A Prototypical Model on Hakka Serial Verb Constructions

This research paper provides a meaning-based account to examining Hakka syntactic constructions that comprise multiple verbs in their scope. The investigation is based on an interdisciplinary approach from the interface of syntax and semantics. The paper is organized into two main parts. The first part of this paper claims that the prototypical construction of the serial verb construction is a syntactic configuration that contains two verbs in the same clause, indicating two interdependent subevents happening at close time intervals. In addition, the paper proposes that greater distance in structural and semantic interdependence between the two verbs forms a gradation deviating from the prototype. In this part, a prototype model, rather than a criterial attribute model, is adopted to define the Hakka serial verb construction (SVC). The second part of paper provides a typological study that classifies the Hakka SVCs into subtypes based on the syntactic structure and the semantic relationship of the component verbs. Syntactic tests are used to test the clausehood of the multi-verb constructions identified in this part.


Hakka Serial Verb Construction
In this section, we first briefly review the wild discussions surrounding the SVC in the literature of linguistics in terms of the basic definition and scope of the syntactic configuration. Then we show that the prototype theory better captures the uncertainty of the construction in comparison with the tradition criterial-attribute model.

Literature on SVC
Serial verb construction (SVC) has undergone quite intensive and extensive discussions by many linguists around the world. An issue that has not yet achieved consensus concerns what should and should not be considered an SVC. The construction can be defined broadly as including any string of verb phrases or clauses juxtaposed together. As in Li and Thompson's (1981) definition of Chinese SVCs, all the following constructions are recognized as SVCs as long as there is no grammatical marker occurring between the two constituents (Li & Thompson, 1981, p. 594): (1) Two or more separate events. (2) One verb phrase serving as the subject or direct object of another verb. (3) Pivotal constructions. (4) Descriptive clauses. Under this definition, an SVC comprises both single and multiple clauses. On the other hand, more linguistic studies have proposed a mono-clausal schema to describe the structure of SVCs. For example, Steward (1963) and Bamgbose (1974) suggested that an SVC is a mono-clause formed from two or more underlying clauses, which may involve meaning change during the process of syntactic transformation. Dai (1990) distinguished three types of SVC: subordination, coordination, and serialization. However, he argued that only serialization forms a true SVC, while subordination and coordination are noted as single-headed and double-headed multiclausal constructions, respectively.
A number of studies have contributed to a categorization of SVC subtypes. Some research involved cross-linguistic investigations, such as Aikhenvald (2006a), who argued that four parameters can be adopted to categorize SVCs. Composition distinguishes between the symmetrical type, in which two component verbs have equal status, and the asymmetrical type, in which the two verbs show a target-modifying relationship. Contiguity distinguishes between the contiguous type, which contains two verbs in a row, and the noncontiguous type, which allows other constituents intervening between the verbs. Wordhood distinguished between one-word and multiword constructions. Inflection distinguishes between single marking and concordant marking. In contrast, the classification of SVCs in other studies was based on investigations of one particular language. Christaller

