Quantifiers, Binding, and Agreement

The main goal of this paper is to argue that Korean pronouns must be phi-feature-compatible with their antecedents, whereas Korean reflexives are not. It is worth pointing out that Korean pronouns are sensitive to the number feature, whereas Korean anaphors are not. A major point to note is that every-type QPs have a Q feature that is plural in its number, whereas which-type QPs have an optional Q-feature that is singular or plural in number. A further point to note is that Korean pronouns are sensitive to phi-features, which is in accordance with Safir’s (2014) hypothesis that “D-bound anaphora must be feature-compatible with its antecedent”. With respect to Korean pronouns, it is worth noting that Safir’s hypothesis does not work for Korean pronouns since they induce a bound variable reading through the phi-feature agreement. Finally, it is significant to note that Korean anaphors are not feature-compatible with antecedents and that they yield a bound variable reading regardless of their phi-features.


Methodology
In this paper, we assume, along with Huang and Tang (1991), Richards (1995Richards ( , 1997, Chomsky (2000Chomsky ( , 2001, Safir (2014), and Sundaresan (2017) that anaphora carries phi-features involving gender, person and number. Anaphors and pronouns have phi-features involving gender, person, and number, but they lack R-feature (referential feature). Thus, in order for anaphora to meet a full interpretation, it has to share phi-features with their antecedents. In this paper, we try to show that the so-called agreement-based mechanism has advantages over Condition A and Condition B, thus eliminating Condition A and Condition B in the Binding Theory. Finally, we aim to examine Safir's (2014) hypothesis that "D-bound anaphora must be feature-compatible with their antecedents" (Safir, 2014, p. 92). According to Safir (2014), "all bound anaphora is achieved by a single, universally available, D-bound anaphora" (Safir, 2014, p. 92): (1) Properties of D-bound anaphora a. "Always feature-compatible: D-bound anaphora must be feature-compatible with its antecedent (informally, this property may be termed antecedent agreement." b. "Always a variable: D-bound anaphora is interpreted as a bound variable regardless of its phi-features". (Safir, 2014, p. 92) In this paper, we will examine whether or not Safir's (2014) hypotheses work for Korean binding.

Q Feature of Every-type QPs
In this section, we wish to argue that every-type QPs have a Q feature that is plural in its number. Let us consider the following sentences: www.scholink.org/ojs/index.php/selt Studies in English Language Teaching Vol. 9, No. 3, 2021 35 Published by SCHOLINK INC.
everyone they-GEN parents-ACC love (Everyone loves his parents.) The reason why we assume that every-type QP has a Q feature that is plural in its number is that (2a) is ungrammatical, but (2b) is grammatical. The singular pronoun ku 'he' is not interpreted as a bound variable, whereas ku-tul 'they' is construed as a bound variable. We attribute the fact that ku 'he' is not construed as a bound variable to the number disagreement between the every-type QP and the pronoun ku 'he'. Then (2a) and (2b) will have the following trees, respectively: (3) *TP loves their [+pl] As indicated in (3), when the quantifier nwukwuna 'everyone' is associated with ku 'he', the pronoun ku 'he' is not interpreted as a bound variable. We attribute the ungrammaticality of (3) to the fact that the quantifier nwukwuna 'everyone' is plural in number, but ku 'he' is singular. Simply put, (3) is ungrammatical due to the number feature disagreement between the QP and the dependent term ku 'he'.
On the other hand, as illustrated in (4), when ku-tul 'they' is linked to the quantifier nwukwuna 'everyone', it induces a bound variable reading, which is possible because both QP and ku-tul 'they' are plural in number. That is to say, (4) is grammatical due to the number feature agreement between the QP and ku-tul 'they'. This in turn indicates that every-type QPs have a Q feature that is plural in number. The following sentences also lend their support to the claim that every-type QPs have a Q feature that is plural in number: It is important to note that the so-called E-type pronoun is not c-commanded by quantifiers. As illustrated in (5), the singular pronoun ku 'he' that is used as the E-type pronoun cannot be associated with the quantifier motuni 'everyone'. On the other hand, the plural pronoun ku-tul 'they' that is used as the E-type pronoun can be linked to the quantifier motuni 'everyone'. This is possible since every-type QPs have a Q-feature that is plural in number. In (5), the E-type pronoun ku 'he' cannot refer to the QP because of the number feature disagreement between the QP and ku 'he'. More specifically, (5) is ungrammatical since the quantifier motuni 'everyone' is plural, but ku 'he' is singular, which results in the number feature disagreement in (5). In (6), on the other hand, the E-type pronoun ku-tul 'they' can refer to the quantifier motuni 'everyone' due to the number feature agreement between motuni 'everyone' which is plural in number and ku-tul 'they' which is plural in number. This is predictable, given the hypothesis that every-type QPs have a Q feature that is plural in number.
