Igbo is a language with three copulas. These are bụ́, dị́ and nọ̀ (Welmers & Welmers, 1969; Welmers, 1973; Emenanjo, 1978, 2015; Nwachukwu, 1983, 1984; Uchechukwu, 2010). The copula bụ́ is the only copula found in specificational and equative contexts (4), but it can also be used in predicational copular sentences, as the examples in (3a) and (4a) show.

The sentence in (4a) can also have an equative reading, in which both DPs flanking the copula are referential. However, there are diagnostics indicating that the post-copular expression in (4a) is a predicate. The first test involves pro-form substitution. While predicates are replaced by what, which probes for a property, referential DPs are replaced by who, targeting a specific individual. As demonstrated in (5), the sentence in (4a) serves as a natural answer to a ‘what’ question, confirming that the post-copular position is occupied by a property-denoting expression rather than a referential one. See also the discussion in Section 3.1.

A second diagnostic for the predicate status of the post-copular expression in (4a) is its behaviour in small clauses (Heycock, 2021). In these constructions, the subject and the predicate form a constituent without the presence of a copula. As shown in (6) the DP ónyé ísí àlà á^!nyị́ can function as the predicate of a small clause selected by the verb mé ‘make’; cf. (15) and (16). Following the standard analysis of small clauses, this suggests that the relation between the two DPs is one of predication rather than identity.

Copulas dị́ and nọ̀ are attested in predicational copular sentences and are not possible in specificational and equative clauses, as the ungrammatical sentences in (7) show.

The copulas dị́ and nọ̀ both occur with PP predicates, but they are distinguished by the animacy of their subjects. The copula nọ̀ is restricted to locative contexts with animate subjects (8a), whereas dị́ is typically used to express location for inanimate nouns (8b). The copula nọ̀ seems to be derived from the verb ‘sit’ (Welmers, 1973).

None of the copulas can be omitted in the different copular sentences exemplified above, whether in root or embedded clauses (9). Furthermore, the different copulas cannot be combined (10).

The evidence that these copular elements are verbal in nature comes from the fact that they can take tense, aspect, and negation morphology (11) like regular verbs in the language (12).3 In example (11), the copulas occur in the canonical position of verbs in the language, that is, after the subject, and they co-occur with all the verbal inflectional morphology.

Table 1 summarizes the properties of the three Igbo copulas discussed in this section, illustrating their distribution across different clause types and their morphological capabilities. The copula bụ́ appears to be the common copula found in predicational, specificational and equative copular sentences. The other two copulas are restricted to predicational clauses. Copulas bụ́ and dị́ are the only copulas that can be found in binominal copular clauses, while dị́ and nọ̀ can take prepositional predicates.

Predicational sentences express a predication relation between a subject and a predicate. The non-referential post-copular expression denotes a property (Higgins, 1979). In Igbo, only nominals and PPs can function as predicates; cf. (13) and (14). Igbo bare nouns can function as predicates and can combine with the copulas bụ́ and dị́.

PPs occurring as predicates can only combine with copulas dị́ and nọ̀ (14).

As already illustrated in (6), a key syntactic diagnostic for PCCs is the possibility of juxtaposing the subject and predicate in a small clause without an overt copula (Heycock, 2021). I illustrate this with the verb make in Igbo, as in (15).4

The difference between the binominal copular sentences above is that unlike the PCC in (15), the specificational clause in (16) cannot appear in a small clause. The data in (17) demonstrate that binominal PCCs with dị́ copula also pattern like PCCs with bụ́ copula with respect to the small clause diagnostics.

In sum, the sentences in (15a) and (17a) are predicational clauses. These sentences also display the canonical order (subject-copula-predicate), where the nominals appearing to the left of the copulas correspond to the subject of predication. The inverse copular sentence in (16a) is a specificational sentence. This explains the ungrammatical structure in the small clause in (16b). The fact that the order in the specificational clause in (16a) is impossible in a small clause is evidence that the post-copular nominal in (16b) is not a predicate.

Igbo has a small class of adjectives which consists of only eight property concepts (Welmers & Welmers, 1969; Carrel, 1970; Welmers, 1973; Emenanjo, 1978, 2015); see (18).

