《转换生成语法导论》07
Chapter 7: θ-Theory
★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★
Subcategorisation in terms of syntactic categories such as NP is called categorial selection (c-selection) and selection in terms of semantic categories is called semantic selection (s-selection). S-selection is largely determined by the inherent meaning (sometimes called conceptual structure) of lexical items.
Just as c-selection operates in terms of syntactic categories, s-selection operates in terms of semantic categories called thematic roles or θ-roles.
The verb decide, as in Mary decided that John should leave, also s-selects two participants, the second of which, i.e. the clause (that) John should leave, is called a proposition.
Suppose that a correspondence relation can be established between θ-roles and syntactic categories such that individual syntactic categories serve as the ‘canonical structural realisation’ (CSR) of specific θ-roles. The CSR of agent, for example, will typically be NP, and the CSR of proposition will typically be a clause (CP). It follows from this situation that c-selection does not need to be stated in lexical entries. The c-selectional properties of lexical items are derivable from their s-selectional properties in terms of the relation CSR.
The relationship between the subcategorisation properties of lexical items and their syntactic representations is mediated by the Projection Principle. We will see below that the relationship between the thematic properties of lexical items and their syntactic representations is also mediated by a syntactic principle called the θ-Criterion.
The Principles and Parameters framework outlined in this part of the book is said to have modular structure.
The semantic aspects of lexical items and their s-selectional properties are often described in terms of terminology borrowed from Logic. Verbs are called predicates. The participants involved in the event denoted by a predicate are called arguments.
Verbs which take two arguments are called two-place predicates, and verbs which take one argument are called one-place predicates, where ‘place’ roughly corresponds to ‘argument’. Information relating to arguments of predicates is called the argument structure (of predicates).
Lexical items are said to assign θ-roles to the arguments they select, subject to certain structurally based conditions to be discussed below.
3a. hit: <1, 2> (argument structure)
<Agent, Patient> (thematic structure)
3b. smile: <1> (argument structure)
<Agent> (thematic structure)
Arguments are represented in terms of Arabic numbers and θ-roles in terms of their individual names. The argument and thematic structures represent the selectional properties of lexical items included in their lexical entries along with other types of lexically determined information.
The patient role is generally understood to imply a change in state, and the theme role to imply a change in location or position. However, it is not clear sometimes whether the θ-role involved should be characterized as patient or theme.
Here, we will drop the term ‘patient’ and subsume the situation of change in state under the definition of ‘theme’.
Some expressions seem to share properties with both arguments and non-arguments. This is the case of the so-called ‘weather-it’, illustrated in (12a&b):
12b. It sometimes rains after [PRO snowing]
13b. It is difficult [PRO to predict their next move]
The weather-it in (12) is similar to the pure expletive it in (13) in that it is apparently a non-referring expression. However, the weather-it differs in that it can function as a controller of PRO, as shown in (12).
Expressions with this type of ambivalent nature are sometimes called quasi-arguments.
Weather-it, for example, can be said to be assigned a ‘weather’ θ-role by the weather verb of which it is a special arguments.
The term ‘quasi-argument’ is sometimes also used to refer to certain phrases in idiomatic expressions such as (14a&b):
14a. John took advantage of Bill.
14b. John kicked the bucket.
The verb kick can be said to assign a special θ-role to the bucket responsible for the idiomatic meaning.
On the other hand, given that the idiomatic meaning is the result of combining kick and the expression the bucket together, it could be argued further that the two items are entered together in the lexicon as a single complex item with the meaning ‘die’.
Wh-phrases Wh-phrases are non-referring expressions. Unlike names such as Mary, they do not pick out a specific individual or entity in a given world (or domain of discourse).
17a. Which problem did Mary solve?
The nature of wh-phrases as non-referring expressions, and therefore non-arguments, raises an interesting problem relating to the argument structure of the verb solve in (17).
Traces of wh-phrases (or wh-traces) are said to have the status of logical variables. Logical variables can generally be assigned a value, i.e. they can be translated as a referential expression in a given domain of discourse.
