A grammatico-pragmatic analysis of the because X construction: Private expression within public expression

Background: This article investigates an innovative use of because, called the because X construction (e.g., because homework). Quantitative and qualitative research as well as research about the historical development of the construction have been conducted. The present article aims to determine what motivates the use of the construction. Methods: Based on the data collected from the literature and online sources, the grammar of the because X construction is described in detail. The construction is then analyzed within Hirose’s (2015) three-tier model of language use. Results: A two-layered expressive structure is proposed: The X-element serves as a private expression, which is a speaker’s expression of thought with no intention of communication, whereas the whole construction functions publicly. The private nature of the X-element consistently accounts for the syntactic categories of the X-element and the restrictions on them observed in the literature. Conclusion: The proposed two-layered expressive structure reflects a metapragmatic function of the construction. A private, subjective expression embedded in a public expression has the function of connecting the hearer to the speaker, and it accordingly brings about a joint attention effect. With such a function, the proposed structure is effective especially (but not exclusively) in online communication because one can strategically indicate closeness or intimacy to others, particularly in an environment where nonverbal means are difficult to apply.


Introduction
The word because in English, which is typically followed by a finite clause or an of phrase, conveys a cause or reason, as exemplified in (1a, b): (1) a. He's not coming to class because he's sick.
b. He's not coming to class because of his sickness.
An innovative use of because has recently emerged that is in use particularly in online communication and colloquial conversation in which a single word or non-clausal phrase directly follows because, as exemplified in (2): (2) I cannot go out with you today because homework/sick.
As single words in various grammatical categories such as nouns, adjectives, and interjections can follow because, the construction is called the because X construction. This article investigates the characteristics of this construction, focusing particularly on the status of the X-element from the perspective of the three-tier model of language use (e.g., Hirose 2015). Specifically, it is claimed in accordance with Kanetani (2017) that the X-element serves as a private expression, namely, a speaker's expression of thought with no intention of communication, although the whole construction is used publicly. 1 Pragmatic effects in response to the proposed structure are also discussed.
The present article is organized as follows. After outlining the research methodology in Section 2, I observe the functional and syntactic properties of the construction as well as the construction's characteristic expressive structure in Section 3. Section 4 reviews the typological characteristics of English from the perspective of the three-tier model of language use. Section 5 investigates various ways in which the X-elements are construed as private expressions. Sections 6 and 7 explore the motivations for these private expressions.

Methods
The analysis in the present article follows a traditional linguistic methodology. First, the grammar of the target constitution is described in detail from both semantic and syntactic points of view based on the grammaticality or acceptability of linguistic data, which are collected from the literature and online sources including tweets, blogs, and corpora. As the because X construction has only recently emerged and come to be recognized, the availability of descriptive data from research papers is limited and the description of its grammar is not sufficient. Therefore, describing the semantic and syntactic characteristics of the construction based on the collected data is essential for analyzing the construction.
After the grammar is described, the construction is analyzed within a certain theoretical framework, and discussions about theoretical implications follow. Specifically, the construction is analyzed by revising and expanding Kanetani's (2016Kanetani's ( , 2017, 2019) account of the because X construction from the perspective of the three-tier model of language use, which 1 In this article, I do not distinguish spoken language from written (and typed) language and consistently use the term speaker. Similarly, I use the terms hearer, addressee, and interlocutor even in contexts of written or computer-mediated communications. claims that the X-element functions as a private expression while the entire construction functions as a public expression. The three-tier model is a grammatico-pragmatic theory developed by Hirose (2013Hirose ( , 2015Hirose ( , 2016) as a natural extension from the notion of private and public expressions (Hirose 2000). While the details of the model will be introduced in Section 5, it should be noted here that the grammatico-pragmatic characteristics of a language can only be defined relatively to other languages. In fact, the three-tier model ascribes differences between languages to different combination patterns of the three tiers of language use, namely, the situation construal, situation report, and interpersonal relationship tiers. Therefore, to highlight the grammatico-pragmatic traits of the English language, it is compared to Japanese and the markedness of the construction within the system of English grammar is emphasized.
However, my earlier analyses left open the question of what motivates the proposed expressive structure of the construction. Therefore, comparing the because X construction with other linguistic phenomena of similar expressive structures (an innovative use of kudasai 'please' in online communication (Naya 2017) and soliloquy insertion in conversations (Hasegawa 2010)), the present article discusses the significance of the proposed structure and a general metapragmatic strategy behind these expressions.
3. The because X construction Kanetani (2015Kanetani ( , 2016Kanetani ( , 2017Kanetani ( , 2019 treats the phenomenon from the perspective of construction grammar, where a construction is generally defined as a conventionalized pairing of form and function (Fillmore et al. 1988; Goldberg 1995; Hoffman and Trousdale 2013, among many others). After briefly reviewing the functional properties of the construction in Section 3.1, I identify the formal properties in Section 3.2 and then describe the form-meaning correspondence in Section 3.3.

Functional properties
In this subsection, I review a semantic property of the because X construction, comparing it with the more general because-clause constructions. Sweetser (1990) claims that a because-clause may be used in the content, epistemic, and speech-act domains, as exemplified in (3a-c): (3) a. John came back because he loved her.
b. John loved her, because he came back.
c. What are you doing tonight, because there's a good movie on. (Sweetser 1990: 77) According to Sweetser (1990: 77), these sentences are interpreted as follows. Sentence (3a) describes a causal relation that holds in the real world; that is, John loving her caused him to come back. In (3b), the causal relation is held in the epistemic domain; that is, the speaker's knowledge about the fact that John came back causes him/her to conclude that John must love her. A speech-act because-clause as in (3c) serves as a motivation for performing a certain speech act such as asking about the interlocutor's plans for the night. Kanetani (2015Kanetani ( , 2016Kanetani ( , 2017Kanetani ( , 2019 observes that the because X construction is skewed toward the content reading. This is confirmed by a survey the author conducted in January 2014 shortly after the American Dialect Society's selection of (the innovative use of) because as its 2013 Word of the Year (see Kanetani 2015). Sentences (4a-e) below were developed for the survey and 24 native English speakers were asked about their acceptability. Of the 24 native speakers surveyed, seven speakers accepted the usage with different degrees of acceptability. The scores shown at the end of the examples are the average scores of acceptability by the seven respondents on a scale of 0 (unacceptable) to 3 (acceptable). The scores of those who did not accept the usage at all were eliminated from the calculations. 2 c. He loved her, because back. (0.71/3.00) 2 The survey was conducted from January to February 2014. The results might be different if a survey with the same sentences were conducted now. Since, as Bohmann (2016: 161) notes, the construction is a "rapidly diffusing innovation", its usage may expand rapidly; accordingly, the acceptability might vary in a different survey. d.
[Looking at a wet ground] It's rained, because ground. (0.00/3.00) e. What do you wanna do on our first evening, because Paris? (0.57/3.00) (Kanetani 2015: 66) In (4a, b), the causal relations hold in Sweetser's (1990) content domain. Sentences (4c, d) exemplify the epistemic because X, and sentence (4e) represents a speech-act because X. The results show that [because X] appears to be acceptable in the content domain but not in the epistemic and speech-act domains.
This functional property might be predictable to some extent. Lakoff (1987) observes that a speech act construction that conveys a statement, like a rhetorical question, may occur in because-clauses when those clauses are in sentence-final position.
(5) a. We should go on a picnic, because isn't it a beautiful day! b. * Because isn't it a beautiful day, we should go on a picnic. (Lakoff 1987: 474) The rhetorical question in the because-clauses in (5a, b), isn't it a beautiful day, performs a speech act conveying the statement that it is a beautiful day. Hence, Lakoff calls these because-clauses performative subordinate clauses. As pointed out in Kanetani (2019), a sentence-final because-clause is not sufficient for a performative subordinate clause to occur.
(6) * He's not going out for dinner because Japanese food, his wife is cooking. (Kanetani 2019: 55) Sentence (6) is ruled out even though the because-clause appears in the sentence-final position. In (6), the matrix negation scopes over the entire sentence, which is characteristic of the content because-clause (cf. Rutherford 1970). The intended interpretation of the sentence is therefore something like it is not because his wife is cooking terrible Japanese food that he is going out for dinner (cf. OK Sam is not going out for dinner because his wife is cooking Japanese food (Hooper and Thompson (1973: 494)). Thus, Kanetani (2019) concludes that it is epistemic/speech-act because-clauses (reasoning because-clauses in his terms) that can be performative. 3 As the sentence-initial position is reserved for content becauseclauses (cf. Hirose 1999; Kanetani 2019), this generalization compensates for but is not incompatible with what Lakoff (1987) says.
As an epistemic/speech act because-clause performs a speech act of its own, the because X construction is naturally incompatible with an epistemic/speech-act reason clause, because the word or phrase that appears in the X-slot cannot perform an independent speech act. In this connection, because of NP (e.g., (1b)), one of the traditional uses, is also restricted to the content domain. An epistemic/speech-act because-clause is not replaceable with a because of phrase, as shown in (7) Lastly, Okada's (2020) discussion on the origin of the because X construction supports this claim. While acknowledging the difficulties in ascertaining when and how a new structure was generated, Okada (2020) considers the because X construction to have developed historically through the following steps. First, a blending occurs of because S and because of NP, yielding the new structure because NP, where the "NP works as a reference point for the conceptually relevant proposition" (Okada,p. 8). Subsequently, "the category restriction of the complement is nullified and elements of any category will appear as far as they work as reference points for the conceptually relevant proposition" (Okada, p. 8). 4 3 Cappelle (2021: 24-25) reports some examples that seem problematic to this generalization. They will be analyzed in future research. 4 Other possibilities of the development of this construction are discussed by  and Kanetani (2019). Both  and Okada (2020) reveal that the structure of the because X construction, though innovative, is not as new as one might think.
If this is correct, that is, if one of the inputs motivating the because X construction is the because of NP construction, it is not surprising that the meaning of the because X construction is skewed toward the content domain.

