Typologizing nominal expressions: the noun phrase and beyond

1 Introduction

This article offers foundations towards a comprehensive typology of the nominal domain, capturing the significant language-internal and crosslinguistic diversity of nominal expressions. This new typology moves away from the traditional approach that is based on the following dichotomy: between languages that have classic, rigidly structured NPs, like English (1), and languages that do not, like Kalkatungu (2), which shows different orders for noun, adjective and demonstrative, and allows discontinuous structures. This dichotomy is often – if perhaps less frequently nowadays – linked with the more general concept of (non-)configurationality, the study of which has its roots in analyses of specific languages, especially from Australia. The presence of ‘problematic’ structures as in (2) led to the idea that several Australian languages do not or do not clearly have NP units as established for, e.g., Germanic or Romance languages (e.g., Blake 1983; Evans 2003: 227–234; Hale 1983; Harvey 1992; Heath 1986; Rijkhoff 2002: 19–22).^[1]

However, and leaving aside controversies around the concept of non-configurationality, there is wide-ranging evidence in the literature that the picture is far more complex than a simple dichotomy suggests, and this is reflected in some analyses which make finer distinctions. For example, Rijkhoff (2002: 19–23), in his crosslinguistic study of the NP, distinguishes between integral NPs (where elements that semantically modify the head are also modifiers syntactically), non-integral NPs (where elements that semantically modify the head are in apposition to it at sentence level), and something in between (where each element is a minor NP; all are in apposition within a larger phrasal structure). Rijkhoff further argues that some languages may have appositional modification generally (referring to several Australian languages), whereas others only have it in specific circumstances (e.g., for any additional modifiers which exceed a given limit of integrated modifiers as in Yimas [Foley 1991: 4, 184, 188], or only for certain types of modifiers as in Korean [Lee 1989: 118]). Krasnoukhova (2012; see also this issue) argues that many South-American languages belong to the last category, i.e., where we see a mix of structures: they have both integral and non-integral NPs (in Rijkhoff’s sense), often depending on the type of modifier, and with structural and pragmatic motivations for using non-integral NPs. Louagie and Verstraete (2016), building on work by McGregor (1989, 1990, 1997a), Croft (2007), Schultze-Berndt and Simard (2012) and others, show that the nominal domain in many Australian languages should be analyzed as involving a range of structures. Finally, also evidence from diachronic work suggests that we are not dealing with a dichotomy, but that syntactic rigidity and complexity may increase or decrease in piecemeal fashion (Carlier and Combettes 2015; Freek Van de Velde 2009; Himmelmann 1997; Mark Van de Velde this issue; Reinöhl 2016a).

The following example illustrates the diversity in the nominal domain within a single language. Here and elsewhere in the article, we use the label of “(nominal) construal” for the different structural realizations of nominal expressions, i.e., expressions consisting of elements sharing a referential (or sometimes other) discourse role (for more on this see Section 2.1). Bininj Kunwok, a language from northern Australia, generally has flexible nominal expressions, which are functionally, but not syntactically, a unit. Evans (2003: 227) argues that “[a]lthough several nominal words pertaining to the same entity are often adjacent, there is rarely evidence that they form part of a syntactic unit; rather they are related paratactically and the relations between them are worked out from pragmatics rather than syntax.” This is illustrated in (3a)–(3c), which shows different orders for the semantic head and its modifiers. In addition, Bininj Kunwok allows discontinuous expressions, which are “particularly common” with measure expressions (Evans 2003: 242), as illustrated in (3d). These are not the only construal types available in the language, however, as Bininj Kunwok also has phrasal structure available in a small part of the grammar. Example (3e) illustrates rigid NP construal with indefinite ‘one’, where the indefinite marker is always found initially (Evans 2003: 244).^[3]

It is clear that we need an alternative typology that deals with the full diversity in the nominal domain. This article lays foundations for such an alternative typology, based on the range of construal options for nominal expressions, some of which do and some of which do not show evidence for phrasal structuring (Section 3.1). We also put a particular focus on language-internal diversity and on motivating factors for the choice between construal types (Section 3.2). To further enrich our understanding of this diversity, we discuss the relation between word class and structure of nominal expressions (Section 4), as well as diachronic developments in the nominal domain (Section 5). Section 6 focuses on some higher-level factors influencing the distribution of construal choices within language corpora, viz. orality, literacy, and communication modes.