The Prototype of SVC
The formal generative approaches usually adopt the criterial-attribute model to define linguistic structure and deal with linguistic categorization, in which a category is defined by a set of features and membership to a class requires the possession of all the properties on the list. However, as discussed in Section 1.1, since there is no agreement among linguists concerning what kind of syntactic configuration should be considered an SVC, one can hardly come up with a complete list of criterial attributes for the SVC that identifies all the members and filters out all the nonmembers. Instead, this paper follows the cognitive tradition and adopts the prototype theory. The theory originated in cognitive psychology in the 1970s (Rosch, 1973(Rosch, , 1975(Rosch, , 1977Rosch & Mervis, 1975) and was adopted by linguists around the early 1980s. Since then, the prototype theory has grown steadily and was established as a central notion, especially in the field of cognitive semantics (Wierzbicka, 1985;Lakoff, 1987;Langacker, 1987). In this paper we are concerned with the theoretical application in the subfield of linguistics. The prototype model proposes a graded categorization in which categorization is regarded as a matter of degree. In the prototype model, some members of a category are considered more central, recognized as the prototypical instances, whereas other members form a gradation from central to peripheral depending on the distance by which they deviate from the prototype.
The SVC label should ideally be general to include all structural patterns that contain more than one verb that share the same grammatical subject in the syntactic scope, wherein each verb indicates a subevent that cooperates with other subevents to form a main event. Figure 1 diagrams the situation. The component subevents are shown as boxes. These boxes are interconnected with semantic relations, which are indicated by association lines. These subevents are enclosed by a solid, thick-line circle to indicate that they collectively form a larger main event. By contrast, those co-occurring verbs found in (4) do not form prototypical SVCs. They are identified as atypical SVC as they possess all of the features listed in (1) but violate some of the features listed in (2). In (4a) the two verbs name two daily chores that Hakka people usually do in the countryside. In (4b) the two verbs are semantically related in that the first verb indicates the manner that describes how the subject talks to someone. In (4c), a cause-effect relationship is identified. In (4d), a propositional saying verb brings another clause as the complement. However, as shown here, at least one of the features listed in (2) is violated. For (4a), the two verbs can switch their position without changing the meaning of the sentence, violating (2a). As to (4b), there are elements intervening between the two verbs, violating (2b). In (4c), a grammatical linker ʒen ("because") is overt indicating the cause-effect relationship of the two verbal components, violating (2c). In (4d), the two verbs are bound in a predicate-argument relation, violating (2e).
(4) a.  (1) and (2). In (5a), the two verbs indicate two events. The time expressions ts'okpuȵit ("yesterday") and kimpuȵit ("today") signal that the two events are proceeding along different timelines. The continuity of a time span is interrupted by the introduction of a time signal. In (5b), the two verbs ts'iuʃui ("swim") and pa ("carry on the back") have different subjects. Any switch of the subject introduces a new event because subject change signals some kind of discourse discontinuity. Therefore, (5a) and (5b) are not serial verb constructions due to violation of both (1) (1) and (2) ctions that ful ng those that v As shown in nonical catego Hakka serial v different categ mponent verbs cles to subdivi tions deviate f erb constructio As mentione To be specific nstruction conta mantic relation earlier in (5) The examples in (6a) and (6b) show that the introduction of a subject signals some kind of discourse discontinuity. The repeated occurrence of the same subject also highlights the discontinuity because the presence of a subject forces a repeated subject role to re-enter into the center of play. Examples (6c) and (6d) show that a time signal may break into the continuity of a time span. In (6c), the two temporal expressions signal that each of the individual subevents is proceeding along different timeline. In (6d), the time expression tatȵit ("everyday") initiates a new timeline for the second subevent. In (6e), the first subevent expresses the mental state of the kid in his receipt of the gift; the second subevent describes his action with the gift possession. The semantic relationship between the two subevents is unclear; that is, the semantic bond of the two subevents is weak. As shown in (6), even though these sentences contain multiple verbs, they are considered instances of non-SVC due to the violations of (1).

Syntactic Tests
Linguists use different criteria to test the clausehood of a syntactic construction. While Haspelmath (2016) follows the proposal of Bohnemeyer el al. (2007, pp. 500−501) and uses an independent negation as a cross-linguistically applicable test for clausehood, he also cites Cleary-Kemp (2015) and van Staden and Reesink (2008), who argue that it is not possible to find the criteria that can be applied to all languages for measuring the independence of clausehood; instead, they argue that the criteria for clausehood is generally language-specific.
The tests used to determine whether a syntactic configuration is mono-clausal or bi-clausal are not universally applicable, but they in general are appropriate to particular languages.
In this paper, we adopt the following criteria to test the independence of clausehood for a serial verb construction, which based on our definition, includes all syntactic configurations that contain multiple verbs, and these verbs share the same subject and indicate subevents that are interdependent to each other. First, we use negation to test monoclausality in the sense that negation has scope over the minimal clause that includes all of these verbs. Second, an SVC is formed on the basis of a complex predicate, which according to Butt (1995), is associated with a functional structure that contains a single subject and a single predicate. Therefore, we use the placement of a subject and the insertion of a conjunction to test monoclausality. Third, it is argued that the component verbs of an SVC should be marked by the same tense, aspect and modality (Aikhenvald, 2006b;Baker, 1989;Brown, 2008;Schiller, 1990), hence we use the placement of an auxiliary to test the scope of a mono-clause. Fourth, we try to switch the order of the verbal complexes to test the interdependence of the two subevents indicated by the verbs. The illustration of syntactic tests is explained in the following Table 1. Result placement of a negation marker mo or moi monoclausality -occur with the first/second verb  mono-clause -occur with both of the verbs  multi-clause placement of a modal auxiliary monoclausality -occur with the first/second verb  mono-clause -occur with both of the verbs  multi-clause placement of another overt subject monoclausality -occur with one of the verbs  mono-clause -occur with both of the verbs  multi-clause placement of a conjunction monoclausality -yes  multi-clause -no  mono-clause the order of verbs interdependence of events -fixed position  strong interdependence -switchable position  weak interdependence

