives such as bueno ‘good’, malo ‘bad’, bello ‘beautiful’, feo ‘ugly’, negation with no is little used in the esTenTen11 corpus. In the few relevant cases, the positive and the negated adjectives are contradictory, not contrary antonyms. To give an example, in (22a) the phrase metales no preciosos denotes the set of metals that do not fall within the range of the predicate metales preciosos, i.e., those metals that are not precious metals in the technical sense. While the adjective precioso ‘valuable’ is scalar in some of its readings, cf. (22b), in which precioso refers to a scale of (beach) quality, and more particularly to a very high value on such a scale, the only sense attested in the corpus with the negated form no precioso appears to be the technical sense, referring to metals that are highly resistant to chemical influences, such as gold or palladium. The negated adjective term no precioso ‘not precious’ thus is genuinely intersective, referring to a technically defined class of metals.

Other examples for no preceding relevant adjectives are given in (23). In both examples, the reading of the negated adjective appears to be intersective rather than genuinely qualifying, expressing non-membership in the set referred to by the adjective: In (23a), una zona no buena ‘a not good area’ refers to an area which does not belong to the set of good neighbourhoods, while (23b) states that the set of places in Costa Rica which do not also belong to the set of beautiful places is very small.

A similar observation is made for English non- by Lieber (2004: 121), and French non- by Dugas (2017: Chapter 2); both elements tend to have contradictory readings, as opposed to the prefix in- with adjectives. Fábregas (2023) points out that Spanish in-, in contrast to no, combines only with gradable adjectives. English non-, tending towards a contradictory reading, contrasts with English un-, which does not show such a preference (Jespersen 1917; Horn 2005).

In the few cases in which no precedes adjectives with a genuinely qualifying semantics, the interpretation is one of contrariety (cf. also Varela Ortega & Martín García 1999: 5021), (24), implying neither one nor the other extreme of the scale denoted by a pair of adjectives holds true.

4.3 Summary

On the semantic side, adjectives negated by no commonly exhibit contradictory negation, whereas a contrary interpretation is seldom attested, aligning with the interpretation found for sentential, i.e., syntactic, negation; cf. De Clercq (2020: 54–56) for English not, non- and un-. On the syntactic side, there is likewise no reason to consider no preceding adjectives as different from constituent negation. Regarding its syntactic placement within the adjective phrase, no has the same distribution as other (negative) adverbs and quantifiers, as in (25).

In sum, no preceding adjectives can be analyzed as a negative adverb, both with respect to its interpretation and its distribution. In the following discussion, Spanish no preceding adjectives will thus play only a minor role.

5.1 Nouns with intersective interpretation

Expressions with no preceding a noun often have an intersective interpretation, in which no expresses contradiction, (26), similar to sentential negation or negation of adjectival constituents.

(26) is an instance of contradiction: The negated noun and its base stand in a relation of complementary antonomy, in that an object either is a member of an association or is a member of the disjoint set of non-members. Other nouns with a contradictory interpretation under negation are nouns – often deadjectival – denoting membership in a group, such as a nation or religion (e.g., los no católicos ‘the non-catholics’, los no indígenas ‘the non-indigenous’) or function (los no propietarios ‘the non-owners’, los no huéspedes ‘the non-guests’). In a study on French, Dugas (2014) therefore calls this type ‘complementary negation’, and Andreou (2017) subsumes the corresponding English forms under the term ‘general negative reading’.

Other nouns for which an intersective interpretation is commonly observed are deverbal agent nouns, (27). Deverbal agent nouns, like deverbal adjectives, are similar to verbs in those respects which are relevant to the interpretation of negation. In (27), the deverbal agent noun fumador ‘smoker’, derived from the base verb fumar ‘smoke’, refers to the set of entities with respect to which a proposition ‘x smokes’ containing the base verb either holds (‘those who smoke’) or does not hold true (‘those who do not smoke’). The interpretation of the noun can, just like that of the verb, either be episodic or dispositional.