While the article draws on literature and data of a variety of languages from different language families and areas, it is somewhat biased towards Australian languages synchronically and towards Indo-European languages diachronically, as these languages have been most prominent in the respective discussions. Other families and areas have not received the same amount of attention; exceptions include Rijkhoff’s (2002) seminal crosslinguistic work and Krasnoukhova’s (2012) study of South-American languages. This bias is partly remedied by the contributions to this special issue, which include studies of Bantu, Papuan, Eastern Caucasian and South-American languages besides Australian and Indo-European ones. We will reference the contributions throughout the article to highlight where they are particularly relevant, and also provide brief summaries of each in the final Section 7. Before discussing individual construal types and their distributions within languages, however, we first come to some conceptual preliminaries.

2 Preliminaries

At the outset, we need to clarify our understanding and use of the term nominal expression (NE): we use it as a general term for one or more elements in the nominal domain which together function as a unit, e.g., together establishing or tracking reference or functioning as a nominal predicate, regardless of the construal type they appear in, and thus also regardless of whether they also form a syntactic unit or not (see Himmelmann 1997: 111, 117–119; see Section 4 on word classes appearing in nominal expressions). We restrict the use of the term noun phrase (NP) to such nominal expressions whose elements belong together not only functionally, but also syntactically. Section 2.1 shows the value of a discourse-functional understanding of NEs as guiding principle given the significant formal diversity. This approach also helps exclude cases where we do not in fact find functional unity. Section 2.2 focuses on nominal expressions that are also syntactic units. We discuss different types of evidence often used for the identification of “classic” NPs and show how most of them are not as straightforward as they are sometimes made out to be.

3 Typologizing the nominal domain

The nominal domain shows rich diversity both within and across languages, which can evidently not be captured by a simple categorization based on a yes-no question regarding NP-hood. We need an alternative and more fine-grained typology based on the range of construals languages have available. This point was already briefly illustrated in the introduction for the Australian language Bininj Kunwok (3). Section 3.1 descriptively introduces some of the construal types that are available crosslinguistically. Section 3.2 focuses on the range of construals available within individual languages and on how these are interrelated. We also take a closer look at factors motivating the choice of a particular construal in a given language. As elsewhere in this article, we here primarily focus on nominal expressions whose sub-constituents are words rather than phrases, as the distribution of words seems generally more formally constrained than that of embedded complex (phrasal) expressions such as relative clauses or adpositional phrases – if there are restrictions, that is. However, we will add a few remarks also on phrasal sub-constituents where appropriate.

4 Word class

This section investigates the relation between word class and construal choice and structure. We here adopt the common typological approach to word classes as form classes based on morphological and syntactic distribution (e.g., Evans 2000; Himmelmann 2017; Sasse 2015; for approaches highlighting relations to referential functions and ontological categories, see Hengeveld [1992] and Croft [2000] respectively). Many studies have discussed the significant diversity in word class systems around the world ranging from just two classes (full words vs. particles) to much more finely partitioned word class systems such as ones found, e.g., in Indo-European languages. The structure of the word class system both in the domain of content and function words can impact on the types of construals found in the nominal domain.^[16]

We now survey the main word classes relevant in the context of this article and how they impact on construal options. A language has a class of nouns if there is a morphosyntactically defined word class which primarily encodes entities, while another form class encodes events and, optionally, states (i.e., verbs) (e.g., Himmelmann 2017; Sasse 2015).^[17] One such language is English, where elements other than nouns cannot normally be used without special formal adaptation to denote entities. For example, an adjective needs an ‘empty’ head one as in the red one (although there are some examples where an adjective is ‘coerced’ into behaving like a noun, e.g., the poor). Verbs can be inserted into the slot otherwise reserved for nouns to carry the discourse function prototypically associated with nouns, i.e., when they are used referentially. However, in order to do so, additional formal means are necessary, e.g., gerund morphology, unless we are dealing with flexible or ‘zero-converted’ elements such as play or dance. Example (39) displays both zero-conversion (the touch) and the use of a gerund (a touching).^[18]

If a language has one morphosyntactic class that primarily expresses entities as well as properties of entities, we speak of ‘nominals’. An example is Warao, which has a flexible class of elements which can function referentially (40a) or attributively (40b).^[19] Other examples include also Vedic Sanskrit (e.g., Example (26)) and some of the Australian languages cited in this article (see Louagie [forthc.] for a survey). Languages for which it has been proposed that they lack divisions among content words include Tongan (Broschart 1997), Mundari (e.g., contributions in Linguistic Typology 9(3)), and Riau Indonesian (Gil 2013).

