The study also highlights the constructs of current linguistic theory, arguing for distinctive features and the notion 'onset' and against some of the claims of Optimality Theory and Usage-based accounts.
The importance of Henk Zeevat's new monograph cannot be overstated. [...] I recommend it to anyone who combines interests in language, logic, and computation [...]. David Beaver, University of Texas at Austin
AUTHOR: Cristofaro, Sonia TITLE: Subordination SERIES: Oxford Studies in Typology and Linguistic Theory PUBLISHER: Oxford University Press YEAR: 2003
Szymon Slodowicz, Centre for General Linguistics, Typology and Universals Research (ZAS), Germany.
The book presents a cross-linguistic analysis of subordination based on a sample of 80 languages. The analysis comprises complement, adverbial and relative relations. The main goal of the study is to present functional correlations between different kinds of subordination relations and individual morpho-syntactic phenomena used to encode them.
Chapter 1 Theoretical premises
Cristofaro's study of subordination is situated within the typological-functional framework and differs form previous works on this subject such as Noonan (1985) or Givón (1990), primarily in the strictly functional definition of the subject of study. The term subordination is used with regard to a particular way to construe the cognitive relation between two events, such that one of them, (&) lacks an autonomous profile, and is construed in the perspective of the other event. (p.2) This functional definition allows for a comparison of a wider range of languages and structures as it was possible with traditional methods. The results the study are implicational hierarchies, primarily the Subordination Deranking Hierarchy (SDH) which orders different subordination relations with respect to the possibility of their use in a language.
Chapter 2 The Notion of Subordination
Cristofaro hopes to avoid the problems encountered by traditional definitions of subordination by defining it in strictly functional terms. Her definition does not refer to any formal properties of the subordinate constructions but refers only to the function subordination fulfills in communication. For the purpose of the investigation Cristofaro adopts the proposal developed by Langacker (1997) in which subordinate sentences designate multiple processes and the profile of one of them is overridden by the other one. The State of Affairs (SoA) without a profile is not asserted, a crucial diagnostic, for this definition of subordination. Cristofaro claims that her definition is form-independent and can be tested via language-independent assertion tests.
Chapter 3 The Coding of Subordination: Parameters for Cross-linguistic Research
The dependent clauses are compared to declarative clauses in isolation with respect to the form of the verb and the way the participants are coded. The variation in form is captured by the notions of balancing (the dependent clause is encoded by the same means as the independent clause) and deranking (the dependent clause is coded differently). The coding of TAM distinctions, person agreement and coding of participants are covered by these notions.
Chapter 4 The cross-linguistic Coding of Subordination: Methodological Premises
The outcome of the analysis of subordination is presented in form of quantified implicational generalizations. The implicational generalizations used in the study are extended with existential and universal quantifiers, in order to make predictions more precise. An implication is regarded to be valid if the number of exceptional languages does not exceed more than one third of the significant cases. The study uses a sample of 80 still spoken and extinct languages. The sample covers all phyla and all language isolates. Languages within each phylum are chosen according to the Diversity Value index, and according to the availability of sufficient resources.
Chapter 5 Complement Relations
Cristofaro uses the classification of the complement taking predicates (CTPs) suggested by Noonan (1985). These classes can be described according to a number of parameters such as the level of clause structure at which the relation is established, predetermination of semantic features of the dependent SoA (such as for example TAM distinctions), and the degree of semantic integration between the two SoAs. The preference of one of the participants of the main SoA for the dependent SoA to come about is claimed to be one further issue in the semantics of CTPs. The result of the analysis of the verb form in complement relations is the Complement Deranking- Argument Hierarchy (CDH):
According to (1), if a deranked form is used at any point on the hierarchy, it is used at all points to the left. This hierarchy is also valid for specific phenomena such as lack of TAM distinctions, use of non-verbal morphology and non-realization of A and S arguments.
It is argued that this hierarchy results from the semantic properties of the CTPs. The first argument in favour is that any information on the verb denoting the dependent SoA may be missing as long as it can be recovered. The recovery of the information is possible when the CTP predetermines it. This would explain why verbs predetermining TAM values or identity of participants of their complements rank higher than those that do not. Further, verbs exhibiting a higher degree of semantic integration rank higher than verbs with a lower degree. Thirdly, the preference for the SoA to occur plays a role in the hierarchy too. CPTs in which one participant shows preference for the dependent SoA to occur rank higher than the CPTs in which no preference is present.
Chapter 6 Adverbial Relations
In this chapter Crisrofaro discusses different adverbial relations (AR) (purpose, before, relation, after relation, when relation, reality condition and reason relations). She defines AR as the case where the dependent SoA describes the circumstances under which the main SoA takes place. AR encode semantic properties different from those of the complements. The analysis of the verb forms used in different ARs with respect to the balancing/deranking distinction leads to The Adverbial Deranking Hierarchy (ADH).
(2) purpose >before, after, when > reality condition, reason
Parameters marking the individual TAM distinctions and lack of person agreement behave slightly differently. With respect to the use of special forms of TAM and person agreement, no hierarchy can be established due to scarcity of evidence but it seems that, when present, they do not violate the ADH. The coding of participants in AR largely follows the NOM-ACC pattern with S or A arguments not expressed and can be aligned as follows:
The tendency not to express the participants is strongly preferred in purpose relations. The coding of arguments as possessors is not a preferred option.
As with complement relations, the alignment of the relations on the hierarchy is explained by the same factors (predetermination, semantic integration etc.), however, some additional arguments are required too. For example, all relations predetermining the time reference of the dependent SoA outrank those relations which do not predetermine this parameter. Relations involving semantic integration (purpose relations) of the two SoA outrank relations not implying any integration (the rest of the AR). The mood value and the preference of the participant for the dependent SoA to be realized seem to play some role too.
The additional factor responsible for the alignment of the ARs is the possibility for the dependent SoA to be encoded as an object (roughly equivalent to Langacker's 'thing'). Relations allowing for the encoding of the dependent SoA as an object should exhibit fewer of the verbal properties. This expectation is in accordance with their position on the ADH.
Chapter 7 Relative Relations
In restrictive relative relations (RR) one participant of the main SoA is identified and further specified by means of the dependent SoA. RRs bear some similarities with the AR in that in neither case is the information strictly required by the mains SoA, and the dependent SoA can be understood as the property of some entity of the main SoA. R.r are still different since they involve no semantic connection between the two SoA.
Description of RRs cannot rely on the parameters used for complement and adverbial relations since all RRs behave uniformly with respect to these parameters. Instead, following Keenan and Comrie's (1977) Accessibility Hierarchy, the syntactic role of the relativized participant in the clause coding the dependent SoA is applied. The result of the analysis of the verb form is the Relative Deranking Hierarchy (RDH) accounting for the balancing/deranking distribution and some individual parameters.
(4) A, S >O > IO > OBL
Lack of aspect and mood distinctions show a slightly different pattern
(5) A, S, O > IO, OBL
In the sample, the majority of languages follow the NOM-ACC pattern in expressing the RRs however, some split alignments are attested. The shared participants always refer to the relativized item.
The role of the relativized item is the most important factor in the explanation of the hierarchies. If a deranked form is used in the relativization of a less accessible role, it will also be used for all more accessible roles. The same is valid for gapping. Cristofaro associates these implications in terms of increased complexity in processing of the relations to the right side of the continuum. The loss of information brought about by the lack of TAM distinctions can be more easily recovered if the RR is easier to process.
Chapter 8 Comparison of Complement, Adverbial and Relative Relations
The global comparison of different types of subordination gives some further support to the proposed hierarchies. The form of the verb used to express the dependent SoA is subject to The Subordination Deranking Hierarchy (SDH):
(6) Phasals, Modals > Desideratives, Manipulatives, Purpose>Perception > Before, After, When, A relatives, S relatives > Reality condition, Reason, O relatives > Knowledge, Propositional attitude, Utterance, IO relatives, OBL relatives
The lack of TAM distinctions and person agreement and the use of case marking/adpositions obey slightly different versions of the SDH, however, no contradictory evidence emerges.
With respect to the lack of overt expression of the participants, subordination relations can be aligned in the Subordination Argument Hierarchy (SAH):
The hierarchies (7) and (8) are motivated by a number of factors introduced in the discussion of the individual types of subordinate relations. These are predetermination, semantic integration, preference, the mood value of the dependent SoA and the ability to construe it as an object. For all relations which predetermine any features of the dependent SoA it is this predetermination that is responsible for the higher ranking of such relations with respect to those relations which do not impose any restrictions.
All relations showing a higher level of semantic integration outrank the relations showing little or no integration. Additionally, the preference and the mood value make their contributions too. One further influence is the possibility to construe the dependent SoA as an object (relations which allow for such construal of the dependent SoA are located on the left side of the hierarchies). Finally, in some cases, the level of clause structure can provide additional explanation (relations located at the lower level of clause structure outrank those located at the higher level). Two further factors have to be mentioned with regard to the position of RRs namely the role of the relativized element and the possibility of construing the SoA as an object.
The general argument which Cristofaro is making on the basis on the comparison of the subordination relations is that morpho-syntactic realisation of different relations is dictated by their semantic properties and that SAH is a valid generalization about the connection of these relations to one another.
Chapter 9 The Coding of Subordination Relations: Functional Motivations Cristofaro explains the implicational alignment of different subordination relations by functional notions of economy and iconicity. Syntagmatic economy seems to account for the loss of features conditioned by predetermination. The influence of semantic integration can be accounted for by the iconicity of independence. The importance of the preference for the independent SoA to occur might also be explained in terms of iconicity of distance where the preferred SoA is construed as less distant from the non preferred one.
To account for some additional facts Cristofaro applies the theory of processes and things presented in Cognitive Grammar which pertains to the cognitive status of the dependent SoA. The ability of the dependent SoA to be realized as an object, and the use of case marking/adpositions among others, seems to support the theory in which the sequential scanning (in short, the distinguishing property of verbs) of the dependent SoA is suspended and summary scanning (the distinguishing property of nouns) is preferred. This line of argumentation can be extended to the absence of TAM distinctions and participant-coding as well.
Chapter 10 Correlations between Individual Morphosyntactic Phenomena The investigation of the mutual relations between different morphosyntactic phenomena used in subordination reveals some interesting correlations. In accordance with the principle of relevance proposed by Bybee (1985), the lack of aspect distinctions is less frequent than the lack of tense and mood distinctions. On the other hand, the use of special forms to express A is more frequent than the use of special forms to express T and M distinctions. Lack of person agreement and the use of case marking/adpositions entails the lack of TAM distinctions or their expression by special forms. Lack of overtly expressed arguments entails lack of person agreement and lack of TAM distinctions. Finally, expressing arguments as possessors entails lack of TAM distinctions and lack of person agreement and the use of case marking. Cristofaro claims that these implications lend support to the hypothesis that processual properties are suspended in the dependent SoAs to different extents. This hypothesis can highlight a connection between the lack of TAM distinctions and the presence of nominal properties. It cannot, however, account for the lack of overtly expressed participants since most of the dependent SoAs is still relational. This issue has to dealt by the principle of syntagmatic economy.
The cross-linguistic investigation of subordination provided by Cristofaro is an impressive piece of work. The contribution to this field of study is significant in many respects. It is to my knowledge the only analysis based on such an extensive sample. This careful analysis pays equal attention to different types of subordination which enables it to bring out numerous similarities between the various strategies used to express the notion of dependency between two SoAs. Both the connections between the type of expressed relation, and the morphosyntactic phenomena used in their coding are investigated thereby providing support for the implicational hierarchies proposed by Cristofaro.
There are, however, some questions left open. The hypothesis that all languages should be in a position to express certain cognitive functions (p. 46), and the definition of the object of research in purely functional terms seems to imply that subordination should be an option universally found in languages. The answer to this question is not given.
A vulnerable point of the investigation is the reliance on the (mostly English) translations with respect to the assertiveness of the complement clauses (p. 41) which runs the risk of misinterpretation of the data. The option of raising as superficially similar, but substantially different from participant sharing is briefly discussed (p. 79-80) but this distinction does not play any role in the rest of the argument.
There is also a question concerning the option of backward control (p.117). Unfortunately, no account is given of why some languages show the unexpressed participant in the subordinate clause (the more widespread option) and some in the matrix. These drastically different options cannot be accommodated by the principles proposed. Apparently, this option is not considered (p.253), information about participants (...) can be specified only once, for the SoA for which it is not predetermined (...) One more question might be asked concerning the theory of iconic motivation: if a deranked form is always a result of iconicity, what happens in the case of subject sentential complements involving deranked verb forms which are not predetermined in many respects?
The final remark has to do less with the arguments developed by Cristofaro as with the theory of Cognitive Grammar in general. As the author herself mentions (p.259), within the Cognitive approaches independent evidence for the claims made about the nature of human perception and information processing still remains somewhat elusive.
Bybee, J. L. (1985). Morphology: A study of the relation between meaning and form. Philadelphia: Benjamins.
Givón, T. 1990. Syntax. A functional typological introduction. Vol 2 Amsterdam: John Benjamins.
Noonan, Michael (1985). Complementation. In Timothy Shopen (ed.), Language typology and syntactic description. Vol 2: Complex constructions. Cambridge University Press, 42-140.
Langacker, Ronald W. 1987. Foundations of cognitive grammar, Vol.1: Theoretical Prerequisites. Stanford, Stanford University Press
Keenan, E. L., and Comrie, B. (1977) Noun phrase accessibility and Universal Grammar. Linguistic Inquiry 8:63-99.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER:
Szymon Slodowicz received an M. A. in General and Comparative Linguistics from the CAU in Kiel, Germany and is currently working as a researcher on the project "Typology of control verbs" affiliated at the Centre for General Linguistics, Typology and Universals Research (ZAS) , Germany.