Sentences like (7) and (8) also lend their support to the claim that every-type QPs have a Q-feature that is plural in number: (7) *Motuni i -ka ku i -uy cip-ul ciessta. Again, the ungrammticality of (7) and the grammaticality of (8) are predictable, given the assumption that every-type QPs have a Q feature that is plural in number. In (7), ku 'he' does not induce a bound variable reading due to the number feature disagreement between motuni 'everyone' and ku 'he'. On the other hand, in (8), ku-tul 'they' are construed as a bound variable because of the number feature agreement between motuni 'everyone' and ku-tul 'they'. Again, this in turn implies that every-type QPs have a Q feature that is plural in number.
Now we wish to point out that the number feature agreement between Korean anaphors and their antecedents is not obligatory. As pointed out by Kang (2013), Korean anaphors exhibit the following property: (9) "The Korean reflexives caki 'self' and caki-casin 'self-self' are structurally or semantically singular NPs whose referent consists of a single entity or more than a single entity". (Kang, 2013, p As exemplified in (10a), the singular anaphors caki 'self' and caki-casin 'self-self' induce a bound variable reading because caki 'self' and caki-casin 'self-self' agree in number with the which-type QP.
As alluded to in (10b), however, the number agreement between Korean anaphors and antecedents need not be obligatory. In (10b), the Korean reflexives caki 'self' and caki-casin 'self-self' can be associated with plural antecedents such as Tom and Bill. We thus conclude that Korean pronouns must be phi-feature-compatible with their antecedents, but Korean reflexives are not.

Q Feature which is Optional
In this section, we wish to argue that which-type QPs have an optional Q-feature that is singular or plural in number. Both singular terms and plural terms are a condition of a bound variable induced by which-type quantifiers. Let us consider the following examples: (11) a. Which way is the post office, please?
b. Which houses are most likely to need a humidifier?
As alluded to in (11a), the which-type QP can take a singular NP as its head. On the other hand, as illustrated in (11b), the which-type QP can take a plural NP as its head. This suggests that which-type QPs carry an optional Q-feature that is singular or plural in number. The same applies to Korean: As illustrated in (13a), the singular reflexive caki-casin 'self-self' can be bound to the which-type QP, which leads to a bound variable reading of caki-casin 'self-self'. As indicated in (13b), on the other hand, the plural reflexive caki-casin-tul 'self-self-pl' can be associated with the which-type QP, which results in a bound variable reading of caki-casin-tul 'self-self-pl'. This in turn implies that which-type QPs have an optional Q feature that is singular or plural in number. We thus conclude that unlike every-type QPs, which-type QPs take singular dependent terms or plural dependent terms as a condition of bound variable anaphora and that a bound variable reading of singular dependent terms and plural dependent terms is available due to the fact that which-type QPs have an optional Q feature which is singular or plural in number.

Pronouns that are Feature-compatible with their Antecedents
As observed earlier, which-type QPs have an optional Q-feature that is singular or plural in number.
We wish to show that Korean pronouns must be phi-feature-compatible with their antecedents. Let us observe the following sentences: (14) Ku i -uy emeni-eykey etten salam i -i he-GEN mother-DAT which person-NOM senmwul-ul ponayssnunya?
present-ACC send-PAST (Which persons sent a present to their mother?) The reason why (14) is grammatical is that the Korean pronoun ku 'he' is feature-compatible with the which-type QP. That is to say, the antecedent etten salam 'which person' has phi-features such as gender (masculine or female), person (third person), and number (singular), and the pronoun ku 'he' carries phi-features such as gender (masculine), person (third person), and number (singular). Thus, the Korean pronoun ku 'he' agrees in phi-features with the which-type QP, which leads to the grammaticality of (14). On the other hand, the reason why (15) is grammatical is that the plural pronoun ku-tul 'they' are feature-compatible with the which-type QP. More specifically, the antecedent etten salam-tul 'which persons' has phi-features such as gender (masculine or female), person (third person), and number (plural) and the plural pronoun ku-tul 'they' carries phi-features such as gender (masculine or female), person (third person), and number (plural). Thus, the Korean plural pronoun ku-tul 'they' agrees in phi-features with the which-type QP, which results in the grammaticality of (15).
