Towards a Unified System of Representation for Continuity and Discontinuity in Natural Language

Prakash Mondal; Ratna Kandala

REVIEW 4 major objections 6 minor 4 references

Towards a Unified System of Representation for Continuity and Discontinuity in Natural Language

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper claims that three rival grammar formalisms can be rewritten as one representation system that handles both continuous and discontinuous syntax.

desk verdict Useful survey, but the claimed unification rests on a Correspondence Principle that can equate any dependency with any functor-argument pair, so the formal result collapses. read the letter →

arxiv 2506.05235 v1 pith:OG5QG564 submitted 2025-06-05 cs.CL

Ratna Kandala , Prakash Mondal This is my paper

classification cs.CL

keywords syntacticdiscontinuityunifiedrepresentationphrasestructuregrammardependencycategorialconstituencyfunctor-argumentrelationsnon-configurationallanguages

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to prove that three apparently incompatible grammar formalisms—phrase structure, dependency, and categorial—are one system of representation when viewed through a single correspondence principle. It applies the principle to discontinuous constructions in German, Croatian, and Kalkatungu, showing that each construction can be derived in all three formalisms and converted in both directions. If true, this would mean constituency, head-dependent, and functor-argument descriptions are notations for the same underlying structure, so continuous and discontinuous syntax need no separate machinery. The wider stake is that the cognitive system may represent both kinds of languages in one format.

What carries the argument

The load-bearing device is the Correspondence Principle, $A(B^*)\lor A({}^*B)\equiv A|B$, which reads: if B depends on A and sits to the left (or right) of A, then the pair can be written as a functor-argument category $A|B$ with neutral direction. The paper supplements this with a dependency valuation function $\delta$ that assigns real values to nodes, and with wrapping, which lets a functor combine with a non-adjacent argument by treating the functor and one argument as a combined lexical form. The derivations then move stepwise: CG cancellations are drawn into PSG trees, using tangled trees when word order requires crossing branches, and each cancellation is rewritten as a dependency relation, producing a unified representation for the sentence.

What would settle it

Take a dependency graph in which one head takes two dependents of the same category on the same side, and try to assign each word a single CG category so that every dependency edge matches a distinct functor-argument application under $A(B^*)\lor A({}^*B)\equiv A|B$; one unmatched edge would be a direct counterexample to the claimed equivalence.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that continuous and discontinuous sentences can be carried through a chain of conversions—PSG to CG, DG to CG, CG to DG, and CG to PSG—so that no formalism is left with an irreducible representation. For a German verb-final construction, a Croatian copular clause with a split noun phrase, and a Kalkatungu clause with a discontinuous noun phrase, the paper derives dependency graphs, CG cancellation steps, PSG rules, and unified diagrams, then concludes that establishing these conversions is tantamount to establishing $\mathrm{PSG}\equiv\mathrm{CG}\equiv\mathrm{DG}$ in their representational descriptions of natural language constructions.

Load-bearing premise

The load-bearing premise is the stipulated Correspondence Principle, which assumes every dependency relation between two words can be rewritten as a functor-argument relation, even though in exceptional steps the word assigned to B on one side of the equivalence can be a different word from the B on the other side.

Editorial extensions

If this is right

A single system of representation can cover continuous and discontinuous structures without adding rules, constraints, or transformations beyond the Correspondence Principle.
Representations in any of the three formalisms can be converted into either of the other two, so a constituency tree, a dependency graph, and a CG derivation become views of the same structure.
The traditional opposition between constituency-based and non-constituency formalisms would be a theoretical artefact, not a property of languages.
The same cognitive representation could serve speakers of fixed-word-order and free-word-order languages, with no need for separate continuous and discontinuous syntax modules.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Read literally, the Correspondence Principle's exceptional steps allow the category B on one side of the equivalence to pick out a different word from B on the other side, which makes the equivalence a matching rule rather than a true identity; its scope should be tested on dependency graphs where no shared category exists.
A natural extension is to run the same PSG to CG to DG chain on English long-distance dependencies and parentheticals; if the chain goes through, the unification extends beyond free-word-order languages.
The paper's choice to avoid type-raising in favor of iterative wrapping could be tested by checking whether all CG derivations for coordinate or shared-constituent constructions remain inside the closure of the wrapping operation.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Dear colleague,

The headline: this paper has a genuinely useful survey of approaches to syntactic discontinuity, but its central claim—that PSG, DG, and CG are one representational system—rests on a principle that can equate almost anything with anything. I don't think the formal result stands.

