The Inclusiveness Condition in Survive-minimalism

The Inclusiveness Condition in Survive-minimalism Minoru Fukuda Miyazaki Municipal University fukuda@miyazaki-mu.ac.jp March 2013 1. Introduction Given a phonetic form (PF) representation! and a logical form (LF) representation ", the computational system of human language, C HL, maps a numeration to (!, "). 1 According to Chomsky (1995, 2000), this mapping procedure is subject to an inviolable principle referred to as the Inclusiveness Condition (IC), which precludes features absent in the numeration from entering C HL. 2 However, adopting the derivational formation of a numeration proposed by Stroik (2009a), we argue that the IC is no longer required as an independent principle under the strictly derivational theory of syntax referred to as Survive-minimalism (Stroik (2009b)). 3 In the standard version of Minimalism, a numeration has been assumed to be constructed according to the following principles: (i) it is determined before C HL starts; (ii) its members lexical items are blindly selected from the lexicon; (iii) they are selected all at once; and (iv) they are unordered. 4 However, Stroik (2009a) accepts none of these principles, instead favoring the idea that the formation of a numeration proceeds piecemeal, which leads to a number of significant consequences, as discussed below. The organization of this paper is as follows. In section 2, after a brief review of the definitions of the IC, we will point out its two implicational requirements. In section 3, following Stroik (2009a), we see that the standard Minimalist assumptions cause serious problems. Then, illustrating the derivational steps of construction building and feature checking, we will show in section 4 how Stroik s (2009a) piecemeal numeration formation solves these problems and derives the effects of the IC. In section 5, on the basis of Survive-minimalist assumptions (Stroik (2009b), Stroik and Putnam (2010)), we will argue that the effects of Chomsky s (2008) no tampering condition are also derivative. Consequently, Survive-minimalism offers a simpler, more restrictive theory of C HL than that provided by Chomsky. 2. The Inclusiveness Condition To begin this discussion, we would like to examine the two versions of the IC 1

mentioned above from derivational perspectives. In an early stage of Minimalist inquiry, Chomsky (1995: 228) states the IC as in (1), which is slightly different from its more recent versions in a significant respect. (1) A perfect language should meet the condition of inclusiveness: any structure formed by the computation (in particular,! and ") is constituted of elements already present in the lexical items selected for N; no new objects are added in the course of computation apart from rearrangements of lexical properties (in particular, no indices, bar levels in the sense of X-bar theory, etc.; see note 7). Later, the term objects in (1) is replaced with the term features, in Chomsky (2000: 113). (2) The Inclusiveness Condition: No new features are introduced by C HL. This is a natural consequence because lexical items are composed of features and thus syntactic objects are ultimately assembled out of features as well (see Chomsky (2000: 100 101)). More importantly, the two versions share the qualification no new, which induces a significant implication that only the features present in the numeration can participate in C HL. Thus, using the Agree-based theory, Hornstein, Nunes, and Grohmann (2005: 74) define the IC as in (3); N indicates Numeration. (3) The LF object " must be built only from the features of the lexical items of N. Two remarks are in order here. First, the implication given above arguably serves as a theoretical basis of the standard Minimalist assumption that the numeration formation must be completed before the derivational computation begins, though, as we will argue in section 4, it is possible to disregard the standard assumption and still maintain its effects. Second, this implication, taken as a principle, induces a further implication that the IC prevents C HL from having direct access to the lexicon, limiting its access domain to the numeration. We will argue in section 3 that the same effects are present under Stroik s (2009a) derivational manner of numeration formation. 3. Premises for Numeration Formation Since Chomsky s (1995) first proposal, it has been assumed in Minimalist syntax that the numeration is thoroughly predetermined prior to the initial stage of syntactic derivation. This means that the lexical items used for the construction of the target 2