These adjectives have been argued (Green & Igwe, 1963; Welmers & Welmers, 1969) to be possible only in direct noun modification, that is, attributively, as shown in the (a)-sentences in (19) and (20), but not in predicate position, as demonstrated by the (b)-sentences.

While these adjectives cannot appear in predicate position, nouns that are derived from the same verbal roots can, as exemplified in (21) and (22).

The data in (20b) show that the adjective ọ́má ‘good’ cannot occur in the post-copular position. What is allowed in this position is the nominal predicate ḿ^!má ‘goodness’ (21). Both ọ́má and ḿ^!má are derived from the same verb root má ‘be good’. Similarly, as illustrated in (19b) and (22) (see also (24)), the difference between the adjective and the noun derived from the verb chá ‘be white’ is tonal, while the adjective has two high tones, the noun bears a low and a high tone. There is further evidence from tone and syntactic distributions for a distinction between nouns and adjectives as categories in Igbo. Regarding the tonal evidence, in an associative construction, where we have a noun modifying another noun, a tone change is observed when bi-syllable nouns with high tones, for example, are combined (Green & Igwe, 1963; Emenanjo, 1978; Déchaine, 1993; Hyman, 2011). This is the downstep tone on the last syllable of the second noun in the post-copular positions in (15). In isolation, the nouns have high tones, but in the noun-noun modifying configuration, the second high tone of NP2 becomes downstep (23a). However, no such tone change is found when a noun and an adjective co-occur (23b).

One piece of distributional evidence comes from other modifications. The noun phrases in the post-copular position in (21) and (22) can be modified just like any other noun in the language while they occur in predicative sentences, as in (24c) and also outside of copular constructions. The sentence in (24c) also shows that relative clause formation is possible, which is a kind of modification that is impossible with adjectives (Emenanjo, 1978; Zamparelli, 2008).

The final argument for the fact that adjectives are not used predicatively in Igbo has to do with the cognate objects that some verbs take. Welmers (1973) notes that half of the adjectives in the language are derived from verbs. As expected, the cognate objects derived from these verbs are the same as the post-copular nouns in (21) and (22) and not the adjectives. The relevant examples are given in (25). The additional data in (26) show examples of cognate objects with their corresponding verbs that do not have adjectives.

Igbo is not particularly special in exhibiting a small class of adjectives that function only in attributive environments. Koopman (1984) reports that in the West African languages Vata and Gbadi, adjectives are only used attributively. Although Igbo attests a small class of adjectives, the language has an open class of property concept nouns and verbs (Nwachukwu, 1983, 1987; Amaechi & Hopperdietzel, 2022); see section 3.1.3 for some details and relevant examples.

Crucially, the inability of adjectives to function as predicates serves as a key premise for the analysis developed in Section 5. Since property concepts in predicative positions must be realized as nominals (or verbs), the structural analysis of Igbo copular clauses is restricted to nominal and prepositional complements. This elimination of the Adjective Phrase (AP) as a potential predicative category supports the central proposal of this paper that the distinction between bụ́ and dị́ reflects a structural difference between nominal small clauses and prepositional small clauses (containing a silent P).

In the next section, I examine the two copulas that permit nominal predicates and set aside the copula nọ̀ from now on as it allows only prepositional predicates. The discussion focuses on the distribution of bụ́ and dị́ attested in binominal copular sentences, which constitute the main concern of this study, and highlights additional distinctions between them.

The additional differences discussed in this section illustrate the various constructions in which copulas appear in the language, as well as the specific forms they take in those contexts. First, inverse copular sentences in the sense of Moro (1997) are possible only with copular sentences involving bụ́ as illustrated in (4), repeated here as (37).6

The data in (37) and (38) demonstrate that both bụ́ and dị́ are possible in the canonical order. This observation supports the view that if inverse sentences are indeed specificational, it is expected, as previously noted, that the copula bụ́ appears in non-predicational sentences. Furthermore, the fact that both bụ́ and dị́ may occur in sentences with the canonical order can be interpreted as meaning that the canonical order can be specificational or predicational (Woodard, 2018; Roy & Shlonsky, 2019; Roy, 2022).