What's new and good: Section 5 reviews Austin & Bresnan's LFG treatment, Dowty's phenogrammar, McCawley's tangled trees, Citko's parallel merge, and Barry & Pickering's dependency constituents. That survey is clear, accurate enough, and will be a handy reference for students. The German example (10) actually works: in every CG cancellation, the functor-argument pair coincides with a head-dependent pair, so the Correspondence Principle is satisfied with the same words on both sides. The authors are also transparent about the exceptional cases, which is more than many papers do.

The soft spot is load-bearing, not cosmetic. The Correspondence Principle is stated as A(B*) ∨ A(*B) ≡ A|B, but then the authors explicitly allow cases where only A or only B is the same on the two sides, with the other variable denoting a different word. That is exactly what happens in the Croatian and Kalkatungu derivations. For example, Croatian Step 3 equates the dependency je(*učionica) with the functor-argument relation Naša\je: the LHS's B is učionica, the RHS's B is Naša. Once that substitution is allowed, ≡ is no longer a relation between the same linguistic objects. It becomes a license to pair any dependency relation with any functor-argument relation that shares one category. No formal semantics for ≡ is given, and there is no restriction on admissible substitutions. The equivalences therefore do not establish a substantive unification; they show that each step can be post-hoc reconciled. The principle is also credited to anonymous works (likely the authors' own prior work), which blocks checking prior art.

A related but smaller issue: the paper claims "PSG→CG, DG→CG, CG→DG, CG→PSG" as if these are novel, but the PSG-DG correspondence is old (Gaifman 1965; Hays 1964), and CG-to-dependency mappings are standard in parsing. The only novel piece is the Correspondence Principle, and it is not well-defined.

For a reader: the survey in Section 5 is the valuable part. The formal derivation in Section 6 does not support the conclusions, and the cognitive speculation in Section 7 goes beyond anything the paper establishes. I would not cite the paper for the unification, but I might point someone to the survey.

Recommendation: send to referees—the question is important enough and the survey worth preserving—but the reviewers should be told to focus on the Correspondence Principle. I expect rejection unless the authors can give a real definition of ≡ and restrict variable substitutions. That is a big lift.

Best, [Name]

Referee Report

4 major / 6 minor

Summary. The paper argues that phrase structure, dependency, and categorial analyses of continuous and discontinuous syntax can be unified into one representational system. It reviews evidence for discontinuous constituents, compares Phrase Structure Grammar (PSG), Dependency Grammar (DG), and Categorial Grammar (CG), and then proposes a "Correspondence Principle" A(B*) ∨ A(*B) ≡ A|B that equates head-dependent relations with functor-argument relations. Using this principle, the paper gives bidirectional conversions among PSG, DG, and CG for one German, one Croatian, and one Kalkatungu sentence, and concludes that these conversions establish PSG≡CG≡DG.

Significance. If the equivalence were established, the paper would offer a significant unification: constituency, dependency, and categorial representations would be notational variants, and discontinuity would become a natural consequence rather than an anomaly. The paper is also useful as a survey of existing treatments of discontinuity, including tangled trees, Parallel Merge, LFG, phenogrammatical structure, and dependency constituents, and it makes its intended conversion steps concrete and checkable for three typologically distinct languages. Those strengths are real. However, the load-bearing Correspondence Principle is stipulated rather than derived, it is not well-defined under the "exceptional cases" admitted in Section 6, and the three worked examples do not amount to a proof of general equivalence. The current manuscript therefore does not establish the advertised unification.

major comments (4)

