Atypical Compound Verb Constructions in Hindi/Urdu The Case of Transitive Polar Verbs with Intransitive Vector Verbs

While a lot of research has already been done on defining compound verb constructions and understanding their semantic and pragmatic usages, there is still a gap when it comes to our understanding of atypical compound verb constructions where transitive polar verbs are paired with intransitive vector verbs and vice versa. In this paper we focus on atypical constructions with two intransitive vector verbs: jānā and baiṭhnā. We collate and review the different meanings that scholars have ascribed to these constructions and present our own analysis. We conclude that the main reason behind using an intransitive vector verb with a transitive polar verb is to reduce the transitivity of the action by implying reduced agency on the part of the Agent-like argument of the sentence. In addition to that, the use of an intransitive vector verb with a transitive polar verb also tells us which argument in the sentence is affected by the action from the point of view of the speaker. However, sometimes atypical constructions are also used by the speaker to express his/ her shock or astonishment over a particular event, thus imposing his/her own ideas about agency, volitionality, and affectedness on the action being observed.


Introduction 1
The compound verb construction in Hindi/Urdu (hereafter H/U) is a widely studied and discussed phenomenon. 2 However, so far, a conclusive understanding of the factors or conditions under which compound verbs are used still evades us. The semantic implications, like 'completeness' or 'perfectivity', of these constructions have already been studied and discussed in detail (see, among others, Pořízka 1967-9;Hook 1974Hook , 1978Hook , 1991Hook , 1993Kachru 1979, 158-9;Butt, Ramchand 2001;Poornima 2012). However, some scholars agree on the fact that there are other factors that govern their usage and these are more pragmatic in nature and depend on the informational context of the action (Kachru 1979;Jagannathan 198, 264-5;Hook forthcoming;Drocco, Tiwari forthcoming). Moreover, compared to the studies concerning the semantic difference between using a compound verb as opposed to a simple verb, there are not many studies on the use of different vector verbs with the same polar verb, especially when a polar transitive verb is compounded with an intransitive vector verb (Drocco 2018). Indeed, as has been pointed out by many scholars (see, for example, Hook 1974Hook , 1978McGregor 1977, 99-105;Caracchi 2002, 163-9;Jagannathan 1981, 266;Shapiro 1989, 145-60;Nespital 1997, vii-xxvii), in compound verb constructions we typically see the use of intransitive vector verbs with intransitive polar verbs and transitive vector verbs with transitive polar verbs. However, the same scholars maintain that there are certain instances where a transitive polar verb is used with an intransitive vector verb and vice versa. In this paper our aim is to study some of the instances when transitive polar verbs are used with intransitive vector verbs, what we are referring to as an 'atypical compound verb construction'. 3 This paper is divided into four sections. In section two we define both typical and atypical compound verb constructions and present a proof of the 'atypicality' of the latter. We also discuss semantic implications of atypical compound verb constructions that use jānā, 'to go' and baiṭhnā, 4 'to sit' as intransitive vector verbs when paired with transitive polar verbs. In section three, we briefly review how different points on the transitivity continuum (Hopper, Thompson 1980;Tsunoda 1985;Lazard 2002;Kittilä 2002;Naess 2007) are realized morpho-syntactically in H/U and we discuss the implications atypical compound verb constructions have on the transitivity of the clause. In particular we show that the use of atypical compound verb constructions can change the transitivity parameters for the main verb. In section four we build further from the analysis of Kachru (1979) and Drocco (2018) and discuss the idea of negative surprise and astonishment associated with the use of atypical constructions with jānā as a vector verb. 5 We conclude with our main findings in section five.

2
What Are Atypical Compound Verb Constructions?
As mentioned above, in atypical compound verb constructions 6 certain transitive polar verbs are paired with intransitive vector verbs and vice versa. Following Montaut (1991, Ch. 5;2004, 125-6) in calling these constructions 'atypical,' we think the main reason behind the use of this nomenclature is the fact that, above all, they are encountered less frequently in texts and spoken language 7 compared to their 'typical' counterparts. This fact can be demonstrated by comparing the search results for typical and atypical compound verb constructions in a Hindi corpus developed by Indian Institute of Technology, Bombay. 8 We will do this for two transitive polar verbs: karnā, 'to do') and kahnā ('to say'). The transitive polar verb karnā is normally seen compounded with denā, lenā, jānā and baiṭhnā, where denā and lenā are transitive and 4 While jānā is commonly paired with intransitive polar verbs in typical constructions, baiṭhnā only occurs in atypical compound verb constructions where it is paired with transitive polar verbs (McGregor 1977;Caracchi 2002, 168; see also Nespital 1997, 936-40).