Formal properties
Let us turn to the formal properties of the because X construction. First, as a consequence of the functional properties observed in Section 3.1, sentences with [because X] behave in the same manner as those with a content because-clause. Both of them allow the reason part ([because X] or the because-clause) to appear in sentence-initial position and to be focalized by an exclusive subjunct such as only and just. 5 Relevant examples of the because X construction are given in (8) and (9) As mentioned in Section 3.1, a sentence-initial because-clause is characteristic of the content reading. Hence, the attested example in (8) indicates that the because X sentences are compatible with the content reading. The [because X] in (9) is focalized by the exclusive just. As Kanetani (2019: chapter 4) claims, focalization is possible for content because-clauses but not for epistemic/speech-act because-clauses. The ungrammaticality of (10) shows that the exclusive just cannot focalize an epistemic because-clause: (10) * It has rained, just because the ground is wet. (Kanetani 2019: 71) Up to this point, it has been shown that the because X construction syntactically behaves like the content because-clause construction and not like the epistemic/speech-act because-clause constructions.
To identify the formal property of the because X construction, it is necessary to consider what syntactic categories are likely to appear in the X-slot. The categorial restriction on the X-element is accounted for by the construction's expressive structure to be proposed in Section 4. Schnoebelen (2014) counts the target construction in tweets and groups all items that have 50 or more occurrences based on their parts of speech. The results are summarized in Table 1. Similarly, Bohmann (2016) examines 805 tweets and summarizes the categories that appear in the X-slot as in Table 2.
The two tables commonly include nouns (or noun phrases), adjectives, and interjections. The most recent corpus survey by Mendes Junior and Mattos (2021) also confirms that the categories occurring frequently in the X-slot are (in descending order) nouns > adjectives > interjections > adverbs > verbs. Thus, it is safe to say that nouns, adjectives, and interjections frequently appear in the X-slot.
In addition to these three categories, several other categories are identified. First, in comparing Tables 1 and 2, it should be noted that Schnoebelen's (2014) "compressed clause" is not the same as Bohmann's (2016) "reduced clause". In Schnoebelen's survey of tweets, the most frequent token is yolo, which is a compression of you only live once. Schnoebelen states that "if you spell it out, because you only live once is actually completely standard (you only live once is an example of a fine full clause). But yolo is a lot like an interjection" (underline added). Bohmann (2016: 161) also distinguishes these compressed clauses from their clausal counterparts and considers these "(semi-)lexicalized, fixed expressions". As special forms generally convey special functions, following Schnoebelen (2014) and , I take compressed clauses as fixed expressions with a function similar to that of interjections. 6 As noted above, the reduced clause in Table 2 is distinguished from the compressed clause in Table 1. Bohmann (2016) defines reduced clauses as finite clauses "often with deleted subjects" (p. 160). There is one thing we must bear in mind in dealing with reduced clauses. Namely, in some cases, a reduced clause follows because while the sentence does not exemplify the because X construction. As Okada (2020) points out, subordinate clauses generally allow the subject and copula to be deleted. Observe (11) That is, Okada claims that the sentences in (11a, b) do not exemplify what we call the because X construction because their (superficial) because X part may be recoverable based on the matrix subject being combined with a copula verb. Incidentally, based on a survey of the Oxford English Dictionary, Okada (2020) reports that this type of "subject + copula" deletion in a because-clause dates as far back as the 16th century. Therefore, the systematic "subject + copula" deletion structure should be distinguished from Bohmann's (2016) reduced clause and eliminated from the analysis (at least for the present purposes); thus, the "reduced clause" is limited to examples such as the following: (13) Bye going to study for English because didn't finish this morning because fell asleep.
(Carey 2013, underlines added) Lastly, "agreement" and "pronoun" are taken as independent categories in Table 1. Bohmann might include them (if at all) in the "other" category. It is necessary to highlight the relatively small number of words in these categories. The agreement words include yes, yeah, no, and a few other similar words; the pronoun is a closed category consisting of only In Kanetani (2017), assuming that a compressed clause is a register-specific equivalent to a clause, that is, a clause used particularly in computer-mediated communication, I eliminated compressed clauses from the analysis. However, given the considerable frequency of compressed clauses and their differences from full clauses (see Section 5.1), I have included them in the present analysis. a small number of members. Nevertheless, agreement words appear far more frequently than pronouns in this construction. In Table 1, "agreement" (12.97%) actually nears "interjection" (14.71%), one of the most frequently used categories. By contrast, pronouns (2.45%) are used far less frequently. This is also supported by McCulloch (2014), who observes that a pronoun is "weird" when used in this construction.  The mechanism by which to account for the (non-)occurrence of these elements will be further discussed in Section 5.

Form-meaning pairing of the because X construction
From the observations given so far, the form-meaning pairing of the because X construction may be described as in (15): (15) [CLAUSE i because X j ] $ [P (evoked by "X j ") is a reason for Q i ] (Modified from Kanetani 2017: 95) In (15), the form of the construction specified on the left side of the double-headed arrow ($) is paired with the meaning specified on the right side; the coindexed elements in the form-pole and meaning-pole represent the form-meaning correspondences. That is, CLAUSE i conveys the propositional meaning Q i ; X j with the meaning represented as "X j " is a word (or phrase) from one of various categories such as those listed in Tables 1 and 2. Since a because X sentence represents a real-world causal relation in the content domain (Section 3.1), the X-element needs to represent a certain propositional content. Therefore, the meaning of the word in the X-slot cannot be taken as a simple denotation of the lexical/phrasal meaning but should be understood as a relevant proposition evoked by it. 7 Note that the form-meaning pairing in (15) is a base-level representation and that there are variations. For example, [because X] may precede the main clause (e.g., (8a, b)); there are also cases-as used in colloquial or online contextswhere the main clause is reduced or omitted (e.g., Early morning gym because fat (Bohmann 2016: 149)), as well as cases where orthographic variations of because (e.g., bc, cuz, and coz) are involved. However, differences between these formal variations are not considered in the present article.
3.4 Two-layered expressive structure: Private expression within public expression Kanetani (2016Kanetani ( , 2017Kanetani ( , 2019 proposes the construction's expressive structure using Hirose's (2000) terms: "private expression" and "public expression". In the present article, I maintain the basic claim that the element in the X-slot prototypically serves as a private expression. Hirose (2000Hirose ( : 1624 proposes two levels of linguistic expressions, called private and public expression: the former is "the level of linguistic expression corresponding to the non-communicative, thought-expressing function of language", whereas the latter is "the level of linguistic expression corresponding to the communicative function of language". Thus, the claim being made here may be rephrased as follows: the element in the X-slot basically has a thought-expressing function with no intention of communication. 7 Okada (2020: 8) accounts for this fact in terms of Langacker's (1993) notion of reference point, saying that "elements of any category will appear as far as they work as reference points for the conceptually relevant proposition".
However, this does not mean that the whole construction functions as a private expression. Schnoebelen (2014) reports that 36% of the tweets investigated involve @-mentions, which indicates that they are aimed at a specific person or persons as a reply (cf. also , and therefore that the because X construction seems skewed toward the "interpersonal". Thus, the expressive structure of the prototypical construction may be illustrated as in (16), with Hirose's (2000) notations of private expression represented in angle brackets with the subscript "Priv" < Priv …> and public expression represented in square brackets with the subscript "Pub" [ Pub …]. 8 (16) [ Pub because < Priv X>] The representation in (16) indicates that the whole message is communicated as a public expression, within which a private expression is encapsulated.