The example in (39) demonstrates that prepositional predicates selecting the copula dị́ are incompatible with inversion, while nominal predicates allow for inversion as seen in (15a) and (16a). This suggests that the inversion observed in (16a) is a category-specific property of nominals, rather than a general feature of all predicative structures. The data in (39) is also not surprising as Igbo does not attest locative inversion that is prominent in Bantu and many other languages (Salzmann, 2011). Given the rigid SVO order in Igbo, the prepositional predicates may only be fronted via focus movement (39c) or some other A-bar dependencies. These data are also relevant to the debate on whether predicational and specificational clauses share a common underlying structure, where the latter is derived by inverting the predicate (Moro, 1997; Mikkelsen, 2005). If specificational clauses are indeed the result of predicate inversion, the ungrammaticality of (38b) and (39b) indicates that this derivation is not unrestricted in Igbo. Specifically, the operation appears to be blocked for the predicates selected by dị́, thereby restricting specificational inversion to the nominals associated with the copula bụ́.

Existential sentences belong to the class of constructions where copulas have been shown to be found (Stassen, 1997; Pustet, 2003; Myler, 2018). In Igbo, dị́ can be used in existential constructions in addition to have-existential that the language attests similar to what is obtainable in many languages (40). Existential constructions, such as There is a book on the table and locative PP predicates like The book is on the table are argued to be closely related because both often involve spatial reference and share structural elements. A common approach in the literature is that existentials and locatives can derive from a common underlying structure (Freeze, 1992).

Finally, in clefts constructions bụ́ copula is found, as illustrated in (41).

Table 2 below summarizes the additional differences highlighted above for the two copulas under discussion in Igbo. We saw in the previous section that bụ́ occurs only with nominal predicates, while dị́ can combine with both nominal and prepositional predicates. This explains why the former is compatible with inversion (specificational sentences) and cleft structures, and the latter is attested in existential and verbal stative constructions, which are both assumed to involve predication.

In the remainder of the paper, I will concentrate on the interpretational differences between the copulas and also provide a syntactic account.

The puzzle is that Igbo, like many other languages such as Spanish, Portuguese, and Modern Irish, has two forms of the copula in predicational constructions. In Section 2, it was noted that dị́ is a predicational copula, while bụ́, which is the copula found in specificational and equative sentences is also attested in predicational constructions. This means that the sole predication vs. identity distinction is insufficient to account for the difference between the two copulas. One possibility of coding the difference between multiple copulas used in predicative structures is in terms of the semantic properties of the predicate, as in the distinction between individual-level and stage-level predicates (Carlson, 1977).7 Individual-level predicates express permanent and stable properties of an individual, while stage-level predicates identify spatio-temporal slices of an individual. For instance, individual-level predicates such as tall are predicated of individuals directly, and stage-level predicates such as sick are predicated of stages of individuals. While predicative nominals with dị́ can have either individual-level or stage-level interpretations, predicational sentences with bụ́ are consistently individual-level. However, the distinction between stage-level vs. individual-level predicates alone does not sufficiently account for the entire distribution of the two copulas. This situation parallels the distribution found in Irish, where the distinction between stage-level and individual-level predicates also fails to fully capture the distribution of the copulas Tá (or Bí) and Is (Carnie, 1995; Doherty, 1996). The discussion of Irish here serves to illustrate that the insufficiency of the stage-level/individual-level distinction in Igbo is not an isolated phenomenon but aligns with patterns in other languages where copula distribution requires a more fine-grained event-structure analysis.8

Carnie observes that the predicates with Is are individual-level predicates, while Tá allows both individual- and stage-level predicates. On this basis, he argues that the distinction between the two copulas cannot be explained in terms of the individual-level vs. stage-level predicate distinction. It is also interesting to note that copula Is is found most productive with nominal predicates and Tá appears with all types of predicates except nouns. The Igbo data in (43) illustrate the fact that dị́ allows for both individual- and stage-level predicates.

The locative modifier in the vehicle supports a stage-level interpretation of sentence (43a) as this allows for a temporal reading; cf. Chierchia (1995); Kratzer (1995); Maienborn (2004), among others. By contrast, with the individual-level reading of (43b), the inclusion of the locative modifier results in a degraded sentence. I will later demonstrate that sentence (43b) is actually ambiguous between an individual-level and a stage-level interpretation.