Using the terminology of Logic, the wh-question (17a) can be translated as in (18), where the expression ‘for which problem’ corresponds to the wh-phrase which problem and the variable x corresponds to the wh-trace. The expression ‘for which problem’ is said to be the operator binding the variable:
18. for which problem x [Mary solved x]
If the answer to the question is ‘Mary solved the Maths problem’, the variable in the object position of the verb is assigned the value MATHS (PROBLEM). Wh-traces are therefore (potential) arguments by virtue of being variables.
19a. Who solved which problem?
The predicate solve has only one argument at SS, namely the wh-trace of who in the subject position. The requirement must therefore hold of a subsequent level of representation. Whatever the exact nature of this level of representation, something must happen that would result in removing the wh-phrase which problem from the object position and leaving a variable trace behind. In other words, the required representation must have the form shown in (20a), where the object wh-phrase is moved to Spec, CP leaving a variable trace behind:
20a. LF: [CP [who]i [which problem]j did [IP ti I [VP solve tj］]
20b. for which person x and which problem y [x solved y]
PF is the ‘visible’ level and LF the ‘invisible’ level. Any reordering (movement) process which takes place prior to or at the level of SS will be ‘visibly’ reflected (at PF). However, any reordering (movement) process which takes place in the mapping from SS onto LF will not be ‘visibly’ reflected.
It follows that movement of the wh-phrase which problem which derived representation (20a) applies in the mapping from SS onto LF.
Movement process which take place prior to or at the level of SS are called overt movement. Movement processes which apply in the mapping onto LF are called covert movements.
21. CP
Spec C’
NP C IP
[+Q]
NP NP
who which problem
The process which moves categories in the mapping from SS onto LF is called Quantifier Raising (QR). Wh-phrases are sometimes called quasi-quantifiers.
Quantifiers (22) and (23) include the expressions every(one) and some(one), called quantifier phrases or just quantifiers:
22a. John suspects everyone.
23a. Mary likes someone.
Quantifiers are non-referring expressions and therefore non-arguments. Everyone in (22) and someone in (23) do not pick out a specific individual but simply define the range of the object argument.
Every(one/thing/body/candidate) is called a universal quantifier and some(one/thing/body/book) an existential quantifier.
If the LF representation of (22a) and (23a) remain as they are in (22b) and (23b), we will have a situation where the two-place predicates suspect and like have only one argument instead of the two arguments specified in their argument structure. The analysis outlined above for wh-phrases in situ carries over to quantifiers, so that quantifiers too undergo QR in the mapping from SS onto LF.
Unlike wh-phrases raised by QR, which are adjoined to Spec, CP, quantifiers raised by QR are adjoined to IP.
Here we will discuss one of the advantages of a QR-based analysis for quantifiers which has to do with sentences with more than one quantifier. This phenomenon is sometimes called multiple quantification.
26a. Everyone suspects someone.
The first interpretation can be paraphrased as ‘everyone has someone whom he/she suspects.’ The second reading can be paraphrased as ‘there is someone whom everyone suspects.’
Using different terminology, in the first reading, everyone is said to have wide/broad scope and someone narrow scope.
As with grammatical relations in general, we expect scope relations to have a structural basis. Scope can be defined as in (27):
27. Scope
The scope of α is the set of nodes that α c-command in the LF representation.
In other words, the fact that such sentences have two possible interpretations follows from the fact that they can have two different LF representations which give rise to two different scope relations between the quantifiers.
In view of the fact that ‘object’ and ‘subject’ are structure-based (functional) terms, it is somewhat inaccurate to use them to refer to arguments/θ-roles in argument/thematic structures. The terms usually used in relation to argument/thematic structures are internal arugment/θ-role and external argument/θ-role.
One way of ensuring that arguments/θ-roles are assigned to appropriate positions is to assume that arguments/θ-roles are specified in lexical entries as to whether they are internal or external. Here, we will adopt the notation in (34) to distinguish between external and internal arguments/θ-roles:
34a. kick: agent <theme>
34b. smile: agent <Ø>
Given the argument/thematic structures in (34a&b), the external argument/θ-role will be mapped onto the subject position, and the internal argument/θ-role onto the complement position.
Lexical heads are said to directly θ-mark their internal arguments and indirectly θ-mark their external argument.
Structurally, internal arguments are sisters to the lexical head which selects them, whereas external arguments are not.
