Anaphora and Merge

The ultimate goal of this paper is to show that binding can be captured in terms of Merge and Transfer. It is well-known that the phi-deficiency view of anaphora is taken to be the predominant view in the Minimalist literature and that an anaphor is assumed to be a nominal that lacks one or more phi-features. We examine the phi-deficiency view of anaphora and argue that animate features, phi-features, and R-features are necessary. Korean ku-casin “he-self” and English himself underspecified for R-features are subject/object/indirect object-oriented and are strictly local anaphors. On the other hand, Korean caki “self” underspecified for phi-features is subject/object-oriented and both locally and non-locally bound. Korean caki-casin “self-self” underspecified for both features (phi-features and R-features) are subject-oriented and strictly a local anaphor. Finally, within the Minimalist work, we show that binding can be captured by animate features, phi-features, R-features, Merge, and Transfer. Transfer provides the governing category and semantic computations, by which binding can be captured.

ku-casin "he-self", and caki-casin "self-self" are licensed after the first Transfer, whereas Korean caki "self" is licensed after the first and second Transfer.

The Phi-deficiency View
The phi-deficiency view of anaphora is the predominant view in the Minimalist literature (Sundaresan, 2017). In what follows, we argue that phi-features-based (gender, number, and person) approaches to binding have difficulty accounting for it. To begin with, let us consider the following sentence: (1) *[Tom-uy catongcha] i -ka caki i -uy cipapey GEN car-NOM self-GEN house in front of cwuchatoyeissta.
is parked (*Tom's car i is parked in front of caki i 's house.) (1) clearly indicates that the antecedent of Korean caki must be [+animate]. The reason why (1) is ungrammatical is that the antecedent of caki "self" is the inanimate NP Tom's car. Exactly the same can be said of (2): (2) *[Tom-uy computer] i -ka caki i -uy cha-ey issta.
GEN computer-NOM self-GEN car-in is (Tom's computer is in caki's car.) (2) is ungrammatical due to the fact that the antecedent of caki "self" is not [+animate]. This in turn indicates that approaches which are based on the three categories (gender, number, and person) of phi-features cannot capture the ungrammaticality of (1) and (2) since Korean caki "self" carries the [+animate] feature. Standard theories classify phi-features into three categories (gender, number, and person), but Korean caki "self" requires only two categories: person and animate features. It is significant to note that there is no restriction on Korean caki "self" with respect to gender and number features: (3) a. Tom i -i caki i -uy pang-eyse cako-issta.
TOP self-GEN room-in sleeping-is And-TOP self-GEN room-in sleeping-is ([Tom and Mary] i are sleeping in caki i 's room.) As alluded to in (3b), Korean caki "self" is sensitive to person, but there is no restriction on caki "self" with respect to number and gender, as illustrated in (3c). However, English reflexives are sensitive to these features: (4) a. Tom i /Mary i blamed himself i / herself i . b. They i blamed themselves i .
This clearly indicates that phi-features (gender, number, and person)-based approaches have difficulty accounting for Korean caki "self". As evidenced by (1), (2), and (3), Korean caki "self" requires two categories such as person and animate features, which is not in accordance with standard theories which classify phi-features into three categories (gender, number, and person). Now let us consider the following example: (5) Tom i said that they i+j went on a picnic.
(5) clearly indicates that there is no agreement in the phi-features of Tom and they, especially in the number feature of Tom and they. However, they can refer to Tom and someone else. This clearly indicates that English pronouns are not sensitive to number and that they require two categories such as person and gender. Exactly the same can be said of Korean. Let us consider the following sentence: (6) Mary i -nun ku-tul i+j -uy chinkwu-lul piphanhayssta.