Roy (2013) observes that the well-known distinction between individual-level vs. stage-level predicates is not capable of capturing the properties of PCCs in languages like French, Spanish, Irish and Russian. She distinguishes three different kinds of PCCs and suggests that the interpretational differences between the different types of these sentences come from the internal syntax of the post-copular predicate. She argues for a 3-way classification of PCCs based on the semantic criteria of maximality and density. Maximal predicates are devoid of perceptible spatio-temporal subpart properties, that is, there are no smaller states that could be proper parts of them. In this sense, when the subject of predication is said to be in a maximal eventuality, it means there exists no broader eventuality that simultaneously contains the subject in a present state. The predicate describes a complete, non-divisible situation rather than a temporally or spatially segmented one. Density relates to whether a predicate holds for every subpart of a particular eventuality or not. While dense predicates have spatio-temporal subpart properties that are non-atomic and homogeneous, non-dense predicates have spatio-temporal subpart properties that are atomic and allow for possible gaps. The three PCCs are (i) defining sentences, which are constructed with maximal predicates; (ii) characterizing sentences involving non-dense predicates; and (iii) situation-descriptive sentences, which feature dense predicates. Roy claims that the three types of predication enumerated above can be distinguished by potential questions as illustrated below.

Defining and characterizing sentences are grouped together in Roy’s system since they ascribe a property to an individual and are not dense. Roy shows that the three-way classification has a double two-way split as schematized in (45).

A defining property, unlike a characterizing property, must be salient enough to ‘define’ an individual as a particular member of a class of individuals. Situation-descriptive sentences, on the other hand, differ from the other two types in that they do not ascribe a property to an individual but describe situations. For example, the French statement in (46a) is a defining sentence that can give rise to ‘lifetime’ effects. The characterizing property of being intelligent in (44b) is non-dense since the state of being intelligent need not hold for the subject at the moment of utterance – it allows for ‘gaps’. Finally, the situation-descriptive property of being sick (44c) is dense since it holds for every subpart of the eventuality that would render the predicate false.

Based on the presence or absence of an indefinite article in the post-copular predicative expression in French (46), Roy argues that both sentences are predicational, and illustrate the different interpretation of non-verbal predicates grammaticalized in French. The variant with the indefinite article (46a) is argued to be a defining sentences, while the bare noun variant in (46b) involves a dense/non-dense predicate and constitutes a characterizing sentence.

Roy shows that while (46a) constitutes a felicitous answer to the question Who is X?, (46b) can only answer the question What is/does X? Additional important differences between the two sentences involve their (in)compatibility with aspectual and temporal modifications. For instance, the indefinite article variant in (46a) triggers lifetime effects in the past tense, while (46b) does not. Moreover, only the bare noun variant can be marked for perfective aspect (47).

Roy further highlights distinctions between characterizing and situation-descriptive sentences in French. For example, characterizing sentences allow for clitic doubling and there are restrictions on lexical categories; see Roy (2013, Chapter 3) for relevant data and discussion. The example in (48) demonstrates that characterizing and situation-descriptive sentences differ in their truth conditions. The nominal predicate in the first conjunct is characterizing, whereas the adjectival predicate in the second conjunct describes a situation. Crucially, the negation of one conjunct (that is, the second in (48)) does not lead to a contradiction. This indicates that one clause may be true independently of the other. For instance, the property of being a drunkard can characterize the subject, Paul, while the state of being drunk may not apply to him at the time of the utterance.

In terms of the syntactic structure of the three non-verbal predicates in (44), Roy argues that both defining and characterizing sentences are projections of a noun, represented in (49a-b) by their functional projections NumP and ClP, respectively. However, situation-descriptive sentences can be any category except nominal. (49a-b) shows the structure and functional projections within the nominal phrase involved in the first two kinds of PCCs. (49c) indicates that the predicate in situation-descriptive sentences can be either adjectival or prepositional. DegP is the projection for adjectives.

Maximal predicates involve a number head, Num, which introduces a maximal quantity of an eventuality, while non-dense predicates involve a classifier head Cl, which is responsible for the atomic structure and divides the eventuality denoted by the non-verbal predicate into subparts. The structures in (49) suggest that maximal and non-dense predicates are nominals, while dense predicates are often AP or PP.