syntactic structure are all preselected from the lexicon before the syntactic derivation starts. This has been widely adopted as a standard Minimalist assumption (see Hornstein, Nunes, and Grohmann (2005: 69-70)). However, there are two more premises behind the standard method of numeration formation: first, that the lexical items are selected all at once, and second, that they are unordered in the numeration. To briefly recapitulate, it has been widely assumed that the numeration formation is based on the following four premises: (i) the numeration is built before syntactic derivation begins; (ii) its members lexical items are blindly or randomly selected from the lexicon; (iii) they are selected all together, so the numeration is built all at once; and (iv) the members are unordered. 5 Thus, in order to derive sentences such as (4), the numeration, indicated as NUM in (5), must be built all at once with seven unordered lexical items preselected from the lexicon. 6 (4) There seems to be someone here. (5) NUM = {there 1, T 1, seem 1, T to1, be 1, someone 1, here 1 } Stroik (2009a: 26 27), however, after noting the premises behind the standard assumption, points out daunting search problems with it. 7 For example, if a numeration consists of ten lexical items, there should be in principle 10!, or 3,628,800, structured strings of words generated from the numeration. Even in the case of (5), we will have 7!, or 5,040, strings of the lexical items. Despite the enormous processing labor involved here, a large majority of the strings crash or are filtered out, and only an extremely few can reach the interfaces as interpretable derivations. This leads Stroik (2009a: 29) to state that standard Minimalism, as it currently conceives of the NUMeration, necessarily produces a grossly crash-rife syntax, not a crash-proof syntax. 8 If we more seriously consider the principle of blind selection for the numeration, stated as premise (ii) above, the situation is in fact even worse than that just described. Let us suppose that our lexicon consists of 10,000 lexical items, from which ten are selected. Since the selection applies blindly and the same items can be selected more than once, in principle there should be 10,000 10 ways to select ten out of 10,000. These naturally include what Hornstein, Nunes, and Grohmann (2005: 71) refer to as crazy numerations such as (6), which never lead to convergent interface-interpretable derivations. (6) NUM = {there 1, T 1, seem 1, T to2, be 1, get 1, yesterday 1, someone 2 } 3

It seems plausible to suppose that most of the numerations constructed in the manners indicated above are of the crazy type, but since there is no principled way to exclude such cases, C HL has to compute all of them. Thus, the random selection for a numeration will impose a further considerable amount of fruitless workload on C HL despite the extremely low probability of successful selection. Another problem with the standard Minimalist assumption is the presence of the look-ahead mechanism. For example, Chomsky (2000) proposes that the formation of a numeration is phrase based, so that the numeration for construction (4) should consist of two sub-arrays, as indicated in (7), provided that T as well as C and v heads a phase. (7) NUM = {{there 1, T 1, seem 1 }, {T to1, be 1, someone 1, here 1 }} The phase-based formulation of a numeration allows us to reduce these costly computational demands to some extent and to remove the possibility of over-generation of constructions such as (8) without recourse to Chomsky s (1995) proposal that Move is more costly than Merge (see also Hornstein, Nunes, and Grohmann (2005: 357)). 9 (8) *There seems someone to be here. However, let us point out that certain premises are implicit under Chomsky s (2000) proposal. First, according to premise (iv) above, the sub-arrays are also unordered in the numeration, but it is implicitly assumed that C HL first computes the second sub-array, {T to, be, someone, here}, and only then the first one, {there, T, seem}. Second, each of the two sub-arrays is implicitly assumed to have a specific phase head. Thus, the phase head of the first sub-array must have T as a tense-checker and the second one must have T to as a non-tense-checker. Without these implicit premises, daunting search problems will once again arise. Putting it differently, the phase-based numeration should have a look-ahead capacity to avoid these problems; in particular, it should be able to anticipate how the derivation will proceed when the numeration is formed through the selection of lexical items and the formation of sub-arrays. Obviously, the look-ahead property of a numeration is inconsistent with the tenets of strictly derivational theories. 4. Numeration in Survive-minimalism In order to circumvent the above-mentioned fruitless generative property of a numeration, Stroik (2009a) renounces the standard Minimalist view of numeration, 4

