Patterns of Relative Clauses in Dagbanli

The article examines the patterns of relative clauses in Dagbanli, a Gur language spoken in northern part of Ghana. It focuses on a range of possible RC patterns, and presents a coherent classification using Vries’s model of RC types. The article argues that Dagbanli has two RC types which are characterized by shared features so “indefinite pronoun” that forms a compound with the nominal root, and maa or la “clause-final determiner.” The first RC type is restricted to cases in which the antecedent has subject function within the RC, and the other RC type occurs only with nonsubjects as relativized head making use of postsubjectival particle ni to mark subordinated clauses. It is proposed that Dagbanli has a postnominal word order of N . . . RC . . . D and allows D-type ([[Subj V Obj] D] RC) in-situ HIRC (Head Internal Relative Clause) as well HERC (Head External Relative Clause). It also presents ŋun “who” and ni “which” as question particles that are used to introduce relative clauses in Dagbanli.


Introduction
According to Saah (2010), "a typical relative clause usually consists of an initial Noun Phrase (NP) followed by the modifying clause which can perform any of the grammatical functions in a sentence" (p. 91). In a natural conversation, following Vries (2002), a typical example of a relative clause would be like the following example as presented in Dagbanli: Example 1 zaŋ ti bia so ŋun zu baa maa teke give 3sg child pro rel steal.perf dog Det. "Give it to the child who stole the dog" In the sentence in Example 1, bi-a "child" is the antecedent, ŋun "who" is a relative pronoun that refers to the antecedent bi-so, and ŋʊn zu baa maa "who stole the dog" is relative clause that postmodifies antecedent bi-so. In a sentence in Dagbanli, a relative pronoun also functions as subject as illustrated in Example 2: Example 2 ŋʊn zu baa maa rel steal.perf dog Det. "Who stole the dog?" The Examples 1 and 2 show the possible clause structure patterns in Dagbanli. It is clear in Example 2 that the RC is an embedded clause that forms part of an NP as noted by Saah (2010) who states that syntactically or semantically, the typical relative clause usually consists of an initial NP (the antecedent or head) followed by the modifying clause. Together, they make up one complex NP, which can perform any of the grammatical functions in a sentence such as subject and object. (p. 91) I agree with Saah that the [N RC] . . . ] type of NP can function as a subject as shown in Example 3a or an object as shown in Example 3b in a sentence in Dagbanli: Example 3 a. bi-so ŋʊn zu baa la ʧaŋ-ja child rel steal.perf dog Det go.perf "The child who stole the dog is gone" b. ti ɲa bi-so ŋʊn zu baa la 3pl see.perf child rel steal.perf dog Det "We saw the child who stole the dog" The Examples 3a and 3b show a postnominal word order [N . . . RC . . . D] in Dagbanli. The article examines the possible patterns of relative clauses in Dagbanli and presents a coherent classification. Previous accounts of RC in Dagbanli include Olawsky (1999), Fiedler and Schwarz (2005), Hudu (2012), Issah (2013), and others.
None of the previous study provides the possible different pattern of RC and outlines the features of RC in Dagbanli. This article shows that there are two RC types in the language which are characterized by shared features represented by so "indefinite pronoun" and maa and la "clause-final determiners." Fiedler and Schwarz explain that the first RC type is restricted to cases in which the antecedent has subject function within the RC and the other RC type occurs only with nonsubjects as relativized heads and makes use of postsubjectival particle ni marking also some other subordinated clauses. I propose that Dagbanli allows D-type in situ HIRCs (Head Internal Relative Clauses) as well as HERC (Head External Relative Clause) as analyzed in the data presented in this article. I also argue that ŋun "who" and ni "which" are particles used to introduce relative clauses in Dagbanli. The data are analyzed using Vries's (2002) model of RC types. The rest of the article presents a brief background of Dagbanli while "Data" section discusses data used in this article. "Relative Clauses" section discusses relative clause; "Parametric Variation" section looks at parametric variation of RC (Vries's model of RC types); "Main Types of Relative Clauses" section discuses types of RC, and "Conclusion" section concludes the article.

