Significant mid-level results of generative linguistics

At the Athens conference in May 2015 (Generative Syntax in the 21st Century: The Road Ahead), the assembled linguists discussed some of the significant mid-level results that have come out of generative linguistics. These are results in the sense that they have been established reasonably well, for a range of languages, and are generally accepted by consensus as being right. They can be assumed as background in the pursuit of other research. This doesn’t mean that they are incontrovertable, but the negation of any of them would be a proposal that would have to be argued for. They are mid-level in the sense that they are general enough to apply to a large sample of languages, and specific enough that it is fairly clear how to test them.

They are grouped (after the fact) into approximately thematic clusters.

  1. Unaccusativity (Robust): There are two classes of monovalent verbs such that the argument in the unaccusative class is predicate-internal, while the argument in the unergative class is predicate-external (in derivational terms, the unaccusative argument originates predicate-internally)
  2. The Agent asymmetry (Robust): NPs bearing Agent roles are higher than NPs bearing other roles in the unmarked structure of the clause.
  3. Passive valence reduction: Agents are the easiest arguments to suppress in valency reduction.


  1. X-bar theory, categories, and headedness
  2. Extended projections: Clauses and nominals consist of a (respectively) verbal/nominal head, dominated by zero or more members of an ordered sequence of functional elements
  3. Cinque hierarchy (Good): There are semantically defined classes of TAM functors that appear in the same hierarchical order in all languages in which they exist overtly.
  4. Cinque hierarchy for adverbs: There are semantically defined classes of adverbs that appear in the same hierarchical order in all languages in which they exist overtly (related to or identical to the TAM hierarchy)
  5. Morphology Mirrors Syntax (Robust): The hierarchy of projections as reflected in free words is the same one that is reflected in morphological structure when morphemes express the same notions as the free words.
  6. CP-DP parallelism (Robust): There are substantive parallels in structure between noun phrases and clauses, most obviously in the case of nominalizations but also detectable in other kinds of nominals (e.g. similarities between subjects and possessors, subject to cross-linguistic variation).
  7. The final over final constraint: It is relatively difficult to embed head-final projections in head-initial ones, compared to the opposite (132 but not *231, where 1 takes 2 as a complement and 2 takes 3)
  8. Cinque’s version of Greenberg’s U20: Only one unmarked order is found prenominally for Dem, Num, and Adj, namely Dem > Num > Adj > N; ordering possibilities increase as N is further to the left in the sequence. The facts suggest (i) a universal hierarchy Dem > Num > Adj > N, where these categories exist, (ii) the possibility of leftward but not rightward movement of projections of N to derive some other orders, and (iii) the absence of such movement of adnominal modifiers alone (e.g. no information-neutral movement of Adj across Num and/or Dem unless it is in a projection containing N) (May generalize to other categories)
  9. Functional Material Doesn’t Incorporate (Robust): Higher functional structure such as determiners and complementizers doesn’t incorporate into superordinate lexical heads.
  10. SOV scrambling: All SOV languages allow a degree of word order freedom (scrambling); VO languages may not.
  1. Movement in general (not restricted to A-bar or A)
  2. Coordinate Structure Constraint: Extraction from a Coordinate Structure is not possible unless it is by Across-the-Board movement (the phenomenon of pseudocoordination has to be distinguished; e.g. “What did you go (to the store) and buy?”; pseudocoordination shows characteristic properties, for example a restricted class of possible left-hand categories (cf. *“What did you walk and buy?”), extraction only from the open-class right-hand member (cf. *“Which store did you go to and buy shrimp?”)
  3. Head Movement Constraint (Robust): Head movement doesn’t cross heads. This cannot be escaped by excorporation: If X moves to Y by head-movement, then X cannot move on, stranding Y. (Clitic movement crosses heads and must be distinguished from head movement proper, i.e. head movement of complements in extended projections to their selecting projections, and of incorporees to their selecting predicates)
  4. Movement is upward (Robust): Movement is upward, landing in higher syntactic positions.
  5. Right Roof: Rightward movement is clause bounded (“the right roof constraint”)
  6. Second position (Robust): There are second position effects which are category-insensitive, i.e. not sensitive to the category of the element in first position, but no second to last effects which are similarly category-insensitive. (This allows for immediately pre-verbal positions in V-final structures)
  7. Syntactic clitic placement (Robust): A major class of clitics (phonologically dependent items) have their location in the surface string determined by purely syntactic principles of the language (i.e. ignoring the phonological dependency).