As far as meaning is concerned, there is evidence which suggests that the θ-role of the external argument is determined by a combination of the meanings of the lexical head and its internal argument.
The external θ-role is sometimes said to be assigned via predication, where the term ‘predicate’ refers to the verb and its internal arguments, i.e. VP. The θ-marking of the external argument is indirect in the sense that it is mediated by VP (or V’).
37c. put: agent <theme, location>
38c. give: agent <theme, goal>
For the purposes of the discussion we will assume that both the direct object and the indirect object are internal arguments of the (complex) verb.
The verb θ-marks PP under sisterhood, and the head P of PP transmits this θ-role to its NP object also under sisterhood. The role of the preposition in this respect is restricted to transmitting to its object the θ-role originally assigned by the verb. The implication is that the preposition involved does not have an argument/thematic structure of its own.
40a. John baked fresh bread (for his guests).
41a. John opened the door (with a credit card).
Remaining with the topic of prepositions, a distinction is sometimes drawn between ‘semantically empty’ prepositions and other prepositions.
43a. Mary’s translation of the book
43b. Mary translated the book.
43c. Mary is fond of John.
Positions can now be classified as to whether they are A-positions (read argument positions) or A’-positions (read A-bar positions), and whether they are θ-positions (read theta positions) or θ’-positions (read theta-bar positions).
At this stage, it should be clear that while all A’-positions are also θ’-positions, not all A-positions are also θ-positions.
Generally, A-positions are the positions where an argument can be base-generated, and A’-positions are the positions where an argument is not base-generated. θ-positions are the positions assigned a thematic role, and θ’-positions are the positions not assigned a thematic role.
47a. *John seems (that) Mary has solved the problem
48a. *There solved a problem
49a. *Mary solved there
50. θ-Criterion
i) Each argument must be assigned a θ-role.
ii) Each θ-role must be assigned to an argument.
As such, the θ-Criterion essentially holds of DS representations, i.e. the level where the argument structure of lexical items are first projected. The θ-Criterion also holds of SS and LF by virtue of the Projection Principle. Thematic structures are lexical properties of lexical items and like other lexical properties they fall under the scope of the Projection Principle.
53a. Mary seems to have solved the problem.
54a. *Mary believes to have solved the problem.
55. θ-Criterion
i) Each argument must be assigned one and only one θ-role.
ii) Each θ-role must be assigned to one and only one argument.
The trace and its antecedent are said to form a (movement) chain. The antecedent, i.e. the raised NP in (53), is the head (of the chain), and the trace is the root/tail (of the chain). The crucial implication of (53) is that the θ-Criterion should be viewed as a condition on the assignment of θ-roles to individual arguments as well as argument chains. As a matter of fact, the θ-Criterion can be exclusively viewed as a condition on chains if the notion ‘chain’ is extended to include non-moved categories such as {Mary} and {the problem} in Mary solved the problem.
A chain which consists of one member, i.e. a chain not derived by movement, is called a trivial chain. A chain which consists of more than one member, e.g. {Mary, t} in (53), is called a non-trivial chain.
The idea that the θ-Criterion holds of chains implies a severe restriction on movement transformations such that only those which move an argument (from a θ-position) to a θ’-position are allowed. Movement of an argument (from a θ-position) to another θ-position will result in the derivation of a chain with two θ-roles in violation of the θ-Criterion.
Our next step is to check whether the other movement processes we have identified so far are consistent with this requirement. The discussion of NP-movement in passives is postponed to a later section in this chapter. The movement processes which affect head categories such as I and V are ignored, as they are not directly relevant to the discussion.
The target of NP-movement, i.e. Spec, IP, is an A-position. A chain derived by movement to an A-position is called an A-chain. Topicalisation, Wh-movement and QR differ in that their target is an A’-position. A chain derived by movement to an A’-position is called an A’-chain. As far as the θ-Criterion is concerned, the relevant member of A’-chains is the variable trace as the head is usually a non-argument.
Raising to Object Consider (59a&b):
59a. John believes Bill to be a genius.
59b. John considers Bill to be a fool.
60a. John believes him/*he to be a genius.
60b. Bill is believed to be a genius.
60c. John (sincerely) believes (*sincerely) Bill to be a genius.