Background of Dagbanli
Dagbanli belongs to the Oti-Volta subgroup, a major branch of Gur languages within the Niger-Congo family of Africa (Naden, 1988;Naden, 1989;Bendor Samuel, 1989;Hudu, 2010). It has been classified by Naden (1989) and Wilson (1970) as belonging to the Gur language family. Dagbanli is spoken mainly in the Northern Region of Ghana by Dagbamba. Hudu (2012) observes that speakers of Dagbanli are called Dagbamba "plural" and Dagbana "singular." Dagbani and Dagomba are the most used forms of Dagbanli and Dagbamba. Dagbanli has three main dialects: the eastern dialect spoken in Yendi and its surrounding villages; the western dialect spoken in and around Tamale and the Nanunli which is also an eastern dialect spoken in Bimbila and its surrounding.

Data
The data in this article are based on nayaɣili "the eastern dialect spoken in Yendi." The data used are both primary and secondary. The primary data were collected as samples of natural conversations by native speakers in the markets, homes, lorry parks, and focus key informant interviews. The interview methods used were rapid and anonymous surveys interviews and sociolinguistic interviews (Labov, 1966;Milroy & Gordon, 2003). My personal intuition as a native speaker complemented the information I put together from the unstructured interview.
To elicit the data, rapid and anonymous surveys were used to collect the initial data, while sociolinguistic interviews were used on one-on-one exchanges between the researcher and the participants to cross check the initial data obtained. Ten participants; five men and five women were selected and they were between the ages of 25 and 50 years. The data were collected from the villages around Yendi. The interview was selected for the data collection because most of the participants who were selected were men and women who could not read nor write. Participants were contacted personally for their consent before the exercise. The secondary data were taken from Dagbanli existing literature Olawsky (1999) and Issah (2013). Vries (2002) states, "to establish what can be called a relative clause, the definitions that make use of the concepts modification or antecedent are obviously too narrow, since there are appositive relatives, head-internal relatives, etc" (p. 1). Vries thus considered the defining properties of relative clause which are both semantic and syntactic as follows:

Relative Clauses
Example 4 i. "A relative clause is subordinated." ii. "A relative clause is connected to surrounding material by a pivot constituent." Vries (2002) explains that "the pivot is a constituent semantically shared by the matrix clause and the relative clause. Often it is a noun phrase. If it appears to be spelled out inside the matrix clause, it can be recognized as an antecedent" (p. In Example 5a, the matrix is ti ɲa "we saw," the noun is Ali, and the RC is ŋun zu baa maa "who stole the dog." Ali is then the pivot in this construction shared by both the matrix and the RC constituents. In Example 5b, the matrix is ti ɲa "we saw," the pivot is teebʊlʊ, and the RC is a-ni da la"which you bought." I propose that Dagbanli relative constructions are characterized by the use of the "relative pronouns" ŋun or ni and clause-final determiners maa/la (see Fiedler & Schwarz, 2005;Hiraiwa, 2007;Sulemana, 2012). Vries (2002) observes that if the pivot is spelled out inside the relative clause, the construction is head internal: [matrix . . . [RC . . . NP . . . ] . . . . In this case, the matrix contains the gap, which is filled by the whole relative construction (as illustrated in Example 5) or-if RC is preposed-by a demonstrative construction (a correlative construction). Vries (2002) views that, "variation concerning the position and content of the gap is expected, since there are different strategies to cope with the dimensionality problem that (4ii) poses (i.e. the pivot must be in two sentences at once)" (p. 2). An additional essential property of relative clauses which Vries (2002) noted, "the semantic θ-role and syntactic role that the pivot constituent has in the relative clause, are in principle independent of its roles in the matrix clause" (p. 2). This is briefly illustrated in Example 6 where paʔ-a "woman" is the pivot and the experiencer in the main clause (paʔ-a la kpia "the woman is dead") and patient in the relative (paʔ-a-so a-ni ŋme la "the woman whom you hit."). Syntactically, it is a subject in the main clause and direct object in the subordinate.
Example 6 paʔa-so a-ni ŋme la kpia woman 2sg.rel hit.perf Det die.perf "The woman whom you hit is dead" In Example 6, the gap in the RC representing paʔa-so "woman" is both semantically and syntactically independent of its roles in the main clause. Languages can restrict the number of available internal roles, that is, they can be scaled differently on a grammatical function hierarchy (Bakker & Hengeveld, 2000;Keenan & Comrie, 1977;Lehmann, 1984, p. 219) cited in (Vries, 2002) who, however, mentioned that in many languages prepositional objects and lower functions are not possible relative positions. There are also languagedependent constraints that have to do with the possibility of recovering the function of the relative "gap" (Givón, 1984). Vries (2002) observes that differences between relative clauses can be found on any imaginable aspect of the construction based on the RC models of sample pattern. He mentions that "231 relative strategies put together from 176 languages around the world. They are compiled from typological data in Peranteau (1972), Keenan and Comrie (1977), Downing (1978), Comrie (1981), Givón (1984), Lehmann (1984), Keenan (1985), Smits (1988), and Culy (1990) cited in Vries (2002, p. 3 In this article, I will briefly illustrate the contrasts between relative clause patterns that can be found in Dagbanli according to the differences mentioned by Vries (2002) models. These models are illustrated in the Examples 7 to 17.