5.2 Deverbal event nouns

Non-intersective nouns that combine freely with no can be classified into three types: first, derived nouns with event interpretation, discussed in this section, second, derived nouns that denote states or properties, and third, nouns referring to concrete or abstract objects; see the next two sections. The corpus contains many occurrences of derived nouns with an event interpretation, (28–29).

In these examples, the NP containing the negated noun refers to a situation in which an event of the kind denoted by the base does not occur. The DP may have episodic reference, as in (28), where it refers to the fact that the government did not disclose important evidence, or it may be non-episodic, as in (29), where it refers generically to a set of situations which is not instantiated (here, a potential occurrence of an event of someone not complying with the rules). A similar interpretation of no is found with nominalized infinitives, (30).

I will not address the question of how to formally account for the interpretation of negation with event-denoting expressions; cf. Fábregas & González Rodríguez (2020) for a recent proposal. What is relevant here is that, in the case of no preceding deverbal nouns with eventive reading, negation has scope over the predication – which in sentential negation corresponds to a finite verb; whereas in the above examples it is lexicalized by a nominalization – and the interpretation is one of contradiction, (31).

Should no preceding deverbal nominalizations thus be classified as contradictory negation, just like sentential negation, as proposed by Varela Ortega et al. (1999: 5021), or is it of a different nature? In work on the French negative prefix non, corresponding in some respects to the Spanish form no under consideration here, Dugas (2014; 2017) refers to this interpretation of non-occurrence as the ‘ontological’ reading, in which case the negated form denotes the non-occurrence of the denoted event or absence of an entity (Dugas 2017: 112). This interpretation is in line with the proposal made by Fábregas & González Rodríguez (2020), who claim that expressions interpreted as an event that does not occur refer to inhibited eventualities. According to Dugas, this interpretation is different from that of prototypical sentential negation, which is contradictory, and comparable to a reading of non-existence (hence, “ontological” negation), absence or privation, which I consider in the following section.

5.3 Derived nouns that denote a state or property

When no precedes a derived – deverbal or deadjectival – noun that denotes a state or property, the interpretation is that of a situation in which the state or property does not hold. In the corpus, no occurs frequently with derived nouns ending in -ancia/-encia, (32), derived from stative verbs and adjectives (RAE-ASALE 2009a: §5.11a), and -idad/-edad, (33), derived from adjectives (RAE-ASALE 2009a: §6.1a).

In these examples, the negated noun no asistencia ‘non-attendance’ denotes the absence of the state of attendance on the part of the participants in the course, and the negated noun no necesidad ‘non-necessariness’ denotes the absence of the property of being necessary with respect to didactic training. Thus, the interpretation is one of privation, denoting a condition of something lacking a state or property. For such expressions the term ‘ontological reading’, i.e., a reading of negation that relates to the existence of an entity – which is negated – as introduced by Dugas (2017: 112), appears indeed appropriate. Even clearer cases of a privative reading can be seen in combinations of no with concrete or abstract non-derived nouns, as shown in the following section.

5.4 Nouns referring to concrete or abstract objects

In the corpus under study here, no appears in few examples with object-denoting nouns. Expressions of this kind often have a rhetorical effect and a particular stylistic markedness. In most cases the interpretation is privative: The object denoted by the noun is absent. An example is given in (34), where the apposition of no lugar ‘non-space’ and falta de lugar ‘lack of a space’ clearly suggests that the intended interpretation is privative.

The occurrence of no with abstract nouns like espacio ‘space’ or tiempo ‘time’ is much more frequent than with concrete nouns. Space or time are attributes that are inherent to many things, which makes it easy to assign a privative interpretation to the combination of such nouns with no, as in (35).

The presence of most concrete objects, in contrast, is accidental: They can be present or absent, hence the marked effect which arises if such nouns are combined with no. Two examples are given in (36–37). Again, what is denoted by the negated noun appears absent, despite being expected by convention to be present.