TOP they-GEN friend-ACC criticized (Tom i criticized their i+j friend.) In (6), the Korean pronoun ku-tul "they" does not match its antecedent for phi-features, especially number and gender features. Facts like these apparently convince us that English and Korean anaphora is against the standard categories of phi-features including gender, number, and person. Furthermore, it is possible for plural anaphors or pronouns to be bound by singular quantifiers: (7) a. Everyone i outwitted themselves i .
b. Everyone i outwitted their i adversary.
c. Everyone i assumes John outwitted them i (Kang, 2019).
In (7), anaphors and pronouns do not match their antecedents for phi-features, especially the number feature. More interestingly, it is also possible for plural anaphors or pronouns to be bound by singular quantifiers: (8) a. Someone i outwitted themselves i .
b. Someone i outwitted their i adversary.
c. Someone i assumes John outwitted them i .
Again, English anaphora does not match its antecedent for the number feature, which clearly indicates that the phi-deficiency view of anaphora in the Minimalist work is on the right track.

R-features
The main goal of this section is to reveal that phi-features (gender, number, and person)-based approaches have difficulty accounting for Korean anaphora. We try to argue that anaphors are by nature non-referential, but they can have their own reference (R-features). R-features (U-features in Richards (1997)) are defined as features which the noun carries. An NP which carries an R-feature has its own reference like Tom. Richards (1997)  (9a) and (9b) are ungrammatical due to the fact that there is no antecedent to match for phi-features.
That is to say, in (9a) and (9b), anaphors must be bound to their antecedents, but there are no antecedents which can bind them, which results in the ungrammaticality of (9) and (9b). Then why is (9c) grammatical? We wish to argue that (9c) is grammatical since Korean caki "self" carries an R-feature. Simply put, Korean caki "self" carries its own reference, namely the hearer. Thus, (9c) is grammatical even though there is no antecedent to match for phi-features. That is why an R-feature other than phi-features in (9c) is necessary. Phi-features play no role in (9c) and R-features obtain their role in (9c). There is another environment in which Korean caki "self" favors the hearer as its referent over its antecedent: Clearly, in (10a), the hearer is preferred over the linguistic antecedent as the referent of caki "self" since Korean caki "self" carries an R-feature. The unavailability of the hearer as the referent of Korean caki-casin "self-self" in (10b) is predicted, given the fact that caki-casin "self-self" does not carry an R-feature. Thus, it is reasonable to conclude that Korean caki "self" requires three systems such as animate features, phi-features, and R-features.

Phi-features and R-features
In what follows, we argue that phi-features and R-features are closely related to the binding behavior of Korean reflexives and English reflexives. We assume that Korean ku-casin "he-self" and English himself with underspecified R-features are subject/object/indirect object-oriented and that they are strictly local anaphors. Let us consider the following example: (11) Tom i -i Bill j -eykey ku-casin i/j -uy pang-ul showed (Tom i showed Bill j ku-casin i/j 's room.) In (11), Korean ku-casin "he-self" can be linked to the subject Tom, which in turn suggests that it is subject-oriented. In addition, ku-casin "he-self" can refer to the indirect object Bill. The coindexation between ku-casin "he-self" and the indirect object Bill indicates that Korean ku-casin "he-self" with underspecified R-features is subject/indirect object-oriented. Now let us consider the following sentence: said (Tom i said to Mary j that Bill k criticized ku-casin *i/*j/k .) The coindexation between ku-casin "he-self" and the embedded subject Bill in (12) implies that ku-casin "he-self" can strictly a local anaphor. In (12), ku-casin "he-self" can refer to the embedded subject Bill, but it cannot refer to the matrix subject Tom. By this contrast it becomes clear that Korean ku-casin "he-self" is strictly a local anaphor. Thus, it is reasonable to assume that Korean ku-casin "he-self" underspecified for R-features are subject/indirect object-oriented and strictly a local anaphor.
Now attention is paid to English himself underspecified for R-features: (13) Tom i believes that Bill j told James k about himself *i/j/k .
The coindexation between himself and the subject Bill indicates that English himself is subject-oriented.