Vries's Model of RC Types
Kind of modification/relation. These are RC forms that show restrictive, appositive, and degree: Example 7 a. Ali ɲa ʧiʧa so ŋʊn bi paasi teesi la Ali see.perf teacher pro rel neg pass.perf exam Det "Ali saw the teacher who did not pass the exam." b. Ali ɲa ʧiʧa maa,ŋʊn bi paasi teesi la Ali see.perf teacher Det rel neg pass.perf exam Det "Ali saw the teacher, who did not pass the exam." c. *Ali ɲa kᴐm din ha be lᴐʔ-ʊ maa ni Ali see.perf water rel pro copula pot Det loc "Ali saw the water that there was in the pot." Example 7a shows RC modification as restrictive without a comma triggered by the pronoun so. The presence of the pronoun calls for the restricted RC to identify who the object is, where Example 7b is appositive with a comma separating the main clause from the RC that postmodifies without restriction. This occurs in the language because the definite article maa identifies the object with or without the RC. The Example 7c is ungrammatical because degree relative relation does not occur in the language. Vries (1998) notes that a relative clause can be semantically restrictive, appositive (nonrestrictive), or maximalizing.
Hierarchical status of RC. These are RC forms that are embedded within DP or it is correlative.
Example 8 Adam ŋʊn da bu-a maa kana kpe name rel buy.perf goat.sg Det come.perf loc "Adam who bought the goat came here." Example 8 shows that the RC is an embedded clause that is found within the DP or the NP "Adam," the subject of the sentence. The relative clause postmodifies the subject "Adam" and forms part of the DP or the NP. This shows the hierarchical status of RC as an embedded clause within DP or NP in Dagbanli. Example 9a shows a headed relative clause in Dagbanli with the presence of a head noun pɨn-ʃɛŋa "presents" postmodified by the RC while Example 9b shows a free relative with the absence of a head noun that is postmodified by the RC. This shows that some RCs in Dagbanli can occur with or without a head.

Presence/absence of relative pronoun
Example 10 a. Azima kaagi ji-li ʃɛli ni da la Azima visit.perf house pro rel buy.perf Det "Azima visited the house which I bought." b. Azima kaagi ji-li la Azima visit.perf house Det "Azima visited the house which I bought." Example 10a shows the presence of a relative pronoun ni in a clause, whereas Example 10b shows its [ni] absence. This means that it is possible to have the presence of a relative pronoun or without it in a clause structure in Dagbanli depending on the context.

