CIENCIAS Y LETRAS

Give a full derivation, from numeration to final step, for the sentence in (1). Label each step for the process involved in that step (e.g. merge, agree) and the principle of principles that require or allow this step (e.g. checking of uninterpretable features, hierarchy of projections, etc.).

(1) John will open Parliament on Wednesday.

Tree 1

Numeration: John [N], will [T], open [V, uN], Parliament [N], on [P, uN], Wednesday [N].

1. Merge (�open�, �Parliament�) checking [uN] on �open�.
2. Merge (υ, output of 1) � Hierarchy of Projections (υ � V).
3. Move �open� to υ.
4. Merge (�John�, output of 2 and 3) checking [uN] on υ (on its intermediate projection υ�).
5. Merge (�on�, �Wednesday�) checking [uN] on �to�.
6. Adjoin (output of 5 to output of 4).
7. Merge (�will� [T], output of 6) � Hierarchy of Projections (T � υ).
8. Move (�John�) to specifier of T position, Merge (�John�, output of 7) � Subject Requirement.

Consider the principle in (2), adapted from the ICVE (p. 198 of the Adger textbook). Please use this formulation, not the one in the textbook.

(2) Parallelism Condition on υP-Ellipsis (or just Parallelism)
      A υP can only be elided if there is structurally identical υP elsewhere in the discourse.

This helps account for the fact that an elided υP always has the same interpretation as some other fully pronounced υP. So, as Adger notes, (3) makes sense, but (4) is strange, as it suggests that Julie's truck increased in virulence last year. (# is used when a sentence is grammatical but odd in meaning.)

(3) Lloyd's car fell over a cliff last year... and Julie's truck did [_υP...] too.

(4) #Computer viruses increased in virulence last year... and Julie's truck did [_υP...] too.

Say, for each of the following sentences, what elided material is. Are these results consistent with parallelism?

(5)
	1. I will eat mango. Gillian will [] too.	(5)	1. I will [υPeat mango]. Gillian will [υPeat mango] too.
	Tree 2
	2. Raffi has made pasta. David has [] too.	2. Raffi has [υPmade pasta]. David has [υPmade pasta] too.
	Tree 3
	3. Brita is making scones. Gillian is [] too.	3. Brita is [υPmaking scones]. Gillian is [υPmaking scones] too.
	Tree 4

These results are consistent with Parallelism because the antecedent υPs are identical as the elided ones in the second sentences: in (5).1., for instance, the elided material is the υP [eat mango], which is the same that appears in the first sentence �I will [eat mango]�.

Now fill in the elided υPs in (6) as they are understood (i.e., give how you would pronounce [] if something were pronounced in this position). Then give the surface from the antecendet υP (that is, how the antecedent &upsilon is pronounced). Based only on comparison of these two forms, does it look like parallelism is respected in these cases?.

(6) I ate mango. Gillian did [] too.

1. [] is understood as: 'eat a mango'
2. Antecedent υP: 'ate a mango'

(7) Raffi makes pesto pasta. David does [] too.

1. [] is understood as: 'make pesto pasta'
2. Antecedent υP: 'makes pesto pasta'

Now give full tree structures for the first sentences in (6) and (7) (i.e., the antecedent sentences). Include all structure, agreement, etc. Explain how these structures show that, despite appearances, parallelism is respected here.

(6)

Tree 5

(7)

Tree 6

In the first sentences, as they are complete sentences, tense features are pronounced in υ: T values [uInfl:] on υ; in (6) �eat� + υ [uInfl: past] is pronounced �ate� and in (7) �make� + υ [uInfl: sing] is pronounced �makes�. In the second sentences, the elided material is understood as the first sentences� υPs without the tense features: [υP eat a mango] and [υP make pesto pasta]. This is due to the fact that υ is elided, so it is impossible to pronounce tense features on it, and do-support is required to carry them. Thus, T is pronounced in elliptic sentences on the auxiliary: in (6) do [past] (�did�); in (7) do [sing] (�does�). Although elided υPs in (6) and (7) and their antecedents do not look like identical in surface, their structure is exactly the same �the only difference is where tense is pronounced; Parallelism, therefore, is respected in these sentences as well.

The following examples are degraded for most speakers. Explain why. Make reference to parallelism in your answer.