and proposes that a numeration is not blindly built all at once but instead piecemeal, as mentioned above. In other words, all LIs [lexical items] enter NUM one at a time, as they are derivationally required (Stroik (2009a: 30)). Thus, the selected items in the numeration are not unordered. These amount to a discarding of the four manners of numeration formation stated in section 3, and the assumption of alternates: (i) a numeration is not built before syntactic derivation starts but instead in the course of derivation; (ii) lexical items are derivationally selected or copied from the lexicon as required; (iii) they are copied one at a time; and (iv) the members of a numeration are ordered. Specifically, the derivational sequence is subject to the hierarchical feature matrices (FMs) of the lexical items of a numeration (Stroik (2009a: 31)). (9) FM = <SUBCAT <IF-selecting <CAT <IF-selected>>>> All the features in the FMs play a major role in concatenating operations such as Merge and Remerge by means of feature checking. SUBCAT is a subcategorization feature that checks a specific category feature indicated as CAT (e.g. D, V, M(odal), etc.). IF-selecting is an interface feature that checks a specific interface feature indicated as IF-selected (e.g. Case, Q(uestion), WH, etc.). Since the checking operation proceeds in the order indicated in FM, the checking operation between SUBCAT and CAT must precede the one between IF-selecting and IF-selected. More importantly, Stroik (2009a) argues that checking features such as SUBCAT and IF-selecting cannot be independently introduced to derivation; instead, their introduction is always motivated by the presence of CAT and IF-selected. For example, the CAT feature introduced in a derivation drives the introduction of a lexical item with the SUBCAT feature in the next stage of derivation. The same ordering restriction applies in the case of IF-selecting and IF-selected features. As a consequence, the first item introduced to C HL must have a CAT feature rather than a checking or selecting feature. In other words, there is a first element in the numeration, and the members of the numeration are derivationally ordered by the FM. To see how Stroik s (2009a) proposal works, let us examine the derivation of (10). 10 (10) Who will Pat see? The first item to be introduced in this derivation is the wh-element who, the FM of which is <CAT-D <WH>>, and which is copied from the lexicon to the numeration and again from the numeration to a derivation. 11 5

(11) NUM = {who<cat-d <WH>>} The hierarchy of the FM requires the CAT-D feature to be checked first, for which it is necessary to introduce a lexical item with the SUBCAT-D feature. Thus, the transitive verb see, whose FM is <SUBCAT-D <<SUBCAT-D <CAT-V>>>, is copied from the lexicon to the numeration and again from the numeration to the derivation for Merge. 12 (12) a. NUM = {see<subcat-d <<SUBCAT-D <CAT-V>>>, who<cat-d <WH>>} b. Merge {see, who} When see and who are Merged, the CAT-D feature of who is checked by the SUBCAT-D feature of see. 13 However, the checked features are not deleted or erased under Survive-minimalism but instead remain intact. The unchecked features then survive, inducing further Copy and (Re)Merge operations in compliance with Stroik s (2009a: fn.12) survive principle, until they are all checked. 14 (13) If Y is a syntactic object SO merged into a syntactic projection headed by X and Y has Y has features incompatible with (i.e., cannot potentially be checked by) the features of X, then Y remains active in NUM. According the hierarchical order of the FM, the next feature to be checked is the SUBCAT-D feature of see, and the lexical item Pat, whose FM is <CAT-D <CASE>>, is copied from the lexicon to the numeration and again from the numeration to the derivation, in which it is Merged with the SO indicated in (12b). 15 (14) a. NUM = {Pat<CAT-D <CASE>>, see<subcat-d <<SUBCAT-D <CAT-V>>>, who<cat-d <WH>>} b. Merge {Pat, {see, who}} As the next step of derivation, the hierarchical order of the FM instructs C HL to introduce a category bearing the SUBCAT-V feature for checking. Thus, the modal will is copied from the lexicon to the numeration and again from the numeration to the derivation. Incidentally, will can have the IF-selected Q feature as well as the CASE-NOM(inative)-selecting feature. The output of Merge is shown in (15b). (15) a. NUM = {will<subcat-v <CASE-NOM <CAT-M <Q>>>>, Pat<CAT-D <CASE>>, see<subcat-d <<SUBCAT-D <CAT-V>>>, who<cat-d <WH>>} b. Merge {will, {Pat, {see, who}}} The next feature to be checked is the CASE-NOM-selecting feature, which is 6