Presence/absence of complementizer
Example 11 a. Azima kaagi ji-li ʃɛli ni da la Azima visit.perf house some rel buy.perf Det "Azima visited the house which I bought." b. Azima kaagi ji-li la Azima visit.perf house Det "Azima visited the house which I bought." The data in Example 11 show that a clause can show the presence/absence of a complementizer and still serve the intended purpose or meaning.

Presence/absence of resumptive pronoun
Example 12 a. bi-so ŋʊn zu baa maa ʧaŋ-ja child rel steal.perf dog Det go.perf "The child who stole the dog has gone" b. *bi-so ŋʊn zu baa maa n ʧaŋ-ja child rel steal.perf dog Det 3sg go.perf "The child who stole the dog has gone" The data in Example 12 show that Dagbanli does not have resumptive pronouns as a feature of relative clauses. Inserting a resumptive pronoun renders a construction ungrammatical as in Example 12b.

Hierarchical position of head
Example 13 a. wahʊ-so doo maa ni da-la ɲala wahʊ viɛli Horse.pro man Det. rel buy.perf Cop horse good "The horse that the man bought was a good horse." b. doo maa ni da wahʊ-so la ɲala waɣ viɛli man Det rel buy.perf horse.pro Det cop horse good "The horse that the man bought was a good horse." The data in Example 13 show that, in Dagbanli, both HIRC (Example 13a) and HERC (Example 13b) occur in clause constructions. This is discussed in Example 21 in this article.

Linear order of head and RC
Example 14 a. n labi bʊkʊ ʃɛli a-ni ti-ma maa 3sg lose.perf book some 2sg.rel give.perf.1sg Det "I have lost the book you gave me" b. a-ni ti ma bʊkʊ ʃɛli maa n labi-li 2sg.rel give.perf 1sg book pro Det 1sg lose.perf.3sg "The book which you gave me is lost." The Example 14a

Inflectional completeness of RC
Example 15 bʊkʊ ʃɛli a-ni ti ma la ʧaʔɨ-ja book pro 2sg.rel give.perf 2sg.obj Det tear.perf "The book which you gave to me has been torn up." Example 15 shows that Dagbanli inflectional completeness of RC occurs where Dagbanli marks for both participial RC and finite RC. It shows that the same lexical verb marks for both participial and finite in the language.

Position of determiner with respect to N and RC (initial, middle, and final)
Example 16 a. n mini doo-so ŋun mi-a maa n tɔʔɨsɨ 1sg conj man-pro rel know.perf Det 1sg.speak.perf "I spoke with the man who knows you." b. *n mini doo-so *maa ŋun mi-a ntɔʔɨ 3sg. conj man Det rel know.perf 1sgspeak.perf "I spoke with the man who knows you." The determiner can only occur in final position with respect to N and RC in a construction as in Example 14a, but Example 14b is ungrammatical because the determiner cannot occur before the relative pronoun. The determiner only occurs in final position in RC but not in both medial and initial. With nouns, it also occurs in word final like the RC.

Position of (case) markers, if any
Example 17 a. bi-so ŋʊn zu baa la ʧaŋ-ja. Child-pro rel steal.perf dog Det go.perf. "The child who stole the dog has gone" (cf. Vries, 2002, p. 3) The position of case marking in Dagbanli is possible because they are used as relative clause markers as illustrated in Example 17. "When a pronoun occupies a position in which we would have expected an unpronounced copy of a whconstituent we refer to it as a resumptive pronoun" Haegeman (2006, p. 364, cited in Sulemana, 2012.