  1. Binding Theory
  2. Principle B (Robust): Pronouns, in the unmarked case, can’t be locally bound (under the same A-position class of locality as for Principle A), but can be bound nonlocally
  3. Principle C (Robust): an R-expression can’t be bound by (systematically corefer with) a c-commanding pronoun
  4. Structure relevant to binding (Robust): The conditions on pronominal reference cannot be stated purely with linear order. The subject-nonsubject distinction plays an important role, especially for Principle A (and B to the extent that it is complementary).
  5. Strong crossover: Coreference is impossible between a pronoun in an argument position and a c-commanding antecedent when the antecedent has moved across the pronoun; i.e. is the head of a filler-gap dependency where the gap is c-commanded by the pronoun. Example: “Who did he say was hungry?” Coreference impossible.
  6. Weak crossover: Coreference is degraded between a pronoun and a c-commanding antecedent when the antecedent has moved across the pronoun; i.e. is the head of a filler-gap dependency where the gap is lower than the pronoun. Example: “Who did his mother say was hungry?” Coreference degraded.
  1. Arguments
  2. Improper movement: A-positions (as diagnosed by case, agreement, and binding) feed unbounded dependencies (e.g. the tail of a wh-movement, relative clause formation, or topicalization chain is in an A-position). Unbounded dependencies preserve case, agreement, and binding configurations, and do not (normally) feed A-positions (i.e. they do not normally increase the possibilities for an element to enter case-agreement-relevant relations, unlike passive, raising, etc.).
  3. Control versus raising (Robust for control): Obligatory control is a subject-to-subject relation (or, in some cases, object-to-subject relation) in which one referent gets thematic roles from two predicates, related to each other by nonfinite complementation; in Raising, the shared argument gets only one thematic role, from the embedded predicate
  4. Structural agreement (Very robust): There is a structural bias affecting agreement such that nominals higher in the clause are agreed with in preference to lower nominals, except where marked case on a higher nominal may disqualify it. (reflected in subject agreement over object agreement)
  5. Grammatical Subject (Robust): There is a distinction between grammatical subject and thematically highest argument (though traditional subject diagnostics may decompose even further)
  6. Diesing’s Generalization (Robust): If uniquely referring DPs (definites and/or specifics; Milsark’s “strong” noun phrases) and weak indefinites with the same grammatical function occupy different positions, then the uniquely referring DPs are structurally higher.
  7. PCC: Languages place strong restrictions on the use of local direct objects when a goal NP is present (NP, or DP, as opposed to PP).
  8. No NCC: There is no number case constraint; languages do not restrict the grammatical number of the direct object when a goal NP is present.
  9. Ergative subjects (Robust): Asymmetries between arguments for purposes of unmarked word order, binding, and control work the same way in nominative and ergative languages. Clause structure in ergative and accusative languages is homomorphic
  10. Null subjects: Many languages allow pronouns to be unpronounced in certain positions under certain conditions. Where possible, these pronouns act much like overt pronouns for e.g. Binding Conditions.
  11. High causatives: In a morphological causative, the new causee will be higher than any argument of the base verb.
  12. Marantz’ Generalization: In benefactive applicative constructions, the new argument will be structurally higher than the base internal argument.
  13. Erg Agreement is dependent on Erg case: No language has a nominative-accusative case system and an ergative-absolutive agreement system, although matched systems are possible, and the opposite mismatch is possible (Bobaljik 2008, and typological sources).
  14. No Active Case: No language has an active system of case marking, whereas active systems of agreement marking are possible. (Baker and Bobaljik in press/in progress, but well documented)


  1. Quantifier Raising
  2. QR: The logical scope of natural language quantifiers (over individuals, times or situations/worlds) does not have to match their surface position. Quantifier scope is co-determined by structural factors (islands, clausal boundaries), logical properties of the quantifier (universal vs. existential) and the form of the quantificational expression (simple vs. complex indefinites).
  3. QR is clause bound: The scope of (expressions corresponding to) universal quantifiers is limited by conditions identical or very similar to the conditions on A-movement (clause bounded, except in restructuring contexts).
  4. Widest scope indefinites: In many languages, morphologically simple indefinites (some books at least one book) may take unbounded scope, even across islands.
  5. Reconstruction: Dislocated quantificational expressions can take scope below their surface position, but no lower than their base position.


A-bar. A-bar phenomena

  1. A-bar Unity (Robust): A class of A-bar (filler-gap) constructions (including interrogatives, relative clauses, focus movement constructions, and operator-variable chains) show unified behavior with respect to locality and configuration.
  2. Successive Cyclicity (Robust): Unbounded dependencies are successive-cyclic, as diagnosed by locality effects.
  3. Covert A-bar dependencies: There are operator-variable relations where the operator is low on the surface that are restricted by the same laws as A-bar dependencies, where the A-bar element is high on the surface. For example, the interpretation of wh-in-situ for selection and scope parallels overt wh-movement in a significant and fairly well-defined class of cases.
  4. Subject-object asymmetry for A-bar: High (preverbal ) subjects are more difficult to extract than low (often postverbal) subjects in a class of cases
  5. Freezing: It’s harder to subextract from subjects and objects that have moved; no language will permit movement out of a moved subject or object but not out of a nonmoved one, under otherwise identical conditions.
  6. Specifier bias in Pied-piping: If you can pied-pipe from a complement then you can pied-pipe from a specifier
  7. Adjunct extraction is hard (Robust): If a phrase is an island for argument extraction, then it is also an island for adjunct extraction
  8. Parasitic gaps (Robustness not known): An A-bar chain can license an otherwise illicit gap in an adjunct
  9. Resumptive pronouns: Resumption is by pronouns (not by dedicated resumptive particles)
  10. Resumptive pronoun island alleviation (Robust): Resumptive pronouns tend to alleviate island effects
  11. Local subject condition on resumption: There is a class of resumption which is incompatible with local subject position
  12. Left-dislocation: Many languages allow one or more kinds of left dislocation, with systematic similarities and differences from A bar movement (e.g. lack of case connectivity)
  13. Intervention Effects (Beck Effects): Covert A-bar chains (i.e. in-situ wide-scope-bearing elements) cannot cross (take scope over) scope-bearing interveners