We discussed in Section 3.1.1 differences between such rigid NP construals where each slot and functional role is more or less tied to a different word class, as opposed to rigid NP construals where members of a word class can occupy several different slots and thus different functional roles. This distributional evidence directly feeds into the classification of, e.g., nouns versus nominals (see also, e.g., McGregor [2013]). Nominals also play a role in numerous languages which display flexible nominal groups and/or discontinuity (e.g., Garrwa or Vedic Sanskrit).

While all languages seem to have a word class covering the reference to entities, not all languages have a word class that expresses properties of entities attributively without special formal adaptation. If a language does have such a class, this might be a dedicated class of adjectives (i.e., when the class primarily covers properties, but not other semantic categories) or a broader class, e.g., the nominals in Vedic Sanskrit and Warao.

Besides such dedicated or flexible word classes used for attribution, there are other structural solutions, like relative clauses, polysynthetic structures, complex nominal expressions or adverbial structures. For instance, several languages allow lexemes from the verb class to modify a nominal head, resulting in a variety of construction types. In Bororo, a relativized construction is used, resulting in a complex nominal expression, as illustrated in (41) (Krasnoukhova this issue). In other languages, modification is encoded outside of nominal expressions. In Tuscarora, for example, one strategy is predicational apposition, where the verb modifies the nominal semantically, but does not form a nominal expression with it, as illustrated in (42) (analysis of Hengeveld et al. [2004: 531–536]; see also Mithun [2000]). In Nivkh, finally, verbs can modify the head in a polysynthetic construction, as illustrated in (43) (analysis of Rijkhoff [2002: 138]). See Krasnoukhova (this issue) for a discussion of word class and attributive modification for South-American languages; see Riessler (2016) on attributive modification in Asian and European languages.

Quantification can also be expressed by various means across languages, either internally or externally to nominal expressions, e.g., by adverbial, preverbal or verbal elements. For instance, in Karo, quantification is expressed by sentential adverbs (which are thus not part of the nominal expression), as illustrated in (44) (Gabas 1999: 51; in Krasnoukhova 2012: 179). In Samoan, the numeral is a predicate in an embedded clause (introduced by the general TAM marker e [genr]), resulting in a complex nominal structure, as illustrated in (45) (Mosel and Hovdhaugen 1992; in Rijkhoff 2002: 169).

We round off this section with some remarks on determiners, specifically articles, and argument-marking adpositions.^[20] We have already commented on the impact of determiners on word order (see Section 3.2.2). More generally, the presence of strongly grammaticalized adnominal determiners, i.e., articles, has been argued to impact on construal choices in the nominal domain. Himmelmann (1997) argues that the grammaticalization of articles can bring about the phrasal organization of nominal expressions, and the same has been argued for the grammaticalization of postpositions in Indo-Aryan (Reinöhl 2016), both of which we discuss in a bit more detail in Section 5. As for internal construal structure, determiners have sometimes been interpreted as impacting on headedness relations, most famously in the context of the NP/DP debate. As we are here primarily concerned with construal choice, rather than with differences in how internal formal and semantic relations between sub-constituents are interpreted in terms of headedness relations, we refer the reader to the dedicated literature on this question (e.g., Abney 1987; Himmelmann 1997: 144–157; Matthews 2007: 27–60; Zwicky 1985, 1993).

5 Diachronic developments

Nominal expressions may over time develop (greater) phrasal rigidity. For some languages and language families it has been shown that this development is tied to the grammaticalization of adnominal function words such as articles or adpositions (Himmelmann 1997; Reinöhl 2016a). Starting points towards rigidity can be reanalyses of syntactically co-ranking constituents as one phrase (e.g., Himmelmann 1997; also called symmetrical groups in Reinöhl 2016a). An example is found in Latin, when demonstratives and nouns, which could occur and refer independently, were reanalyzed as one NP as a side-product of the grammaticalization of the demonstrative into an article. Another example of symmetrical groups are adverbs and local case-marked nominals which developed into prepositional phrases in most branches of Indo-European (Hewson and Bubenik 2006; Reinöhl 2016b). In other cases, we are dealing with head-dependent relations or asymmetrical groups as starting points, as in the development of phrasal nominal expressions in Indo-Aryan (Reinöhl 2016a, 2016b). Old and Middle Indo-Aryan relational nouns, adverbs and participles combining with possessors and other types of arguments developed into the rigidly organized nominal expressions of modern Indo-Aryan languages (Hindi, Punjabi, Nepali etc.), taking the form of postpositional phrases. This is illustrated by the development of the Old/Early Middle Indo-Aryan adverb upari ‘above’ into the Hindi postposition par ‘on’ in examples (46)–(47). In modern Indo-Aryan languages, postpositional phrases appear not only as adjuncts, but also by default in core argument roles, see example (48).