Prototypical SVCs
A prototypical SVC is mono-clausal. It contains at least two verbs in one clause. The two verbs are often contiguous, and their relative order is fixed. The verbs must share the same subject. They are semantically related, and they collaborate to contribute to one event.
I. Loi/hi-constructions Dai (1990) defines syntactic constructions that involve loi and hi as the "verb serialization" type, projecting a sequence of two verbs V 1 V 2 , and V 1 is either the verb loi ("come") or hi ("go"). Some examples are provided in (7), in which loi and hi directly precede another verb, functioning to introduce a purposive phrase.

II. Serial Resultative Constructions
When two verbs are juxtaposed, the second one often indicates the result of performing the action of the first verb. The V 1 V 2 construction is sometimes identified by Chinese linguists as a compound in which V 2 acquires the lexical status as a grammatical particle. However, there are also evidences showing that the V 1 V 2 pattern is not always inseparable (Matthews, 2006;Wu, 1992); that is, they can be identified as individual verbs. Some examples of this kind are shown in (9).

III. Shared Object Constructions
Some linguists regard "shared objecthood" as a significant feature for defining SVCs (Baker, 1989;Stewart, 2001). In Mandarin Chinese, we do find many instances where juxtaposed verbs function like a syntactic compound and that they share not only the same subject but also the same object. Examples are given in (11) In this section, we see that the application of the above syntactic tests for all subtypes of prototypical SVCs has shown the same results, which are summarized in Table 2. As shown in the table, the negator and the auxiliary only appear before the first verb and they have scope over the sequence of the two verbs. Only one subject is allowed to be present, and the insertion of a conjunction it not allowed. The ordering of verbs is fixed. The two verbs cannot switch their position with one another. These tests show a strong monoclausality for all of the subtypes that are identified into the prototypical SVC category.

Non-Prototypical SVCs
Non-prototypical SVCs also contain multiple verbs, but the constructions are relatively weaker in terms of their monoclausality. The syntactic tests often show contradictory results since syntactic constructions that are classified into the non-prototype do not possess all of the features in (1) and (2). As shown in this section, some tests argue for a mono-clausal analysis while some argue for a bi-clausal analysis regarding the same syntactic construction.
This section identifies seven subtypes of Non-prototypical SVCs. The first three subtypes are argued to be bi-clausal. The syntactic tests give evidence by showing that for these subtypes, each verb can have its own subject; besides, an overt connector is almost always possible. The other four subtypes, contrastively, have stronger monoclausality as most of them allow only one subject, one negative particle, and one modal auxiliary for the verb sequence.

I. Constructions with an Overt Connector
Some serial verbs are combined by an overt connector. The connector may indicate different kinds of semantic relationship. For example, in (13), a cause-effect relation, a resultative, and a purposive relation are identified because each sentence contains an overt grammatical indicator that directly points out the semantic relation between the two component verbal constructions. In (22a), the connector ʒen indicates the cause-effect relation; in (22b), the complementizer tet brings a resultative clause to explain the degree how those beasts have been scared; in (22c), the complementizer loi introduces a purposive phrase to the preceding phrase. The purpose for the attempt to change the method is to run another round of competition.
The following (14) shows that for each kind of syntactic test, there are always syntactic configurations that are compatible with the meaning associated with the syntactic process.  Vol. 11, No. 2;2021 ……kaiteu ʒatʃ'u ts'iu hak tet p'inmiaŋ tseu la.
The co-occurring verbs in the examples (13) have potential to be negated at the same time; each has potential to take an individual auxiliary, to include an overt subject in their own structure, to be connected with an overt conjunction, and in some instances they can even switch their position. All these argue for a bi-clausal analysis for this subtype of construction.

