Phasing out: On some concepts and consequences of phase-based minimalism

 

Jacek Witkoś

Adam Mickiewicz University, Poznań

 

The aim of this presentation is to critically review the gains and costs of adopting the last version of Noam Chomsky’s minimalist enterprise. It is concluded, following a detailed presentation of several aspects of the phase-based minimalism, that despite a few arguments in their favour, phases encounter a multitude of both conceptual and empirical problems.

The core of Chomsky’s (1998, 1999, 2001) current approach is illustrated by the following principles:

 

(1) Matching and Agree (Chomsky 1998:38):

  1. Matching is feature identity

  2. D (P) is the sister of P

  3. Locality reduces to 'closest c-command'

 

(2) Phase Impenetrability Condition (Chomsky 1999:10)

 

The domain of H is not accessible to operations at ZP, but only H and its edge.

Where: ZP is the least strong phase and the edge consists of specifiers.

 

Yet another significant aspect of the latest system is that the overt and the covert cycles in the derivation have collapsed into one, the single cycle system: there is only one cycle, LF and PF representations are built incrementally and backtracking is severely constrained. This solution calls for multiple application of operation Spell-out, which distinguishes between the material that can be sent to the LF-interpretive mechanisms (LF-interpretable features) and the material that can be sent to the PF interpretable mechanisms (PF-interpretable features and specified/valued uninterpretable features).

 

The advantage of the definition in (1) above is that problematic aspects of both the spec/head relation and the head/head relations as checking configurations can be avoided.

 

Another interesting argument in favour of phases is presented in Fanselow (2003), which discusses interesting cases of topicalisation in German. Here, the constituent moved to the preverbal position in V-2 contexts is only a subpart of the topic of the sentence:

 

(3)

a. War er fromm?

Was he religious?

b. In der Bibel gelesen hat er nur selten.

He has only rarely read in the Bible.

c. In der Bibel hat er nur selten gelesen.

 

In example (3c) in the Bible need not be an independent topic but can stand for the VP topic, thus examples (3b) and (3c) can have the same information structure; in the Bible serves as a proxy topic for the VP (Pars Pro Toto). Fanselow’s argument for the close, clause internal, cyclic interplay between syntax and phonology is based on the observation that the movable proxy of the VP topic in German is not identified morphologically but it is identified via the Nuclear Stress Rule (NSR), a device of the PF component. Thus, the syntactic proxy topicalisation in (3c) cannot apply before the phonological rule of NSR scans the vP constituent for the most deeply embedded phrase (here, [in der Bibel]), assigns stress to it and hereby designates it as a feasible proxy topic. Only then can the syntactic component of grammar resume its operation and move the proxy topic to [spec,C]. The vP as a phase seems like a good idea.

 

However, challenges to Chomsky’s definition of the phase seem to outnumber the advantages.

 

Chomsky’s chief argument for phases is based on the idea that without them the economical Merge over Move algorithm has to be suspended in complex expletive constructions. Yet, Hornstein and Witkoś (2003) and Witkoś (forthcoming) present an alternative view of expletive constructions:

 

(4) a. There is someone wondering whether someone is here

b. Someone is wondering whether there is someone here

 

(5) [DP [DP there] someone]

 

The view of expletive constructions in (5) does not require phases and separate sub-arrays; if there does not directly merge into Spec TP, the derivations of (4a-b) are not comparable.

 

There is also a problem with motivating Long Distance Movement:

 

(6)

a. Which book do you think Susan lost?

b. [CP [which book]i … [vP ti … v[+EPP] … [CP ti … [vP ti … v … ti ]]]]

 

Śmiecińska (2003: 63-64) challenges the minimalist rationale for such derivations, especially the reason for the [+EPP] feature on matrix v.

 

Additionally, there may be violations of the Phase Impenetrability Condition in (2). Long Genitive of Negation (LGoN) in Polish, as observed in Błaszczak (2001) and Witkoś (2003b) clearly violates the PIC:

 

(7)

a. Maria nie chce wypić kawy.

Maria not wants drink coffee-GEN/*ACC

‘Maria doesn’t want to drink coffee.’

b. ... [NegP Neg-P [VP ... v … [CP [TP PRO[+φ] [vP tPRO …v … DPOB-G[+φ]]]]]]]

 

Contrary to what it promises, the phase-based system does not exclude the derivational ‘look-ahead’:

 

(8) Multiple Wh-movement constructions in English:

a. Who knows what?

b. [TP T [vP who v [VP V what]]]

(9)

a. Who1 did Peter tell t1 [PRO to read what2]

b. Peter told this student [PRO to read what].

c. [vP PRO v [VP V what]]

d. [vP [what]1 [v’ PRO v [VP V t1]]]

e. [CP [what]1 C [TP T [vP t1 [v’ PRO v [VP V t1]]]]]

f. [vP Peter v [VP who [V’ V [CP [what]1 C [TP T [vP t1 [v’ PRO v [VP V t1]]]]]]]

 

At the local stage (9c) the Wh-phrase can either move or not, or to be more precise, following Chomsky (1998:45) either the entire Wh-phrase or only its abstract head is moved. But how can the decision be reached locally within the embedded vP?

 

Other arguments are also provided, showing that the strict notions of vP and CP phases need to be revised. In conclusion, while the head/complement configuration is useful for feature checking, derivations may be better off without phases, and even more so, if the ‘look-ahead’ is unavoidable.

 

 

Home | Abstracts