The three-tier model of language use
In Section 3.4, the two-layered expressive structure was proposed, as shown in (16), in which the element in the X-slot crucially serves as a private expression while the entire construction may function as a public expression. This section considers what it means that a private expression is used in a public expression in terms of the three-tier model of language use (Hirose 2015).
As mentioned in Section 2, the three-tier model is a grammatico-pragmatic theory proposed inter alia by Hirose (2015) as a natural extension from the deconstruction of the speaker into the private self as the subject of thinking and the public self as the subject of communicating (Hirose 2000). According to the three-tier model, language use comprises the three tiers listed in (17a-c), and "languages differ as to how the three tiers are combined, according to whether their basic egocentricity lies in the public self or the private self" (Hirose 2015: 123).
(17) a. situation construal tier: the speaker as private self construes a situation, forming a thought about it.
b. situation report tier: the speaker as public self reports or communicates his construed situation to the addressee.
c. interpersonal relationship tier: the speaker as public self construes and considers his interpersonal relationship with the addressee. (ibid.: 123) In this model, the English language is characterized as follows: "In English, a public-self-centered language, the situation construal tier is normally unified with the situation report tier, to which is added the interpersonal relationship tier […]. The unification of situation construal and situation report means that one gives priority to the outside perspective from which to report a situation and linguistically encodes as much as is necessary to do so. Thus, even when the speaker himself is involved in a situation as a participant, the reporter's perspective places his self as a participant on a par with the other participants; hence comes objective construal. On the other hand, the fact that the situation report tier is not unified with the interpersonal relationship tier means that one can assume an unmarked (or neutral) level of communication which does not depend on any particular relationship between speaker and addressee, a level where the speaker and the addressee are assumed to be linguistically equal, being in a symmetrical relationship." (ibid.: 123-124, underlines added) To highlight the characteristics of English, let us compare them with those of Japanese, which is described as follows: "In Japanese, a private-self-centered language, the situation construal tier is normally independent of the situation report tier and the interpersonal relationship tier […]. Thus, in construing a situation, the speaker can freely place himself in the situation and view it from the inside; also, he does not need to linguistically encode what is already given in his consciousness; hence comes subjective construal. On the other hand, the situation report tier is unified 8 I have added "prototypical" in the present version. My earlier works (Kanetani (2016(Kanetani ( , 2017(Kanetani ( , 2019 and the first version of the present article) did not distinguish prototypical instances from more extended ones. Bohmann (2021: 32) reports some instances whose X-elements superficially do not seem to serve as private expression (e.g. because you know, because this fever); however, as discussed in Section 7, they are motivated by the two-layered expressive structure proposed here.
with the interpersonal relationship tier, which means that in reporting a situation to someone, the speaker must always construe and consider his relationship with the addressee, defining himself and the addressee in terms of that relationship. Thus, in situation report, interpersonal relationship is linguistically encoded as much as possible, and there is no unmarked level of communication neutral to interpersonal relationship." (ibid.: 124-125, underline added) The crosslinguistic difference in the unification pattern of the three tiers is illustrated in Figures 1 and 2, where the bold faces indicate the tiers in which the unmarked deictic center is located, i.e., the default position where deictic expressions are interpreted. 9 Figure 1 illustrates that the situation report tier, where the unmarked deictic center is located, is unified with the situation construal tier in English. Therefore, English speakers need to construe the situation objectively as they report it to others. Conversely, Figure 2 shows that, in Japanese the unmarked deictic center is located in the situation construal tier, which is independent of the unification of the situation report and interpersonal relationship tiers, allowing Japanese speakers to express the situation as they construe it.
Ide (2006) also neatly describes the typological difference. According to Ide, while Japanese speakers tend to view themselves as participants in the situation described, English speakers tend to take the perspective of an omniscient narrator and overview the entire speech event from the outside. The different perspectives of the two languages are illustrated in Figures 3 and 4. 10 As illustrated in Figure 3, an English speaker sees and describes herself on stage as others see her. A Japanese speaker, illustrated in Figure 4, plays the role of a participant on stage as well as the narrator. The speaker, who is embedded in the situation, has to linguistically encode only what is necessary and leaves other elements (me, this book) unspoken. Thus, to describe the same situation, the Japanese speaker would say as follows, with the parenthetical elements not necessarily being expressed. 11     In short, English speakers prefer to take an objective perspective from outside of the situation while Japanese speakers prefer to take a subjective perspective from the inside. This observation is compatible with the three-tier model, because the default deictic center in English is the public self, located in the situation report tier (as shown in Figure 1), while that in Japanese is the private self, located in the situation construal tier (as shown in Figure 2). Another important typological difference Hirose (2015) puts forward as a consequence of the three-tier model is that the unmarked mode of expression is public expression in English and private expression in Japanese. Consider the following contrast: The unification of the situation construal and situation report tiers in English means that the utterance today is Saturday has a performative structure (cf. Ross 1970), as in (21): Note that all of the Japanese sentences in (19) and (20) convey one and the same propositional content, today is Saturday. The unmarked sentence in (19), however, functions as a private expression by itself; hence, it is incompatible with the performative clause, as shown in (22). Instead, various expressions sensitive to the interpersonal relationship, such as the unmarked sentence-final particle yo, the (super) polite form of the copula desu or degozaimasu, etc., are employed to make the expression public.
With the typological characteristics of English in mind, let us consider the fact that the X-element represents the speaker's private expression. According to the three-tier model, an English speaker essentially takes a reporter's (or an objective) perspective and linguistically encodes as much as is necessary to do so; hence, the unmarked mode of expression in English is public expression. The expressions in the X-slot, on the other hand, exhibit the speaker's subjective construal in that the speaker does not linguistically encode what is already given in his/her consciousness, which is characteristic to languages like Japanese. This claim is in line with Bergs's (2018) argument that the because X construction is subjective compared with the content causal because-clause construction in the sense of Traugott and Dasher (2004). Crucially, it is not the whole construction that is subjective but only the part that deviates from the norm of the English language. Presumably, the subjectivity Bergs observes in this construction is related to the subjective nature of the X-element. In the following sections, I examine the privateness and subjectivity of the X-element.

X-element as private expression
It was observed in Section 3.2 that nouns, adjectives, interjections (including compressed clause), agreement words, and reduced clauses frequently appear in the X-slot. In Section 3.4, the two-layered expressive structure for the because X construction was proposed, which is reproduced here as (23): In the following subsections, I examine the "privateness" of the expressions that frequently appear in the X-slot and the "publicness" of those that rarely appear.

Interjections and compressed clauses
Interjections frequently appear in the X-slot as in (24): (24) a. That feeling you get when you finish an essay and you just want to cry because yay.
b. Admittedly, not in the UK yet, because aargh. (Carey 2013) Interjections are described as "purely emotive words" (Quirk et al. 1985: 853) that "have expressive rather than propositional meanings" (Huddleston and Pullum 2002: 1361); that is, they are used to express, rather than to communicate, the speaker's emotion. Therefore, interjections by nature may serve as private expressions with no intention of communication.
How then is the conveyed message understood by the hearer? Padilla Cruz (2009) examines cases where a subordinate clause is replaced by an interjection as in (25): He explains that "the hearer could recover the missing clause using contextual and/or encyclopedic information" (Padilla Cruz 2009, pp. 190-191) and that the meaning of sentence (25) may be understood as something in (26a-c) or the like: (26) a. She is so beautiful that I like/love her.
b. She is so beautiful that I have fallen in love with her.
c. She is so beautiful that I would very much like to marry her. (Padilla Cruz 2009: 190) It is important to note that specific emotions are mapped onto each interjection; for example, aargh is used to express "fear, anger, or other strong emotion" (Oxford Advanced Learner's Dictionary, 8th edition [OALD 8 ]), and yay is used to show that one is "very pleased with something" (OALD 8 ). Thus, the utterance because aargh may be construed as because something extremely bad happened. Therefore, the lexical information as well as the "contextual and/or encyclopedic information" plays an important role in recovering the message.
My earlier analysis fails to distinguish the roles of a speaker and hearer, and only identifies a metonymic relation between the semantic content of an interjection and that of a clause (Kanetani 2015). However, by using an interjection, the speaker merely expresses an emotion with no intention of communicating; it is the hearer who attempts to understand the utterance in question based on the contextual, encyclopedic, and/or lexical information (cf. Padilla Cruz 2009; Kanetani 2016).
As observed in Section 3.2, compressed clauses such as yolo and ilysm (a compression of I love you so much) have a similar function to that of interjections. In fact, they exhibit certain features distinct from their clausal counterparts. For instance, the compressed clause yolo is pronounced in an exclamatory tone, conveys specialized meanings, and can be converted into the verb yoloing ( https://www.youtube.com/watch?v=MKT3DaClfvY [retrieved on July 2, 2021]; see also Bohmann 2016: 161), which are shared features with interjections and not with finite clauses. Because of the meaning of the clausal counterparts (e.g., yolo for you only live once), compressed clauses may convey more specific meanings close to clauses than simple interjections. In the present article, however, compressed clauses used in this construction are profitably analyzed in a similar way to interjections (cf. Schnoebelen 2014, Bohmann 2016). 13