These examples are similar to a famous example discussed by Horn (2005), “un-hit of the week” (Horn 2005: 340), referring to a situation in which a hit almost happens; it is expected to occur, but then does not. According to Horn, an “unX is not structurally a member of the category X, but it shares a significant functional status with Xs” (Horn 2005: 340).

5.5 Stereotype negation

In the literature on lexical negation in English, and more recently in French, another interpretation is discussed, illustrated with an English example in (38).

Andreou (2017) labels this interpretation as “stereotype negation”, a term credited to work by Bauer & Lieber & Plag (2013). In stereotype negation, the referent of the negated noun lacks some of the qualities which stereotypical referents of that noun usually have (Andreou 2017: 1). The term “stereotype” is used here in the sense of Putnam (1975: 150) as the “idea of the characteristics of a normal member of the kind” denoted by a term. This idea is not new, though. Horn (2005: 343), in his analysis of the 1970s Seven Up Uncola advertising campaign, describes this reading as referring to “a bad example of the category, lacking a functionally significant property” of the referents of the base noun. As to French, this reading is also mentioned by Dugas (2015), (39).

In Spanish, this qualifying interpretation appears marginal at best with no and is not easy to find in the corpus. To be sure, it is not possible to gain negative evidence from a corpus study, but it is however notable that the Spanish negated noun no cola, in some respects comparable to the English expression uncola cited by Horn as an example for this reading, is indeed attested a couple of times in the spTenTen11 corpus, but not with an interpretation of stereotype negation, but with an intersective interpretation, expressing non-membership in the class of colas, (40), i.e., with a contradictory interpretation.

According to (40), the class of carbonated soft drinks consists of two subclasses: the class of colas and the class of those drinks which are not colas. The expression no cola presupposes the class of colas in some way as the stereotypical exemplar of the class in its entirety, since all other exemplars are referred to as those which are not instances of the stereotype. However, the negative connotation found with English or French negated nouns, as in (38–39), appears to be absent here.

Rare are examples in which a negative connotation is clearly visible, and where the interpretation seems to be that of a bad example of a category, (41–43).

One can only speculate why the interpretation of stereotype negation appears to be less frequently attested with Spanish no than with English and French non-. If we take it that syntax is the realm of compositional semantics, while non-transparent or context-dependent meanings are found more often for lexical items than for syntactic phrases – possibly because shorter expressions have a higher probability of acquiring some sort of stable mental representation, i.e., of being stored –, this fact may indeed be taken as an indication that Spanish no, even when occurring within the noun phrase, is more closely tied to syntactic negation than its French and English prefixal counterpart non-.

5.6 Summary

The preceding discussion has shown that negation of nouns with intersective meaning tends to be contradictory, expressing that the referent of the DP is not a member of the set denoted by the noun. Conversely, when dealing with nouns referring to events, states, properties, or objects, the interpretation leans towards privation: the denoted eventuality, property, or object is absent, or its existence is negated. Lastly, stereotype negation – a non-compositional and non-transparent kind of interpretation – is relatively rare, though attested in Spanish. Assuming that contradiction is the interpretation of sentential negation, whereas privation is more characteristic of lexical negation (Horn & Wansing 2022), Spanish no preceding nouns shares semantic similarities with syntactic negation, since a contradictory interpretation predominates in these contexts. Yet, as it entails a sense of privation in many other cases, it also shares similarities with lexical negation.