While some studies focus on the very early stages of phrasality in a language or language family (such as Reinöhl 2016a), others describe the development of additional slots in the phrasal template. Thus, it may be argued that Latin already possessed some rigidity in the nominal domain in the form of prepositional phrases, but that the syntax of nominal expressions complexified through the grammaticalization of articles and further determiners (Carlier and Combettes 2015; Himmelmann 1997). Freek Van de Velde (2009, 2010 studies how the Dutch NP accumulated structure over its history in the form of more and more slots for different types of determiners and modifiers.

Another path towards further NP-internal slots is proposed by Mark Van de Velde (this issue) for some Bantu languages, where appositional modifiers following NPs, marked by a prefixed augment, have in some languages become reanalyzed as part of the NP. Based on marked word order, as well as prosodic and other evidence, Van de Velde argues that the right-edge modifiers are appositional for instance in Bemba (49). By contrast, they are integrated in Haya (50), as evidenced for example by the fact that the order in (50) is standard for NPs. Van de Velde argues that the integration of appositives is the third stage in a diachronic scenario, where the first stage involves the emergence of a nominalized, appositional construal.

Krasnoukhova (this issue) tentatively proposes another diachronic pathway further expanding the NP, viz. where predicates expressing properties develop into attributive modifiers in integrated NPs for some South-American languages. Support for this hypothesis is found, first, in the fact that in many South-American languages, predication is the preferred (or even only) way of semantically qualifying a reference. In some languages, predicative structures involving juxtaposition of a subject noun and a descriptive predicate may be reanalyzed as head-attribute nominal expressions, although more evidence is needed to further substantiate this (Krasnoukhova this issue). Second, many elements expressing properties are verbal in nature, so that complex structures are needed to express modification within a nominal expression (such as relativization). This is illustrated in (51), where the property word ‘good’ is relativized to modify the head noun ‘dog’. This may historically derive from a sequence of paratactic clauses with a topicalized subject: ‘dog, this one (is) good, it dies’. In fact, this interpretation is still synchronically available in a closely related language (Krasnoukhova this issue).

As for the integration of new syntactic material (in whichever of the pathways described in the previous paragraphs), Himmelmann (this issue) argues that a reanalysis necessarily requires that the elements in question can be prosodically chunked together, with prosodic phrasing reflecting what belongs together informationally (Chafe 1994).

Nominal expressions are also known to lose complexity in the form of syntactic slots. A primary pathway involves the reanalysis of syntactically independent forms as clitics or affixes. With regard to nominal expressions, this type of development is particularly well known from the development of postpositions into case markers (see e.g., Lehmann 2002 [1982]). One such change in progress can be observed in Indo-Aryan. The previously syntactically and phonologically independent postpositions have become phonologically dependent clitics. As for their syntactic status, authors disagree on whether they are currently still best analyzed as postpositions or are best treated as case affixes (for the former analysis see Reinöhl 2016a, for the latter Butt and Ahmed 2011). Another example is found in the emergence of construals with nominalized modifiers in apposition to the phrase containing their semantic head in Bantu languages, as discussed in detail in Van de Velde (this issue).

Finally, contact influence, more specifically the presence of a significant number of L2 learners, has been argued to play a role not only in morphological simplification, but also in an increase in analytic patterns, both in the nominal and in other syntactic domains (e.g., DeLancey 2014 on Tibeto-Burman languages; Lupyan and Dale 2010; Trudgill 1989, 2011). While analytic does not necessarily mean rigid, a concomitant rigidification of nominal expressions is often implicitly or explicitly assumed. For instance, the rise of phrasal nominal expressions in Indo-Aryan has been argued to be due to contact with Dravidian and other language groups of the sub-continent (e.g., Thomason and Kaufman 1988: 141–144). It can be difficult to prove such contact scenarios, however, and the case of Indo-Aryan is no exception. At this point, it appears that contact may have been involved in the fixation of pre-existing patterns, but did not introduce altogether novel structures (see Reinöhl 2016a: 119–122 for discussion).