This in turn suggests that which-type QPs carry an optional Q-feature that is singular or plural in number. Ku 'he' or ku-tul 'they' agrees phi-features with a single entity or plural entities induced by As observed earlier, every-type QPs have a Q feature that is plural in number. The reason why (16) is ungrammatical is that the Korean pronoun ku 'he' does not agree in the number feature with its antecedent. More specifically, the every-type QP has phi-features such as person (third person), number (plural), and gender (masculine or female) and the pronoun ku 'he' carries phi-features such as person (third person), number (singular), and gender (masculine). Thus, the ungrammaticality of (16) is attributed to the number disagreement between the every-type QP and the pronoun ku 'he'. This in turn indicates that Korean pronouns are sensitive to a number feature. On the other hand, the reason why (17) is grammatical is that the pronoun ku-tul 'they' agrees in phi-features with the every-type QP. That is to say, the every-type QP has phi-features such as gender (masculine or female), person (third person), and number (plural) and the pronoun ku-tul 'they' carries phi-features such as gender (masculine or female), person (third person), and number (plural). We take this as indicating that ku-tul 'they' agrees in phi-features with the every-type QP, hence the grammaticality of (17).
A number of approaches to the Binding Theory have made crucial reference to pronouns. An interesting property is exhibited by the Korean singular pronoun ku 'he' and the Japanese singular pronoun kare 'he'. It is a well-known fact that unlike the English pronoun he, the Japanese singular pronoun kare 'he' cannot be interpreted as a bound variable (Hoji, 1991(Hoji, , 1995(Hoji, , 1997, Aoun and  Hornstein (1986,1991), Higginbotham (1992) among others). Thus, the following sentence with the intending reading is ungrammatical, as indicated in (18): (18) *Daremo i -ga kare i -ga tukut-ta omotya-o everyone-NOM he-NOM make-PAST toy-ACC kowasi-ta.
break-PAST (Everyone i broke the toy that he i had made.) (Hoji, 1991, p. 287) The following quotation clearly shows that the Japanese singular pronoun kare 'he' cannot induce a bound variable reading: (19) a. Kare 'he' cannot yield a bound variable reading.
(Saito & Hoji, 1983, Aikawa, 1989 b. Kare 'he' must be A'-free. (Aoun & Hornstein, 1986) c. Kare 'he' must be operator free. (Katada, 1991) The above statements are the same though differing in explanatory flavor. However, these statements cannot account for why kare 'he' cannot be a bound variable. We wish to argue that the reason why (18) is ungrammatical is that daremo 'everyone' is plural in its number and kare 'he' is singular. That is to say, the Japanese singular pronoun kare 'he' does not agree in number with the QP, which results in the ungrammaticality of (18). The following sentences lend their support to our claim:  Aikawa (1989) points out that the Japanese plural pronoun karera 'they' can be a bound variable. Our agreement-based approach predicts why (20) is ungrammatical, whereas why (21) is grammatical. We wish to argue that daremo 'everyone' is plural in its number since it involves a group of people and kare 'he' is singular in number. That is, the pronoun kare 'he' does not agree in number with the every-type QP, which results in the ungrammaticality of (20). On the other hand, the plural pronoun karera 'they' in (21) is phi-feature-compatible with the QP, which leads to the grammaticality of (21).