6.1 Spanish no as a building block of words

Let us first consider the position of no with respect to “separable prefixes”, termed modifiers of N° here (Section 2.2). In general, prepositional N° modifiers like sin ‘without’ or pro ‘pro’ tend to be closer to the base than adverbial N° modifiers, such as casi/cuasi ‘almost’ (Varela Ortega & Martín García 1999: 5005), but sequences of two adjacent adverbial N° modifiers are attested in Spanish (Varela Ortega & Martín García 1999: 5005). Two relevant forms which are listed in the Diccionario de la lengua española (Real Academia Española 2014) as “prefixes” are ex (‘ex’) and cuasi (‘almost’), and expressions in which they precede no are sporadically attested in the esTenTen11 corpus, (45–46). More occurrences were found with cuasi than ex preceding no, despite the latter having a higher absolute number of occurrences.

With respect to its distribution, no thus appears to pattern with other modifiers of N°, many of which – but not all – are homonymous with adverbs or prepositions. In the same position certain other adverbs can occur, such as también ‘also’, entonces ‘then’ and a few others, (47). According to the Nueva gramática de la lengua española (RAE-ASALE 2009b: § 13.8i), even certain morphologically derived adverbs like actualmente ‘currently’ are acceptable in this position, at least with a small class of rôle-denoting nouns like presidenta ‘president’.

Let us briefly note that no can also occur internal to sequences that are considered compounds by some scholars, such as [ N N ] expressions, which have compound status according to Varela Ortega (2012), (48), or [ N de N ] expressions, treated as compounds by Kornfeld (2011), (49). In (48), the non-head nouns have an intersective reading and the reading of no is that of contradiction. The structure in (49), a type which is often found with eventive nouns, in contrast, has a reading of non-occurrence, which above was classified as privation.

In this respect, the distribution of no is again the same as that of other modifiers, as shown in (50–51).

Finally, in some expressions no and the base noun form a unit which, as a whole, serves as the semantic base for suffixation (RAE-ASALE 2009b: § 48.2s), (52). In other words, the derivational suffix that is attached to the word that follows no has scope over the complex expression built by no and the root, as in (52b) and as in the English gloss of (52a), rather than the other way around, as in (52c).

Hence, as the data in this section show, no can be closer to a base than a “separable prefix” or N° modifier, it can occur within compound words and it can be in the scope of a derivational affix, indicating that no interacts with other building blocks of words.

6.2 The interaction with negative polarity items

At the same time, even when no occurs as what can arguably be considered a building block of words, it also interacts with phrasal syntax, as is visible in the fact that it can license negative polarity items (NPI). In this respect, no contrasts with Spanish negative prefixes, (53).

As pointed out for English by Williams (2007: 19), negative elements which are parts of words, such as un- in unhappy, have scope over their hosts as well as over constituents that are part of the argument structure of their host, while negative elements which are parts of constituents larger than a word, such as sentential negation, can have scope over larger stretches of a sentence. In (53), the NPI ningún occurs in a constituent which is an argument of the noun asistencia, and hence no adheres to Williams’ constraint on the scope of negative elements that are parts of words. More interesting with respect to the question of whether no is a building block of words or of phrases are therefore structures in which the NPI is not contained in the argument structure of the host of no. Such expressions can indeed be found in the corpus, (54–55).

In (54), the PP containing the NPI en ningua de sus manifestaciones is (arguably) not an argument of the negated noun, but an adjunct to that noun, (55 has a similar structure). While further and more systematic research appears to be called for on negative concord with different types of negative prefixes and prefix-like elements, in Spanish it seems clear that no contrasts with negative prefixes in this respect. Crucially, Spanish no can license NPIs within a larger syntactic domain than negative prefixes, with its scope extending beyond the arguments of its host.

6.3 Semantic transparency and compositionality

In studies on the demarcation between morphology and syntax, certain differences between the objects of morphology and of syntax have been claimed to exist: Syntactic constructions tend to be semantically transparent and their semantics tends to be compositional, in contrast to the objects of morphology, i.e., complex words, which are said to be sometimes semantically intransparent and tend to have a non-compositional semantic interpretation (cf. Haspelmath 2011: 35–36; Siddiqi 2014 for overviews on previous studies on the topic). To be sure, this does not mean that all objects of morphology are listed and no object of syntax is listed (Williams & Sciullo 1987). In the literature on negation in English or other languages, it has likewise been claimed that sentential negation shows greater semantic transparency than affixal negation (e.g., Horn 1989: 280; Varela Ortega & Martín García 1999: 5021; De Clercq 2020: 52). This is illustrated for Spanish in (56).