located higher in the FM than the other unchecked features in the numeration. In this case, there is no need to apply Copy, because the numeration has the lexical item Pat, whose CASE feature is still unchecked. 16 Thus, the survive principle is invoked and Pat is copied from the numeration to the derivation for Remerge. (16) a. NUM = {will<subcat-v <CASE-NOM <CAT-M <Q>>>>, Pat<CAT-D <CASE>>, see<subcat-d <<SUBCAT-D <CAT-V>>>, who<cat-d <WH>>} b. Remerge {Pat, {will, {Pat, {see, who}}}} At this stage, there are three features unchecked, and the CAT-M feature is located higher than the other two IF-selected features in the FM. Therefore, a complementizer C with the SUBCAT-M feature is copied from the lexicon to the numeration and again from the numeration to the derivation. 17 Incidentally, C can also check the WH feature as well as the Q feature, and thus its FM is something like <SUBCAT-M <Q <WH>>>. 18 Then, C is Merged with the SO indicated in (16b). (17) a. NUM = {C<SUBCAT-M <Q <WH>>>, will<subcat-v <CASE-NOM <CAT-M <Q>>>>, Pat<CAT-D <CASE>>, see<subcat-d <<SUBCAT-D <CAT-V>>>, who<cat-d <WH>>} b. Merge {C, {Pat, {will, {Pat, {see, who}}}}} The two unchecked features are in turn checked through Remerge operations. First, the modal will is copied from the numeration and Remerged with the already constructed SO shown in (17b) for the Q feature. (18) a. NUM = {C<SUBCAT-M <Q <WH>>>, will<subcat-v <CASE-NOM <CAT-M <Q>>>>, Pat<CAT-D <CASE>>, see<subcat-d <<SUBCAT-D <CAT-V>>>, who<cat-d <WH>>} b. Remerge {will, {C, {Pat, {will, {Pat, {see, who}}}}}} Second, the wh-element who is copied from the numeration and Remerged with the SO indicated in (18b) for the WH feature. (19) a. NUM = {C<SUBCAT-M <Q <WH>>>, will<subcat-v <CASE-NOM <CAT-M <Q>>>>, Pat<CAT-D <CASE>>, see<subcat-d <<SUBCAT-D <CAT-V>>>, who<cat-d <WH>>} b. Remerge {who, {will, {C, {Pat, {will, {Pat, {see, who}}}}}}} 7

Because there is no element in the numeration with an unchecked feature, the derivation terminates; then, as a well-formed derivation, it is sent to the interfaces for interpretation. Thus, the movement or displacement property of human language is explicable by means of strictly local operations such as Merge and Remerge. Survive-minimalism also offers a crash-proof syntax that circumvents Brody s (2002) criticism of derivational theories as well as the daunting search problems in a successful manner. Let us recall here the discussion provided in section 2. Two implications were induced from the definitions of the IC: (i) all the objects or features used in C HL must be present in the numeration, and (ii) C HL has direct access to the numeration but only indirect access to the lexicon. We should notice here that even under Stroik s (2009a) piecemeal approach to the formation of a numeration, these two effects are obtainable. First, since the numeration is formed piecemeal, as required in conformity with the FMs, the features participating in C HL are all present in the numeration and other features cannot be present. Second, as indicated in this section, in the (Re)Merge operation, features or lexical items are first copied from the lexicon into the numeration and then from the numeration into a derivation. These copy operations apply one at a time. Thus, C HL has access to the lexicon through the numeration, which means that it has direct access only to the numeration. If the above-provided arguments are on the right track, the IC will be a derivative or secondary notion under Survive-minimalism. Its conceptual necessity will be undermined, and hence, in this case, it should not be adopted as an independent principle constraining the derivational computation of human language. 5. The No Tampering Condition Given Stroik s (2009a) piecemeal numeration formation, it is possible to dispense with the IC as redundant. In fact, Survive-minimalism provides us with further implications with regard to Chomsky s (2008) no tampering condition (NTC), by which the IC can be motivated. Chomsky s (2008: 138) introductory statement of the NTC is shown in (20). (20) A natural requirement for efficient computation is a no-tampering condition (NTC): Merge of X and Y leaves the two SOs unchanged. If so, then Merge of X and Y can be taken to yield the set {X, Y}, the simplest possibility worth considering. Merge cannot break up X or Y, or add new 8