[Section 6, Correspondence Principle, and (11c) Step 3 / (12c) Steps 2, 4, 5] The asserted equivalence A(B*) ∨ A(*B) ≡ A|B is not well-defined because the variables A and B are allowed to denote different words on the two sides. In (11c) Step 3, the LHS Aux(*N) instantiates B=učionica while the RHS Det\Aux instantiates B=Naša; in (12c) Step 2, B is ṯuar-Ø on the LHS and maḻṯa-Ø on the RHS; and in (12c) Step 4 both A and B change, with A=japacara-tu on the RHS but A=kuḷa-ji on the LHS. The paper explicitly licenses these "exceptional cases" immediately after stating the principle, but it gives no formal semantics for ≡ and no restriction on admissible substitutions. Under this licensing, the sign can equate any dependency pair with any functor-argument pair that shares one category, so the DG↔CG derivations are vacuous. Footnote 5's claim that the principle "filters out" unwanted relations is also untenable under this reading.
[Section 6, (10c)-(10d) and (11b)-(12b)] The DG↔CG conversions are circular with respect to the claim of unification: each conversion step applies exactly the Correspondence Principle that is the target of the proof. For example, (12b) says "by using the correspondence principle, we have that ḻaji(*ṯuar-Ø) ≡ ḻaji/maḻṯa-Ø," and (11b) says the same for Aux(*N) ≡ Det\Aux. Since the principle itself already asserts the equivalence between head-dependent and functor-argument relations, these steps cannot count as independent derivations of that equivalence; they are instances of it.
[Section 6, (10a), (11a), (12a) and the CG → PSG subsections] The PSG↔CG conversions are not formal translations. In the PSG→CG direction, CG categories are assigned to terminal nodes by hand and cancellation lines are drawn through a pre-existing PSG tree; in the CG→PSG direction, a PSG tree is inferred from each CG cancellation. No rule is given that maps arbitrary PSG rules or arbitrary CG categories onto one another, and the choice of lexical category assignments is unconstrained (for example, maḻṯa-Ø is assigned Det2 in (12) without a procedure for deriving this assignment). The three sentences therefore illustrate parallel annotation, but they do not establish an equivalence between the two formalisms.
[Section 6, final paragraph, and Section 7] The claim that "establishing PSG->CG, DG->CG, CG->DG and CG->PSG is tantamount to establishing PSG≡CG≡DG" overgeneralizes from the three sentences (10)–(12). Even if each derivation were internally valid, no inductive, constructive, or formal proof shows that the conversions hold for all, or even for a characterized class of, continuous and discontinuous constructions. At most, the paper gives existence illustrations for three data points.

minor comments (6)

[Section 5, paragraph on Baker's UTAH] There is a typo "a a one-to-one mapping"; the duplicated article should be deleted.
[Section 4.2, formal definition of DG] The formal definition ends with "(see also." followed by nothing; this incomplete citation should be completed or removed.
[Section 6, footnote 6] The page range "154–15" should be "154–155."
[References and Section 2] The text cites "Anonymous 2014" and, in Section 6, "Anonymous 2022, Anonymous 2023, Anonymous 2024," but none of these works appears in the reference list; since the Correspondence Principle is attributed to the latter three, this is not merely a formatting issue.
[Section 6, Figures 16 and 25] The text says that crossing lines are drawn in standard PSG trees for expository purposes, but standard PSG trees prohibit crossing branches; the figures should be labeled as expository/tangled trees or the convention should be clarified in the caption.
[Section 6, (12), footnote 7] Footnote 7 says "the author in the original source" without identifying who that author is; it should name Van Valin 2001.

Circularity Check

3 steps flagged · score 8.0 of 10

Correspondence Principle already asserts the DG↔CG equivalence, so the claimed unification is stipulated; exceptional steps swap variables to make equations balance.

self definitional [Section 6, 'The Correspondence Principle' paragraph (after the introduction of examples (10)-(12))]
"Below is The Correspondence Principle (Anonymous 2022, Anonymous 2023, Anonymous 2024) proposed to unify the direct head-dependent relation and the functor-argument relation between any two given words A and B: A(B*) ˅A(*B) ≡ A|B ... In exceptional cases, only either A or B tends to be the same on LHS and RHS, and the other category can vary across sides."

The principle is stated as an equivalence, not derived. The paper's Section 6 then uses it to perform every DG→CG and CG→DG conversion, saying 'by using the Correspondence Principle' at each exceptional step. Consequently, the conclusion PSG≡CG≡DG is an application of the stipulated equivalence, not an outcome derived from independent definitions. The clause allowing A or B to vary across sides further makes the equivalence a license to pair arbitrary dependency and functor-argument relations that share one category.
self citation load bearing [Section 6, 'The Correspondence Principle' paragraph (same as above)]
"Below is The Correspondence Principle (Anonymous 2022, Anonymous 2023, Anonymous 2024) proposed to unify the direct head-dependent relation and the functor-argument relation between any two given words A and B."

The load-bearing premise of the entire unification is supported only by citations to works authored by 'Anonymous' in this anonymized submission. No proof of the principle is supplied here, and no external, machine-checked or independently verified derivation is cited. Since the principle is exactly the equivalence that the paper claims to establish, the argument reduces to an unverified self-citation chain.

1 more flagged steps

other [Section 6, (11b) Note for Croatian Step 3; cf. (12b) Notes and (12c) Steps 2, 4-5 for Kalkatungu]
"Note: In step 3, by using the correspondence principle, we have that je(*učionica) ≡ Naša\je. In other words, Aux(*N)≡Det\Aux. We can also express it as: A(*B)≡B\A. Since Naša and je do not participate in any (direct) dependency relation as seen in Figure 27, the functor-argument relation is constructed through Aux and N in step 3."