As argued in Section 4, the interpretation of no preceding adjectives is contradictory, and only in few cases – if at all – is it one of contrariety; other interpretations of negation seem unattested with adjectives. Accordingly, no with adjectives in (56a–b) tends towards compositional, contradictory interpretations, in contrast to the forms with negative prefixes in (56a–b). Likewise, under negation with no, all readings of the base tend to be preserved, while under affixal negation certain readings of the base may be unavailable, (56c). Under negation by no the base mortal continues to allow both of its readings, as no mortal can mean both ‘not mortal, not destined to die’ and ‘not deadly, not having the potential to cause dead’, but under negation by in- as in inmortal, only the first meaning is available.

It should however be noted that the domains of in- and no are complementary, as in- attaches only to adjectival bases that denote a scale (Fábregas 2023), in contrast to no, which occurs more often with relational adjectives than with gradable adjectives (Section 4). This fact alone may explain, at least to some extent, the differences in semantic transparency found for the two types of negative expressions.

Section 5 has shown that the interpretation of no preceding nouns is often contradictory, but it can also be privative. In this respect, no competes with des-, which – when combined with a nominal base – can likewise have a privative interpretation. As shown by (57), the degree of semantic transparency is comparable for both types of expression.

Hence, with respect to semantic compositionality, a certain contrast between negative prefixes on the one hand and constituent and sentential negation on the other hand, with no preceding nouns somewhere in between, appears to be discernible, but this contrast might be an epiphenomenon of other factors (such as the specific semantic constraints imposed by prefixes like in- and des- on their bases) rather than being indicative of a clear difference between lexical and syntactic negation.

6.4 Summary

In summary, Spanish no preceding nouns exhibits semantic differences from other negative prefixes such as des- or in-, so that expressions with no tend to be semantically more transparent than expressions containing negative prefixes. Likewise, no can occur closer to the base than “separable prefixes”, within what may be construed as complex words, and no is found internal to compound words. Furthermore, no sometimes belongs semantically closer to the root than a derivational suffix. At the same time, the interaction between negation and negative polarity items extends over a larger domain of the phrase for no than for negative prefixes. Spanish no preceding nouns thus demonstrates properties characteristic of both building blocks of words and building blocks of phrases.

7.1 The representation of negation in functional structure

Lexical-Functional Grammar approaches to the representation of negation in f-structure are, according to Przepiórkowski & Patejuk (2015), of two types: Negation can be represented by means of a feature Neg of the Adj(unct) function (e.g., Laczkó 2014; 2015; 2021; Dalrymple et al. 2019: 68; Bögel & Raach 2020), or by means of a binary feature such as ±Neg(ation) or ±Pol(arity) (e.g., Przepiórkowski & Patejuk 2015; Bond 2016). Under the first option, the occurrence of more than one instance of negation with different scopal properties within a phrase can better be accounted for, whereas the second option is better suited for analyzing phenomena of negative concord and of negation expressed by non-concatenative means such as tone (Bond 2023).

Negative prefixes in general as well as negation in Spanish have so far not been analyzed in the framework of LFG or LRFG. In what follows, I sketch a preliminary analysis of Spanish no preceding nouns and adjectives; negative prefixes like des- or in- are not considered in detail. Since Spanish no can co-occur with negative prefixes, giving rise to multiple instances of negation in a phrase, no seems best represented in f-structure as a feature neg(ation) of the function Adj(unct), corresponding to a constituent which is an adjunct to another head or phrase, (58).