features to them. Therefore Merge is invariably to the edge and we also try to establish the inclusiveness principle, dispensing with bar levels, traces, indices, and similar descriptive technology introduced in the course of derivation of an expression. Summarizing the NTC as in (21), Narita (2011: 16) argues that the NTC, like the IC, is a specific manifestation of more general principles of computational efficiency. (21) No elements introduced by syntax are deleted or modified in the course of linguistic derivation. We agree with Narita (2011) that the NTC is not primitive but derivative. However, we would like to argue that it is redundant because, as we will readily show, the Survive-minimalist syntax has already incorporated what the NTC imposes on C HL. For ease of exposition, let us first summarize the main contentions of (20) and (21) in the following way: once an element is introduced to C HL, (i) it cannot be deleted; (ii) it cannot be modified; (iii) a new feature cannot be added to it; and (iv) in consequence, Merge always applies to its edge position (irrespective of whether it is Internal Merge or External Merge). Since the third contention manifests itself as the IC, which has already been discussed in section 4, we would like to argue that the other three are derived from basic Survive-minimalist assumptions. First, as exemplified in (10) (18), neither feature deletion nor copy deletion is employed under Survive-minimalism (see also Stroik (2009b)). In Survive-minimalism, all features are interface-compatible and interface-identified, so there is no need to delete them. If uninterpretable features are deleted in the course of derivation, how can we expect a child to learn these features from data without them? Second, as Brody (2002: 22 23) argues, the derivations create opaque objects whose internal elements and composition is not accessible to any further rule or operation, and thus no derivational process can operate inside of the Merged syntactic objects. Since Survive-minimalism lays the groundwork for nonrepresentational derivational theories, the internal structure of SOs naturally remains as a sanctuary through derivational computation. Thus, there is no way to modify syntactic objects that have already been introduced to C HL. On the other hand, the standard version of Minimalism cannot but assume conditions such as the NTC because, equipped with look-ahead and look-back mechanisms such as phase-based formation of a numeration and the phase 9

impenetrability condition, it is not a strictly derivational theory (see Brody (2002)). Lastly, it follows from the second argument just given that (Re)Merge is always applied to what Chomsky (2008) calls the edge position rather than to the inner position. In fact, Stroik and Putnam (2010) argue that every application of (Re)Merge introduces a new item to the left-periphery. 19 (21) Merge s Uniformity of Directionality (MUD): All (Re)Merge operations are left-merge operations. These arguments suggest that the NTC effects as well as the IC effects emerge as a corollary of fundamental Survive-minimalist assumptions. Therefore, there is no conceptual necessity to adopt the NTC as an independent principle. 6. Concluding Remarks Any version of Minimalism must adopt and assess its notions and grammatical apparatuses in terms of conceptual necessity. We have argued that neither the IC nor the NTC has a solid conceptual ground as a condition on C HL under Survive-minimalism. We hope that the arguments provided here have shown that Survive-minimalism paves the way to an optimal solution to the question of the design of the human faculty of language, through a conceptually simpler and purely derivational approach to C HL with a crash-proof syntax. References Boeckx, Cedric (2012) Syntactic Islands: Key Topics in Syntax, Cambridge University Press, Cambridge, the UK. Brody, Michael (2002) On the Status of Representations and Derivations, Derivation and Explanation in the Minimalist Program, ed. by Samuel David Epstein and T. Daniel Seely, 19 41, Blackwell, Oxford. Broekhuis, Hans and Wim Klooster (2010) Merge and Move as Costly Operations, ReVEL, Special edition, Number 4, 155-182. Chomsky, Noam (1975) Reflections on Language, Pantheon, New York. Chomsky, Noam (1995) Categories and Transformations, The Minimalist Program, 219 394, MIT Press, Cambridge, MA. Chomsky, Noam (2000) Minimalist Inquires: The Framework, Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik, ed. by Roger Martin, David Michaels, and Juan Uriagereka, 89 155, MIT Press, Cambridge, MA. 10