Content words
Content words include nouns, adjectives, verbs, and adverbs, constituting the proposition being conveyed. However, it is the hearer who recovers the proposition containing these words as its parts, with the aid of the PART FOR ALL-more specifically, WORD FOR CLAUSE-metonymy. The speaker, on the other hand, simply chooses the most salient word from a clause conveying the meaning of "P" in (15) as a reason (see also footnote 7). What is "salient" may be something that pops into the speaker's mind at the time of utterance; therefore, the word represents the speaker's private expression to the extent that he/she does not need to place others at the center of his/her consciousness. Nouns and adjectives are open-set content words whose primary function is "to carry the meaning of a sentence" and hence "typically carry the burden of the semantic content of utterances" (Cruse 2011: 267f.). Apart from nouns and adjectives, adverbs and verbs are also open-set content words that can be used in this construction (Okada 2020; Mendes Junior and Mattos 2021). The use of verbs and adverbs in this construction may also be accounted for on the same ground as the use of nouns and adjectives.
Recall that pronouns are not used (McCulloch 2014) or may be used only rarely (Schnoebelen 2014; see Table 1).
(27) ?? I can't go to the party because you.
This observation is compatible with the present proposal that the X-element represents the speaker's private expression.
In terms of Hirose's (2000) dichotomy between private and public expressions, English personal pronouns are primarily defined as public expressions (cf. Hirose 2000Hirose , 2015. 14 In this regard, it is worth quoting Benveniste (1971: 224f.) here: "[C] consciousness of self is only possible when it is experienced by contrast. I use I when I am speaking to someone who will be a you in my address. It is this condition of dialogue that is constitutive of person, for it implies that reciprocally I becomes you in the address of the one who in his turn designates himself as I." (underline added) In short, only in relation to others can the personal pronoun be defined and used; that is, pronouns cannot be used in the absence of others. 15 This makes personal pronouns unsuitable X-elements, because the slot requires a private expression.
McCulloch (2012) proposes another intriguing restriction on the nominal category. She observes that the noun that follows because should be a bare noun, i.e., a noun with no determiner, as in (28) This restriction also indicates the private nature of the X-element. According to Quirk et al. (1985: 253), "when used in discourse, noun phrases refer to the linguistic or situational context. The kind of reference a particular noun phrase has depends on its determinative element, i.e. the item which 'determines' it". In other words, determination is necessary in a discourse for the speaker to allow the hearer to identify the type of reference. Put differently, unless the speaker has an interlocutor in mind, determination is not necessary in Quirk et al.'s sense. Against this restriction, Cappelle (2021: 27) lists some examples of nouns modified by adjectives (e.g., because bad breath/ because young adults/ because free speech). As Cappelle notes, they may be accepted if "[an adjective] forms a close lexical unit with the noun". For example, young adults refer to people in their late teens or early twenties, and not adults who is young, free speech refers to a right rather than a kind of speech. That is, they are seen as word-like units and are excluded from the restriction. 7 Okada (2020: 8) accounts for this fact in terms of Langacker's (1993) notion of reference point, saying that "elements of any category will appear as far as they work as reference points for the conceptually relevant proposition". 14 Hirose (2000Hirose ( : 1630 argues that due to the lack of a special word for private self in English, "personal pronouns, which are primarily defined as public expressions, are diverted to represent the private self, depending on whether the subject of the private expression in question is the first, second, or third person". 15 Jordan (1989) reports that autistic children use proper names, instead of the first person pronoun, for self-reference. Cappelle (2014) observes an unusual language use by Jerom, a character in the Flemish comic book Suske en Wiske. Cappelle relates Jerome's failure in the proper use of pronouns and his closed eyes as his defensive attitude of cutting off the sight of others' existence while speaking. These studies also suggest that social interaction with or the assumed presence of others is a prerequisite for the appropriate use of pronouns. 16 This seems to rule out because essay. However, McCulloch explicitly says that "the because + noun construction really must consist of a bare noun, not a noun with a determiner or an adjective". Therefore, assuming that she simply misplaced the asterisk, I consider because my essay/an essay/this essay to be ruled out, but because essay to be ruled in.
Bohmann (2021: 32) also reports some noun phrases with determiners (e.g., because the bad acting/because this fever), which bear special (euphemistic) meanings. These examples will be discussed in Section 7.
So far, two restrictions on the nominal category that indicate the privateness of the X-element have been discussed. While other content words may be analyzed in the same way as nouns because they are subjectively selected as possible salient constituents of the corresponding clause, an additional comment is needed on verbs, which are only rarely used in this construction. An example with a verb is given in (30) In summary, the restrictions on nouns and verbs may be reduced to the lack of what generative linguists call functional categories corresponding to the D-and T-heads, respectively. Konno (2012) points out that the lack of a functional category is related to the lack of hearer-orientedness (cf. also Konno 2015). For example, a Mad Magazine sentence (e.g., Him wear a tuxedo?!) that is used to "express surprise, disbelief, skepticism, scorn, and so on, at some situation or event" (Akmajian 1984: 2) cannot be embedded in a verb of communicating such as tell (e.g., ??Mary told him "Him wear a tuxedo?!" (Konno 2015: 146)) . Konno thus views the construction as having an exclusively private function. Notably, the verb in a Mad Magazine sentence is bare and hence lacks tense. Following Konno (2012, 2015), we may posit that the bare nouns used in the because X construction and the bare verb sleep in (30) exhibit the speaker's private expression. 19

Agreement words
This subsection considers agreement words, as in (32): (32) "So I guess you're okay that it's you then?" he says, and Nick grins because yeah. "Very okay." (GLoWbE) In (32), because is followed by yeah, an agreement word, in which the speaker asserts only the polarity of the propositional content with the other details being underspecified. The word yeah in (32) affirms the proposition that 17 Although I do not discuss whether the verb reset is used in the infinitival form or the past form, we can tell that the sentence expresses a specific event in the past because of the temporal expressions for 8 and for 930, and the auxiliary verb could in the preceding sentence. 18 ) should be treated as reduced clauses following because. Example (ic), in which a finite clause follows because, arguably does not exemplify the because X construction. Due to the lack of my before stomach, however, one may argue that this should not count as a canonical use of because but rather as a because X construct. If sentence (ic) were treated as an instance of the because X construction, it should be discussed under the reduced clause, as the omission of my may be parallel to the omission of I in diary English (e.g., (35)). Note in passing that the subject of the main clause I is also omitted in (ic). Only those that are neither reduced clauses nor instances of regular deletion should be treated as representing the pattern of [because + verb]. 19   suggests that as with nouns, "bare" adjectives seem to be preferred for this construction. By "bare" is meant adjectives with no premodifier like very (e.g., because (?very) unexpected (Bergs 2018:48)). However, I leave it for future research as to how this tendency may be related to the privateness of the X-element. I did not consider what may indicate the privateness of the adverbs used in this construction, either. As pointed out by Capplle (2021), this tendency may not be explained solely by the private nature of the X-element. This is also left for future research. he is okay. In this way, agreement words such as yes or yeah affirm certain propositions that lie behind the words, while disagreement words such as no deny them. To maintain this claim, let us observe Nakau's (1994) hierarchical structure of a proposition, as illustrated here (33) As shown in (33), a full proposition consists of the four strata PROP1-PROP4. The lowest layer, PROP1, consists only of the combination of the predicate and its argument(s). Added over PROP1 are the aspectual, tense, and polarity operators, yielding more complex and composite propositions. As the polarity operator is placed at the outermost layer in (32), the proposition that exists behind the agreement words corresponds to PROP4. As with the interjections, the hearer may recover the missing part, PROP3 in this case, by using contextual and/or encyclopedic information.
Framed in Nakau's (1994) model (33), the content words, such as nouns and verbs, used in the bare form correspond to part of PROP1, either PRED or an ARG, with no tense or aspectual operator attached. Crucially, either a content word as part of PROP1 or an agreement word as part of PROP4 may serve as a reference point to evoke a full proposition (for the use of Langacker's (1993) term reference point, see Okada 2020). Generally, when both the speaker and hearer are assumed to be cooperative, the speaker should make his/her contribution as informative as is required (Grice 1975). The use of a private expression, however, does not assume the existence of a hearer, and hence the speaker can express a situation as he/she construes it.
Therefore, the X-elements observed in Sections 5.1-5.3 are representations of the speaker's private expressions. Interjections merely reflect how the speaker takes a certain situation; content words typically appearing in the bare form and agreement words refer to part of a proposition that the speaker constructs in response to the situation construal. Thus, the speaker encapsulates these elements in the X-slot and leaves the remainder of the relevant propositional content unspecified.

Reduced clauses
This subsection examines reduced clauses, which indicate privateness in a different way from the other cases observed in In (35), the subject I is omitted both in the matrix clause and in the adverbial clause. Hirose and Hasegawa (2010: 67) account for the distribution of this construction as follows. As we will see in detail in Section 5, English is by default a public-self-centered, other-oriented (hence, highly objective) language. However, when used in a special context like a diary where communication is not intended, the language need not linguistically encode what the speaker presupposes about him-or herself, exhibiting self-orientedness (or high subjectivity). In this sense, the reduced clauses in (35) may be regarded as representations of private expressions that are restricted to specific registers such as a diary, which is not aimed at a hearer/reader. Likewise, the reduced clauses that follow because in (34) may be considered private expressions.
Some reduced clauses used in the because X construction omit subjects other than I, as exemplified in (36a, b): (36) a. idk if i am bc () depends on baby (Twitter; cited from Bohmann 2021: 32, underline and parentheses added) 22 20 The abbreviations used in (33) are as follows: PROP1-4 = propositions1-4; POL = polarity; TNS = tense; ASP = aspect; PRED = predicate; ARG 1-n = arguments selected by PRED. 21 The parentheses are used to indicate where the pronoun I is omitted. Although the omitted subjects in (36a, b) are not I but it and you, respectively, their omissions may be accounted for along the same line. Haegeman (2019) observes that "the default interpretation in a diary style text is that the null subject is referring to the writer, who is the topic of the diary, but other referents are possible, once these have been made suitably accessible in the context", as in (37) Seeing the string of because why not in (38a), one might consider why not to be a rhetorical question conveying a statement, i.e., Lakoff's (1987) performative subordinate clause (see Section 3.1). However, if the example were taken as a performative subordinate clause, the construction would no longer instantiate the because X construction to begin with. Notice the privateness of the entire sentence, indicated by the omission of subject in the main clause. The expression why not in this example should therefore be considered a fixed expression used to agree to a suggestion rather than a rhetorical question (e.g., Let's eat out this evening. -Yes, why not? (Swan 2016: 630)). In Swan's example, why not is used with yes, thus enhancing the speaker's agreement expressed by yes. That is, the speaker in (38a) clearly endorses the idea that he/she should watch old videos even though the speaker neither explicitly says yes nor replies to any explicit "suggestion". Thus, as with more prototypical agreement words like yes, discussed in Section 4.3, a certain proposition (to be affirmed) exists behind the expression why not. In short, the speaker's agreement to a proposition may be realized as yes in some cases and as why not in other.
More problematic is example (38b). The expression you know is a discourse marker used when a speaker "tell [s] the hearer (or pretend [s]) that they share the same piece of information" (Swan 2016: 301), and hence is by definition an addressee-oriented expression. Thus, this example seems a counterexample to the present analysis. As Swan's (2016) explanation suggests, the speaker does not necessarily tell the hearer but can merely pretend that the reason that the speaker has in mind is shared with the hearer. In other words, the speaker may use the expression without considering the hearer's knowledge state, because pretending is not a matter of communication but a form of behavior. That said, the use of the second person pronoun you in the private expression is not trivial. Even though you know may be used privately, this example should be considered to be a non-prototypical example of the because X construction. This example will be further investigated in Section 7.