5 See also Kachru, Pandharipane 1980, 119-21 as regards the use of baiṭhnā as vector verb. 6 We think that it is not necessary here to introduce the reader of what is a H/U compound verb construction and thus to the meaning of polar and vector verb. See the wellknown definition offered by Hook (2001, 101), recently adopted in Drocco (2018, 266) and Drocco, Tiwari (forthcoming).
7 While we have provided evidence of the infrequent usage of compound verbs in texts, the claim about that being the case for spoken language as well is based on the authors' own experience speaking, learning and teaching H/U and on the opinion of the majority of H/U scholars (see references in the text). One of the anonymous reviewers of the paper thinks that "such combinations are used to generate particular kind of nuances. I would say that in spoken Hindi the combination is used more than in written Hindi". jānā and baiṭhnā are intransitive. 9 If we look at the frequency of the occurrence of these pairs in the simple perfect, kar diyā, kar liyā, kar gayā and kar baiṭhā in the corpus we get the following results: 10 ii After excluding 29 instances of the conjunctive participle stem + kar baiṭhnā ā, for example, ākar baiṭhā (came and sat).
iii After removing occurrences of the conjunctive participle and the verb mukar jānā, and 4 strings that were wrongly pulled up in the search results. iv After excluding 16 occurrences of the conjunctive participle and one occurrence of the plural kar baiṭhīṁ.
As is evident from these search results, kar diyā and kar liyā are encountered over 60 and 20 times more (respectively) when compared to kar gayā and over 200 and 70 times more compared to kar baiṭhā. While these results are based on one corpus, they are indicative of the relative infrequency of atypical constructions.
The two compound verbs kar jānā and kar baiṭhnā commonly occur in the perfect because the specific meanings they convey, the realization of 'going beyond' or 'making a mistake', 11 are probably possible only after the fact. However, we are also including search results for the simple future for comparison. Also, note that the instances of kar denā and kar lenā (cf. table 2) in the simple future are also relatively much fewer. This could be because one of the semantic implications of using compound verbs is emphasizing the perfective aspect (cf. § 1), which, while possible in the simple future (it will be done/completed), is less frequent. 9 For differences in meanings amongst these different pairings see Snell, Weightman ([1989] 2010, Hook (1974Hook ( , 1978 and Drocco, Tiwari (forthcoming).
10 We have removed the results for the masculine plural simple perfect forms as the search results were also returning strings for masculine singular forms on the corpus.
11 As we will see below, this is the specific meaning that jānā and baiṭhnā add respectively when adopted as intransitive vector verbs of transitive polar verbs.  We can repeat the same search for kahnā, 'to say' which is also transitive. The polar verb kahnā is normally only paired with denā, as a transitive vector verb. This is due to the fact that the object of the verb 'to say', which are the things said, are meant for the listener.
Comparing the occurrences of kah denā, kah jānā and kah baiṭhnā, we get the following results in simple perfect (cf.   (Nespital 1997, 716). Some of the search results have this meaning instead of 'to say', but this meaning of kahnā is not relevant for the present paper. ii After excluding 3 results that did not contain the search string, one result that was repeated and one other where kah gayā can be seen as a verb in combination, kah-kar gayā (see footnote 13). Also, as mentioned in footnote 18, some of the search results also had kahnā in the sense of telling someone to do something. iii After excluding 2 results that did not contain the search string. iv Excluding one occurrence of the plural kah dīṁ, one repeated search result and three results where kah gaī can be seen as a verb in combination, kah-kar gaī (see footnote 13). v After excluding 8 occurrences that did not contain the search string and one occurrence of the plural kah baiṭhīṁ.  Thus, we can see that atypical constructions occur less frequently with respect to typical compound verb constructions.
One of the reasons why these constructions are less frequently encountered is because they encode a very specific meaning. 12 Let us consider the vector verb baiṭhnā, 'to sit' for example. It is adopted in this function when the Agent-like argument of the sentence commits (or might commit in the future) what is often seen by the speaker as an action with an undesirable result, that is to say, a mistake or something done foolishly (Snell, Weightman [1989] 2010McGregor 1977, 102-3;Kachru, Pandharipande 1980, 119-21;Nespital 1997, 936-40;Caracchi 2002, 168 While all scholars agree on the semantic nuance expressed by the vector verb baiṭhnā we just discussed, the same is not true for jānā, 'to go' when it is compounded with transitive polar verbs. 13 Accord-12 Another reason that can partially account for the infrequent usage of atypical constructions, especially with the vector verb jānā, is that the set of polar transitive verbs that can form meaningful pairings with jānā is limited. For this last statement see, for example, Nespital 1997, 554-9 and Drocco 2018. More research is needed to understand the precise reasons behind this fact.