Roy argues that the two-way split shown above is grammaticalized in languages such as French, Russian, and Spanish. In Spanish, for instance, the copula ser is found in both defining and characterizing sentences, while the copula estar is attested in situation-descriptive sentences. Roy (2013) also considers the grammaticalization pattern exhibited in Irish (and Scottish Gaelic). She posits that Irish does not share the three-way distinction between dense/non-dense/maximal predicates as found in the other languages she considered. Rather, Irish has maximal and dense predicates, and characterizing sentences have habitual reading involving generic quantification over the event. For instance, while Spanish allows for the existential binding of the eventuality variable introduced by the predicate in copular sentences, Irish allows for the generic/habitual binding. In both cases the operator, either existential or generic, quantifies over dense eventualities.

In Section 4.1, I demonstrated that the distinction between the two copulas found in Igbo predicational sentences cannot be fully accounted for by the traditional individual-level versus stage-level predicate distinction. In this section, I apply Roy’s 2013 three-way classification to Igbo PCCs, showing that all three types play a crucial role in interpreting Igbo predicational copular constructions. Unlike the languages discussed in Roy’s framework, I propose that Igbo grammaticalizes this ternary distinction in predicational sentences through a division between maximal and non-maximal predicates. Consider the contrast in (45) and (50).

In Igbo, the maximal (defining) predicates use the copula bụ́, while the non-maximal (characterizing and situation-descriptive) predicates occur with the copula dị́. The data in (51) exemplify the three-way distinction with the different copulas. The relevant diagnostics are discussed below.

Characterizing sentences with dị́ copula and having an individual-level predicate reading can be rendered with bụ́ plus the noun ónyé ‘person’ demonstrating that both defining and characterizing predicates ascribe a property to an individual (52b-c). In this way, they are different from situation-descriptive sentences, which do not ascribe properties but describe a state or situation.

The difference between the defining sentence in (52b) and the characterizing one (52a) is that the defining sentence predicates a salient or inherent property of the subject, which defines that subject as a whole and assigns it to some category (Roy, 2013). Put differently, a defining sentence identifies an individual as a member of a class based on an intrinsic property that is not time-bound. A piece of evidence to support this is that nouns denoting subsets of humans, such as man, woman, combine with bụ́ and not dị́. In Igbo, these nouns are treated as essential properties (class membership).9

Based on the potential question test exemplified in (44) above, we also see differences between these three PCCs types in Igbo, as illustrated in (51). While (51a) (repeated here as (54b)) is a felicitous answer to the question in (54a), the characterizing and situation-descriptive sentences in (54c) and (54d) are not.

In Igbo, the defining sentences can also be questioned with gí^!ní ‘what’ as in (55). I take it that this demonstrates that the sentences are predicational sentences requiring what that asks for a property (Higgins, 1979; Mikkelsen, 2005); see also Section 3. In this case, asking for a property in the post-copular position requires the use of the focus marker kà that occurs in non-local subject questions; cf. (54a) and (55a).10

For wh-questions about characterizing and situation-descriptive predicates, the use of the simple wh-pronoun gí^!ní ‘what’ is not possible, as shown in (56a). A different wh-question strategy involving a relative clause is used (56b-c); cf. Goldsmith (1981) and Amaechi (2020). The inability of gí^!ní ‘what’ to question predicates in dị́ clauses is not accidental but follows from the structural proposal defended here. Since gí^!ní is a nominal wh-word, it targets the nominal defining properties associated with bụ́. In contrast, characterizing and situation-descriptive predicates, which I argue involve a silent prepositional structure (see Section 5), behave syntactically like manner or locative adverbials. Consequently, they require the manner interrogative (56b-c), effectively treating the property not as a nominal entity but as a state of being.

Each of these PCC types will be discussed in detail in the following subsections. The discussion will focus on showing the nominal nature of these post-copular predicates and how each behaves differently.

Following the literature on the syntax of PCCs, I assume that the copula be found in predicational copular sentences is a semantically vacuous verbal element whose presence is only required to bear tense features; see Heycock (2021) for a recent overview. The copula is treated as a raising verb that takes a small clause as a complement.12 Bowers (1993) analyzes the small clause as the projection of a functional head, Pred that takes the predicate as its complement and the subject as its specifier. This is represented in the asymmetric structure in (68). The assumption is that APs, NPs and PPs require Pred° to function as predicates (Baker, 2003). Given that the subject precedes tense and aspectual elements in Igbo, the underlying subject generated within the small clause A-moves to Spec,TP for the requirement of the edge feature that Spec,TP be filled.