Obviously, whether and how languages develop under contact is influenced by numerous factors: besides population size, contact duration and number of L2 learners, these include geographic, socio-economic and political factors, as well as social structure (e.g., marriage traditions and kinship systems) and, importantly, social identity and language ideology (see e.g., Evans 2019; Thomason and Kaufman 1988; Trudgill 2011 for discussion of some of these factors). Thus, Meakins et al. (2019) for example show that it is not a preference for simplification which steers the adoption of a particular expression in the mixed language Gurindji Kriol, but a bias towards one of the source languages (Kriol) over the other (Gurindji). This resulted in the selection of several complex variants over simpler ones. Other research has shown that diversity, whether syntactic or on other linguistic levels, can remain remarkably stable and even increase under intense, even multilingual contact (e.g., Evans [2019] reporting on languages in southern New Guinea and northern Australia).

6 Orality, literacy, communication mode

A fuller understanding of diversity in the nominal domain needs to take into account not only crosslinguistic and language-internal evidence of construal types, but also their language-internal distribution. We have already touched on the impact of information structure and narrative flow on construal choice (Section 3.2.2). Here, we round off the discussion by mentioning some ways in which the higher-level parameters of orality and literacy, as well as of communication mode, impact on the types of nominal expressions attested.

Nominal expressions in spoken language are affected by the segmentation of language into intonation units as opposed to into sentences (e.g., Chafe 1994; Himmelmann et al. 2018). English and morphosyntactically similar languages show an average of only about 4–5 words per “substantive”, i.e., content-supplying intonation unit (e.g., Chafe 1994: 63–64; Himmelmann this issue). Accordingly, nominal expressions in spoken language also tend to be simpler in comparison to ones found in written language, with extreme cases of the latter found e.g., in legal, administrative, political or academic genres (see e.g., Maas [2010: 106, 112, et passim] on the frequency of attributes and relative clause in German writing vs. speaking). Afterthoughts and other types of add-ons also seem characteristic of oral language production, whereas writing tends to encourage longer, integrated nominal phrases. Other studies elaborate on discourse constraints on what content is found in intonation units, and consequently in nominal expressions. Chafe’s (1994) “one new idea constraint”, developed on the basis of spoken data from Seneca (Iroquoian) and English, holds that nominal (or verbal) expressions containing two or more content words contain maximally one term which expresses a new referent, state or event, or the expression is an entrenched collocation (see also Himmelmann this issue). It is an open question whether there is a link to the finding reported for some languages that there is a limit on NP complexity (see e.g., Louagie and Verstraete 2016: 50; Olsson this issue; Rijkhoff 2002: 330, this issue).

In analyzing orality/literacy effects, many authors have pointed out the need to consider not only the ‘medial’ distinction between oral and written language, but also differences in terms of ‘conceptual’ orality and literacy, i.e., language structure that is typical, but not exclusive to oral and written language respectively (e.g., Chafe 1994; Koch and Oesterreicher 1985; Ong 1982; Tannen 1982). Thus, e.g., a prepared speech may be medially oral, but at the same time show signs of conceptual literacy, having been carefully “worked over” (Chafe 1994: 43), shaped into full sentences, with less repetitions, no or fewer false starts, longer and more complex sentence structure etc.

Orality has also been studied intensively for societies lacking (wide-spread) writing, so-called ‘primary orality’ (Ong 1982). In the context of the present article, this research tradition is particularly relevant given that many of the indigenous languages reported on here are either still or have until recently been cultures of primary orality. While early work in this domain focused on historical attestations, and particularly on historical oral art forms (e.g., the seminal work by Parry [1928] on the Homeric epics, and similar studies following in his footsteps), more recent research has expanded this approach to living cultures (e.g., contributions in Carlson et al. [2011]). However, it is still not widespread in typological studies to analyze syntactic structures and variation in detail from the perspective of (primary) orality and literacy.