Generalization
From the Parametric variation of RC, the data show that Dagbanli is an subejct verb object (SVO) language which does not permit all the possible patterns of RC as marked ungrammatical in some of the patterns. According to Fiedler and Schwarz (2005), Dagbanli has two RC types at its disposal which share the following features: Example 18 (i) The head is represented by so "an indefinite noun class pronoun" that forms a compound with the nominal root or it is used alone, (ii) Determiner máá (sometimes lá) that is added to the end of the RC. (p. 125) The data above agree with Fiedler and Schwarz's assertion in Example 18 as illustrated on Examples 1, 3, 6, 7, 12, 13, and 16. It is confirmed from the data that the determiner can only occur in RC clause final as shown in Example 16 and almost in all the clauses in the data. Fiedler and Schwarz (2005) further explain . . . One of the RC types is restricted to cases in which the antecedent has subject function within the RC. Apart from the two features mentioned above, it is formed with the help of a disjunctive pronoun in the subject slot which follows the head and fulfills here the function of a relative pronoun. The other RC type occurs only with non-subjects as relativized heads and makes use of post-subjectival particle n(i) marking also some other subordinated clauses. In this head-internal RC type the head is either retained in its postverbal slot (cf. the circumnominal ex. 20a) or it is moved to the initial position of the relative clause (cf. the postnominal ex20b). (Fiedler & Schwarz, 2005, p. 125).
They presented the data below as evidence to support their explanation:

Non-subject
Example 20 a. a-ni ŋme do-so maa nya-la n zo 2sg-CNJ hit man-IND:CL DET COP-FM 1sg friend "The man whom you have hit is my friend." b. do-so a-ni ŋme maa nya-la n zo man-IND:CL 2sg-CNJ hit DET COP-FM 1sg friend "The man whom you have hit is my friend." (cf. Fiedler & Schwarz 2005, p. 126) The data presented in Examples 6, 7. 12, 13, 14, and 16 confirm the generalization by Fiedler and Schwarz (2005;p. 125). Hiraiwa (2007) noted that "[C]ross-linguistically, HIRCs are roughly classified into two types: the nominalization-type and the D-type" (see Hiraiwa, 2005, p. 2). He explained the nominalization-type HIRCs are relative clauses where the entire relative clauses are nominalized by a nominalizing suffix or a complementizer. On the contrary, the D-type HIRCs use an external determiner instead of a nominalizer/C. This is illustrated in Example 21: Dagbani: (Peterson, 1974) a. [n ni puhi saan-so la] chaŋya 1Sg. ni greeted stranger-Spec.Id D has-gone "The stranger who I greeted has gone." (cf. Hiraiwa, 2007, p. 3) b.

Main Types of Relative Clauses
According to Vries (2002), there are four main types of relative clauses as illustrated below: Each type has a headed and a free variant which has been shown for postnominal relative clauses in Example 14. The data in Examples 14,19,20,and 22 show that Dagbanli has possible RC patterns that are postnominal and prenominal in a sentence. This is exemplified extensively in Lehmann (1984, cited in Vries, 2002. Some important absolute and scalar differences between the four types are summarized in Tables 1 and 2. As illustrated in (14) and (22) above, circumnominal relatives and correlatives have an internal head. The former type is nominalized, i.e. it is a DP (see e.g. Culy, 1990)-hence there can be an external case marker or determiner. Thus only correlatives are bare sentences, which are almost always leftadjoined to the matrix clause. (Vries, 2002, p. 5) "Prenominal relatives show strong nominalization phenomena often with nominalizing affix, there can be temporal and modal limitations, etc. " (Vries, 2002, p. 5) e.g. (15) above. Although postnominal relatives are the most common, the other types occur in different language families across the world (Vries, 2002, p. 5). This is illustrated in Dagbanli in Examples 14 and 20.

RC in Dagbanli
RC in Dagbanli is constructed with an obligatory head and a relative pronoun. Olawsky (1999) observes that another type of subordination in Dagbanli is the relative construction and explains that the relative pronouns are of complex nature, as they consist of two elements; the first drawn from the set of indefinite pronouns whereas the second element is taken from the set of emphatic pronouns. Dagbanli RC has the following salient features illustrated in Example 24:  The Examples 25a and 25b show both head/antecedent NP (bi-a "child," su-a "knife") and animate subject bi-a as well as inanimate subject su-a in Dagbanli RC. They also show obligatory relative clause markers ŋun and ni and both have clause-final determiners maa in Example 25b or la in Example 25a. Example 25a shows indefinite noun class pronoun so for animate subject while Example 25b presents ʃɛli for inanimate subject.

