Rackowski and Richards (2005) propose that phases can be 'unlocked' to become transparent for extraction if they first Agree with a higher probe, which may then attract a goal from within the phase's domain. Provided that phases count by default as potential movable interveners due to their featural make-up (Abels 2003), then prior Agree with the phase allows the probe to skip the phase boundary, as per Richards' (1998) Principle of Minimal Compliance in (1). Thus, in many languages, extraction out of CPs or DPs requires the matrix verb to (φ-)agree with the containing phase.
(1) Once a probe P Agrees with a goal G, P can ignore G for the rest of the derivation.
Although the research program of Phase Unlocking has been prolific (see a.o. van Urk & Richards 2015; Branan 2018; Halpert 2019; Hedding & Yuan 2023; Ershova 2024; Lee & Yip 2024), it still raises important questions: (i) Are unlocking and successive cyclicity both required to escape a phase or not? (ii) How is unlocking compatible with the Phase Impenetrability Condition (Chomsky 2000) and spell-out? (iii) How does the Principle of Minimal Compliance allow to ignore an intervening locality domain? (iv) Is unlocking an exceptional primitive operation or part of a broader strategy to undo opaqueness?
To address these issues, I propose the condition in (2), for which I will provide novel empirical evidence from Austronesian Malagasy and Tagalog:
(2) Unlocking of phase XP requires remerge of XP to the edge of the probing head H, before an XP-embedded goal YP can subextract to a(n outer) SpecHP.
Thus, Agree is not sufficient for unlocking; it is only movement of the containing phase into the projection of the probing head that makes that phase transparent for extraction, yielding a multiple-specifier configuration. I suggest that (2) can ultimately provide an answer to all four open questions above.
In the end, (2) will also support the special status of phases as absolute cyclic locality domains, arguing against Halpert and Zeiljstra (2026), who subsume all phase-like locality effects to Relativized Minimality (Rizzi 1990).