Privateness and content causal relation
In Section 3, I claimed that the meaning of the because X construction is restricted to Sweetser's (1990) content causal relation. After pointing out in Section 4 that the X-element exhibits a characteristic of a private-self-centered language like Japanese, I observed in Section 5 that the X-element serves as private expression. In this section, I consider how these facts are intertwined along with Kanetani's (2017) view of a speech act unit as a small discourse. While Sweetser (1990) proposes the three domains in which because functions, Kanetani (2019) claims that the epistemic and speech act because-clauses should be grouped together and that the distinction of the content causal relation from epistemic/ speech-act causal relation is crucial. The division is based on how speech act units are formed. A sentence in the content domain performs one speech act as a whole, whereas two independent speech acts are performed in the epistemic/speechact domain. Compare the following sentences, where the arrows (" and #) indicate intonation patterns: (39) a. Is the ground wet because it has rained?" (Kanetani 2019: 46) b. Has it rained,å because the ground is wet.# (ibid.: 54) Sentences (39a, b) are interrogative sentences with a content because-clause and an epistemic because-clause, respectively. The rising intonation appears at the end of the sentence in (39a), which indicates that the scope of the question encompasses the whole sentence, thus performing a single speech act: The rising intonation in (39b) appears at the end of the main clause. That is, the question scopes over the main clause, while the because-clause independently performs a statement speech act (see Section 3.1; cf. also Lakoff 1987). Thus, the distinction of the epistemic and speech act becauseclauses may be reduced to the kind of speech act (e.g., a statement, question, or imperative) being performed in the main clause (Kanetani 2019).
Given the differences in speech act unit formation between the content because-clause construction and the epistemic/ speech-act because-clause construction, I recast each speech-act unit here as a "small discourse" to account for the relation between the content reading and the private nature of the X-element in the because X construction. Namely, the content because-clause construction consisting of one speech act unit can be taken as a discourse, which starts with the situation described in the main clause and ends with the situation described in the because-clause. By contrast, in the epistemic/speech-act because-clause construction, there exist two paratactic discourses that are independent of each other. For instance, the content causal sentence the ground is wet because it has rained depicts (as it were) one scene, whereas the epistemic causal sentence it has rained, because the ground is wet depicts two separate scenes, one about raining and the other about the ground being wet. The obligatory comma intonation between the main clause and the epistemic/speech-act because-clause (Sweetser 1990) symbolically represents the discourse boundary or the scene shift.
Together with the notion of small discourse, let us consider the because X sentence in (2), repeated here as (40): (40) I cannot go out with you today because homework/sick. (= (2)) Since the sentence describes a content causal relation, its discourse structure is also assumed to be the same as that of the content because-clause construction. Specifically, the sentence delivers a discourse on the speaker being unable to go out because of his/her homework/sickness. At the beginning of the discourse, the speaker takes the reporter's perspective, placing his/herself in a situation as a participant, just as with the speaker in Figure 3. That is, the speaker starts the discourse with the unmarked mode of expression in English, playing the role of a narrator, who observes the situation objectively standing on a par with the hearer. As the discourse progresses, however, the speaker switches his/her perspective to a perspective from the inside, describing the situation as a participant, as if he/she jumped into the situation and fused with his/herself on stage, just like the speaker illustrated in Figure 4.
What then makes the speaker switch perspectives? As seen in Section 4, the because X construction itself has a public function (Schnoebelen 2014; Bohmann 2016). Therefore, by starting with the unmarked mode of expression in English, the speaker indicates the publicness of the expression while avoiding the abrupt occurrence of a private expression. Then comes a private expression which indicates his/her own thought expression. As such, it naturally follows that the epistemic/speech-act causal relations are restricted. As mentioned earlier in this section, an epistemic/speech-act becauseclause introduces a new discourse independent of the main clause, even though the because-clause appears after the main clause. Thus, the speaker needs to start the new discourse in the unmarked mode.
Some examples may seem problematic for the small-discursive account. Consider the following examples: (41) a. NSF cancels new political science grants because … politics.
b. Because distance, since we know how fast light travels, if we know how far away a star is, we can also tell how old it is by knowing how long it would have taken to get there. (= (8)) In (41a), the main clause subject, the NSF (National Science Foundation), is a third person and is not identical with the speaker who jumps into the situation toward the end of the discourse. In (41b), on the other hand, the [because X] part appears sentence-initially, so it seems difficult to maintain the idea that the speaker switches his/her perspective as the discourse progresses.
Let us first consider example (41a). The actual tweet is linked to a blog written by the same person, where a detailed explanation is given, as in (42): (42) A couple of weeks before the deadline for new grant proposals in political science were due, the NSF has canceled the program, at least for this grant cycle. No explicit reason was given, but everyone knows why it happened. Back in March, Congress passed the Coburn Amendment to the Continuing Appropriations Act of 2013, which limits political science funding to research that "promotes national security or the economic interests of the United States." … .
(http://www.preposterousuniverse.com/blog/2013/08/05/national-science-foundation-cancels-call-fornew-political-science-grant-proposals/ [retrieved on July 2, 2021, underlines added]) The first section underlined in (42) "the NSF has canceled the program, at least for this grant cycle" makes virtually the same statement as the main clause in (41a). The other underlined part states the reason: "Congress passed the Coburn Amendment to the Continuing Appropriations Act of 2013, which limits political science funding to research that 'promotes national security or the economic interests of the United States'". However, as is clear from the sentence between the two underlined parts, the reason is the author's opinion. In short, in (41a), using the word politics, the author represents the NSF's intention and explains it from his own point of view. Thus, in (41a), the author starts the discourse with the narrator's perspective and then presents his own private expression while maintaining the narrator's perspective without being fused with any participant in the situation.
The other case we need to consider is (41b), where the [because X] precedes the main clause. When a because-clause appears in sentence-initial position, it is contextually presupposed. To confirm this, consider the following dialogue: (43) A: Why is the ground wet? B: #Because it has rained, the ground is wet.
In (43), speaker B's response to A's question is anomalous. The response should assert the reason for the ground being wet; nevertheless, the sentence-initial because-clause indicates that the reason is contextually presupposed. By the same token, sentence-initial [ because X] may be considered to be contextually presupposed. In other words, a sentence-initial [because X] may be used only within a context where X is established as (part of) a topic. This can be confirmed by seeing the actual context of use, as COCA allows us to check the context in which the sentence is used. Sentence (41b) appears during an interview on the performance of a space telescope, where the interviewee talks about how distant galaxies can be resolved into individual stars with the telescope. To this extent, sentence (41b) causes no abruption.