The meaning of no, represented here – for the sake of simplicity – as the Pred feature of the f-description in (45), is that of negation, with the value ‘neg’. Its precise interpretation is computed on the level of semantic structure with reference to other f-structure constituents. Thus, the f-description in (58) holds for all instances of negation considered here, regardless of the formal expression by no or by negative prefixes like in- ‘in-, un-’ or des- ‘dis-’. A partial lexical entry for Spanish no, without any distributional information, though, is given in (59), corresponding to the f-description in (58) above.

What (59) means is that Spanish no spells out the semantic Pred feature ‘neg’, which on the level of semantic structure (s-structure) corresponds to the operator of negation. Its precise interpretation is computed in s-structure, not in c-structure or f-structure, depending, among other factors, on the semantic type of its complement, which can be, e.g., an eventuality, a scalar property, or an entity. Whether negation in a phrase containing no is interpreted as contradictory, contrary or privative is thus a matter for semantics. This analytical choice obviates the assumption of different c-structure or f-structure representations for different interpretations of negation. I postulate but a single representation of Spanish no in vocabulary structure, in contrast to De Clercq (2020), who assumes different interpretations for different positions of negation in constituent structure. The next two sections discuss how to represent the distribution of no.

7.2 Constituent structure

Previous sections have shown that no preceding adjectives has a distribution which is similar to that of an adverb, (25). The distribution of no preceding nouns, in contrast, is different from that of most adverbs, but analogous to “separable prefixes”, termed N° modifiers here. As noted in (47) above, in the same position a small number of adverbs can likewise occur. A description of the distribution of no thus requires (at least) two different c-structure rules, one in which no is – like most other adverbs – adjoined to a maximal projection, and another one in which it is – like “separable prefixes” – adjoined to a lexical head.

Complex expressions of the first type, in which no is adjoined to a maximal projection over which it has scope, as in instances of constituent negation, such as (60c), can be generated by the (annotated) c-structure rule in (60a), corresponding to a c-structure tree as exemplified in (60b). The annotation ↓∈(↑Adj) expresses that the constituent of type Adv (referred to by “↓”) has the function of an adjunct of the node to which it is adjoined; this adjunct is part of (“∈”) the set of adjuncts of the nodes by which the node in question (“↓”) is dominated (referred to by “↑”), and XP (here: AP) is the head (hence the annotation “↑=↓”) of the complex expression containing the Adv adjunct.

The c-structure rule in (60a) is a general rule introducing adverbial adjuncts. For the second type of structure considered here, in which no immediately precedes a noun, a different c-structure rule is required. Note that this is also a rule needed on independent grounds, as the same kind of rule will generate structures containing “separable prefixes”, such as cuasi ‘almost’, and adverbs immediately preceding a function noun within the DP, as in (47) above. The relevant c-structure rule is given in (61a), corresponding to the c-structure tree in (61b). This rule generates expressions with no, “separable prefixes” and adverbs like también ‘also’ preceding a noun. The symbol ‘X’ stands for the category of these three types of expressions.

Let us term the category of formatives like cuasi or ex, ‘modifier of N°’ rather than “separable prefix”, as proposed in Section 2.2. Like no, they function as adjuncts. With respect to f-structure, let us further assume that the function of cuasi ‘almost’ is that of a quantificational N° modifier, denoted here by the abbreviation Quant, while ex ‘ex’ is a temporal N° modifier, referred to here by the type Temp in the (partial) lexical entries given in (62a). cuasi or ex, expressing functions like Quant or Temp, are thus analyzed as modifiers that are adjoined to words, i.e., to constituents of the category N°, by the c-structure rule in (62b), which is the revised version of (61a) above.