The equation relates a dependency between je and učionica (Aux(*N)) to a functor-argument relation between Naša and je (Det\Aux). The variable B denotes učionica on the LHS but Naša on the RHS. The same device is used in Kalkatungu: V(*N2) ≡ V/Det2 pairs ḻaji(*ṯuar-Ø) with ḻaji/maḻṯa-Ø; N1(Det1*) ≡ Det1/Adj pairs kuḷa-ji(Ṉa-ci*) with Ṉa-ci/japacara-tu; V(N1*) ≡ Det1\V pairs ḻaji(kuḷa-ji*) with Ṉa-ci\ḻaji. Thus the 'equivalence' is not a relation between the same two words; it is a hand-constructed match that makes the conversions balance, so the derivation is circular.

full rationale

The paper's claimed unification (PSG≡CG≡DG, Section 6) rests on the Correspondence Principle A(B*) ∨ A(*B) ≡ A|B, introduced in Section 6 and attributed only to 'Anonymous 2022, 2023, 2024.' The principle already states, as a stipulated equivalence, that any dependency relation between A and B is the same as a functor-argument relation between A and B. Every DG→CG and CG→DG step ('by using the Correspondence Principle') is therefore an application of the very equivalence the paper claims to establish. The exceptional clause then permits A or B to denote different words on the two sides, which the paper uses in Croatian Step 3 (Aux(*N) ≡ Det\Aux, with B = učionica on the LHS but Naša on the RHS) and in Kalkatungu Steps 2, 4 and 5 (V(*N2) ≡ V/Det2, N1(Det1*) ≡ Det1/Adj, V(N1*) ≡ Det1\V). In each of these, the LHS dependency is between one pair of words and the RHS functor-argument relation between a different pair that merely shares a category; hence the conversions are hand-matched to balance rather than derived from a well-defined equivalence relation. No independent formal derivation or benchmark is supplied, and the only cited support is the authors' own anonymized prior work. The conclusion is therefore forced by the stipulation rather than demonstrated.

Assumptions & free parameters 2 free parameters · 4 assumptions · 2 invented entities

The claim rests on the Correspondence Principle, which is an ad hoc stipulation rather than a theorem. The worked examples require hand-picked category assignments and per-step choices of which word B denotes. The paper also assumes that DG relations are order-free and that N/S categories suffice. No independent entity is introduced beyond the notation.

free parameters (2)

B-variable assignment in exceptional Correspondence Principle steps = e.g., Croatian Step 3: RHS B = Naša/Det, LHS B = učionica/N; Kalkatungu Step 4: RHS A = japacara-tu/Adj, LHS A =…
When functor and argument do not share a direct dependency, the paper lets B (or A) denote different words on the two sides of the equivalence; this is a free choice made per example and is what makes the equations balance.
Lexical category assignments for the worked sentences = den:N2, lesen:V2, zu:Infinitive, versucht:V1, haben:Aux, Sie:N1; etc.
The CG derivations only work because each word is assigned a category by hand; no independent criterion or lexicon is provided.

assumptions (4)

ad hoc to paper The Correspondence Principle: A(B*) ∨ A(*B) ≡ A|B
Stated in Section 6 with no derivation; it asserts the equivalence the paper aims to prove.
ad hoc to paper Every functor-argument cancellation corresponds to a head-dependent relation, with the functor as head.
Used in every CG-to-DG step; when no direct dependency exists, the paper substitutes a different dependent B on the LHS.
domain assumption All categorial analyses can be carried out using only the categories N and S.
The authors state they use only N and S; this restricts the framework and is not justified.
domain assumption Dependency relations are intrinsically order-free and can be represented without linear order.
Borrowed from DG literature (Tesnière, Debusmann); the paper relies on this to identify one dependency graph for all word orders.

invented entities (2)

The Correspondence Principle (as a formal relation)
purpose: Bridges dependency notation and categorial functor-argument notation.
No falsifiable prediction; it is stipulated and cited only to anonymous prior works.
Neutral direction functor-argument relation A|B
purpose: Expresses that either A or B can be the functor in a categorial relation.
Defined informally; no independent syntactic or semantic evidence is provided.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Towards a Unified System of Representation for Continuity and Discontinuity in Natural Language." pith.science (2026). https://pith.science/paper/OG5QG564