II. Complement Clause
There are constructions that involve two verbs, one functions to provide an explanation or context that further explains the other as shown in (15) In (15a), the verb of saying ham ("yell") requires a clausal complement to further describe the content of speech. In (15b), the clause "be very disgusted at her" functions as the complement of the verb pian ("become"), explaining how the subject has changed her attitude toward her daughter in law. In (15c), the thinking verb ti ("know") also requires a clausal complement "be deceived by the centipede" to further explain the information the chicken knows. In all of the aforementioned cases, the two verbs are connected through the process of complementation.
For this subtype of SVC, the bi-clausal status is as obvious as the previous subtype according to very similar evidence as shown in (16). Kaikuŋ ts'iu voi ti voi pun ŋ ̣ kuŋ ts'uŋ tsap'ian hi le.

Overt Subjects 1 subject 
Gi ham ts'in ʒanvoŋ.

subjects 
Gi ham gi ts'in ʒanvoŋ.
As shown in (16), the two clauses can both be marked by an individual negator, can take separate auxiliaries, can take an individual overt subject, can be intervened by a complementizer kong, arguing for a bi-clausal analysis for the constructions at issue.

III. Coordination of Events
In some cases two verb phrases are juxtaposed and parallel to one another, conjoining two subevents through coordination, as shown in (17) In (17a), two parallel constructions, hold hands and hold feet, form a unit; another pair of constructions, dance and sing, form another unit. Each unit consists of two coordinated events, and the second unit is set as the purpose for the event indicated by the first unit. They hold their hands and feet in order to dance and sing. All four of the subevents happen simultaneously and cooperate to construct a happy, cheerful wedding scene. In (17b), the two subevents, "put on makeup" and "dress up", indicate the membership on the list that the bride was requested to do. The two subevents are expressed by a series of two verb phrases. The following (18) (18), all of the syntactic tests suggest a bi-clausal analysis for the coordination, including negation, auxiliary, subject, conjunction, and verb ordering tests.

IV. Manner/Instrument-Act
Two subevents often cooperate to form a major event. As shown by the two sentences in (19), each of the first verbs indicates the tool or the manner by which the subject carried out the action of the second verb. For this type of construction, the subevents indicated by the two verbs collaboratively contribute to one major event. The results imply a mono-clausal analysis by showing that the negative marker moi must appear before the first verb, and the scope must extend over the entire clause. We cannot simply negate the subevent indicated by the second verb. Similarly, the restriction against an overt auxiliary or an overt subject for each of the verbs argues for a strong interdependence between the two subevents. By contrast, the conjunction and the ordering tests suggest a bi-clausal analysis for the same constructions in (19).

V. Constructions Showing Repetition of Movement
We also find SVCs consisting of two verbs denoting actions that occur repetitively one after another on the same timeline. Examples are shown in (21). In (21a), the actions "swim swiftly here" and "swim swiftly there" are expressed by two consecutive verb phrases. The two phrases are juxtaposed next to one another to show a continuous, repetitive movement. In (21b), the two verb phrases "ask detailed/long questions" and "ask simple/short questions" again indicate two actions that happen repetitively targeting at the same person. The following (22) presents the results of the proposed syntactic tests.
Negation 1 st verb  mo ts'iu loi ts'iu hi; mo mun tʃ'oŋ mun ton 2 nd verb  *ts'iu loi mo ts'iu hi; *mun tʃ'oŋ mo mun ton both verbs  *mo ts'iu loi mo ts'iu hi; *mo mun tʃ'oŋ mo mun ton Auxiliary 1 st verb  oi ts'iu loi ts'iu hi; oi mun tʃ'oŋ mun ton 2 nd verb  *ts'iu loi oi ts'iu hi; *mun tʃ'oŋ oi mun ton both verbs  *oi ts'iu loi oi ts'iu hi; *oi mun tʃ'oŋ oi mun ton Overt Subjects 1 subject  gi ts'iu loi ts'iu hi; gi mun tʃ'oŋ mun ton 2 subjects  *gi ts'iu loi gi ts'iu hi; *gi mun tʃ'oŋ gi mun ton Conjunction ts'iu loi ʒu ts'iu hi mun tʃoŋ ʒu mun ton Verb Ordering *ts'iu hi ts'iu loi; *mun ton mun tʃoŋ The syntactic tests argue for a strong monoclausality for this subtype of SVC. As shown above, only one negator and one auxiliary is allowed, and they must precede the entire VP sequence if they occur. The two verbs must share one overt subject. Verb ordering is usually fixed. The only test that claims a bi-clausal analysis allows the possible insertion of a coordinating conjunction between the two phrases.