Since there is no evidence that elements like cuasi, ex or no change the morphosyntactic features of the expression with which they combine, we can assume here that they are adjoined to words rather than to roots. In this respect, they contrast with certain other Spanish prefixes, such as Spanish sobre- ‘over’ in sobre-volar ‘to overfly’, where the derived form differs from the base with respect to argument structure (Gibert-Sotelo 2021: 248). The analysis proposed here for cuasi and ex as modifiers of N° is in line with the approach to analyzing certain prefixes as adjuncts – but others as head or specifiers – developed by Fábregas & Krämer (Fábregas & Krämer 2020; Fábregas 2024), going back to ideas by Di Sciullo (2005); cf. also the discussion in Gibert-Sotelo (2021, §6). Spanish no preceding nouns can be analyzed in the same way: It is a modifier of N°.

If formatives like no, cuasi or ex are adjuncts, they are expected to allow for iteration, restricted only by conditions on interpretation. An expression like la cuasi no censura ‘the almost non-censorship’ can thus be represented by the c-structure given in (63). Note that while Lexical-Functional Grammar subscribes to the basic tenets of X’-structure, admissible c-structures are less restricted than in certain other formalisms, allowing, e.g., immediate domination of X° by XP (cf., e.g., Dalrymple et al. 2019: 97–99 for discussion).

In (63) the Quant-modifier cuasi has scope over the Neg-modifier no, which is reflected in the relative embeddings in the (partial) f-description in (64).

Note that the c-structure in (62b), adjoining the modifier to the lexical head, is also compatible with the observation that nouns modified by no can occur within compound structures, as in (48–49) above.

To summarize the discussion in this section, expressions containing Spanish no preceding phrases (syntactic negation) and expressions containing Spanish no preceding lexical heads (lexical negation), are generated by two different c-structure rules, given in (60a) and (62b), respectively. Both rules are required on independent grounds in the Spanish grammar, the first to account for modification of phrases by adjuncts like adverbs, the second to account for modification of N° by “separable prefixes” and a restricted set of adverbs, like también ‘also’ in (47).

7.3 Vocabulary structure

Is it possible to account for this bipartite distribution of Spanish no in instances of syntactic and lexical negation by postulating only a single lexical representation? This would require that no is assigned to a distributional category which allows insertion in both the rule in (60a), in which no functions as an adverb, i.e., as a building block of phrases, and (62b), in which no functions as a building block of complex words, classified as a prefix by some scholars, and termed ‘modifier of N°’ here.

Alternatively, two lexical entries could be assumed for no, one for no as an adverb – a building block of phrases – and one for no as a modifier of N° – a building block of words. It might then be said that the first element belongs to phrasal syntax, while the second pertains to word syntax or morphology. Both elements are homonymous, i.e., spelled out by the same form no in Spanish – in contrast to languages like French or English, where different forms are used. Two arguments can be put forward against such an approach: First, the contrast between no and other negative prefixes with respect to negative concord, see Section 6.2, shows that a clear separation between no as a building block of words and of phrases is not appropriate. Second, the same question – one or two lexical entries – arises with respect to other modifiers of N° considered here, such as cuasi ‘almost’, but also for adverbs like también ‘also’ and others that can immediately precede a noun, (47). If two lexical entries are assumed for no, then one should also assume two representations for cuasi ‘almost’, también ‘also’ and many other formatives.

In what follows, another possibility is therefore explored, that of leaving the c-structure category of no unspecified. The claim made here is that a single lexical entry is sufficient to account for the bipartite distribution of no. The representation proposed for no in (59) above postulates that the Spanish lexicon contains a formative which is verbalized as no, expresses the negation operator and has the function of an adjunct of the type Neg. Crucially, in (59) no information on syntactic distribution is associated with this element.

But if there is no information on distribution contained in the lexical representation, how is it possible to prevent overgeneration, i.e., insertion of no in positions where it is not found in Spanish? Two aspects of the formalism adopted here, i.e., LRFG, are crucial for such an approach to be feasible: First, lexical insertion – conceived of as verbalization of a piece of structure and meaning – is constrained not only by distributional information, but also by information on the function and meaning of a vocabulary item, see Section 2.1. Second, if stored items with overlapping specifications compete for lexical insertion, the most specific item compatible with a given context is chosen (Asudeh et al. 2023: 32; Asudeh & Siddiqi 2023: 890), by a mechanism which is similar to DM’s subset principle (Halle 1997; Embick & Noyer 2007: 298).