The Relative Complementizer
English, for example, uses "case-marked" relative pronouns which "are derived historically from case-marked interrogative pronouns such as who, whom, when, etc" (Givón, 1993, p. 12, cited in Saah, 2010. Issah (2013) notes that Dagbanli uses the particles ŋun/ni to introduce RC. The particle comes after the head NP, and it selects a sentence/clause as its complement. This is analyzed as a relative complementizer (see Saah, 2010). The relative complementizer is compulsory in Dagbanli because when it is deleted, the sentence is rendered ungrammatical. This is shown in Example 26: Example 26 a. bi-asoŋʊn kana kpe maa da-la loori child pro rel.cop come.perfloc Detbuy.perf car "The child who came here bought a car" b. *bi-so kana kpe maa da-la loori child come.perf loc Det buy.perf car "The child who came here bought a car" The ungrammatical sentences in Example 26b provide evidence that Dagbanli has an invariant relative clause complementizer whose function is to introduce the RC as illustrated in Example 26b.

Restrictive Versus Nonrestrictive Relative Clauses
In English, the two types of relative clauses differ in both their morphosyntax and semantics. Phonologically, nonrestrictive relative clauses are "pronounced with a comma intonation, i.e., with pauses after the head and the relative clause. Restrictives do not have this comma intonation (Comrie, 1981 Example 27a has a nonrestrictive relative clause as indicated by the commas. Fabb (1990, p. 57, cited in Saah, 2010) also asserts, "an RR[C] modifies its host, while an NRR[C] does not," and that the NRRC "has no syntactic relation to its host/antecedent." Watters (2000, p. 225, cited in Saah, 2010 asserts that the distinction between restrictive and nonrestrictive relative clauses "is generally not marked in African languages." The closest we can come to an appositive reading of relative clauses is when they are extraposed. Saah (2010, p. 106) concludes that Akan relative clauses are essentially restrictive in nature, this is confirmed in Aboh and James (2010, p. 28), and in Dagbanli one of the RC types is restricted to cases in which the antecedent has subject function within the RC as in Example 28 Example 28 do-so ŋun chaŋ maa ɲɛla n zo man-IND:CL 3sg.DISJ go DET COP-FM 1sg friend "The man who has left is my friend." (cf. Fiedler & Schwarz 2005, p. 125)

Conclusion
The article explores the possible patterns of relative clauses in Dagbanli and presents a coherent classification. It shows that the two RC types in Dagbanli are characterized by features represented by so "an indefinite noun class pronoun" and clause-final determiners maa and la. It shows that Dagbanli allows D-type in situ HIRC as well HERC as presented in the data. It also has a word order of [N . . . RC . . . D]. The article illustrates the contrasts between relative clause patterns in Dagbanli according to the differences between relative clauses that can be found on any imaginable aspect of construction based on the RC models described in Vries (2002). From the Parametric variation of RC, the article shows that Dagbanli is an SVO language which does not permit all the possible patterns of RC. The article discussed RC formation in Dagbanli and concludes that RC in the Dagbanli has all the possible features discussed in Saah (2010) except one feature-a resumptive pronoun in the relativized position. I aso argue that ŋun "who" and ni "which" are particles used to introduce relative clauses in Dagbanli. It also discussed possible relevant RC patterns in Dagbanli that reflect Vries's (2002) approach adopted in this article. It is shown that the Position of Determiner with respect to N and RC, the determiner can only occur in final position respect to N and RC in a construction but the determiner cannot occur before the relative pronoun. Almost all the models tested are relevant in Dagbanli showing the presence of head headed/free relatives, relative pronoun, and complementizer except the presence of resumptive pronoun which seems not to occur in the language.

Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding
The author(s) received no financial support for the research and/or authorship of this article.