Significance of the X-element serving as private expression
Thus far, I have claimed that the because X construction has a two-layered expressive structure without considering its motivations. In this section, I examine the meaning of the two-layered structure from metapragmatic strategy perspectives. Some phenomena with the two-layered expressive structure in Japanese have been reported in the literature. First, Naya (2017) investigates an innovative use of kudasai 'please' in social networking services, as in (44), which is distinguished from its canonical use with the te-conjunctive form ore-no tooan-o tensakushite kudasai 'please correct my answer(s)'.
(44) Ore-no tooan-o tensakushiro kudasai 1. Sg-Gen answer-Acc correct. Imp please 'Please correct my answer(s).' (Naya 2017: 63) Naya regards the imperative form tensakushiro 'correct IMP ' as bearing a private function, which is turned into a public expression by adding kudasai 'please'. That is, the sentence is interpreted as expressing dual messages, as in (45): the speaker's wish as expressed by the private expression and the speaker's request as expressed by the public expression.
(45) I say to you that I strongly wish someone to correct my answers.
(Naya 2017: 74) Since the construction is used exclusively in the environment of online communication, Naya argues that its use is motivated by this environment, in which some can fulfill the speaker's wish but others cannot. Those who can correct the speaker's answers may understand the sentence as an indirect request, while those who cannot may understand it simply as the speaker's expression of a wish. Naya claims that the use of sentence (44) reflects a metapragmatic strategy of taking various users into consideration. That is, while avoiding being too polite so that the request may not threaten the positive face of the members of the social networking service community, the speaker also indicates negative politeness by indirectly requesting the correction (cf. Brown and Levinson 1987).
Second, Hasegawa (2010: chapter 5) observes that soliloquy sometimes appears in conversations. Observe the following dialogue between a teacher, indicated by H (Higher social status), and a student, indicated by L (Lower social status): Because of their different social statuses, L is required to use honorifics. Nevertheless, in her second turn, Hee, sensei demo soonandaa, she uses the plain form. Moreover, the exclamatory interjection hee and the exclamatory sentence-final particle daa indicate that the utterance is understood as soliloquy ( Hasegawa 2010: 160). As soliloquy is inserted in a conversation, this may be a case where a private expression appears within a public environment at a discourse level. According to Hasegawa, L in (46) strategically uses the soliloquy in her last turn. While she needs to indicate deference, the use of honorifics necessarily indicates psychological distancing as well. When she wishes to simultaneously express deference and intimacy, she "may temporarily quit the on-going dialogic discourse and switch to soliloquy" (Hasegawa 2010: 162). Hence, Hasegawa views this strategy as a "metapragmatic shift" by which to mitigate the psychological distancing.
In both cases, the use of a private expression within a public forum may be related to the indication of intimacy or solidarity. Although they are examples in Japanese, the same is true in English. Hirose (2013) observes that sentence (47b) conveys a greater sense of closeness than (47a). According to Hirose (2013: 24), by omitting the subject I, the speaker of sentence (47b) "is describing the situation not from the perspective of an outside reporter, but from that of an inside participant, which he imposes on the addressee; this results in the speaker bringing the addressee closer to him". Thus, Hirose points out that the subjectless sentence is not compatible with a respectful form of address, as shown in (47b). This might seem contradictory to the fact that the soliloquy in (46) indicates the speaker's intimacy to the addressee of a higher social status. However, the soliloquy in (46) serves as a private expression, which is not intended to be communicated to but to be overheard by the hearer. In contrast, because of the address terms, (47b) must be considered a public expression addressed to sir or Professor Brown. Therefore, they are essentially distinguished from each other. Hirose (2013) observes that (47b) is a subjective utterance compared with (47a), just like sentences typical of diary English, as in (35). Crucially, without the address terms, the subjectless utterance hope you like it marks friendliness, and to this extent, the use of soliloquy to be overheard in a conversation and the subjectification by omitting the first-person subject both contribute to shortening the distance between the speaker and hearer.
Generally, by using subjective expressions, the speaker brings the hearer closer to him-/herself (Hirose 2013). Private expressions are considered to be highly subjective because they may be used regardless of others (cf. Hirose 2016). In other words, the speaker as a private self reveals his or her "bare" thoughts. The preface to Hasegawa's (2010) monograph describes the act as follows: "revealing one's thoughts without interpersonal linguistic devices is a manifestation of trust". By being attracted to the speaker who reports a situation from the inside, the hearer also has to see the situation from the same perspective as the speaker-namely, the perspective from the inside. Accordingly, a sense of the involvement in the situation may be virtually shared with the hearer, yielding an effect of joint attention. Note that joint attention effects are observed not only in the context of early-stage language acquisition (e.g., Tomasello 1988) but also in certain linguistic uses among adult speakers (e.g., Cheshire 1996). It is also worthwhile pointing out a similarity between some elements in the X-slot (e.g., bare nouns) to what Quine (1969) termed observation sentences, like "Dog!", where joint attention, or empathy in the Quinean terms, is essential for the understanding thereof. Therefore, indicating closeness or intimacyand hence inviting the hearer's empathy-in this manner is a shrewd strategy to get along with others, particularly in the environment of online communication, where nonverbal information such as facial expressions and paralanguage is not available (cf. Naya 2017). 24 Given the joint attention effect, example (38b), repeated here as (48), may be further elaborated.
(48) Kinda don't wanna do this morning shift because you know. (= (38b)) As observed in Section 5.5, the discourse marker you know may function as a private expression, since the speaker can merely "pretend" that the hearer also knows what the speaker has in mind (cf. Swan 2016: 301). Thus, the speaker may use you know without considering the hearer's actual knowledge state, i.e., whether the hearer really knows the reason or not. However, to the extent that the speaker is conscious of the hearer when using you know, sentence (48) should be considered an extended instance sanctioned by the joint attention effect, which is entrenched in the because X construction. Therefore, in response to Bohmann's (2021: 33) question whether the because X construction should be treated as one unified category or a category of more or less prototypical and extended instances, this article takes the latter position. Once the construction is entrenched, apparent public expressions may appear in the X-slot, as they are coerced into private expressions by the construction (cf. coercion by override (Audring and Booij 2016)). Combined thus with the construction's joint attention effect, the meaning of the discourse marker invites the hearer's attention to the speaker's thought and hence makes it possible for the speaker to share with the hearer an unstated reason, which may or may not be different from the hearer's understanding. Because of the private nature of the X-element, the speaker does not have to care about whether he or she can appropriately communicate the intended propositional content with the expression. Another discourse marker reported in Mendes Junior and Mattos (2021) (because see?) may be explained along the same line.
The following examples reported in Bohmann's report, which contain definite noun phrases, should also be treated as extended instances sanctioned by the construction's joint attention effect.  (49a, b); crucially, however, they do not simply refer to bad acting in general and high temperature but are used as euphemisms. 25 With the definite article the in (49a) and the demonstrative this in (49b), the speakers indicate that the euphemistic senses are shared with the hearers. Euphemisms shared between the interlocutors in such a way are similar to in-group jargons in that the use of general words are understood in specific senses between the interlocutors. Brown and Levinson (1987: 103ff.) view the speaker's use of in-group jargons as a positive politeness strategy to claim "common ground" with the hearer and as indicating that "he [= the speaker] wants to 'come closer' to H [= the hearer]" (p. 103).
McCulloch's (2014) example with a definite noun in (50) also exemplifies this characteristic. 24 Unlike Naya's (2017) example of innovative use of kudasai 'please' in Japanese, the because X construction is not restricted to online communications, but as most examples have been documented from online resources, particularly from Twitter (Bohmann 2016), the environment may well affect or motivate the construction. 25 The tweets may be seen by many and unspecified people. Thus, due to the construction-external factor, the speakers may avoid using more straightforward expressions despite the constructional specification of the X-element to be a private expression. (50)

I can't pass in my homework because THE DOG (McCulloch 2014)
Example (50) is based on the well-known English expression the dog ate my homework, a phrase used (humorously) by children to make a poor excuse for their failure to turn in homework assignments. As the stale phrase is well-recognized in the speech community, the speaker can claim common ground and highlight friendliness with the hearer by using the definite noun the dog.
In summary, the expressive structure represented in (16) is entrenched in the construction by the use of its prototypical instances. 26 Because of the privateness of the X-element, the construction bears a function to indicate the speaker's closeness or intimacy to the hearer; once established as such, the construction extends its use and accommodates apparent public expressions in the X-slot (e.g., (48), (49a, b), (50)). Even though the forms of the X-elements indicate the publicness, the expressions are interpreted in accordance with the construction's function. Put differently, the construction sanctions such apparent public expressions as long as they do not conflict with the constructional function. Cappelle (2021: 28) observes that "in many of the because X cases, the speaker conveys something like 'I know that I'm presenting the reason in a very condensed, almost cryptic way, but I'm sure you know what I mean by it'" (underline added). That is, using the because X construction, most speakers believe that mutual understandings with the hearers are established about the reason. The discourse marker you know in (48) well represents this characteristic; the definite noun phrases in (49a, b) and (50) indicate that the information is shared between the interlocutors. Although the speaker may condense the information regardless of the hearer's actual knowledge state, the speaker may also indicate that the condensed information is understood by the hearer. The co-operative hearer has to decode the "cryptogram"; as a result, a joint attention effect is reinforced. As such, just because some X-elements look like public expressions does not mean that the construction extends its use to an unlimited extent. The X-elements sanctioned by the construction are limited to those whereby the speaker can indicate his/her assumption that the condensed information is shared with the hearer.

Conclusion
This article has claimed that the X-element in the because X construction represents the speaker's private expression while the whole construction functions as a public expression. To account for the two-layered expressive structure, using the notion of small discourse, I have argued that the speaker who starts the discourse with the unmarked mode of expression in English switches his/her perspective to the inside perspective as the discourse progresses. Although these basic claims are in line with Kanetani (2017), the present article has emphasized the significance of the two-layered expressive structure and given an answer to the question left open in the earlier work: what effects are brought about by the proposed two-layered expressive structure. By encapsulating private expression within the public expression, the speaker becomes involved in the situation while avoiding the abruptness. At the same time, by using a subjective expression, the speaker also brings the hearer closer to him-/herself (cf. Hirose 2013), so that the hearer feels intimacy toward the speaker (cf. Hasegawa 2010; Naya 2017). The speaker may utilize the constructional function established as such and strategically embed in the X-slot apparent public expressions such as discourse markers (e.g., you know, see?) and definite nouns (e.g. the bad acting, this fever), indicating that he or she assumes that the hearer knows what the speaker means by them. As a result, the hearer is forced to construe these expressions as being shared and to be involved in the speaker's situation construal.
I conclude this discussion by comparing the present argument with the treatment of the construction in the previous studies. Bohmann (2016) views the densification of information as a motivation (cf. Biber and Finegan 2001), while Okada (2020) characterizes the X-element as a reference point for the conceptually relevant proposition. These observations seem to account for essentially the same mechanism from different directions. As a reference point, the X-element facilitates the comprehension of the target propositional meaning (cf. Langacker 1993). To do so, the proposition is condensed into the most salient element, which may function as a "keyword" that best represents the proposition. The present conclusion that the X-element is subjective enough to attract the hearer closer to the speaker and accordingly bring about a joint attention effect is compatible with the previous views. As noted in Section 7, the X-element is similar to an observation sentence used when the speaker expresses a situation as he/she construes it. When the speaker provides the X-element as a keyword, the hearer empathetically viewing the situation from the situationinternal perspective is ready to recover the intended propositional content based on it. This conclusion also directly supports Bergs's (2018) observation that the construction is subjective. However, the hearer's commitment toward what is being said, as well as the speaker's commitment, is crucially involved.