13 Another interesting instance of atypical usage of jānā as a vector verb is often seen when it is used along with kahnā. The different meanings discussed in this paper that are typically associated with using jānā do not explain all the instances where kah jānā is encountered. For these cases sometimes it is possible to look at kah jānā as a verb

3.
vah kām kar ga-y-ā. (*us ne) 3SG.DIR work.M do go.VV-PERF-M.SG 'He did the work unintentionally'. 14 (adapted from Pandharipande 1981, 169) However, as we just said, not all of them agree on this additional semantic implication of jānā when used as intransitive vector verb of transitive polar verbs (see, among others, Nespital 1997, 556-9;Montaut 1991Montaut , 2004. For example, Shapiro simply maintains that: in combination (i.e. kahkar jānā) and not as a compound verb. However, it is important to point out that this is not always the case. To illustrate this last point, see the examples in (a) and (b)  In sentence (a), if we analyze kah ga-e haiṁ as kahkar ga-e haiṁ, 'have said and gone', it does not have the same meaning. The latter sounds as if the subject said something and then left (or passed away) intentionally in that particular order. But in reality, the original compound form, by which we mean the meaning expressed by kah ga-e haiṁ in (a), is less volitional or planned when it comes to how the two actions kahnā and jānā were temporally realized. These constructions are typically used when the Agent-like argument is no longer around to confirm what he/she said or is deceased. Other verbs like karnā and banānā are also sometimes seen used with jānā in this way.
In sentence (b), however, kahkar ga-ī haiṁ comes close to the meaning of kah ga-ī haiṁ. Interestingly, sentence (a) is semantically in between 1. 'Verbs in Combination', where both verbs retain their full meaning and 2. 'Compound Verb Construction', where the vector verb loses its original meaning. While more research is needed on this topic, we believe that this could be one of the ways in which compound verbs slowly evolved and became grammaticalized (cf. Hook 1991).
14 This translation and interpretation of unintentionality expressed by karnā + jānā is from Pandharipande (1981, 169). We do not fully agree with this interpretation (see below). This is also the view of one of the reviewers of this paper.  Shapiro 1989, 146) Starting from these different and inchoate ideas about the use of jānā as a vector verb of (some) transitive polar verbs, Drocco (2018, 276-82) advanced the hypothesis that this particular construction encodes the negative attitude of the speaker/narrator about some unintended, unwanted, unpleasant, and irrevocable effect/result of the action expressed by the transitive polar verb. He explains in detail this particular use of jānā when compounded, especially, with ingestive verbs khānā, 'to eat' and pīnā, 'to drink'. Here is an example from his paper (taken originally from Montaut 2004):

5.
maiṁ zarurāt se zyādā khā ga-y-ā, peṭ 1SG.DIR need than more eat go.VV-PERF-M.SG belly phūl ga-y-ā, jhapkī lag gaī. swell go-PERF-M.SG doze be attached go.VV-PERF.F 'I ate (gulped) more than needed, my belly swelled up, I fell asleep'. (adapted from Montaut 2004, 126) According to Drocco (2018, 278), in (5) jānā is used as an intransitive vector verb for the transitive polar verb khānā, because the speaker's intention (a 1st person singular pronoun) is to emphasize a particular negative affect of the act of eating too much on the Agent-like argument, that is, on himself. In particular Drocco continues giving the following explanation: With the use of zarurāt se zyādā, "more than needed", the action of khānā, "to eat" results in an unintended, unwanted, unpleasant, unexpected, and irrevocable effect. This result is in fact explicitly mentioned in the same sentence with the expression peṭ phūl gayā, jhapkī lag gaī, "(my) belly swelled up, (I) felt sleepy". (2018,278)  Therefore, it must be noted that while previously scholars have mainly emphasized the semantic implications of the atypical usages of jānā and baiṭhnā, we think that it is also important to pay attention to the fact that these meanings are associated to the speaker's perception of the event (who, sometimes, can be the Agent-like argument of the sentence, as in the last example). This is because the speaker also has at his/her disposal at least one alternative phrasing of the event which does not use this construction 15 and thus the fact that the speaker chooses to use an atypical construction also shows his/ her attitude or opinion about the action in question. While this may not appear to be an important distinction to make at this stage, we will discuss its implications in section four.