An obvious way of reconciling the fact that the NP in question is the external argument (subject) of the verb of the embedded clause with the fact that it has object-like properties is to assume that it is base-generated in the external argument position (subject position) of the embedded clause and subsequently moved to the object position of the root verb.
This instance of NP-movement is called Subject to Object raising or Raising to Object.
At the moment, note that the root verbs which occur in the constructions discussed belong to a limited class of verbs called believe-type verbs.
65a. Mary gave the book to John.
66a. Mary gave John the book.
(66) is an instance of a phenomenon known as Dative Shift found with many verbs which take two internal arguments.
67a. Mary sent the letter to John.
67b. Mary sent John the letter.
68a. Mary donated the money to the charity.
68b. *Mary donated the charity the money.
69a. Mary transmitted the message to John.
69b. *Mary transmitted John the message.
Transformations are usually not sensitive to individual lexical items, given that they operate on (classes of) categories. In view of this, it is unlikely that the Dative Shift pattern is derived by a transformation, as the transformation would have to be made sensitive to individual lexical items to exclude sentences such as (68b) and (69b).
Although the by-phrase is somehow associated with the external argument of the passive verb, it is unlikely that it receives the external θ-role of the passive verb in the way external arguments receive theirs from active verbs.
73a. solve: [+V; -N] (=V)
agent <theme> →
73b. solved: [+V; -N] (=V)
Ø <theme>
The derived passive form differs from the base active form in that its argument/thematic structure lacks an external argument/θ-role.
74a. The island was uninhabited.
74b. The performance was uninterrupted.
The complex forms uninhabited and uninterrupted in (74a&b) are said to be instances of adjectival passives.
The externalization of the internal argument of adjectival passives takes place in the lexicon. The internal argument is mapped directly onto the subject position of the sentence, as shown in (80b&c).
81a. The ball was kicked (by Mary).
81b. The room was unoccupied (*by Mary).
82a. The ball was (intentionally) kicked (by Mary).
82b. The room was (*intentionally) unoccupied.
83b. the ball was kicked [PRO to make a point]
84b. the room was unoccupied (*[PRO to make a point])
The external argument of verbal passives is said to be merely ‘suppressed’ or ‘suspended’, meaning not mapped onto Spec, IP rather than completely eliminated.
85a. John broke the vase.
85b. The vase broke.
In (85b) the verb resembles verbal and adjectival passives, although, unlike verbal and adjectival passives, it does not seem to display a morpheme which could be related to this property. The verb is said to be transitive in (85a) and unaccusative in (85b) for reasons that will be clear in Chapter 8.
Two major questions arise in relation to transitive/unaccusative pairs. First, what is the status of the external argument in the unaccusative member of the pair (85b)? Is it eliminated as in adjectival passives or simply suspended as in verbal passives? Secondly, which of the two members of the pair is derived and which is basic?
86a. The vase broke (*by John).
86b. The vase broke (*intentionally).
86c. The vase broke (*to prove a point).
Unaccusative verbs are incompatible with a by-phrase (86a), with an agent-oriented adverb (86b) and with Control of the PRO subject of a purpose clause (86c).
The analysis outlined in (89) implies that the semantic structure of the transitive break is more complex than it appears, so that example (85a) above means something like ‘John caused the vase to be broken’. This reading is sometimes called the causative reading, and the lexical process shown in (89) is called causativisation.
90a. Greek translates easily.
90b. Bureaucrats bribe easily.
This implies that the derivation of nouns from verbs affects the argument structure of the base verb such that the arguments somehow become optional.
The NP in (99a) is sometimes referred to as a nominal passive:
99a. The city’s destruction (by the barbarians)
Lexical categories are said to s-select a certain number of arguments to which they assign a corresponding number of θ-roles.
Non-arguments such as wh-phrases and quantified phrases, called operators, undergo movement in the mapping from SS onto LF called QR.
QR confirms the idea that LF is a syntactic level in much the same way as SS and DS.
The structural representation of thematic structures of lexical items is subject to a syntactic condition called the θ-Criterion.
The Theta Criterion essentially holds at LF, where thematic relations are determined, but by virtue of PP it is expected to hold also at the levels of S-structure and D-structure.