Data availability
All data underlying the results are available as part of the article and no additional source data are required.
In this article, Masaru Kanetani presents his original analysis of clauses with because X, where X is usually a single contentful word. Sentence (1) below, from the English Web 2008 corpus (enTenTen08) is a relatively early example: (1) Sometimes we have to do all our work over again because flood.
Given its apparent novelty and idiosyncrasy, this construction has attracted some attention in the recent linguistic literature (cf., e.g., Kanetani  perhaps not yet widely known, nor is it particularly easy to grasp for the non-initiated. Kanetani's paper is commendable in its effort to explain it, not just in words but also with the help of visuals, which I thought was nice. For the benefit of the reader of this report, the essence of Hirose's proposal, which Kanetani draws on, is that languages like English and Japanese differ in how the speaker tends to see herself in a speech situation and how she communicates her thoughts to others. Put simply, in English, the speaker ordinarily sees herself objectively, as a 'public self' alongside the other speech participants. This leads to general explicitness in the use of pronouns and accessible NPs, as the speaker is supposed to be aware of the hearer in the outer world and hence should be as clear as possible. At the same time, the speaker does not need to consider her social relationship with the hearer when reporting a situation, which results in few honorifics or other politeness markers. By contrast, in Japanese, the speaker typically looks at the reported situation from within, muttering as it were to herself, which leads to the frequent omission of situationally and contextually retrievable elements (hence fewer pronouns and discourse-familiar NPs); yet, though perceiving herself as a 'private self', the speaker does need to take into account the nature of her relationship with the hearer. This necessitates the encoding of interpersonal aspects, for instance in the form of honorifics, which Japanese speech abounds in.
As said, it is against the background of this typological distinction that Kanetani treats the because X construction. He claims that, in keeping with the default communicative mode of English, the matrix clause of the construction is public but that the X-element itself is private, thereby giving a glimpse of how the speaker construes a situation for herself, much as in the default mode of Japanese soliloquy-like speech (cf. Hasegawa 2010 on soliloquy phenomena in Japanese and in English). Of course, despite expressing the speaker's thoughts, the X-element is ultimately communicated. What is new to Kanetani's present paper, compared to his previous publications on the subject, is that it seeks to provide a (meta)pragmatic, discursive motivation for this publicprivate shift within a single sentence. While I am not sure whether the term 'metapragmatic' is warranted --this concept would merit being defined and explained in the text --Kanetani's proposal makes sense to me. He argues that by slipping into the private, expressive mode, the speaker hopes to connect with the addressee. So, almost paradoxically, by using inward speech (the X-element), the speaker reaches out to the hearer, trying to bond with him.
The text is well-written, is appropriately structured, and has a well-stocked and up-to-date reference section. The argumentation is coherent and the conclusion is more than sufficiently original. As the paper does not use any new empirical material compared to earlier research by the author or by other authors, the availability of the underlying data is rendered unnecessary. However, as is also noted by Bohmann in his report, this has the downside that the examples used in the text may come across as rather conveniently suiting the analysis put forward. Nevertheless, Kanetani also discusses a few examples that seem to be problematic at first sight (namely examples (38a-b) in the manuscript) and manages to offer an explanation for them. In view of the above unmistakable merits, I recommend the paper be approved for indexing and that only minor changes be made to it.
In what follows, I will spell out some points where I may not fully agree with Kanetani or where I think some clarifications may be in order, without this entailing that the author should revise his analysis in any major way. In at least one case, my points may even lend further support to the crux of his analysis.

Specific points
(Page numbers refer to the PDF version of the article).

Speech acts in the main clause
On p. 4, Kanetani writes about (3c), reproduced below, that "[a] speech-act because-clause as in (3c) serves as a motivation for performing a certain speech act such as asking about the interlocutor's plans for the night." (italics added to Kanetani's original) (3c) What are you doing tonight, because there's a good movie on.
The italicized part seems to suggest that no speech act is performed in the main clauses of (3a) and (3b), while it could be argued that declarative clauses such as John came back and John loved her also perform a speech act, namely an assertive one. In what sense, then, is the function of the because-clause in (3c) any different from its function in (3b), where it provides a reason for making a statement (John loved her, because he came back)? Kanetani does treat Sweetser's (1990) 'speech act' (e.g. (3c)) and 'epistemic' (e.g. (3b)) uses together later on in his article (section 6), so this suggests the difference may not be that important to him.

Speech acts in the subclause
The discussion of example (6) on p. 6, reproduced below, is to my mind problematic for a couple of reasons.
(6) *He's not going out for dinner because Japanese food, his wife is cooking. (Kanetani 2019: 55) First, because the example is so utterly bizarre, I found it hard to understand what it is trying to prove. It is meant to illustrate that an unusual speech act is not acceptable in a because-clause when this is a cause-expressing adjunct that plays a role in the propositional contents of the whole clause. I am not suggesting that the author should explain more explicitly, as he does in Kanetani (2019), that this example is used to show that you can't insert a topicalization construction in the because-clause. After all, I suspect the topicalization construction is not acceptable here on irrelevant, largely information-structural grounds. Moreover, to be convincing, the asterisked example would have to be contrasted with acceptable examples in which the same or a similar because-clause has a use in the epistemic/speech-act domain. It would therefore actually be better to use a different kind of performative subordinate clause, in Lakoff's (1987) terms, one that would be more readily interpretable and information-structurally appropriate, for instance: Here again, the because-clause applies to the content domain. Empirical research is likely to confirm, though, that most performative because-clauses appear with the epistemic/speech-act function, as in the following example (where the because-clause functions epistemically): (v) Kay Daly must have eaten a whole bag of lemons because boy, is she sour! (GloWbE) In any case, examples (i)-(iv) undermine Kanetani's claim that a content because-clause always forms one speech-act unit together with the main clause.
In these examples, the speaker simultaneously provides information about why the main clause situation obtained and expresses her emotional involvement about this cause.
If it is recognized that in the content domain, a because-clause can perform its own speech act independently from the speech act of the main clause, then this should mean, as Bohmann also pointed out in his report, that the X in the because X construction may also really perform a speech act of its own.
As I see it, this should not run counter to the central claim that X is private, quite on the contrary.
To the extent that private language is more 'unfiltered' and a more direct representation of the speaker's emotions, we actually expect it to have a broader range of speech-act functions than just assertions. The frequent use of interjections and exclamative abbreviations (e.g. Ugh!, Duh!, OMG!, WTF!) can then be more readily explained.

Playing with perspectives
The use of OMG (short for Oh my God) just mentioned can also combine with a bare noun, as in this example from the web: What is interesting about this example is not just the combination of Omg and a bare noun, nor that the because X part can have its own speech-act value (note that the exclamation mark belongs to the because X segment, not to the host sequence, which isn't even clausal), but that this segment takes the perspective of the you in the title. We get the viewpoint of an imaginary speaker who can be pictured to utter something like, "I just can't listen to this song together with my mom and dad, because Omg sex!" In example (iii) above, the perspective in the performative because-clause may also be that of the third-person subject in the main clause, not that of the speaker. I think Kanetani should acknowledge this possibility. As it is, he provides an explanation for (38a), which also has a thirdperson subject but in which the because X part is (correctly) argued to represent the speaker's viewpoint. Example (vi) illustrates that the viewpoint can also be someone else's, or perhaps more correctly, the viewpoint can be one that belongs to both someone else and the speaker, which is in line with the idea of solidarity argued for by Kanetani.

The boundaries of the construction
I came across the following example, from the BNC: (vii) but the reason that I got up to ten stone is because christmas Is this an example of the because X construction? If this isn't an erroneous utterance, it has the form of a specificational copular pattern "The reason that ... is because X", which is different from the pattern discussed in the text. I do not think it would be possible to use that instead of because, while this is fully acceptable if because introduces a clause: (viii) a. but the reason that I got up to ten stone is {because/that} I've been engorging myself on mince pies, turkey stuffing and eggnog over the Christmas holiday b. but the reason that I got up to ten stone is {because/*that} christmas

Focalized because X
The examples given in (9), quoted below, are suspect: (9) a. Living people bother you because angry. Ghost make trouble only because sad, lost, confused. (COCA) b. If a society needs a large, powerful law enforcement establishment, then there is something gravely wrong with that society; it must be subjecting people to severe pressures if so many refuse to follow the rules, or follow them only because forced. (Corpus of Global Web-Based English [GloWbE]) Example (9a) is just weird. It's from a novel where a character clearly uses non-native English. Indeed, the whole context is this: "Living people always more trouble than ghost," Kwan continued. "Living people bother you because angry. Ghost make trouble only sad, lost, confused." I thought of Elza, pleading for Simon to hear her. "Ghost, I know how catch," said Kwan. "My third auntie teach me how. I call ghost-'Listen me, ghost!'-one heart speaking each other." Example (9b) is by a competent speaker of English (namely the Unabomber). Here only because forced is similar in structure to only {when/if} forced. It is close to the kind of structure that Kanetani excludes from the analysis ("Therefore, the systematic "subject + copula" deletion structure should be distinguished from Bohmann's (2016) reduced clause and eliminated from the analysis (at least for the present purposes)" (p.7). This is a reduction of only because they are forced. I wonder if there are any real examples of a focalized because X.
By the way, as for the included type of reduced clause, it might be interesting to comment on the recursivity apparently allowed by the construction when presenting (13) Observe that the underlining should probably include this morning.
In fact, though, I'm no longer sure this is an example of the because X construction: note that the main clause, too, is elliptic (I'm is dropped), and so the whole utterance might illustrate private-like language use.