In addition, coindexing English himself with the object James is also acceptable. From this it is clear that English himself is also object-oriented. In (13), on the other hand, the himself-binding by the embedded subject Bill implies that English himself is strictly a local anaphor. Thus, it is reasonable to conclude that Korean ku-casin "he-self" and English himself underspecified for R-features are subject/object/indirect object-oriented and strictly local anaphors.
Now we assume that Korean caki "self" with underspecified phi-features is subject/object-oriented and not local. Let us consider the following sentence: (14) John i -i Bill j -ul caki i/j -uy pangey katwuessta.
NOM ACC self-GEN in room kept (John i kept Bill j in caki i/j 's room.) In (14), Korean caki "self" can be associated with the subject John and the object Bill. From this it is clear that caki "self" is subject/object-oriented. Let us observe the following example: (15) Tom i -i [Mary j -ka caki i/j -lul kwachanhayssta]-ko malhayssta.
NOM NOM self-ACC overpraised-COMP said (Tom i said that Mary j overpraised caki i/j .) The coindexation between caki "self" and matrix subject Tom indicates that caki "self" can be non-locally bound, whereas the coindexation between caki "self" and the embedded subject Mary implies that caki "self" can be locally bound. We thus conclude that Korean caki "self" underspecified for phi-features is subject/object-oriented and both locally and non-locally bound.
Finally, let us turn our attention to caki-casin "self-self" underspecified for both features (phi-features and R-features). Caki-casin "self-self" underspecified for both features is subject-oriented and strictly a In (16), caki-casin "self-self" cannot refer to the matrix subject John and the object Mary. It can only refer to the embedded subject Bill. This indicates that caki-casin "self-self" underspecified for both features (phi-features and R-features) is a local and subject-oriented anaphor. We thus conclude that Korean ku-casin "he-self" and English himself underspecified for R-features are subject/object/indirect object-oriented and strictly local anaphors, that Korean caki "self" underspecified for phi-features are subject/object-oriented and not local, and that caki-casin "self-self" underspecified for both features (phi-features and R-features) is a local and subject-oriented anaphor.
(17) Transfer: Transfer the minimal structure containing the finite complementizer to phonological and semantic computations. Once a structure has been transferred, it is no longer accessible to further syntactic computation. Chomsky (2019b) argues that "while no firm conclusions can be drawn, it is plausible that Merge and Transfer are rooted in principles of efficient computation". In this section, following Chomsky (2019a/b), we define a syntactic unit as follows: In this section, we provide the following theorem of anaphors: (19) a. Korean ku-casin "he-self" and English himself are licensed after the first Transfer. b. Korean caki "self" is licensed after the first and second Transfer.
c. Korean caki-casin "self-self" is licensed after the first Transfer.
Korean ku-casin "he-self" and English himself must match their antecedents for phi-features since they are non-referential. Note that the anaphors ku-casin "he-self" and himself underspecified for R-features are subject/object/indirect object-oriented and local. Korean ku-casin "he-self" and English himself's phi-matching takes place after the first Transfer since they are strictly local anaphors. Transfer provides the governing category which is the minimal structure containing the finite complementizer that and it provides semantic computations, by which binding is captured. On the other hand, Korean caki "self" underspecified for phi-features is subject/object-oriented and both locally and non-locally bound. Thus, phi-matching can take place after the first and second Transfer since Korean caki "self" is both locally and non-locally bound. When it comes to Korean caki-casin "self-self" underspecified for both features (phi-features and R-features), it is strictly a local anaphor and subject-oriented. In this case, phi-matching takes place after the first Transfer since Korean caki-casin "self-self" is strictly a local anaphor. Note that z must be licensed by y since z underspecified for R-features is strictly a local anaphor. This is possible since z and y are inside of the minimal structure containing the finite complementizer that.
