Which properties should be regarded as the hallmarks of parentheticality? How does a parenthetical interact with its host or root clause? – Is there any such interaction at all? What is the discourse status of parenthetical content? And how can we explain that parenthetical material systematically escapes the scope of structurally higher operators? These are just some of the questions that the literature on appositives and other constructions categorized as ‘parenthetical’ have been concerned with in the past decades, and that have recently been the subject of much discussion. Todor Koev's new book Parenthetical Meaning offers new perspectives on these and related questions, and thereby attempts to clear up some of the mysteries that research on parentheticals has struggled with for many years.
Koev's investigation into parentheticals starts from the question of what constitutes parenthetical meaning, and how this kind of meaning is related to the meaning expressed by the root clause in (or attached to) whose syntactic structure the parenthetical appears. The intricate factor here lies in the seemingly dual nature of parentheticals; they exhibit a degree of independence, yet simultaneously display instances of interpretive interplay with the root clause. Chapter 1 (pp. 1–16) is mainly concerned with two issues. The first is how to discern parenthetical meaning from other components that contribute to sentence meaning, such as entailment or presupposition. While acknowledging that parentheticals share certain similarities with entailments, such as introducing discourse-new information, and presuppositions, such as having non-at-issue content or projecting under various types of embedding, Koev suggests that parentheticals have a unique distinctive feature: their own illocutionary meaning encoded by an own force operator. The second issue addressed in chapter 1 has to do with the question as to which constructions qualify as parenthetical, and whether the members of this seemingly inhomogeneous group share some crucial characteristics. Koev differentiates between two classes of parentheticals, contingent upon the presence and nature of their interaction with the root clause's interpretation. PURE parentheticals, which encompass as-parentheticals (1a), clausal parentheticals (1b) and nominal appositives (1c) as well as appositive relative clauses (1d), do not affect the interpretation of the root clause; their contribution is limited to commenting on its content. Although parentheticals belonging to this class can occur embedded, they fall out of the scope of operators appearing higher in the structure.
(1)
(a) John is a crook, as we all know.
(b) Carl (he is one of my cellmates) used to wear a bolotie.
(c) Marlon Brando, my favorite actor, won the Oscar in 1955.
(d) Florence, who lives next to me, owns twelve cats.
In contrast, IMPURE parentheticals, such as utterance adverbs (2a), slifted parentheticals (2b) and antecedents of so-called biscuit conditionals (2c), are not easily embeddable and they have an effect on the root clause's interpretation.
(2)
(a) Frankly, Skylar looked gorgeous in that dress.
(b) The lawyer, Jack said, had an affair with the judge.
(c) If you are hungry, there are biscuits on the sideboard.
Despite their distinctive features, pure and impure parentheticals share the crucial property of illocutionary independence from the root clause.
The foundational principles of the formal framework employed in the book, specifically update semantics, are introduced towards the end of the introductory chapter using a simple declarative for exemplification. Chapter 1 concludes with an outline of six puzzles from the literature on parentheticals, which recur throughout the book. The first puzzle concerns the projectional behavior of parentheticals, i.e. the fact that parentheticals systematically escape the scope of structurally higher operators. In a similar vein, the second puzzle relates to the question as to why a quantificational antecedent cannot bind a pronominal element occurring within the parenthetical. The third, at-issueness puzzle, has to do with why parenthetical content is not suitable for answering the current QUESTION UNDER DISCUSSION (henceforth: QUD). The remaining three puzzles are specific to impure parentheticals – in particular, it is unclear why these parenthetical constructions are not syntactically embeddable (embeddability puzzle), why they obligatorily only create upward-entailing environments (polarity puzzle) and how modification over a slifted parenthetical reduces the possibility of hedging (hedging puzzle).
In chapter 2 (pp. 27–56), Koev discusses if and how parenthetical constructions affect the ILLOCUTIONARY FORCE of the root clause. Illocutionary force is understood as a conventionalized semantic effect on the context which depends on the respective clause type. For the account of the relationship between parenthetical and root clause carried out in this book, just two of these types (and the effects they impose) are relevant – declarative and interrogative. The approach to illocutionary force adopted here is operator-based, meaning it relies on the idea that illocutionary force is encoded at a syntactic level (as opposed to a pragmatic level, as occasionally proposed in the literature) through the presence of force operators. Koev defines two force operators, DECL (declarative) and INT (polar interrogative), that bind the propositional content of their domain. This, in turn, allows parentheticals and root clauses to be illocutionarily independent from one another. Their contents remain separate, resulting in root clauses and (pure) parentheticals behaving akin to two independent sentences with respect to how, for example, anaphoric links can be established. Hence, a parenthetical can have a different illocutionary force (due to the presence of a respective force operator) than the root clause, and consequently, parenthetical content such as the proposition expressed by the appositive relative clause in (3a) need not be part of the question expressed by the root clause, and vice versa for the clausal parenthetical in (3b).
(3)
(a) Did Florence bury her black cat, who suddenly passed away last Tuesday, in her garden last night?
(b) Boloties (didn't they go out of fashion in the 1990s?) are Carl's favorite accessory.