★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★
★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★
Subcategorisation in terms of syntactic categories such as NP is called categorial selection (c-selection) and selection in terms of semantic categories is called semantic selection (s-selection). S-selection is largely determined by the inherent meaning (sometimes called conceptual structure) of lexical items.
Just as c-selection operates in terms of syntactic categories, s-selection operates in terms of semantic categories called thematic roles or θ-roles.
The verb decide, as in Mary decided that John should leave, also s-selects two participants, the second of which, i.e. the clause (that) John should leave, is called a proposition.
Suppose that a correspondence relation can be established between θ-roles and syntactic categories such that individual syntactic categories serve as the ‘canonical structural realisation’ (CSR) of specific θ-roles. The CSR of agent, for example, will typically be NP, and the CSR of proposition will typically be a clause (CP). It follows from this situation that c-selection does not need to be stated in lexical entries. The c-selectional properties of lexical items are derivable from their s-selectional properties in terms of the relation CSR.
The relationship between the subcategorisation properties of lexical items and their syntactic representations is mediated by the Projection Principle. We will see below that the relationship between the thematic properties of lexical items and their syntactic representations is also mediated by a syntactic principle called the θ-Criterion.
The Principles and Parameters framework outlined in this part of the book is said to have modular structure.
The semantic aspects of lexical items and their s-selectional properties are often described in terms of terminology borrowed from Logic. Verbs are called predicates. The participants involved in the event denoted by a predicate are called arguments.
Verbs which take two arguments are called two-place predicates, and verbs which take one argument are called one-place predicates, where ‘place’ roughly corresponds to ‘argument’. Information relating to arguments of predicates is called the argument structure (of predicates).
Lexical items are said to assign θ-roles to the arguments they select, subject to certain structurally based conditions to be discussed below.
3a. hit: <1, 2> (argument structure)
<Agent, Patient> (thematic structure)
3b. smile: <1> (argument structure)
<Agent> (thematic structure)
Arguments are represented in terms of Arabic numbers and θ-roles in terms of their individual names. The argument and thematic structures represent the selectional properties of lexical items included in their lexical entries along with other types of lexically determined information.
The patient role is generally understood to imply a change in state, and the theme role to imply a change in location or position. However, it is not clear sometimes whether the θ-role involved should be characterized as patient or theme.
Here, we will drop the term ‘patient’ and subsume the situation of change in state under the definition of ‘theme’.
Some expressions seem to share properties with both arguments and non-arguments. This is the case of the so-called ‘weather-it’, illustrated in (12a&b):
12b. It sometimes rains after [PRO snowing]
13b. It is difficult [PRO to predict their next move]
The weather-it in (12) is similar to the pure expletive it in (13) in that it is apparently a non-referring expression. However, the weather-it differs in that it can function as a controller of PRO, as shown in (12).
Expressions with this type of ambivalent nature are sometimes called quasi-arguments.
Weather-it, for example, can be said to be assigned a ‘weather’ θ-role by the weather verb of which it is a special arguments.
The term ‘quasi-argument’ is sometimes also used to refer to certain phrases in idiomatic expressions such as (14a&b):
14a. John took advantage of Bill.
14b. John kicked the bucket.
The verb kick can be said to assign a special θ-role to the bucket responsible for the idiomatic meaning.
On the other hand, given that the idiomatic meaning is the result of combining kick and the expression the bucket together, it could be argued further that the two items are entered together in the lexicon as a single complex item with the meaning ‘die’.
Wh-phrases Wh-phrases are non-referring expressions. Unlike names such as Mary, they do not pick out a specific individual or entity in a given world (or domain of discourse).
17a. Which problem did Mary solve?
The nature of wh-phrases as non-referring expressions, and therefore non-arguments, raises an interesting problem relating to the argument structure of the verb solve in (17).
Traces of wh-phrases (or wh-traces) are said to have the status of logical variables. Logical variables can generally be assigned a value, i.e. they can be translated as a referential expression in a given domain of discourse.