To sum up, we have seen that jānā and baiṭhnā, when used with transitive polar verbs, have very specific meanings and that, in part, could explain why they are encountered less frequently in the language. Also, as we have reported above, while there is a general agreement among scholars when it comes to the semantic implications of baiṭhnā as a vector verb, this is not the case with jānā.
In the next section we will try to understand how the different semantic meanings associated with the usage of jānā and baiṭhnā in atypical constructions are related to the effect they have on the transitivity parameters of the clause/sentence they are used in.

Reduced Transitivity of Atypical Compound Verb Constructions
While the grammatical categories of 'transitive' and 'intransitive' are centuries-old, 16 the linguistic understanding of what 'transitivity' is has become more refined in the decades after Hopper and Thompson's seminal paper (1980). Linguists now view 'transitivity' as a spectrum, where a certain verb under a certain construction is 'more' transitive, while the same verb under a different morphosyntactic setting can be 'less' transitive (Tsunoda 1985;Lazard 2002;Kittilä 2002;Naess 2007, Ch. 2). Different languages have different ways of encoding information about transitivity seen as a scalar notion. In H/U we typically see it expressed in one of the following main ways: 15 For example, another way in which (1) can be reported is: usne kaisā kām kar diyā.
In this example the transitive polar verb karnā, 'to do' is compounded with the transitive vector verb denā, 'to give' and, for that reason, the Agent-like argument is followed by the ergative postposition ne. For our explanation on why baiṭhnā, instead of denā, is used in (1)  1. The use of the postposition ne in the perfect with transitive verbs (Hock 1985, 250;Bashir 2016, 450-3); 17 2. Indirect constructions with the subject followed by the postposition ko that are used for encoding the experiencer (for an overview on this topic see Masica 1991, 346-56 andHock 2016); 3. The use of passive derivation in negative sentences to express the inability to do an action by the agent (see Masica 1991, 356-8, but especially Pandharipande 1979); 4. The use of intransitive/anticausative verbs with the Agent-like argument expressed by adding the postposition se, as an Instrument through which the action happened; 18 5. The use of compound verbs (Drocco 2018). 19 The first three features just listed have been widely studied in H/U and Indo-Aryan linguistics in general. On the contrary, the transitivity implications for compound verbs have so far only been studied by a few scholars. 20 Thus, as said above, the goal of this section is in part to fill this gap.
In the previous section we mentioned that the speaker's choice of using a polar transitive verb compounded with an intransitive vector verb is linked to his perception of the event. In these cases, the speaker perceives the result of the action done by the Agent-like argument of the sentence as either: • unwanted, unintended contrasted against what is wanted, intended in accordance with a socio-cultural norm or a personal habit, or • a mistake made by the Agent-like argument of the sentence.
This specific attitude of the speaker is realized in H/U by means of compounding transitive vector verbs with jānā and baiṭhnā respectively. The resulting atypical constructions taken into examination here are very interesting, because they change the syntactic status 17 Some linguists question the view that ne can be seen as a marker of transitivity (see, among others, Mohanan 1994).
18 An example to illustrate this usage is the following sentence found in the web: ek din mujh-se kap ṭūṭ gayā, 'One day the cup broke 'through' me'. This kind of meaning can also be expressed by the simple verb ṭūṭnā and similar verbs. Cf. https://www.bhaskar. com/news/JHA-RAN-HMU-MAT-latest-ranchi-news-034003-736146-NOR.html.
19 A small category of verbs in H/U can be used both transitively and intransitively, like khonā, INTR 'to get lost', TR 'to lose (something)' and badalnā, INTR 'to be changed, to change', TR 'to change (something)'. When these verbs are paired with jānā, the resulting compound form becomes definitively intransitive (Snell, Weightman [1989] 2010Caracchi 2002, 165-6).
20 And this is true also for the use of intransitive/anticausative verbs with the Agentlike argument followed by the postposition se. of a transitive construction to an intransitive one, even if the result is still a bi-actant construction that can be translated as a transitive clause in other languages such as English. According to some scholars (see, among others, Pandharipande 1981, 168-9 andKachru 1981, 182-3, 186-7), the intransitive status of this particular construction is proven by the fact that the use of an intransitive vector verb with a transitive polar one: • changes in the perfective the canonical case-marking of an original ergative construction with the Agent-like argument followed by the postposition ne and the Object-like argument in agreement with the verb (if not followed by the ACC postposition ko) to a non-ergative one; • moreover, the passivization of this particular type of construction is not possible.