Ong (1982: 36–57) mentions a variety of characteristics that are typical of primary orality. Some of these apply particularly to oral art forms, whether recorded or written down at some point, such as the Homeric epics, the Jewish Bible, or the living traditions in areas of former Yugoslavia (Carlson et al. 2011; Ong 1982). Among other things, this includes a discourse preference for paratactic over hypotactic structures (e.g., Ong [1982: 36–37] comparing the syntactic structures of the original Jewish bible with modern renditions).^[21] Also some more content-oriented or performance-related characteristics mentioned by Ong (1982) impact on the internal structure of nominal expressions. For instance, Ong highlights the role of collocations in the “aggregative” style favored in oral art forms: “[o]ral folk prefer, especially in formal discourse, not the soldier, but the brave soldier; not the princess, but the beautiful princess; not the oak, but the sturdy oak” (1982: 38). This observation ties in with findings of an abundance of epithets in nominal expressions in some oral art forms (e.g., Gonda [1959] on the Rigveda). At the same time, this field remains very much in need of further investigation; for instance, oral art forms in Australia do not seem to be particularly characterized by complex nominal expressions.^[22]

Turning to communication modes, the (non-)interactional character of story-telling may also affect the shape of nominal expressions. We here restrict ourselves to one striking example to illustrate the importance of communication mode. While modern day European narratives are often produced by single-party storytellers, Australian Aboriginal narratives are often produced in a multi-party setting (e.g., Hill 2018; McGregor 2004; Walsh 2016). Hill (2018) shows that, in multi-party stories in Umpila, reference is often co-constructed, and may include numerous repetitions and elaborations. For example, initial reference in multi-party Umpila narratives is often not expected to lead to immediate recognition of the correct referent, but is often semantically more general to invite and encourage co-telling. This is illustrated in (52), where little information is given by the first speaker to identify the referents, which invites the other speakers to contribute more details, like the number of people involved and where they are from.^[23]

7 Overview of the contributions

This section summarizes the contributions to the special issue. They include both general articles and studies focusing on a particular language or area. The articles overlap in different ways, making it hard to order them thematically. We discuss them in alphabetical order.

Nikolaus Himmelmann develops a new, crosslinguistic framework for understanding the relationship between prosodic and syntactic phrasing in building larger units, whether in the nominal domain or elsewhere. Rather than considering prosodic phrasing as derivative of syntactic phrasing, as is often assumed, Himmelmann argues that these constitute two alternative strategies of relating words to each other. On the one hand, syntactic phrasing proper is “prosodically robust” in the sense that it does not depend on an alignment with prosodic structuring. For instance, noun phrases are commonly uttered within single intonation units, but they can also be spread over two or more intonation units. On the other hand, “prosody-dependent” constructions are such structures where prosodic phrasing is not accidental to the construction, but determines its meaning composition. Typical examples are afterthoughts or appositions. While independent, the two organizational principles are connected insofar as the development of prosodically robust constituent structure or “phrase structure proper” presupposes that the syntagm in question is sufficiently often – if not always – produced as a prosodic unit.

Olga Krasnoukhova studies the different nominal construals that are attested in a genetically diverse sample of 65 South American languages, focusing on one functional role, viz. that of attributive modification involving properties. She shows that both within and across these languages, property concepts are expressed by roots belonging to a range of different word classes, with a majority involving verbs, and others involving adjectives, nouns, adverbs or flexible elements which may be used both for establishing or modifying reference and for predication. The variety of word classes involved leads to considerable diversity in available construal types. These include for instance appositional modification with property lexemes that are nouns (‘the good one, the chicken’), complex noun phrases involving relativization of verbal property lexemes (‘the clay which is white’) or possessive constructions with property lexemes that are nouns (‘the house’s smallness’), simple noun phrases (‘the small lake’), and compounds (‘big-head’). Krasnoukhova also suggests a diachronic pathway from predication to attribution in noun phrases.

Yury Lander investigates noun phrase structure in Tanti Dargwa, an East Caucasian language. Noun phrases are typically head-final, while modifiers show flexible ordering with respect to each other, their order possibly determined by factors like information structure and heaviness. In other words, there are no dedicated syntactic positions for modifiers with descriptive semantics (such as adjectives) or those with determining semantics (such as demonstratives). Nonetheless, there are morphological differences between these elements in terms of the attributive suffixes they may or may not take. In addition, only descriptive modifiers but not determiner-like elements may appear in an alternative NP construal type, viz. where they follow the nominal head. Lander focuses on two types of modifiers which show characteristics of both descriptive and determining elements: contrastive modifiers, and a set of modifiers including possessor NPs, ‘other’ and ‘every’. These elements are never fully like either descriptive or determining modifiers with respect to the above-mentioned morphological and distributional features, and a functional account in which they alternatively have attributive or determiner roles does not successfully account for them. For instance, possessors may occur in post-head position like descriptive modifiers, but they may do so without the attributive suffix attached, i.e., showing a formal property of determining modifiers. Lander suggests that determiners form a category which is grammaticalizing but has not reached a stage of obligatorily showing certain characteristics.