(8)

1. *Gillian has made pasta. David is [] too.
2. *I am eating mango. Gillian has [] too.

1.

Tree 7

2.

Tree 8

On the one hand, the ungrammatically of (8) could be the consequence of an ill-formed VP-ellipsis construction: the auxiliary, which carries the tense, in the elided sentences must be the same as the one in the fully pronounced ones or, if there is only the main verb, �do� appears. In (8)1., for example, the verb in the elided υP [made pasta] has a valued uninterpretable inflectional feature Perf that Agrees with the Perf feature on the perfect auxiliary �has�; however, in the second sentence �David is [made pasta] too� the auxiliary is the progressive one, but it cannot Agree with the main verb �make� because it has already checked its uninterpretable inflectional feature (so it has already been deleted) and valued it. The same lack of Agreement, in this case between the perfect auxiliary and the verb already valued as progressive, happens in (8)2.: *I am eating a mango. Gillian has [eating a mango] too.

As a result of this interpretation, sentences in (8) respect Parallelism as the elided υP is structurally identical as the one in the antecedent sentence. The problem is, in the VP-ellipsis constructions, the auxiliary does not Agree with the main verb, which is elided, and that is why they are degraded for most speakers.

On the other hand, the ungrammaticality could be due to Parallelism is not respected: in (8)2, for instance, grammatical elided υP [eaten a mango] has not a structurally identical antecedent υP; the only υP in the sentence �[eating a mango]� has a different structure: the main verb has its uninterpretable inflection feature valued as a Prog because of the Agreement with the progressive auxiliary whereas the elided one has it valued as Perf, agreeing with the perfect auxiliary �has�. The problem in this case is that it does not respect Parallelism.

First consider the following from p. 151 of the Adger textbook. (R-Expressions are "full NPs" like John and the woman. Other NPs, like pronouns and anaphors are not R-expressions.)

(9) The R-Expression Generalization (or, Condition C)
      An R-expression cannot be coreferential with an NP that c-commands it.

This helps to explain why he and John cannot corefer in (10). (As usual, subscripts are used to annotate coreference possibilities.)

(10) He_j/*i thinks that John_i won't go to the party.

But, his and John can be coreferent in (11), because his does not c-command John, as we see in the tree below (I have left out υP projections here.)

(11) His_j/*i mother thinks that John_i will go to the party.

Tree 9

Now consider (12). This example is well-formed for most speakers under a variety of meanings, including one with coreference as shown here.

(12) Mary₂ admires John₁, but he₁ thinks Sally₃ doesn't [].

Give two possible structures for the elided material in []. Include subscripts showing coreference. (Non-native speakers should ask me or another native speaker for these judgements.)

The elided material in (12), as it is understood (always referring to John₁), could be:

(12)

1. Mary₂ [υP admires John₁], but he₁ thinks Sally₃ doesn�t [υP admire John₁].
2. Mary₂ [υP admires John₁], but he₁ thinks Sally₃ doesn�t [υP admire him₁].

Now consider what Condition C and Parallelism predict about these structures. Do these proposed grammatical principles make the correct predictions in this case? Explain your answer.

Parallelism predicts (12)1. because the elided υP has the same structure as the antecedent υP �the only difference, as it has been analysed above, is tense is carried by auxiliary �do��, so we have the sentence:

(13) He₁ thinks Sally doesn't admire John_*1.

Tree 10

According to Condition C, an R-expression �like �John�� cannot be coreferential with a NP that c-commands it. As �he� is c-commanding �John� in (13), this R-expression cannot refer to it and that is why (13) is ungrammatical. However, if Condition C is taken into account, is the elided material in (12)2., which gives the grammatical sentence (14):

(14) He₁ thinks Sally doesn�t admire him₁.

In this one, the pronoun can be bound by �he� because, although it is c-commanding it, they are not in the same VP (as the Pronoun Generalisation (revised) in the last Problem Set established). Nevertheless, the υP [admires John] and the elided υP [admire him] are not structurally identical, so, (12)2. contradicts Parallelism.

Parallelism and Condition C in (12), therefore, contradict each other and the predictions they make about the elided material are only correct if the other principle is not taken into account.

• Adger, D., (2004) Core Syntax. A Minimalist Approach, New York: Oxford University Press.

Publicado el 04/09/2008

Por Melania S�nchez Masi