Questions & explanations
1. Give an example of a sentence with gap degree 1. Is it projective or non-projective?
Consider the sentence 'The dog that I saw barked'. The dependency from 'saw' to 'dog' crosses over 'that' and 'I', but 'that' depends on 'dog' and 'I' depends on 'saw', so all words between are descendants of 'dog' or 'saw', so gap degree 0. For a non-projective example, in German '... weil er das Buch gelesen hat' (because he the book read has), 'hat' depends on 'gelesen', but 'er' and 'das Buch' are between. 'er' depends on 'gelesen', but 'das Buch' depends on 'gelesen' as well, so gap degree 1? Actually, all are descendants, so projective. A true non-projective: '... dass er das Buch gelesen hat' where 'hat' depends on 'gelesen' but 'er' and 'das Buch' are between, and 'das Buch' depends on 'gelesen', but 'er' depends on 'gelesen' too? Wait, in German subordinate clauses, the auxiliary goes to the end, so 'hat' depends on 'gelesen', and 'er' and 'das Buch' are between, but they are descendants of 'gelesen'? Actually, 'er' is subject of 'gelesen', so it depends on 'gelesen', and 'das Buch' is object, so also depends on 'gelesen'. So all are descendants, so projective? No, because t
2. Why is 'Which book did you read?' good but '*Which book did you read the author of?' bad?
In 'Which book did you read?', the trace after 'read' is governed by the verb, so it is properly governed. In '*Which book did you read the author of?', the trace is the object of the preposition 'of' inside the noun phrase 'the author of which book'. The preposition 'of' governs the trace, so that part is fine. However, the trace is inside a noun phrase that is itself inside a clause. The ECP also requires that the trace be properly governed in its position. But here, the noun phrase 'the author of trace' is an island, and the trace is not properly governed because the noun phrase blocks government from the verb. Actually, the ECP is about proper government of the trace itself. The trace is governed by 'of', so it is properly governed. But the sentence is still bad due to Subjacency. So the ECP alone does not rule it out; Subjacency does. The ECP is about the trace's government, not about bounding. So this question might be misleading. Let me correct: The ECP is not the main reason for this ungrammaticality; it's Subjacency. So I'll adjust: Actually, the ECP is satisfied because the
3. Give an example of a sentence that can be generated by a context-free grammar but not by Gaifman's dependency grammar (if possible).
Gaifman's grammar is equivalent to context-free grammar, so it can generate all context-free languages. However, non-projective dependencies (like in 'What did you see?') are not allowed in Gaifman's projective trees. But the sentence itself is still context-free. So, any context-free sentence can be represented by a projective dependency tree? Actually, some sentences require non-projective trees, but they can still be generated by a context-free grammar. Gaifman's grammar cannot produce non-projective trees, but the same sentence can be represented projectively by reordering? This is tricky. For example, 'Whom did you give the book to?' is non-projective. Gaifman's grammar would need to treat 'to' as a dependent of 'give' with 'whom' as dependent of 'to', which is projective if we allow 'to' to be head. So maybe every context-free sentence has a projective dependency representation? Actually, some languages require non-projective structures. But Gaifman's equivalence holds for projective dependency grammars. So non-projective sentences cannot be generated by Gaifman's grammar. Exam
4. Give an example of a well-nested non-projective structure with gap degree 2.
Consider a sentence with words 1 to 7. Suppose word 1 depends on word 7, word 2 depends on word 6, and word 3 depends on word 5. The gaps: between 1 and 7, words 2-6 are not all descendants of 7? Actually, if 2 depends on 6, and 3 depends on 5, and 4 is something else, then the gaps might be nested. For instance, let's construct: words: A B C D E F G. Dependencies: A depends on G, B depends on F, C depends on E. The gaps: for A-G, the words B,C,D,E,F are between. Among them, B and C are not descendants of G? Actually, B depends on F, which depends on G? Not necessarily. If F depends on G, then B is descendant. To have gap degree 2, we need two separate gaps. For example, if A depends on G, and B depends on C, and D depends on E, with order A B C D E F G, then gaps: between A and G, B and C are a gap (since B and C are not descendants of G), and D and E are another gap. If these gaps are nested (one inside the other), it's well-nested. For instance, if the first gap is inside the second, like A [ B [ C ] D ] E F G? Hard to construct without crossing. A simpler example: In German, '...
5. What is the Double Object Constraint? Give an example of a sentence that violates it.
The Double Object Constraint (DOC) says that in a double object construction (like 'give someone something'), both objects cannot be pronouns of the same person or number? Actually, DOC often refers to a restriction that the two objects cannot both be full noun phrases? Wait, let's define clearly: In some languages, you cannot have two non-pronominal objects in a double object construction. Example violation: 'I gave the man the book' is fine, but in some languages you must use a prepositional phrase. In English, DOC is less strict. A better example: 'I gave him it' is awkward because both are pronouns, but not ungrammatical. Perhaps the constraint is about ordering: the indirect object must be animate. Let's provide a clear example: In English, 'I gave the book the man' is ungrammatical because the indirect object must come first. So the constraint is that the indirect object (recipient) precedes the direct object (theme) when both are NPs.
6. Compare 'What did you think that John bought _?' and '*What did you wonder whether John bought _?' Which is grammatical and why?