Using the terminology of Logic, the wh-question (17a) can be translated as in (18), where the expression ‘for which problem’ corresponds to the wh-phrase which problem and the variable x corresponds to the wh-trace. The expression ‘for which problem’ is said to be the operator binding the variable:
18. for which problem x [Mary solved x]
If the answer to the question is ‘Mary solved the Maths problem’, the variable in the object position of the verb is assigned the value MATHS (PROBLEM). Wh-traces are therefore (potential) arguments by virtue of being variables.
19a. Who solved which problem?
The predicate solve has only one argument at SS, namely the wh-trace of who in the subject position. The requirement must therefore hold of a subsequent level of representation. Whatever the exact nature of this level of representation, something must happen that would result in removing the wh-phrase which problem from the object position and leaving a variable trace behind. In other words, the required representation must have the form shown in (20a), where the object wh-phrase is moved to Spec, CP leaving a variable trace behind:
20a. LF: [CP [who]i [which problem]j did [IP ti I [VP solve tj］]
20b. for which person x and which problem y [x solved y]
PF is the ‘visible’ level and LF the ‘invisible’ level. Any reordering (movement) process which takes place prior to or at the level of SS will be ‘visibly’ reflected (at PF). However, any reordering (movement) process which takes place in the mapping from SS onto LF will not be ‘visibly’ reflected.
It follows that movement of the wh-phrase which problem which derived representation (20a) applies in the mapping from SS onto LF.
Movement process which take place prior to or at the level of SS are called overt movement. Movement processes which apply in the mapping onto LF are called covert movements.
21. CP
Spec C’
NP C IP
[+Q]
NP NP
who which problem
The process which moves categories in the mapping from SS onto LF is called Quantifier Raising (QR). Wh-phrases are sometimes called quasi-quantifiers.
Quantifiers (22) and (23) include the expressions every(one) and some(one), called quantifier phrases or just quantifiers:
22a. John suspects everyone.
23a. Mary likes someone.
Quantifiers are non-referring expressions and therefore non-arguments. Everyone in (22) and someone in (23) do not pick out a specific individual but simply define the range of the object argument.
Every(one/thing/body/candidate) is called a universal quantifier and some(one/thing/body/book) an existential quantifier.
If the LF representation of (22a) and (23a) remain as they are in (22b) and (23b), we will have a situation where the two-place predicates suspect and like have only one argument instead of the two arguments specified in their argument structure. The analysis outlined above for wh-phrases in situ carries over to quantifiers, so that quantifiers too undergo QR in the mapping from SS onto LF.
Unlike wh-phrases raised by QR, which are adjoined to Spec, CP, quantifiers raised by QR are adjoined to IP.
Here we will discuss one of the advantages of a QR-based analysis for quantifiers which has to do with sentences with more than one quantifier. This phenomenon is sometimes called multiple quantification.
26a. Everyone suspects someone.
The first interpretation can be paraphrased as ‘everyone has someone whom he/she suspects.’ The second reading can be paraphrased as ‘there is someone whom everyone suspects.’
Using different terminology, in the first reading, everyone is said to have wide/broad scope and someone narrow scope.
As with grammatical relations in general, we expect scope relations to have a structural basis. Scope can be defined as in (27):
27. Scope
The scope of α is the set of nodes that α c-command in the LF representation.
In other words, the fact that such sentences have two possible interpretations follows from the fact that they can have two different LF representations which give rise to two different scope relations between the quantifiers.
In view of the fact that ‘object’ and ‘subject’ are structure-based (functional) terms, it is somewhat inaccurate to use them to refer to arguments/θ-roles in argument/thematic structures. The terms usually used in relation to argument/thematic structures are internal arugment/θ-role and external argument/θ-role.
One way of ensuring that arguments/θ-roles are assigned to appropriate positions is to assume that arguments/θ-roles are specified in lexical entries as to whether they are internal or external. Here, we will adopt the notation in (34) to distinguish between external and internal arguments/θ-roles:
34a. kick: agent <theme>
34b. smile: agent <Ø>
Given the argument/thematic structures in (34a&b), the external argument/θ-role will be mapped onto the subject position, and the internal argument/θ-role onto the complement position.
Lexical heads are said to directly θ-mark their internal arguments and indirectly θ-mark their external argument.
Structurally, internal arguments are sisters to the lexical head which selects them, whereas external arguments are not.