A fair amount of chapter 2 deals with slight modifications of the underlying mechanisms in order to capture the divergent ways in which impure parentheticals affect their root clauses, thus capturing the observation made above that impure parentheticals may interact with the root clause.
Chapter 3, ‘Scopal properties’ (pp. 57–89), investigates the question as to why parenthetical implications systematically escape the scope of higher operators, and why parenthetical content systematically projects to the overall sentence level. Koev discusses in detail five previous approaches to this issue, and highlights some theoretical challenges they are faced with. The derivation of parenthetical projection proposed in the book is based on three main assumptions. First, lexical predicates can take propositional variables as their arguments, and propositional operators can, in turn, bind these variables. Second, clausal nodes in a syntactic structure are headed by such propositional operators – importantly, as mentioned earlier, root clause and parenthetical bear their own separate force operators. Lastly, propositional operators can bind into the content of another propositional operator, but must not themselves be bound. Force operators, then, obligatorily attribute the content in their scope to the speaker of the respective utterance. Since they occupy a top node in the syntactic structure, parenthetical content always projects. So the two components of the sentence which the speaker of (4) commits herself to under Koev's analysis are, first, that Florence does not own any dogs (i.e. the complement of Florence owns a dog), and second, that Florence is an elderly person.
(4) Florence, who is an elderly person, doesn't own any dogs.
However, some constructions that could otherwise be considered parenthetical seem to fall out of this paradigm – such as one-modifiers (5a), double perspective appositives (5b), shifted appositives (5c) and appositive relative clauses like (5d) which obligatorily take narrow scope.
(5)
(a) Skylar wants to buy a dress, a red one.
(b) Mary was surprised that John, a notorious casanova, got married recently.
(c) Florence must have gone mad. Yesterday, she claimed that her cats are all aliens, [who will snatch her to their home planet one day]Florence.
(d) If tomorrow Carl calls the prison officer, who in turn calls the prison warden, we will get into serious trouble.
These apparent parenthetical constructions are not taken into further consideration, as they are, according to Koev, not parentheticals proper. Hence, he limits himself to merely briefly outlining ways in which they should be analyzed instead.
This is a consequential decision, however. To take nominal appositives as an example, Onea & Ott (Reference Onea and Ott2022) argue that examples such as (6) not only violate the at-issueness constraint of parentheticals, but also the binding constraints – albeit that would not contradict Koev, given that he suggests in chapter 4 that parentheticals are underspecified with regard to targeting the QUD. Onea & Ott call such constructions REFORMULATING NOMINAL APPOSITIVES (henceforth: R-NAPs), and suggest that these are syntactically entirely independent illocutions. Moreover, R-NAPs are argued to elliptically include material from the host, as shown in (6b). Such an analysis suggests that the linear interpolation of the elliptical speech act into the host speech act is essentially governed by pragmatic and processing principles that amount to facilitating the discourse structural interpretation of the emerging sequence.
(6)
(a) A friend of mine, Skylar, wants to buy a new dress.
(b) [Illoc1: A friend of mine wants to buy a new dress.] [Illoc2: Skylar wants to buy a new dress.]
Koev does not explicitly discuss such cases, but his only analytical option to handle them (at least as far as we can reconstruct) would be to treat them tantamount to (5a), i.e. to assume some version of the one operator, thus making R-NAPs restrictive modifiers instead of true appositions. In (7a) we suggest one possible analysis along these lines, which in turn would be roughly equivalent to (7b), a case of restrictive modification.
(7)
(a) A friend of mine, an identical to Skylar one, wants to buy a new dress.
(b) A friend of mine who is identical to Skylar wants to buy a new dress.
Progressing further, parenthetical content is generally considered to be non-at-issue, meaning it does not address the current QUD. However, Koev demonstrates in chapter 4 (pp. 90–114) that parentheticals sometimes can address the QUD, as exemplified in (8), where one of the questions raised by A is answered by the root clause in B's reply, the other by the parenthetical.
(8)
A: Who did you meet last night at the pub and what did he drink?
B: I met Peter, who drank whiskey on the rocks.
Rather than treating the discourse status of their content as non-at-issue per default, Koev argues that parentheticals are underspecified in this respect, and compete with the status of the root clause for counting as an answer to the QUD. This competition is uneven, however, since root clauses are conventionally marked for being at-issue. In more technical terms, Koev introduces two operators (R-DECL and R-INT) for root clauses and two (P-DECL and P-INT) for parentheticals. These account for declaratives and interrogatives respectively. The key difference between R-DECL and P-DECL is that only the former imposes a QUD-relevance requirement with respect to its propositional content, the latter remaining underspecified. Regardless of its discourse status, parenthetical content is predicted to always project, because it occurs in the scope of its own force operator, (typically) P-DECL. Thus, discourse status is suggested to be based not only on linguistic marking but also linked to conversational inferences.