The two copulas exhibit a clear distinction in category selection, as discussed in Section 3. While one copula combines only with nominal predicates, the other allows for both prepositional and nominal predicates. For the latter, I argue that the copula selects a PP structure; in cases where the predicate appears to be nominal, it is actually introduced by a null preposition. Building on Bower’s predication analysis while permitting certain relevant features to percolate up to PredP level (Svenonius, 1994; Hegedűs, 2013), I argue that the difference in the spell-out of the verbal copula depends on whether the small clause is nominal or prepositional in nature. In this way, the small clause is not category-neutral as suggested by Bowers (1993). Hence, PredP in (68) is of two kinds illustrated in (69) and (70): nominal and prepositional.13 When PredP is nominal V spells-out as bụ́ and when it is prepositional, the copula is dị́. The idea of different types of PredP has been proposed in the literature. For instance, Baker (2003) proposes that there are different Preds for predicate nominals and adjectives in languages such as Bambara, Ewe, Chinese, Hixkaryana, Mundari, Vai, and Ika. Adger & Ramchand (2003) posit two types of PredPs, one with an event variable and one without. Hartmann (2008, Ch.2) suggests different types of PredP based on their complement: one where the complement is a predicate and the other where the complement is a complex DP. The proposal stated here is one possibility for accounting for the difference in the copulas. The second possibility, which I will come back to later is that there is only one Pred head, but it is lexicalized differently depending on the type of complement it is constructed with. This means that the difference is only morphological.

One piece of evidence in favour of the different PredPs is that coordination of the two post-copular nominal predicates of the two distinct copulas is degraded. Coordination of different syntactic categories as shown in (71) is one of the main arguments for the existence of a functional head in small clauses; see Matushansky (2019) for arguments against this evidence for PredP.

The result of the coordination diagnostic shows that the two PredPs, that is, a nominal with bụ́ and the prepositional one with dị́, are not well-formed. The example in (72) is an attempt to conjoin the post-copular noun phrases in (51a) and (51b). A potential confound for the test in (72a) is that the combination of bụ́ and ọ̀chá is grammatically permissible in Igbo; however, it yields an equative interpretation (e.g., ‘Ibe is [the color] white’ or ‘Ibe is the embodiment of whiteness’). In such a case, both the subject and the post-copular expression are referential DPs. Since the goal of the coordination test is to evaluate predicational aspect (the attribution of a property to an individual), the equative parse is irrelevant to the current argument.

Coordination of the nominal predicates that combine with a particular copula is licit, as in (73).

The nominal predicate in (69) is not a full-fledged DP that occurs in argument positions but rather a predicative nominal, as in Roy’s NumP found in defining sentences; cf. (49). I propose that the prepositional PredP in (70) contains a null preposition in binominal copular sentences with dị́. Whether we have a post-copular noun phrase or a PP, the structure is one with a PP as the complement of the copula. The overt post-copular nominal in these PCCs is the complement of the null P(reposition). Characterizing and situation-descriptive sentences with dị́ copula in (51b)–(51c) will have the structure in (74).

The postulation of a null P head in the structure is motivated by the assumption that all contexts in which the dị́ copula appears exhibit a relational component in their interpretation. To enable a nominal such as whiteness to function as a predicate where the subject is an individual, an additional layer of meaning must be introduced. I contend that this supplementary meaning is contributed by a prepositional element. Crucially, the null P is present only when the prepositional predicate does not express location. This indicates the difference between dị́ with an overt nominal and prepositional predicates, as in (75).14 The claim that I make is that even though we find dị́ on the surface with a nominal, the copula is never constructed with nominals as there is a null P in the structure.