So, the main goal of this section is to answer the following question: Given the fact that using an intransitive vector verb with a transitive polar verb changes the transitivity parameters for the latter, how does this fact relate to the speaker's perception of the underlying action? In particular, how does an atypical construction with jānā and baiṭhnā bring about the different semantic nuances associated with their use as previously discussed in section 2. The sentence in (6) with vector verb jānā is a good starting point to answer these questions: In (6), the Agent-like argument who probably does not eat four/five rotī normally because that may be too much for him, ended up eating these many because he found the combination with the sauce tasty. When jānā is used with transitive polar verbs, it often gives the sense of the Agent-like argument 'going beyond' and this can mean, specifically, 'going beyond a certain norm or habit', thus performing an action or realising a result that was, at first, unwanted and sometimes even unpleasant. It is important to point out and to add that, in many instances, this can be linked by the speaker's perception that the Agent-like argument 'lost control'. This is clear in sentence (6) 21 Indian bread. where, while the Agent-like argument was intentionally eating, he lost control over how much he wanted or originally intended to eat, as he came under the control of his senses and was no longer completely in control of the action he was performing. Thus, when jānā is compounded with a transitive polar verb, we see that the reduced grammatical transitivity of the construction is, in this case (cf. 6), linked to the Agent-like argument's reduced agency (as discussed in § 2). However, the loss in agency perceived by the speaker, when using an atypical construction with jānā, could also result from the fact that the Agent-like argument is forced to do something either by another person or by the circumstances in which he finds himself. This can be seen in one of the examples mentioned in Drocco's paper (2018, 279):  (Premchand [1936(Premchand [ ] 1966 Thanks to the fact that this example is taken from a bigger text, we can look at the context in which this sentence was said. In (7) the Agent-like argument (unhoṁne, ve) is Paṇḍit Oṃkārnāth. Because of his religious observances he has never had a glass of liquor before. But in this particular situation, when he is in a modern setting, he cannot refuse it. Thus, in the context of chapter six of Godān (Premchand [1936(Premchand [ ] 1966, where the sentence is reported, the Agent-like argument can be seen as not completely volitional while engaging in the act of drinking liquor, because he was forced by the circumstances to do an action that was unpleasant and not 'right' according to his convictions. The aspect of 'losing control' as discussed in the case of sentence (6) above, is also seen when jānā is paired with other transitive polar verbs like kahnā, 'to say'. The latter is often used with jānā when the Agent-like argument is being quoted by the speaker as having said something in excess of perhaps what they had intended to say. Consider the following example: In (8) the Agent-like argument ended up saying something they should not have. However, in this particular example the impact of doing that is quite low. But look at example in (9) In (9), in contrast to (8), it is clear that the Agent-like argument, who is also the speaker of the sentence, realizes that he said something he probably should not have. However, unlike (8) where the Agentlike argument was let off the hook by the speaker, in (9) we can see that he is filled with remorse.
This last observation leads us to another important and correlated aspect associated to the use of intransitive vector verbs like jānā or baiṭhnā with a transitive polar verb. In an atypical compound construction, the focus of the speaker is often not on the resulting state of the Object-like argument, as normally happens in typical transitive clauses, but on the consequences of the action on the Agent-like argument. Thus, the transitivity of the sentence is 'lower' not only because of reduced agency, but also because the action affects the Agent-like argument. Indeed, this is obvious in (9), where the speaker, who is himself the Agent-like argument, regrets what he has done and possibly foresees a negative consequence linked to that in the future.
However, it must be noted that for ingestive verbs (Naess 2007, Ch. 4), where the normal effect of the action is already on the Agent-like argument, the impact of using jānā instead of lenā as a compound verb, is to make the effect on the Agent-like argument 'neutral' or even more 'negative' (see sentence in 6) as opposed to 'beneficial' which is often the implication of using lenā. This is evident in (10)  In this sentence, the 'mouse', who is also the speaker, is not so much concerned by the benefit the cat might derive from eating him. He is rather worried about the fact that he might get eaten and it would not be good for him. Consequently, using jānā with khānā in this sentence takes the focus away from how the action of eating would impact the Agent-like argument, the 'cat', positively.
Thus, we have seen that in cases where jānā is used as an intransitive vector verb along with a transitive polar verb, the speaker is able to express his perception of the action and/or the result of the action, that it was originally unwanted, due to the fact that jānā implies a loss of agency either because the Agent-like argument was forced to perform the action or because he lost control during the performance of the action.