Note that daremo 'everyone' is plural in number since it involves a group of people and karera 'they' are plural in its number. The following Korean sentences lend their support to our analysis:  (23), the every-type QP has phi-features involving gender (masculine or female), person (third person), and number (plural) and the plural pronoun ku-tul 'they' carries phi-features involving gender (masculine or female), person (third person), and number. The plural pronoun ku-tul 'they' agrees in phi-features with the every-type QP, which results in a bound variable reading of (23). As observed, our agreement-based approach predicts why the Korean singular pronoun ku 'he' and the Japanese singular pronoun kare 'he' cannot be construed as a bound variable, along with every-type QPs. Now let us turn our attention to Safir's (2014) hypothesis that "D-bound anaphora must be feature-compatible with its antecedent" (Safir, 2014, p. 92). To be brief, Safir's (2014) hypothesis is on the right track since Korean pronouns are sensitive to phi-features. More specifically, whether ku 'he' and ku-tul 'they' induce a bound variable reading or not depends on the phi-feature agreement between QPs and them. Hence, they must be feature-compatible with their antecedents. Now let us turn to Safir's (2014) second hypothesis that "D-bound anaphora is interpreted as a bound variable regardless of its phi-features" (Safir, 2014, p. 92). It is significant to note that Korean pronouns can yield a bound variable reading through the phi-feature agreement between QPs and them. Simply put, Safir's (2014) second hypothesis wrongly predicts that Korean pronouns induces a bound variable reading regardless of their phi-features. We thus conclude that Korean pronouns are sensitive to phi-features and thus they are feature-compatible with their antecedents.

Anaphors
In what follows, we try to show that Korean anaphors are not feature-compatible with their antecedents.
Additionally, we argue that Korean anaphors give rise to a bound variable reading regardless of their phi-features. Let us consider the following sentences: (24) Caki i -uy emeni-eykey etten salam i -i self-GEN mother-DAT which person-NOM senmwul-ul onayssnunya?
presant-ACC send-PAST In this example, the Korean anaphor caki-casin 'self-self' carries phi-features involving gender (masculine or female), person (third person), and number (singular), which in turn indicates that caki-casin 'self-self' agrees in phi-features with the which-type QP, hence the grammaticality of (25).
This seems to suggest that Korean anaphors must be feature-compatible with their antecedents.
presant-ACC send-PAST (Which person sent a present to his mother?) In (26), the Korean plural reflexive may induce a bound variable reading, along with the which-type QP.
In this example, the antecedent etten salam 'which person' has phi-features involving gender (masculine or female), person (third person), and number (singular) and the Korean plural reflexive ku-casin-tul 'he-self-pl' has phi-features involving gender (masculine), person (third person), and number (plural). This indicates that ku-casin-tul 'he-self-pl' is not feature-compatible with its antecedents. In (26), ku-casin-tul 'he-self-pl' does not agree in its number feature with the which-type QP. Despite this, ku-casin-tul 'he-self-pl' gives rise to a bound variable reading, which is against Safir's hypothesis that "D-bound anaphora must be feature-compatible with its antecedent" (Safir, 2014,  p. 92). Likewise, the Korean plural reflexive caki-casin-tul 'self-self-pl' induces a bound variable reading, along with the which-type QP, despite the fact that caki-casin-tul 'self-self-pl' does not agree in its number feature with the which-type QP. This in turn suggests that our data are against Safir's (2014) hypothesis that "D-bound anaphora must be feature-compatible with its antecedent" (Safir, 2014, p. 92). Additionally, it is worth noting that Korean plural anaphors yield a bound variable reading regardless of their phi-features, which is in accordance with Safir's (2106) hypothesis. Now let us turn our attention to every-type QPs. Let us observe the following sentences: In (28), the Korean reflexive caki 'self' may not yield a bound variable reading. It must be noted that the hearer may be favored over the antecedent as the reference of caki 'self'. In (29), on the other hand, the Korean reflexive caki-casin 'self-self' refers to the linguistic antecedent. In (28), caki 'self' does not agree in its number with the every-type QP, hence the degraded grammaticality of (28). However, the Korean reflexive caki-casin 'self-self' yields a bound variable reading, along with the every-type QP. Despite this fact, the Korean reflexive caki-casin 'self-self' does not agree in its number with the every-type QP. Note that nwukwuna 'everyone' is plural in its number since it involves a group of people and caki-casin 'self-self' is singular in its number. It is noteworthy that this is against Safir's (2014) hypothesis that "D-bound anaphora must be feature-compatible with its antecedent" (Safir, 2014, p. 92). It is worth pointing out that our observation supports Safir's (2014) second hypothesis that "D-bound anaphora is interpreted as a bound variable regardless of its phi-features" (Safir, 2014, p. 92 In (30a), caki 'self' occurs non-locally with respect to the every-type QP and this example is perfect with the intended reading. The grammaticality of (30a) indicates that non-local caki 'self' can give rise to a bound variable reading. In this example, the antecedent nwukwuna 'everyone' has phi-features involving gender (masculine or female), person (third person), and number (plural) and caki 'self' has phi-features involving gender (masculine or female), person (third person), and number (singular). This in turn indicates that caki 'self' induces a bound variable reading, despite the fact that it does not agree in its number with the every-type QP. This implies that our observation is against Safir's (2014) hypothesis which "D-bound anaphora must be feature-compatible with its antecedent" (Safir, 2014, p. 92). However, our data support Safir's (2014) second hypothesis that D-bound anaphora is interpreted as a bound variable regardless of its phi-features. As observed, non-local caki 'self' induces a bound variable reading regardless of its phi-features. Likewise, in (30b), caki-casin 'self-self' appears non-locally with respect to the every-type QP and this sentence is perfect with the intended reading.