VI. Constructions Showing Immediate Result
Each of the sentences in (23) contains at least two verbs, and the second verb shows the immediate result that takes place as soon as the action indicated by the first verb has been performed. In (23), the action of the first verb causes an immediate impact as denoted by the second verb. In (23a), the jump instantly causes the death; in (23b), the movement happened right after he carried the ghost onto his shoulder. The syntactic tests are shown in (24).  (24), the negator and the auxiliary are allowed to mark either the first or the second verb, but they do not simultaneously mark both of them. Here we note that if both verbs are negated, the second verb no longer shows the immediate result triggered by the impact of the first verb. Instead, they are bound in a resultative relationship. The interpretation is "he didn't jump into the sea; therefore, he didn't die," in which the first verb does not have a spontaneous, immediate impact on the second verb. The results in (24) also show that the insertion of an overt conjunction is possible, which argues for a bi-clausal analysis. However, the negation, the auxiliary, the subject, and verb ordering tests all speak for a strong interdependence relationship between the two subevents, suggesting a mono-clausal analysis.

VII. Resultative and Purposive Constructions
We can identify two types of cause-effect relations in Hakka SVCs. In the first type, the event indicated by the first verb leads to the event indicated by the second verb, as in (25a) In (25a), the trade results in an income of thirty-six dollars. In (25b), the purpose for the three persons to go to the pigsty is to see the pigs. In both instances, the second verb indicates either the result or the purpose of performing the action of the first verb. The results of syntactic tests are provided in (26). The negator and the auxiliary can occur with either the first or the second verb, and in many cases, they can occur with both verbs at the same time. In addition, the conjunction may appear between the two verbs, suggesting a bi-clausal analysis for the construction. Other syntactic tests, including subject insertion and verb ordering, show a strong interdependence between the two subevents, arguing for a mono-clause analysis.

Discussion-Modify the Prototype Model
Finally, we summarize the results of syntactic tests for each of the aforementioned subtypes of SVCs as shown in Table 3 below.
Based on the results shown in Table 3, we have the following discussions. The discussion of each point will not be attempted in-depth. We will leave them for the matter of future research.  Vol. 11, No. 2;2021 some cases the semantic relationship between the two verbs is unclear, while in other cases two subjects are present for each verb.

Conclusion
While disagreement on the definition, properties, and classification of SVC is abundant, this paper seeks to provide a foundation that discusses the construction from different perspectives.
First, Hakka SVC can be investigated with a meaning-based cognitive approach. This paper pursues a categorization of different subtypes of SVCs based on the semantic relationship between the verbs involved.
Since the co-occurring verbs in SVC predict two semantically related subevents, we can classify the constructions into different subtypes based on their semantic relationship.
Second, this paper defines SVCs by proposing a prototype model for their construction. In Chinese languages, it is difficult to give a precise definition concerning what kind of linguistic construction can be considered an SVC. This is because Chinese languages, including Hakka, allow the subject to be dropped in many situations, especially in the context where the user is telling a story. Therefore, in this paper we adopt the prototype model to define SVC and try to be as general as possible to include all structural patterns containing two co-occurring verbs in sequential clauses and a shared subject. We also define what should be counted as a prototypical SVC, which is a syntactic configuration that contains two verbs in the same clause; further, the two verbs share the same grammatical subject, and they indicate two highly structurally and semantically interdependent subevents that can be attested by syntactic tests. In addition, this paper also proposes that a greater distance in syntactic and semantic interdependence between the co-occurring verbs forms a gradation deviating from the prototypical SVC.