Let us first replace the lexical representation proposed for Spanish no in (59) above, given as a list of phonological and functional information, by a (partial) description of the exponence function ν (Asudeh et al. 2023) for Spanish negation, (65).

(65) means that the exponendum ⟨ [ (↓ AdjType) = Neg ] ⟩, i.e., a (partial) functional description which contains an adjunct of the type Neg, is mapped onto the exponent no. To be sure, not all instances of negation are lexicalized by no in Spanish. As discussed in this article, alternative exponents are negative prefixes like des- ‘dis-’ or in- ‘in-, un-’ or the negative modifier sin ‘without’, as in (18) and (44b). Negative pronouns (nadie ‘no one’), determiners (ninguno as in ninguna cosa ‘no thing’) or adverbs (nunca ‘never’) are to be mentioned, as well. In spelling out the exponendum in (65a), these forms are in competition with each other. The analysis proposed here rests on the intuition that no is the default expression for negation, used only if competing expressions are not compatible with a given context. LRFG provides a formal mechanism for representing this intuition. More precisely, in LRFG, the mapping of an exponendum to an exponent is governed by two constraints: MostInformative and MostSpecific. For the phenomena under study here, the following informal characterizations of the two constraints shall be adopted; see Asudeh & Siddiqi (2023: 890) for a formal representation of MostInformative_c and Asudeh et al. (2023: 32) for MostSpecific. From a set of competing vocabulary items, MostInformative_c choses the item with the longest list of c-structure categories on the input side, i.e., the item that has the most restricted distribution. As to MostSpecific, it choses, again from a set of competing items, the item whose output vocabulary structure, or exponent, contains more features, meaning that, e.g., an exponent which contains more features has a more specific phonological or morphological structure or imposes more specific (e.g., phonological) conditions on its context of insertion than a competing exponent.

Assuming the exponence function for Spanish no in (65) above, it is easy to see that on the input side, the exponendum is maximally uninformative, lacking any distributional information, in contrast to, e.g., the negative prefix in- ‘in-, un-’, combining primarily with adjectival bases (Fábregas 2023) or des- ‘dis-’, preferring verbal bases (Gibert-Sotelo 2018). On the output side, the exponent no also carries fewer specifications than competing exponents, in that it does not show any kind of phonological or morphological alternation, in contrast to, e.g., the prefix in-, see (7) above. Furthermore, unlike negative affixes like des- and in-, no is not bound, so that the output side of its ν-function does not contain any information about a host (in the morphological sense), in contrast to negative affixes which, being affixes, attach to a host of a specific kind, defined with respect to syntactic, semantic or prosodic features, where this information is represented by means of appropriate values for the attribute Host represented in the output of the ν-function corresponding to a given affix (cf. Asudeh et al. 2023 for details). Hence, MostSpecific will assure that no is inserted if and only if in- cannot be inserted – or, for that matter, any other negative exponent which is more specific than no.

A comprehensive analysis of all negative exponents of Spanish is beyond the scope of this study. It should nevertheless have become clear that exponents for negation competing with no have more restricted syntactic distributions (such as negative pronouns like nadie ‘no one’) or impose additional constraints on their host (such as negative affixes like in- ‘in-, un-’). The exponent no is the default exponent of negation in Spanish, used only in contexts with which none of the more specific negative exponents is compatible. The distribution of no is not lexically specified, which is why it can be used in more than one context. In instances of syntactic negation, it patterns with adverbs, in instances of lexical negation, it patterns with other modifiers of N°, such as ex ‘ex’.