Bruno Olsson surveys NP structure in Coastal Marind against the wide-spread background assumption of particularly simple NP structures among Papuan languages. Olsson argues that Coastal Marind essentially allows for two construal types: a tight left-branching structure with a single modifier, besides looser configurations, including discontinuity, which are more seldom employed and only in pragmatically marked contexts. Olsson also shows that the only construal option for attributive qualification is an adjective-noun compound, which has both word-like and phrase-like properties. The second part of the article focuses specifically on discontinuity and observes that it is largely conditioned by information-structural factors similar to those observed in Schultze-Berndt and Simard (2012) for the Australian language Jaminjung, viz. argument focus and sentence focus. He identifies a third function of discontinuity, viz. directing attention. Olsson makes the novel proposal that an underlying condition for discontinuous and other looser construals in Coastal Marind is high ranking of the nominal expression on an “aboutness” scale (i.e., where the utterance provides more information about the referent of the nominal expression).

Eva Schultze-Berndt investigates a hitherto understudied function of discontinuous nominal expressions: they may be a strategy for marking thetic constructions (a subtype of sentence focus constructions). She builds her argument both on the phenomenon of extraposition from subject NP/DP and on discontinuity in a number of Australian languages including Jaminjung/Ngaliwurru, Ngarinyman, and Wagiman. Whereas weight could account for extraposition phenomena, it can be ruled out as a factor triggering the sort of discontinuity she discusses for Australian languages. Instead, sentence focus is a function plausibly signaled by both of these construction types given the contexts in which they appear. Moreover, Schultze-Berndt argues that discontinuity is an iconic strategy reflecting the “informational integration” of the clause as all-new. Like other thetic constructions, discontinuity thus involves detopicalization, i.e., the dissolution of the bipartite topic-comment structure.

Mark Van de Velde investigates patterns of word order, agreement, and prosody in nominal expressions in Bantu languages, all of which show typologically unusual characteristics. Van de Velde specifically draws attention to NP-internal word order patterns which deviate from typological expectations, such as noun-demonstrative-adjective order. He also discusses patterns of widespread semantic agreement in nominal expressions and illustrates the significant variety of agreement marker paradigms across the Bantu languages. Finally, he notes that several languages show evidence for prosodic breaks within the noun phrase, such as the blocking of high tone spreading across the elements of the phrase, or penultimate vowel lengthening appearing in the middle of a phrase. Van de Velde argues that all these characteristics may be explained by two processes of morphosyntactic change that are recurrent across the Bantu family. The first involves modifiers such as adjectives or numerals which are nominalized and used in apposition to the noun phrase they semantically modify. The second development is that in some languages, appositional modifiers are integrated into the noun phrase. Van de Velde thus not only surveys the diversity of nominal construals in Bantu languages, but also suggests a novel pathway of change for nominal expressions.

8 Conclusions

This article lays the foundations for a new typology of how nominal expressions are structured both language-internally and crosslinguistically, and of what motivates the distribution of different construal options. In approaching this vast field of study we have placed particular foci on some issues which have not been in the center of attention, but which we believe play important roles in shaping the variation space of the nominal domain. Particular emphasis is placed on language-internal diversity. Languages differ significantly with regard to the range of construals that are available, and with regard to the role that a particular construal plays within its linguistic and communicative system. Taking one allegedly “basic” construal type as representative of a language, as is done in traditional and more recent large-scale typologies, vastly under-represents the actual data. In understanding construal choices, we discussed a variety of motivating factors including lexical, semantic, and information-structural ones. In order to enrich our understanding of nominal expressions, we adopted several further perspectives. Thus, we discussed the impact of word classes on the structure of nominal expressions. We also considered ways in which nominal expressions may change over time in terms of their structural realizations, how they build up or loose rigidity and/or complexity. We rounded off the discussion by studying the impact of orality, literacy, and communicative modes shaping construals and choices between them.