As far as meaning is concerned, there is evidence which suggests that the θ-role of the external argument is determined by a combination of the meanings of the lexical head and its internal argument.
The external θ-role is sometimes said to be assigned via predication, where the term ‘predicate’ refers to the verb and its internal arguments, i.e. VP. The θ-marking of the external argument is indirect in the sense that it is mediated by VP (or V’).
37c. put: agent <theme, location>
38c. give: agent <theme, goal>
For the purposes of the discussion we will assume that both the direct object and the indirect object are internal arguments of the (complex) verb.
The verb θ-marks PP under sisterhood, and the head P of PP transmits this θ-role to its NP object also under sisterhood. The role of the preposition in this respect is restricted to transmitting to its object the θ-role originally assigned by the verb. The implication is that the preposition involved does not have an argument/thematic structure of its own.
40a. John baked fresh bread (for his guests).
41a. John opened the door (with a credit card).
Remaining with the topic of prepositions, a distinction is sometimes drawn between ‘semantically empty’ prepositions and other prepositions.
43a. Mary’s translation of the book
43b. Mary translated the book.
43c. Mary is fond of John.
Positions can now be classified as to whether they are A-positions (read argument positions) or A’-positions (read A-bar positions), and whether they are θ-positions (read theta positions) or θ’-positions (read theta-bar positions).
At this stage, it should be clear that while all A’-positions are also θ’-positions, not all A-positions are also θ-positions.
Generally, A-positions are the positions where an argument can be base-generated, and A’-positions are the positions where an argument is not base-generated. θ-positions are the positions assigned a thematic role, and θ’-positions are the positions not assigned a thematic role.
47a. *John seems (that) Mary has solved the problem
48a. *There solved a problem
49a. *Mary solved there
50. θ-Criterion
i) Each argument must be assigned a θ-role.
ii) Each θ-role must be assigned to an argument.
As such, the θ-Criterion essentially holds of DS representations, i.e. the level where the argument structure of lexical items are first projected. The θ-Criterion also holds of SS and LF by virtue of the Projection Principle. Thematic structures are lexical properties of lexical items and like other lexical properties they fall under the scope of the Projection Principle.
53a. Mary seems to have solved the problem.
54a. *Mary believes to have solved the problem.
55. θ-Criterion
i) Each argument must be assigned one and only one θ-role.
ii) Each θ-role must be assigned to one and only one argument.
The trace and its antecedent are said to form a (movement) chain. The antecedent, i.e. the raised NP in (53), is the head (of the chain), and the trace is the root/tail (of the chain). The crucial implication of (53) is that the θ-Criterion should be viewed as a condition on the assignment of θ-roles to individual arguments as well as argument chains. As a matter of fact, the θ-Criterion can be exclusively viewed as a condition on chains if the notion ‘chain’ is extended to include non-moved categories such as {Mary} and {the problem} in Mary solved the problem.
A chain which consists of one member, i.e. a chain not derived by movement, is called a trivial chain. A chain which consists of more than one member, e.g. {Mary, t} in (53), is called a non-trivial chain.
The idea that the θ-Criterion holds of chains implies a severe restriction on movement transformations such that only those which move an argument (from a θ-position) to a θ’-position are allowed. Movement of an argument (from a θ-position) to another θ-position will result in the derivation of a chain with two θ-roles in violation of the θ-Criterion.
Our next step is to check whether the other movement processes we have identified so far are consistent with this requirement. The discussion of NP-movement in passives is postponed to a later section in this chapter. The movement processes which affect head categories such as I and V are ignored, as they are not directly relevant to the discussion.
The target of NP-movement, i.e. Spec, IP, is an A-position. A chain derived by movement to an A-position is called an A-chain. Topicalisation, Wh-movement and QR differ in that their target is an A’-position. A chain derived by movement to an A’-position is called an A’-chain. As far as the θ-Criterion is concerned, the relevant member of A’-chains is the variable trace as the head is usually a non-argument.
Raising to Object Consider (59a&b):
59a. John believes Bill to be a genius.
59b. John considers Bill to be a fool.
60a. John believes him/*he to be a genius.
60b. Bill is believed to be a genius.
60c. John (sincerely) believes (*sincerely) Bill to be a genius.