After the first Transfer, z cannot be matched by x for phi-features since x is a female and outside of the minimal category containing the finite complementizer that. Only the first Transfer provides the governing category of Korean ku-casin "he-self". Exactly the same can be said of w. Z cannot be matched by w for phi-features since w is outside of the minimal structure containing the complementizer that which the first Transfer involves. Thus, z can only be licensed by y after the first Transfer. In (23), the minimal structure containing z and y is the governing category of z and z is locally bound to y in the minimal structure. The minimal structure is what the first Transfer provides.
NOM ACC self-GEN room in kept (John i kept Tom j in caki i/j 's room.) Note that Korean caki "self" underspecified for phi-features is subject/object-oriented and that it can be both locally and non-locally bound. In (25), z is bound to the subject x and the object y since Korean caki "self" is subject/object-oriented. In this case, phi-matching takes place after the first Transfer. However, the following sentence shows that phi-matching takes place after the first and second Transfer. Let us consider the following sentence: As observer earlier, Korean caki "self" underspecified for phi-features is subject/object-oriented and can be both locally and non-locally bound. Thus, the Korean anaphor caki "self" (z) is matched by the embedded subject y as well as the matrix subject w. In the minimal structure containing the complementizer that, y binds z, which in turn indicates that z is bound to y after the first Transfer. On the other hand, z is also matched by w which is outside of the minimal structure containing that since it can be non-locally bound. This is possible after the second Transfer. It is worth pointing out that phi-matching between x and z cannot take place since Korean caki "self" underspecified for phi-features is subject/object-oriented. We thus conclude that Korean caki "self" matches its antecedent for person and animate features after the first and second Transfer and that the antecedent can be inside and outside of the minimal structure containing the complementizer that. Now let us consider the derivation of (28): As indicated in semantic computations, there is no antecedent to match for phi-features since xs are not y. Yet, caki "self" can refer to a discourse participant, namely the hearer. It is worth noting that cases like (31) is acceptable due to the fact that Korean caki "self" carries an R-feature. In (31), Korean caki "self" is self-licensed after the first Transfer. Here, Transfer provides the governing category of caki "self" and its semantic computations. However, the minimal structure containing caki "self" is not necessary since there is no antecedent to match for phi-features. We thus conclude that caki "self" can refer to a discourse participant since it carries an R-feature. As observed earlier, Korean caki-casin "self-self" underspecified for both features (phi-features and R-features) is subject-oriented and strictly a local anaphor. This in turn indicates that caki-casin "self-self" must match its antecedent for person and animate features after the first Transfer since caki-casin "self-self" is strictly a local anaphor. In (33), the first Transfer provides the minimal structure containing the complmentizer that, which is the governing category of caki-casin "self-self".
Caki-casin "self-self" is bound to Tom in the minimal structure after the first Transfer. It must be noted, however, that caki-casin "self-self" cannot be bound to Mary in the minimal structure after the first Transfer. This is due to the fact that caki-casin "self-self" underspecified for both features (phi-features and R-features) is subject-oriented. It is worth pointing out that the second Transfer provides the whole sentence where caki-casin "self-self" cannot be bound to the matrix subject John. This is because caki-casin "self-self" is strictly a local anaphor. This in turn suggests that only the first Transfer provides the minimal structure which is the governing category of caki-casin "self-self". We thus conclude that caki-casin "self-self" is licensed by a subject in the minimal structure containing the complementizer that after the first Transfer. The whole sentence which the second Transfer provides is not the governing category of caki-casin "self-self" since caki-casin "self-self" is strictly a local anaphor.
Next, it must be noted that Korean anaphors and English anaphors must be c-commanded by their antecedents: (34) *Tom i -uy tongsayng-i caki i /caki-casin i /ku-casin i -ul GEN brother-NOM self/self-self/he-self-ACC pinanhayssta.
blamed (Tom i 's brother blamed self i /self-self i /he-self i .) The ungrammaticality of (34) is due to the fact that Korean anaphors are not c-commanded by the antecedent Tom in this sentence. Exactly the same can be said about the English reflexive himself: