Docs / Studies / Studies.Collins2026

Reconstruction and the New Minimalism

Collins, C. (2026). Biolinguistics, 20, e22899

In the framework of Marcolli, Chomsky and Berwick, Internal Merge replaces a tree $T$ by the moved term $\beta$ and the contraction quotient $T/^c\beta$, then merges the two, and the quotient is what the semantic interface receives. Its contracted leaf has no structure, so in Pictures of himself, John loves the subject does not c-command the anaphor, though under the copy theory of movement it does; negation likewise loses the quantifier it scopes over. The quotient alters one of the two objects merged, which MERGE leaving copies never does. Collins concludes that the framework cannot capture reconstruction. That holds of a condition read off the final object: a condition evaluated at some earlier stage binds the anaphor with traces alone. With the copies kept, a derivation has a reading for each choice of the chains read low: read low, the antecedent binds the anaphor in the fronted phrase's lower copy, and (15) has both its scope readings through the composition engine. Both cut the quantifier phrase from the copy object, and the low reading is no term of the coproduct the book sends to the semantic interface, which leaves a trace at every cut.

Implementation notes

  • The condition checked is the c-command half of Principle A. The coindexation is given by the paper's example and its local domain is the clause, and the argument turns on c-command only.
  • The derivations suppress the projections the paper suppresses in its tree for the fronted sentence, so the fronted phrase merges with the clause directly.
  • The trace leaf carries the head of the moved term, where the book's label is the term itself. The paper's point is that c-command does not read labels, and the definition here does not.
  • That a condition at an earlier stage recovers the binding is not the paper's claim. It applies the derivational binding condition Cole and Hermon use, and says nothing about whether such a condition is available to the framework the paper criticizes.
  • The reading with no chain read low is the book's contraction with a binder the book does not have. The low reading deletes the higher copy, where the book removes only the lower one, so the two scope readings are the copy theory's, not the framework's.

References

Reconstruction for Principle A

Reconstruction for scope

The readings of one derivation

One derivation of (15) has both scope readings through the composition engine: read low, not every player scored, the reading the paper describes in (16); read high, every player failed to.

theorem scope_readingsL222–232

Built from , ,

The readings as cuts

The readings of (15) both cut the quantifier phrase from the copy object. The high reading is a term of Δ^c, cutting the lower copy, and the low reading a term of Δ^d, cutting the higher copy. The low reading is no term of Δ^c, whose cuts each leave a trace.

theorem scope_readings_cutL244–272

Built from ,

No Tampering

MERGE leaving copies keeps every term of the workspace, the clause among them. MERGE leaving the contraction quotient loses the clause, which is the violation of No Tampering that Collins blames, and which the workspace of Collins and Stabler does not incur.

theorem no_tamperingL276–284

Built from ,