Furthermore, the fact that the complement of dị́ is a PP provides a principled explanation for the inability of copular clauses containing the copula to undergo inversion. This structural constraint, as discussed in Section 3.2, underscores the syntactic incompatibility between PP complements and inversion in this context. The postulation of a null preposition is motivated by broader patterns in Igbo syntax. First, the copula dị́ is fundamentally locative, as evidenced by its requirement for overt PPs in locative constructions (75b). I argue that its combination with property concept nouns preserves this selectional requirement via a null P head. Independent evidence for a null preposition exists outside the copular domain. In Igbo motion verb constructions, locative goals frequently appear without an overt preposition (76), suggesting that a null prepositional head is independently available in the grammar to license nominals in locative/predicative roles (Nwachukwu, 1987; Uwalaka, 1988). Furthermore, the overt preposition nà, as illustrated in (75b), serves as the phonological counterpart to the silent P. While nà is preferred for concrete locations, the null variant is specialized for property concepts in characterizing and situation-descriptive contexts.

An important question that may arise at this point concerns the correlation between this null P and the property concept nouns, such as whiteness, length, and shortness discussed in the previous section which combine with the copula dị́, but are incompatible with definitory properties like (young) man, child or relative (see example (62)). I propose that this incompatibility may be related to the fact that property concept nouns denote a state. Roy (2010), following Grimshaw’s 1990 work on deverbal nominalizations, argues that deadjectival nominals fall into two classes, viz., state-nominals and quality-nominals. The former describes a state, while the latter names a quality that someone or something may possess. Important for the present discussion is that state-nominals must always have an overt external argument introduced by a preposition. This is in the form of a de-phrase in French exemplified in (77a). The absence of a de-phrase in (77b) gives rise to a quality reading. Note that the noun popularité ‘popularity’ in the example is ambiguous between a state-nominal and a quality-nominal interpretation.

Another piece of evidence for the underlying prepositional nature of the post-copular expression with dị́ is the idea that the copula is what is attested in existential and stative constructions; see Section 3. For existential, at least, an approach in the literature is that these involve a locative structure (Freeze, 1992). The claim is that cross-linguistically, the difference between locative sentences and existential sentences involves a difference in surface word orders; both orders are derived from the same base-structure.15 See also Myler (2018) for a recent account of how existentials fit into copula systems. While Igbo exhibits a have/be existential alternation (using dị́ copula and nwé ‘have’), I suggest that this does not necessitate a split-structure analysis that would undermine the null P framework. Following the approach that have is not a primitive lexical item but is instead derived from the copula be plus an incorporated preposition (Freeze, 1992; Kayne, 1993), I propose that both dị́-existentials and nwé-possessives originate from the same prepositional small clause architecture. The distinction is purely morphological, where nwé represents the realization of the copula when it incorporates a specific possessive/existential P-head, while dị́ is the spell-out of the copula when the P-head (locative or manner) remains independent or fails to trigger this specific incorporation.

Consequently, a silent ‘have’ analysis for predicational copular dị́ sentences (i.e., treating ‘Lily is beautiful’ as ‘Lily has beauty’) is untenable for Igbo. Under this framework, such a possessive structure would require the incorporation of the possessive P, which would obligatorily trigger the realization of the verb nwé. The persistence of dị́ in these PCCs signals that the underlying P is not possessive, but rather locative or manner-like. This conclusion is independently confirmed by interrogative evidence. As shown in (78), constructions involving nwé ‘have’ even when the object is an abstract quality like ǹdìdì ‘patience’, are questioned using the nominal wh-word gí^!ní ‘what’.

In contrast, as demonstrated in Section 4.3, predicates with dị́ (characterizing and situation-descriptive clauses) reject gí^!ní ‘what’ and instead require the manner interrogative kèdụ́ ètù ‘how’. This morphosyntactic divergence confirms that dị́ PCCs are not covert possessives, but distinct P-based structures.

Alternative structures for the binominal PCCs considered in this study are discussed in this section. The analysis developed here is a one Pred head account. Unlike the proposal in Section 5.1, which posits two distinct Pred heads (nominal vs. prepositional), this alternative assumes a single Pred head across all PCCs. The apparent differences between bụ́ and dị́ are derived from elsewhere in the structure—specifically, from the internal composition of the predicate phrase rather than from the Pred head itself. This approach follows the line of analysis proposed by Roy (2013); see also Geist (2019), where the relevant contrasts in PCCs do not lie in the PredP but rather in the complement of the Pred head. Roy argues that the semantic distinctions between the three PCC types map unto syntactic structures, as represented in the structures in (49), repeated here in (79).