The grammaticality of (30b) suggests that caki-casin 'self-self' induces a bound variable reading. In this example, the antecedent nwukwuna 'everyone' has phi-features such as gender (masculine or female), person (third person), and number (plural) and caki-casin 'self-self' carries phi-features such as gender (masculine or female), person (third person), and number (singular). The grammaticality of (30b) indicates that caki-casin 'self-self' induces a bound variable reading, despite the fact that it does not agree in number with the every-type QP. This implies that (30b) is against Safir's (2014) first hypothesis that D-bound anaphora must be feature-compatible with its antecedent. However, (30b) lends its support to Safir's (2014) second hypothesis that "D-bound anaphora is interpreted as a bound variable regardless of its phi-features" (Safir, 2014, p. 92).

A True Anaphor
In what follows, we briefly illustrate that the Korean reflexive caki-casin 'self-self' is a true anaphor. In (32a) and (32b), caki 'self' and caki-casin 'self-self' occur locally with respect to the every-type QP.
However, the occurance of caki-casin 'self-self' is much more natural. In (32a), we can only see degraded grammaticality for local binding of caki 'self' as compared to that of caki-casin 'self-self'.
We take this as indicating that caki 'self' competes with caki-casin 'self-self' only when it acts as a bound variable. When its antecedent is an every-type QP, it becomes ungrammatical. That is, caki-casin 'self-self' competes with caki 'self' but caki 'self' loses. Simply put, caki-casin 'self-self' always induces a bound variable reading regardless of its phi-features. Thus, this fact supports Safir's (2014) second hypothesis that "D-bound anaphora is interpreted as a bound variable regardless of its phi-features" (Safir, 2014, p. 92). In (32b), caki-casin 'self-self' gives rise to a bound variable reading, despite the fact that it does not agree in number with the every-type QP.

Conclusion
The main goal of this paper is to argue that Korean pronouns must be phi-feature-compatible with their antecedents, whereas Korean reflexives are not. This paper argues that Korean pronouns are sensitive to the number feature, whereas Korean anaphors are not. In section 3, we have shown that every-type QPs have a Q feature that is plural in its number, whereas which-type QPs have an optional Q-feature that is singular or plural in number. In section 4, we have contended that our agreement-based analysis explains why the Korean singular ku 'he' and the Japanese singular pronoun kare 'he' cannot be interpreted as a bound variable. In addition, we have shown that Safir's (2014) hypothesis that "D-bound anaphora must be feature-compatible with its antecedent" (Safir, 2016, p. 92) is on the right track since Korean pronouns are sensitive to phi-features. Additionally, we have contended that Safir's (2014) second hypothesis that "D-bound anaphora is interpreted as a bound variable regardless of its phi-features" (Safir, 2014, p. 92) does not work for Korean pronouns since they induce a bound variable reading through the phi-feature agreement. In section 4, we have argued that Korean anaphors are not feature-compatible with antecedents and that Korean anaphors give rise to a bound variable reading regardless of their phi-features. In section 5, we have maintained that caki-casin 'self-self' competes with caki 'self' only when it acts as a bound variable. When its antecedent is an every-type QP, local caki 'self' may not induce a bound variable reading. That is, caki-casin 'self-self' competes with caki 'self' but caki 'self' loses.