TL;DR
This paper argues that traditional explanations of cleft sentences (like "It was X that Y") as simply "focusing" constructions are insufficient. By analyzing a large corpus of real-world language, the authors propose a multi-level model that integrates syntax, semantics, and pragmatics to explain why and how speakers choose clefts, offering a more nuanced understanding of their discourse function.
What Problem Does This Paper Address?
Linguists have long been intrigued by cleft constructions, such as "It was the mango that Martin ate" (it-cleft), "What Martin ate was the mango" (wh-cleft), and "The mango was what Martin ate" (reverse wh-cleft). While it's generally agreed that these structures are used for emphasis or to highlight specific information, existing explanations often fall short. Common theories suggest clefts signal "given" vs. "new" information, provide unique serial ordering options, or act as "focusing" devices. However, the authors point out that these explanations struggle to account for the full range of observed usage and often rely on decontextualized examples. This paper aims to bridge this gap by developing a more comprehensive, multi-level model that explains the decision-making process behind using clefts in natural discourse.
Methods
The study is primarily based on a corpus analysis of 587 naturally occurring cleft constructions in both written and spoken discourse. The authors employ a two-stage model to understand cleft usage:
- Stage 1: Cleft vs. Non-Cleft Decision: This stage investigates the factors that lead a speaker to choose a cleft construction over a simpler, non-cleft alternative.
- Stage 2: Cleft Type Selection: This stage examines the factors that influence the choice between the three main types of clefts (it-clefts, wh-clefts, and reverse wh-clefts).
The analysis integrates insights from syntax, semantics, and pragmatics, moving beyond single-level explanations.
Datasets
The primary dataset for this study is a corpus of 587 naturally-occurring cleft constructions from written and spoken discourse. The authors also reference data from the Survey of English Usage spoken data for a detailed analysis of 150 examples of each cleft type. No new datasets were introduced.
Key Insights and Ideas
- "Focusing" is Not Enough: The paper strongly refutes the idea that clefts are solely "focusing" constructions. The corpus data reveals significant counterexamples to this simplistic view, showing that the relationship between syntax, accent, and information status is far more complex.
- Multi-Level Interaction is Crucial: The authors demonstrate that understanding cleft function requires considering the interplay of syntactic, semantic, and pragmatic factors. No single level of analysis can fully explain their usage.
- Presupposition and Assertion Dynamics: Clefts syntactically mark two statuses of propositional content: presupposition and assertion. The presupposition, induced by the cleft structure, contains an existentially quantified variable that needs to be instantiated by the asserted proposition.
- Uniqueness and Novelty are Key: For a cleft to be acceptable, the instantiation of the presupposed variable must be unique or maximal within the discourse context, and crucially, it must be novel. This explains why constructions like "John won and it was John who won" are unacceptable.
- Syntactic Constraints Shape Choice: Different cleft types have distinct syntactic constraints on what can serve as their "head" (e.g., wh-clefts can take bare infinitives, it-clefts can take "NP + not until"). These constraints play a significant role in determining which cleft type is syntactically viable.
- Information Ordering Preferences: While not the sole determinant, the serial ordering of information matters. Wh-clefts, for instance, may be preferred when the presupposed information is lengthy, as they can present it earlier.
Key Results
- Rejection of Simple "Given/New" Mapping: The study found that clefts do not reliably map "given" information to the complement and "new" to the head (strong version) or segregate them syntactically (weak version). A significant percentage of clefts contained new information in the complement, and prosodic nuclei (indicating new information) appeared in both head and complement positions.
- Accentual Patterns are Varied: Contrary to some theories, the appearance of nuclear accents in clefts does not reliably correlate with specific syntactic positions predicted to be "focal." Many clefts had nuclei in multiple positions or in positions not predicted by focus theories.
- Differential Presupposition Capabilities: The three cleft types have different "profiles" regarding the types of information they can convey as presuppositions. Notably, wh-clefts cannot presuppose "new" information in context, while it-clefts can, leading to preferences in certain discourse situations.
- Syntactic Head Constraints: The paper provides specific examples illustrating how syntactic differences between cleft types restrict the constituents that can function as their heads, thereby influencing the choice of cleft.
Limitations and Future Work
The authors acknowledge that their proposed two-stage decision procedure is a "small step towards an implementable algorithm." They suggest that the identified factors could form the basis for "preference weightings" to guide syntactic choice in a given context.
Future work could involve further developing these preference weightings into a more concrete computational model for generating cleft constructions. The paper also implicitly calls for more corpus-based descriptive work to underpin such decision procedures, emphasizing that syntactic choice is not solely a pragmatic or stylistic phenomenon but a complex interplay of linguistic factors.
Summary updated