Chomsky, Noam (2004) Beyond Explanatory Adequacy, Structures and Beyond: The Cartography of Syntactic Structures, ed. by Adriana Belletti, 104 131, Oxford University Press, Oxford. Chomsky, Noam (2008) On Phases, Foundational Issues in Linguistic Theory: Essays in Honor of Jean-Roger Vergnaud, ed. by Robert Freidin, Carlos P. Otero and Maria Luisa Zubizarreta, 133 166, MIT Press, Cambridge, MA. Collins, Chris (2002) Eliminating Labels, Derivation and Explanation in the Minimalist Program, ed. by Samuel David Epstein and T. Daniel Seely, 42 64, Blackwell, Oxford. Frampton, John and Sam Gutmann (2002) Crash-Proof Syntax, Derivation and Explanation in the Minimalist Program, ed. by Samuel David Epstein and T. Daniel Seely, 90 105, Blackwell, Oxford. Hornstein, Norbert, Jairo Nunes, and Kleanthes K. Grohmann (2005) Understanding Minimalism, Cambridge University Press, Cambridge, the UK. Moro, Andrea (1997) The Raising of Predicates: Predicative Noun Phrases and the Theory of Clause Structure, Cambridge University Press, Cambridge, the UK. Müller, Gereon (2011) Constraints on Displacement: A phase-based approach, John Benjamins Publishing Company, Amsterdam/Philadelphia. Narita, Hiroki (2011) Phasing in Full Interpretation, Doctoral dissertation, Harvard University. Seely, T. Daniel (2006) Merge, Derivational C-Command, and Subcategorization in a Label-free Syntax, Minimalist Essays, ed. by Cedric Boeckx, 182 217, John Benjamins Publishing Company, Amsterdam/Philadelphia. Stroik, Thomas S. (2009a) Locality in Minimalist Syntax, MIT Press, Cambridge, MA. Stroik, Thomas S. (2009b) The Numeration in Survive-minimalism, Towards a Derivational Syntax: Survive-minimalism, ed. by Michael T. Putnam, 21 37, John Benjamins Publishing Company, Amsterdam/Philadelphia. Stroik, Thomas S. and Michael T. Putnam (2010) Syntactic relations in Survive-minimalism, Exploring Crash-Proof Grammar, ed. by Michael T. Putnam, 143 165, John Benjamins, Amsterdam and Philadelphia. 11