Additional/Complementary semantic effects
I think the notion of "joint attention" should be explained a bit more fully: how is this established by a private expression and why don't all forms of communication, private or public, establish joint attention on what is being referred to? What does the "hearer's commitment", mentioned in the last sentence of the conclusion, really consist of? Is it a matter of being 'open' to the speaker's privately construed situation? So, how do these semantic (or pragmatic, or even 'metapragmatic' -to be explained more fully!) effects arise? For the speaker to use language that is metapragmatic, I assume she has to use language that explicitly deals with pragmatic effects, which I'm not sure is going on here. The pragmatics of language use have to be discussed via language.
My impression is that in many of the because X cases, the speaker conveys something like 'I know that I'm presenting the reason in a very condensed, almost cryptic way, but I'm sure you know what I mean by it.' Or in simpler terms, 'The reason is X. Need I say more?' There's a sense of 'nuff said' about the X. Especially if X is realized by a noun (with or without modifier), there's a sense of 'You and I know that X inevitably leads to the sort of situation reported in the main clause'.

Suggestions for minor reformulations
p. 3: "After outlining the research methodology in section 2, section 3 observes the semantic and syntactic properties of the construction." → dangling participle, so perhaps: "After outlining ... in section 2, I deal, in section 3, with the semantic and syntactic..." or some other sort of solution.
p. 3: "the description of grammar" → "the description of its grammar"? p. 3: "from a perspective of the three-tier model of language use" → "from the perspective..."? p. 4: "the real-world" → no hyphen needed, I suppose p.5: "The rhetorical question in the because-clauses in (5a, b), isn't it a beautiful day, performs a state speech act conveying that it is a beautiful day." → is "state speech act" the term we need, rather than "statement speech act" or something like that?
p. 5: "this generalization compensates for but is not incompatible with what Lakoff (1987) says" → I'm not sure I understand what is meant here by the underlined sequence; in any case, given the comments above, the paragraph in which this appears may have to be rewritten.
p. 6: "the focalization is possible for content because-clauses but not for epistemic/speech-act because-clauses" → drop "the"? p. 7: "Reduced clauses also need to be considered." → a little odd that this isn't elaborated further at this point in the text or that there is no reference to where it is elaborated.
p. 10: "English personal pronouns are primarily defined as public expressions that can be diverted to represent the private self" → this by itself is not quite clear enough p. 10, footnote 8: "Jerome" → in the original (Flemish) version, the name is rendered as "Jerom".
p. 15/16: "which makes the unmarked mode of expression in English public expression" → I found this hard to process. (Once I saw how it's to be parsed, it was crystal clear, but it wasn't at first) p. 16: "Crucially, it is not the whole construction but only its part that deviates from the norm of the English language." → suggested change: "Crucially, it is not the whole construction that is subjective but only the part that deviates from the norm of the English language." p. 18: I see both the spelling tooan and touan (right before and in example (41)) → should there be two different spellings here?
p. 18: "L in (43) strategically uses the soliloquy" → maybe indicate that this happens in the last turn In the references:

Axel Bohmann
Albert-Ludwigs-Universität Freiburg, Freiburg, Germany This article proposes a novel grammatical-pragmatic analysis of a recent innovation in English, viz. the use of because followed by a complement other than a prepositional phrase headed by of or a finite clause. This usage has come to be referred to as "because X" in the literature. Citing examples from the author's own previous work as well as other sources, and occasionally additional corpus data, the author proposes an analysis along the following lines: because X only occurs with real-world causality meanings (as opposed to epistemic causality or speech-act justification, following Sweetser's (1990: 77) ternary distinction).

○
The element in the X-slot serves as a "private expression" (following Hirose's (2000) distinction between "public" and "private expressions") and hence is atypical for a language like English, in which "the unmarked mode of expression is public expression" (p. 14, page numbers refer to the PDF version of the article) ○ This private expression is embedded within the larger structure of a public expression (the superordinate clause of because X) in order to allow English speakers to switch to a more private mode of expression in the same "small discourse" and thus not to violate the default expectation of public expression.

○
The pragmatic motivation for this two-layered structure is to create a sense of immediacy and involvement while avoiding "abruptness" that allegedly would result from starting an utterance in the mode of private expression.

○
The article is well-written and clearly structured. The sections themselves are internally coherent, and each section articulates a clear argument. One suggestion I have in terms of structure is to establish the theoretical background earlier and with more detail. Being unfamiliar with Hirose's models, I did not always feel like I could understand the criteria by which expressions are categorized into "private" and "public", nor whether these two levels exist to some extent in all utterances or are strictly categorical, mutual opposites. Given that these categorizations are central to the arguments in the paper, but that the theory behind them is, to my knowledge, not widely known, it would really strengthen the paper if the author could provide more clarity here.
The author is clearly an expert on the because X construction, which shows in the careful consideration of all the existing literature on the topic. A comprehensive range of research in three different languages is not only cited, but actively addressed throughout the paper. The author is evidently very careful in achieving a synthesis of all that has been said on the construction so far and is doing a very thorough job in this regard. At times, my feeling is that the author treats the evidence in the literature a bit selectively to suit his line of argument. For example, on p. 10 the paper makes a case that a noun in the X-slot "should be a bare noun" and cites McCulloch (2012) in support of this.  is also cited, who does present premodification of elements in the X-slot as potentially questionable, but is more careful about this claim than the paper under review gives the impression. Finally, whereas both Schnoebelen (2014) and  are discussed in detail elsewhere, both of these contain examples of premodified nouns in the X-slot that the paper is silent about. I would invite the author to address such evidence more actively rather than ignore it.
A similar point relates to method. Coming from a quantitative-empirical background, I sometimes have a difficult time accepting general conclusions about what is or is not permissible based on isolated examples, some of which are made up rather than being attested usage. Especially with a rapidly diffusing innovation, there is reason to be skeptical of native (or non-native) speaker intuitions as reliable sources of evidence. Isolated corpus examples can offer illustrative evidence, but to make strong claims about what can and cannot occur in the X-slot, more systematic evidence would be desirable. Here are a couple of claims from the paper where I think some of the limited data I collected on the topic in the past provide problematic, if not counterevidence. (Examples are drawn from the corpus data for . I would be happy to make the data available upon request): p. 5: "As an epistemic/speech act because-clause performs a speech act of its own, the because X construction is naturally incompatible with an epistemic/speech-act reason clause, because the word or phrase that appears in the X-slot cannot perform an independent speech act." → " I need to just show you the pic cause hahahahaa" (Twitter, Australia) → "I'm still gonna read all of it anyways though because why not" (Twitter, Canada) → "Watching old @user videos because why not." (Twitter, USA) → "taeyeon is so lucky because goddammit" (Twitter, Singapore) → "Kinda don't wanna do this morning shift because you know" (Twitter, USA) ○ The claim, mentioned above, that a noun in the X-slot should be bare: → "I'd turn my light off purely because the bad acting" (Twitter, Australia) → "In that case i need sex then cos this fever" (Twitter, UK) ○ "no examples with a past or third person singular present verb are reported" (fn 11, p. 11) → "idk if i am bc depends on baby" (Twitter, NZ) → "@user @user oil cos wanted it for the carve up" (Twitter, UK) → "Having to squeeze my giant feet into my moms tiny shoes because forgot shoes" (Twitter, USA) → "Can't sleep cause stomach hurts" (Twitter, USA) ○ I do not believe the counterexamples necessarily invalidate the general argument the paper is making, but they indicate that perhaps the claims need to be hedged a bit. If the author is willing to give up treatment of because X as one unified construction and instead differentiate between more or less protoypical uses, then I think the argument still holds up (although once again, a more thorough presentation of Hirose's model would help).
Despite these reservations at the methodological level, I find a lot of merit in the analysis. On p. 19, the author writes on the insertion of private expressions: "By being attracted to the speaker who reports a situation from the inside, the hearer also has to see the situation from the same perspective as the speaker -namely, the perspective from the inside. Accordingly, a sense of the involvement in the situation may be virtually shared with the hearer, yielding an effect of joint attention." This, in my view, is a very elegant summary of some of the prototypical effects because X is enlisted for. Arguments in a similar direction have been made about quotative BE like, which the author may find helpful (e.g. Coupland 2007: 186-187).
the language. See also the attested example given above with "purely because the bad acting". p. 6: "Up to this point, it has been shown that the because X construction syntactically behaves like the content because-clause construction and not like the epistemic/speech-act because-clause constructions." I think there is a confusion of levels of analysis here. Becauseclause is accorded multiple functions for the same formal construction, whereas this level of variation does not seem to be granted to because X. The argument then is that a) because X (also) occurs in contexts which are only permissible in the content reading, and b) therefore ALL of because X entails context reading. If one is willing to consider the possibility that there are multiple because Xs, then the argument becomes problematic.