@misc{pith2026250605235,
  author       = {Pith},
  title        = {Pith review of: Towards a Unified System of Representation for Continuity and Discontinuity in Natural Language},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OG5QG564}},
  note         = {Machine review of arXiv:2506.05235}
}

read the original abstract

Syntactic discontinuity is a grammatical phenomenon in which a constituent is split into more than one part because of the insertion of an element which is not part of the constituent. This is observed in many languages across the world such as Turkish, Russian, Japanese, Warlpiri, Navajo, Hopi, Dyirbal, Yidiny etc. Different formalisms/frameworks in current linguistic theory approach the problem of discontinuous structures in different ways. Each framework/formalism has widely been viewed as an independent and non-converging system of analysis. In this paper, we propose a unified system of representation for both continuity and discontinuity in structures of natural languages by taking into account three formalisms, in particular, Phrase Structure Grammar (PSG) for its widely used notion of constituency, Dependency Grammar (DG) for its head-dependent relations, and Categorial Grammar (CG) for its focus on functor-argument relations. We attempt to show that discontinuous expressions as well as continuous structures can be analysed through a unified mathematical derivation incorporating the representations of linguistic structure in these three grammar formalisms.

Figures

Figures reproduced from arXiv: 2506.05235 by the authors.

**Figure 18.** A DG graph corresponding to (10) 6 In this regard, one may also take note of the treatment of verbal complexes in German with discontinuous constituents for a different analysis (see Müller 2002: 136–138, 154–15) [PITH_FULL_IMAGE:figures/full_fig_p041_18.png] view at source ↗

**Figure 27.** A dependency graph of (11) The following steps correspond to those in the CG derivation in [PITH_FULL_IMAGE:figures/full_fig_p050_27.png] view at source ↗

**Figure 34.** A dependency graph of (12) The following steps correspond to those in the CG derivation in [PITH_FULL_IMAGE:figures/full_fig_p056_34.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

4 extracted references · 4 canonical work pages

[1935]

Studia Philosophica

Die Syntaktische Konnexität. Studia Philosophica. 1, 1 -27, transl. Syntactic connexion in S. McCall, Polish Logic. Oxford 1967, 207–231. Austin, Peter and Joan Bresnan

work page 1967
[1983]

Structure and word order in Kalkatungu: The anatomy of a flat language

“Structure and word order in Kalkatungu: The anatomy of a flat language”. Australian Journal of Linguistics 3(2): 143–175. https://doi.org/10.1080/07268608308599307 Blevins, J ames P

work page doi:10.1080/07268608308599307
[1989]

Long-distance dependencies, constituent structure, and functional uncertainty

“Long-distance dependencies, constituent structure, and functional uncertainty”. In Alternative Conceptions of Phrase Structure, edited by Mark Baltin and Anthony Kroch, 17-42. Chicago: Chicago University Press. Reprinted in Mary Dalrymple, Ronald M. Kaplan, John Maxwell, and Annie Zaenen (eds.), 1995, 137 -

work page 1995
[1999]

Domains in Warlpiri

“Domains in Warlpiri”. In Sixth International Conference on HPSG–Abstracts. 04–06 August 1999, 101–106. University of Edinburgh. Dowty, D avid R

work page 1999

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1935]

Studia Philosophica

Die Syntaktische Konnexität. Studia Philosophica. 1, 1 -27, transl. Syntactic connexion in S. McCall, Polish Logic. Oxford 1967, 207–231. Austin, Peter and Joan Bresnan

work page 1967

[2] [1983]

Structure and word order in Kalkatungu: The anatomy of a flat language

“Structure and word order in Kalkatungu: The anatomy of a flat language”. Australian Journal of Linguistics 3(2): 143–175. https://doi.org/10.1080/07268608308599307 Blevins, J ames P

work page doi:10.1080/07268608308599307

[3] [1989]

Long-distance dependencies, constituent structure, and functional uncertainty

“Long-distance dependencies, constituent structure, and functional uncertainty”. In Alternative Conceptions of Phrase Structure, edited by Mark Baltin and Anthony Kroch, 17-42. Chicago: Chicago University Press. Reprinted in Mary Dalrymple, Ronald M. Kaplan, John Maxwell, and Annie Zaenen (eds.), 1995, 137 -

work page 1995

[4] [1999]

Domains in Warlpiri

“Domains in Warlpiri”. In Sixth International Conference on HPSG–Abstracts. 04–06 August 1999, 101–106. University of Edinburgh. Dowty, D avid R

work page 1999