The first sentence is grammatical because 'that' does not create an island; you can extract 'what' from the complement clause. The second is ungrammatical because 'whether' creates a wh-island, a weak island. Weak islands block extraction of adjuncts but allow arguments? Actually, 'what' is an object argument, and wh-islands often allow argument extraction. But in this case, 'whether' blocks extraction of 'what'? Wait, 'What did you wonder whether John bought _?' is actually marginal for some speakers. Standardly, wh-islands allow argument extraction, so it should be okay. However, many find it degraded. Better example: 'What did you wonder whether John bought _?' is often considered acceptable. So a clearer contrast: 'How did you think that John fixed the car _?' vs '*How did you wonder whether John fixed the car _?' because 'how' is an adjunct, blocked by wh-island.
7. How does the Case Filter relate to the fact that we say 'I saw him' not '*I saw he'?
In 'I saw him', the pronoun 'him' is in object position and gets accusative Case from the verb 'saw'. If we said '*I saw he', the pronoun 'he' has nominative Case, but the verb assigns accusative Case. The Case Filter requires that the noun phrase get the correct Case assigned by its governor. 'He' does not match the Case assigned, so the sentence is bad. The Case Filter is not just about having Case, but about having the right Case. Actually, the Case Filter says every overt noun phrase must have Case, but it does not specify which Case. However, Case assignment is determined by the position. In object position, the verb assigns accusative, so the noun must have accusative form. So '*I saw he' is bad because 'he' has nominative instead of accusative. The Case Filter works with Case theory to ensure the correct morphological form.
8. Explain how mixed languages differ from creoles. Give an example of a mixed language and one syntactic feature that shows mixture.
Mixed languages arise from intense language contact where two languages blend, often with one providing the grammar and the other the vocabulary, unlike creoles which simplify. An example is Michif, spoken in Canada, which combines Cree verbs and French nouns. Syntactically, Michif uses Cree verb morphology (e.g., subject agreement prefixes) but French noun phrases with French determiners and adjectives. For instance, 'li pchi fish' means 'the little fish', where 'li' is from French 'le', 'pchi' from French 'petit', but 'fish' is from English? Actually, Michif uses French for noun phrases and Cree for verbs. So the syntax is a mix: the verb phrase follows Cree word order (verb-initial), while noun phrases follow French order (determiner-adjective-noun). This shows a clear split.
9. Given the sentence 'Mary gave Susan her book', use dependency grammar to decide who 'her' refers to.
The dependencies: 'gave' is root, 'Mary' is nsubj, 'Susan' is iobj (indirect object), 'book' is dobj, and 'her' is poss (possessive) dependent of 'book'. The possessive 'her' could refer to Mary or Susan. Dependency grammar shows that 'her' is attached to 'book', not directly to a person. To resolve, we often use gender and number: both Mary and Susan are female. But a common rule: the possessive pronoun often refers to the subject (Mary) if there is no other clue. However, in some theories, the indirect object (Susan) is also possible. Dependency paths: 'her' → poss → 'book' → dobj → 'gave' → nsubj 'Mary' and iobj 'Susan'. Both are equally close. So additional knowledge (e.g., world knowledge) is needed. Dependency grammar provides the structure but not the answer alone.
10. In the sentence 'I showed her them', is there a violation? Explain.
Yes, 'I showed her them' violates the Person-Case Constraint. Here, 'her' is a third person indirect object, and 'them' is a third person direct object. The PCC typically restricts first/second person indirect objects with third person direct objects, but some versions also restrict third person indirect with third person direct? Actually, the classic PCC is about first/second indirect with third direct. So 'I showed her them' is fine in many dialects? Wait, let's be accurate: The PCC often applies when indirect object is 1st/2nd and direct is 3rd. 'Her' is 3rd, so no violation. But some languages have a stronger constraint. For simplicity, we focus on the classic: 'I showed you them' (you=2nd, them=3rd) is bad. So 'I showed her them' is acceptable in English.
11. Extend the Principle of Unambiguous Binding to explain why in 'The teacher said that the student hurt himself' the reflexive is clear, but in 'The teacher said that the student hurt herself' it is not.
In 'The teacher said that the student hurt himself', 'himself' clearly refers to 'the student' because 'himself' is masculine and matches 'student' (assuming male). The Principle of Unambiguous Binding is satisfied because there is only one possible referent. In 'The teacher said that the student hurt herself', 'herself' is feminine, but 'student' could be male or female; if 'student' is male, then 'herself' has no referent, and if 'student' is female, it refers to 'student'. But if the teacher is female, 'herself' might also refer to 'teacher', creating two possibilities. The principle says this is bad because the reflexive is not unambiguous. So the sentence is only good if 'student' is clearly female and the teacher is male, otherwise it is confusing.
12. What is differential subject marking (DSM)? Give an example from a language that marks subjects differently based on animacy.
Differential subject marking (DSM) is when the subject of a verb is marked with different cases depending on properties like animacy or definiteness. For example, in Hindi, animate subjects of transitive verbs in the perfective aspect take ergative case, but inanimate subjects take nominative case. So 'The boy (animate) opened the door' is 'Ladke ne darwaaza khola' with 'ne' ergative, while 'The key (inanimate) opened the door' is 'Chaabii ne darwaaza khola' – actually inanimate also takes 'ne' in Hindi, so a better example is from Spanish where animate objects are marked with 'a' but subjects are not. For DSM, consider languages like Kannada where animate subjects may take a different case marker than inanimate ones in certain constructions.