A fully fledged formal account is presented in chapter 5 (pp. 115–40). It starts with a useful, comprehensive introduction to a version of compositional dynamic semantics. Koev then shows how the derivation of the parenthetical constructions discussed in the book works out in detail within the framework, on the basis of the illocutionary operators defined earlier. In a nutshell, the essential idea of the compositional system is that semantic terms incorporate what one may call a ‘glue’ variable. This variable may carry or inherit referential specifications, such as the evaluation of some referential index to a certain entity, to higher levels of syntactic representation until illocutionary operators enter the picture and ensure the transformation of pure propositional content into speaker commitments (and more generally discourse content), further restricting the range of possible contexts. Thereby, the illocutionary operators defined have different effects on the discourse representations, albeit some of these effects essentially hinge on general principles, such as, for example, the notion of being syntactically ‘closed’. Overall, the compositional system is not only well presented, but also very straightforward for simple cases. For complicated cases involving binding, scope interactions, bare nouns, etc. questions emerge which, however, the author rightfully ignores.
The remainder of this last chapter revisits the six puzzles on parentheticals introduced in chapter 1. For instance, the proposed solution to the projection puzzle boils down to the presence of a separate force operator in the parenthetical structure – since force operators cannot be bound by higher operators and attribute the content in their scope to the speaker, parenthetical content always projects. As a further consequence, the presence of a force operator on top of the parenthetical also impacts binding. Specifically, a structure headed by its own force operator is assumed to be syntactically ‘closed’ – that is to say, it must not contain any instances of unbound pronouns. The binding puzzle is thus resolved. A solution to the at-issueness puzzle was just presented in chapter 4, where it is stated that parentheticals are undefined for their discourse status, and given that they compete with the root clause in this respect, parenthetical content usually ends up being non-at-issue. Since impure parentheticals, such as utterance adverbs, somehow contribute to the illocutionary component of the root clause, it follows that they are not easily (if ever) embeddable.
In sum, there is much reason to praise the book, as it is not only comprehensive and clear, but it also offers a truly ingenious analysis of parentheticals with a straightforward, but not too demanding formal framework. However, it is not obvious that the theoretical analysis, elegant as it may be, is the last word on this matter. While some other minor worries could be articulated, e.g. about the derivation of slifted parentheticals or the analysis of biscuit conditionals as compared to extant literature, we will only focus on the critical decision to treat what Onea & Ott dubbed R-NAPs as mere restrictive modifiers. The crucial problem is not that seemingly similar constructions, in terms of overt form or prosody, are treated as a distinct class based on their fine-grained semantic properties, as this is standard practice in theoretical linguistics. The problem is that R-NAPs exhibit systematic case connectivity with the host as exemplified in the German examples in (9). It is simply not clear whether Koev's system has sufficient resources to explain the accusative on den Albert in (9a), and likewise the dative on dem Rex in (9b) under the assumption that these are more or less simple restrictive modifiers, especially if analyzed along the lines suggested in (7), above.
(9)
(a) Ich habe meinen Freund, den/*der Albert, angerufen.
I have my-acc friend-acc the-acc/*-nom Albert called
‘I called my friend, Albert.’
(b) Maria musste ihrem Hund, dem/*der Rex, einen Maulkorb anlegen.
Maria must her-dat dog-dat the-dat/*-nom Rex a muzzle put-on
‘Maria had to muzzle her dog, Rex.’
While Onea & Ott are not the only ones who have discussed this problem (indeed, Potts Reference Potts2005 provides a detailed analysis), they have taken the contrast between cases with and without case-connectivity to question the general constraints of parentheticals regarding binding at at-issueness. Thus, being able to handle R-NAPs (in whatever way) seems to be essential if Koev's theory is to remain valid for the entire class of parentheticals. If, however, R-NAPs are parentheticals after all, as suggested by Onea & Ott, Koev's analysis only applies to certain parentheticals, and does not cover the entire empirical picture. Having said this, even under the analysis of Onea & Ott, nominal appositives would have their illocutionary independence as in Koev's system. However, as opposed to Koev's analysis, an additional specialized illocutionary operator for parentheticals would no longer be necessary, since linear interpolation would be a mere matter of pragmatic optimization.
While this needs to remain somewhat speculative here, we wonder whether a unification of the two approaches is possible. After all, given the compositional system provided by Koev, it is not obvious to us that the low-adjunction of appositives plays a clear explanative role in his own theory. The theory is really carried by the independent illocutionary operators. On the other hand, Onea & Ott do not provide a theory for a whole range of other constructions discussed by Koev. Finally, Koev does not sufficiently address the discourse relation between parenthetical constructions and the host. While we deem that he is correct in assuming underspecification of at-issueness at the level of the illocutionary operator, when it comes to the QUD addressed by the root clause, the question remains open whether and what (other) question the parenthetical answers and whether there can be constraints on these questions, i.e. not even a parenthetical can answer entirely random questions. If so, some pragmatic embedding akin to the one provided by Onea & Ott will be necessary for Koev's theory as well. Thus it seems to us that the two approaches have the potential to complement each other.
In conclusion, this book represents a noteworthy scholarly achievement that not only offers a lasting contribution to the field, but also serves as an excellent entry point for junior researchers exploring the semantics and pragmatics of parentheticals. By leaving a number of questions unresolved, it will undoubtedly catalyze additional research on parentheticals.