An obvious way of reconciling the fact that the NP in question is the external argument (subject) of the verb of the embedded clause with the fact that it has object-like properties is to assume that it is base-generated in the external argument position (subject position) of the embedded clause and subsequently moved to the object position of the root verb.
This instance of NP-movement is called Subject to Object raising or Raising to Object.
At the moment, note that the root verbs which occur in the constructions discussed belong to a limited class of verbs called believe-type verbs.
65a. Mary gave the book to John.
66a. Mary gave John the book.
(66) is an instance of a phenomenon known as Dative Shift found with many verbs which take two internal arguments.
67a. Mary sent the letter to John.
67b. Mary sent John the letter.
68a. Mary donated the money to the charity.
68b. *Mary donated the charity the money.
69a. Mary transmitted the message to John.
69b. *Mary transmitted John the message.
Transformations are usually not sensitive to individual lexical items, given that they operate on (classes of) categories. In view of this, it is unlikely that the Dative Shift pattern is derived by a transformation, as the transformation would have to be made sensitive to individual lexical items to exclude sentences such as (68b) and (69b).
Although the by-phrase is somehow associated with the external argument of the passive verb, it is unlikely that it receives the external θ-role of the passive verb in the way external arguments receive theirs from active verbs.
73a. solve: [+V; -N] (=V)
agent <theme> →
73b. solved: [+V; -N] (=V)
Ø <theme>
The derived passive form differs from the base active form in that its argument/thematic structure lacks an external argument/θ-role.
74a. The island was uninhabited.
74b. The performance was uninterrupted.
The complex forms uninhabited and uninterrupted in (74a&b) are said to be instances of adjectival passives.
The externalization of the internal argument of adjectival passives takes place in the lexicon. The internal argument is mapped directly onto the subject position of the sentence, as shown in (80b&c).
81a. The ball was kicked (by Mary).
81b. The room was unoccupied (*by Mary).
82a. The ball was (intentionally) kicked (by Mary).
82b. The room was (*intentionally) unoccupied.
83b. the ball was kicked [PRO to make a point]
84b. the room was unoccupied (*[PRO to make a point])
The external argument of verbal passives is said to be merely ‘suppressed’ or ‘suspended’, meaning not mapped onto Spec, IP rather than completely eliminated.
85a. John broke the vase.
85b. The vase broke.
In (85b) the verb resembles verbal and adjectival passives, although, unlike verbal and adjectival passives, it does not seem to display a morpheme which could be related to this property. The verb is said to be transitive in (85a) and unaccusative in (85b) for reasons that will be clear in Chapter 8.
Two major questions arise in relation to transitive/unaccusative pairs. First, what is the status of the external argument in the unaccusative member of the pair (85b)? Is it eliminated as in adjectival passives or simply suspended as in verbal passives? Secondly, which of the two members of the pair is derived and which is basic?
86a. The vase broke (*by John).
86b. The vase broke (*intentionally).
86c. The vase broke (*to prove a point).
Unaccusative verbs are incompatible with a by-phrase (86a), with an agent-oriented adverb (86b) and with Control of the PRO subject of a purpose clause (86c).
The analysis outlined in (89) implies that the semantic structure of the transitive break is more complex than it appears, so that example (85a) above means something like ‘John caused the vase to be broken’. This reading is sometimes called the causative reading, and the lexical process shown in (89) is called causativisation.
90a. Greek translates easily.
90b. Bureaucrats bribe easily.
This implies that the derivation of nouns from verbs affects the argument structure of the base verb such that the arguments somehow become optional.
The NP in (99a) is sometimes referred to as a nominal passive:
99a. The city’s destruction (by the barbarians)
Lexical categories are said to s-select a certain number of arguments to which they assign a corresponding number of θ-roles.
Non-arguments such as wh-phrases and quantified phrases, called operators, undergo movement in the mapping from SS onto LF called QR.
QR confirms the idea that LF is a syntactic level in much the same way as SS and DS.
The structural representation of thematic structures of lexical items is subject to a syntactic condition called the θ-Criterion.
The Theta Criterion essentially holds at LF, where thematic relations are determined, but by virtue of PP it is expected to hold also at the levels of S-structure and D-structure.
★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★★
还没人转发这篇日记