The extended nominal syntactic structure assumed is as in (80); cf. Borer (2005), Roy (2013), Geist (2019). The D head is responsible for referentiality, and the DP projection is absent in predicate nouns.16 The Num head is the locus of maximality, and its function is for quantization, realized as the indefinite article in languages like French and German as well as numerals (Roy, 2013; Geist, 2019).17 The Classifier merges with the NP, creating a ClP. Cl is responsible for introducing atomic (i.e., making countable) structures (Borer, 2005). The NP is projected by the noun or nominal root introducing the descriptive content.

For Igbo, one could assume that post-copular NPs in defining sentences project NumPs (81a), while those in characterizing ones are ClPs (81b). Finally, the post-copular nominal in situation-descriptive sentences projects a PP (in surface it might be a nominal, but there is a silent P) without classifier and number layers (81c). This aligns with the structure of situation-descriptive sentences, which, according to Roy, cannot be bare nouns since bare nouns are unable to function as predicates.

Since bụ́ and dị́ are both attested in PCCs, the argument is that they are allomorphs of the same morpheme: bụ́ when XP_pred is NumP and dị́ elsewhere (ClP or PP). Assuming Distributed Morphology’s late insertion of lexical items, where the phonological expression of syntactic terminals is in all cases provided in the mapping to PF (Halle & Marantz, 1993, 1994; Embick & Noyer, 2007; Embick, 2015). We will have distinct Vocabulary Items that track whether the complement of Pred head is the more specific NumP or not. This account has the explanation that language variation reduces to morphological differences of what is lexicalized and how.

Given that the choice of copula is based on the syntactic nature of the post‑copular predicate, this approach implies that the copula must be able to see inside the small clause and select a subconstituent that conditions its allomorphy. This, however, introduces a locality problem: the morphology of the higher copula would depend on information that is too deeply embedded in the derivation, in a way that standard locality conditions do not permit. Assuming that PredP is a phase (den Dikken, 2006; Hartmann, 2016), the phase impenetrability condition (Chomsky, 2001) blocks operations outside the phase from accessing material inside the complement domain of the phase head. In other words, the copula cannot “look into” the predicate phrase if that phrase is contained within a completed PredP phase. Ways to avert this locality problem have been proposed in the literature; in each case, an additional assumption is required to permit relations that would otherwise be phase‑bounded. These include phase extension (den Dikken, 2006, 2007), and proposals that the phase boundary may be partially transparent (Baker, 2015; Kalin, 2020; Smith, 2020). In the configurations in (81), PredP takes a NumP, ClP, or PP complement, and the copular V is merged as the sister of PredP. The morphological form of V (bụ́ or dị́) is determined by the category‑defining head inside the complement of Pred. For V to select the correct exponent, the morphology would have to ‘see’ down into PredP, past Pred, into its complement, and then down to the relevant category‑defining head. In this configuration, V is not locally adjacent to that head as Pred intervenes, and the predicate phrase is a full projection. One way to alleviate this locality problem would be to allow bụ́ to span across the relevant structure (82a). By contrast, the two‑Pred account proposed in the previous section offers a structural solution that resolves the locality issue without resorting to elaborate spanning machinery. By positing distinct Pred heads (Pred_N vs. Pred_P), the copula V no longer needs to look down into the complement. Rather, V locally selects the specific Pred head. The morphological form of V is determined by its immediate sister (the specific PredP). This approach provides a more transparent mapping between syntax and semantics while keeping the morphological component simple.

Data from cleft constructions exemplified in (41) and (83) provide further evidence that the distribution of the copulas is determined by categorial selection rather than spans. Igbo attests two different cleft constructions, one formed via focus movement (83a) and the other via relativization (83b). The copula bụ́ and never dị́ is the verb that selects the CP/FocP in the former and the DP in the latter (Amaechi, 2020).

Following Delahunty (1984) and Szabolcsi (1994), CPs and DPs share a nominal-type distribution. Under the two Pred heads analysis proposed here, bụ́ is the default copula when selecting a nominal-type category (NP, DP, or CP). By contrast, a span approach would require additional, independent rules of exponence beyond those already proposed in (82) to account for the cleft data. Put differently, the cleft facts support an analysis in which the copular alternation follows from categorial selection.