Notes 1 For expository purposes, we will ignore the distinction between numerations and lexical arrays in this paper. See Chomsky (2004: 107) for this distinction. 2 Müller (2011: 127) argues that the IC is violable with regard to the introduction of edge features, but Boeckx (2012: 70, 146) points out problems with this view. As far as we can see, since Müller s (2011) theory accommodates look-ahead and look-back mechanisms such as Spreading of / / and the phase impenetrability condition, respectively, it cannot serve as a purely derivational theory despite this self-profession. 3 With regard to the necessity of a numeration, see Stroik (2009a: 29 30). For the opposite view, see Broekhuis and Klooster (2010). 4 Since the members of a numeration are assumed to be unordered, problems will arise irrespective of whether they are lexical items or phases. 5 The all-at-once construction of a numeration is arguably subsumed under an extensional as opposed to an intensional approach in the sense of Chomsky (1975: 120). On the other hand, Stroik s (2009a) piecemeal numeration construction can be subsumed under the latter approach. 6 The example and its numeration are quoted from Hornstein, Nunes, and Grohmann (2005: 336) with minor modification. The subscript numbers indicate how many times lexical items are selected from the lexicon. If we follow Chomsky s (1995: 225) notation, (5) is indicated as (i), but for our purpose, nothing is contingent on these notational differences. (i) NUM = {(there, 1), (T, 1), (seem, 1), (T to, 1), (be, 1), (someone, 1), (here, 1)} 7 The situation is not ameliorated even if we assume a feature-match condition on Merge or, as we will argue below, a phase-based numeration formation. See Stroik (2009a: 27 28). 8 We do not review arguments for crash-proof syntax in the present paper. Readers should refer to Frampton and Gutmann (2002) and Stroik (2009b). 9 Stroik (2009: 28) argues that phase-based numeration formation does not contribute to a radical reduction of unsuccessful derivations. With regard to the difference between (4) and (8), Moro (1997) argues in support of the predicate-inversion analysis. Thus, it may well be that neither the cost calculation of Move and Merge nor phase-based numeration formation is relevant to these cases. 10 Our illustration of the derivation of (10) is based on that of Stroik (2009a: 31 33). Under Survive-minimalism, operations are restricted to Copy, Merge, and Remerge (or External Merge), and there are no operations such as Select and Move (or Internal Merge). For the arguments against Move (or Internal Merge), see Stroik (2009b). 11 The FMs of lexical items are indicated in the numeration. For expository purposes, we follow Stroik (2009a: 31 33) in ignoring the Case feature checking of who. We also do not indicate the projections of SOs derived by (Re)Merge. For a more detailed illustration of derivation, see Stroik and Putnam (2010). Readers may find that the Survive-minimalist method of structure building is well compatible with the label-free syntax argued for by Collins (2002) and Seely (2006). 12 The embossed notation indicates checked features. 13 The Locus Principle, discussed by Collins (2002) and Seely (2006), is similar to Stroik s (2009a) piecemeal numeration formation only with regard to SUBCAT and CAT feature checking. 14 A more precise definition of the survive principle is given in (i), based on Stroik 12

(2009b: 45): (i) If Y is an SO in an XP headed by X and Y has an unchecked feature incompatible with (i.e., cannot potentially be checked by) the features of X, Y must Remerge from the WorkBench with the next head Z that c-commands XP. 15 As already indicated, a checking operation regarding SUBCAT and CAT features must occur before one regarding IF-selecting and IF-selected features. 16 A natural question here is why an extra lexical item is not copied from the lexicon to the numeration to check the CASE feature in this case. Stroik and Putnam (2010: 155) argue that the Remerge operation is more economical than the Merge operation because the former makes use of items already available in the numeration, while the latter requires an extra Copy operation. We would like to speculate that this Remerge-over-Merge property manifests itself because C HL can have direct access only to a numeration. Thus, in computational processes, C HL invariably looks at the numeration first, and instructs it to copy a necessary item from the lexicon only if there is no appropriate candidate for Merge in the numeration. If so, the cost calculation of Copy is irrelevant to the preference to Remerge over Merge. 17 The CAT-C feature of this C is not indicated in its FM. This is because its major role is to terminate the derivation. If C has the CAT-C feature, it will serve as a head of a complement clause to be selected by a verb or an adjective with a SUBCAT-C feature. Thus, the Survive-minimalist way of syntactic computation clarifies how the derivation ends as well as how it starts. See Stroik and Putnam (2010: 151 152) for discussion of categories of this kind. 18 Stroik (2009a: 33) assumes that the Q feature is located higher than the WH feature, though both of them are IF-selected features. 19 Thus, the extension condition effects can be expected as well. 13