Let us now consider the case of baiṭhnā. As we pointed out in section 2, baiṭhnā is used as a vector verb when the speaker thinks that the Agent-like argument made a mistake (sentence (1) above). The reason behind the fact that this point of view of the speaker is verbalized through the use of the intransitive vector verb baiṭhnā is because, according to the speaker, the Object-like argument of the sentence is sometimes the result of an unconscious state of mind of the Agent-like argument. As a consequence, the speaker confers reduced volitionality 26 to the Agent-like argument, as mistakes are (hopefully) non-intentional. Let us see this with an example:  In this sentence, the Agent-like argument (the weavers) seem to have lost their ability to think rationally (they have gone mad) and have already demonstrated that they could do something as extreme as killing someone. So, the speaker is worried they might do something even more extreme in this altered mental state which they or others would regard as a mistake later. Hence, according to the speaker, the Agent-like argument is not fully aware of the actions they are currently performing and might perform in the future. 28 To better understand the different possible attitudes of the speaker towards the action performed by an Agent-like argument of a clause compare the constructions below: 29
In (12) the speaker chooses to use the transitive vector verb denā because according to the speaker, regardless of whether the Agent-like argument was aware or unaware of his mistake while making it, he is still responsible for it to an extent because he caused it. It is possible that the speaker thinks that the Agent-like argument causes the mistake intentionally. On the contrary, in (13) and (14), the speaker chooses to use the intransitive vector verbs jānā and baiṭhnā, because he wants to emphasize the fact that the Agent-like argument made the mistake unintentionally and perhaps he regrets it. It is also important to note that the volitionality in these cases is reduced 27 https://bit.ly/2DpAx0h. only when we are looking at the mistake as the direct object. It is not a comment on the state of the mind of the Agent-like argument with respect to the work he was intentionally performing. He could have been engaged in his task with full volition and awareness, but the same cannot be said about the mistake he made, because that was not intentional (Kittilä 2005).
While (12) is in contrast to (13) and (14) when it comes to the Agentlike argument's parameters of volitionality, we can see that (13) and (14) also differ with respect to how the speaker perceives the action impacting the Agent-like argument. In (13), where an anticausative construction is used, the focus of the speaker is on the mistake that is made, while in (14), the focus shifts back to the Agent-like argument and the impact the mistake might have on him. While both the sentences are reporting that a mistake was made, and to reiterate, this mistake was unintentional, in (14) the speaker is more concerned about the impact the mistake has (or could have) on the Agent-like argument or somebody else. If he realizes its impact, he is already feeling very sorry (or should feel so according to the speaker). If he doesn't realize it yet, he most likely would in the future and would probably feel contrite at the point. So, while in (13) he is just an 'instrument' that leads to the mistake coming into existence, in (14) he does have to shoulder some responsibility for it and 'pay for it' in some sense. Let us look at another example that brings out this aspect of baiṭhnā:
'He blurted out (kah baiṭhā) in a hurry -I want that we elope from here and go to some other place and get married there. The thoughts that Rajiv wanted to express, he expressed them, but the introduction to these thoughts that he had come prepared with, who knows where that got lost? His thoughts proved to be completely dry and meaningless and that was still okay; however, they also felt uncouth and strange when he (himself) heard them'. (adapted from Avguṁṭhan by Ravindranath Tagore) 30 As is clear from the context of the sentence, the speaker is reporting that the subject ended up saying something contrary to what he had planned, and this caused him a lot of embarrassment and perhaps also regret. Thus, in this sentence, we can see both the aspects of loss in volitionality and affectedness of the Agent-like argument coming into play.

323
A final remark is in order about the volitionality of the Agent-like argument in atypical constructions with baiṭhnā. While the examples we have seen so far clearly have Object-like arguments that can be classified as a mistake, often, whether something is/was a mistake or not can only be determined from the context or after the fact. The word 'love' (pyār), which is the Object-like argument in (16), need not always be a mistake and, in fact, is not so, when used in an affirmative sense as in maiṁ tumse pyār kartā hūṁ, 'I love you'. However, due to the fact that it is used along with baiṭhnā, and because of the context (it was one-sided), it is clear that the speaker and the Agentlike first person argument thinks of this as a mistake. Also notice that since this is a narration, it is quite likely that he concluded that it was a mistake not in the moment of falling in love, but only later when he realized that it was one-sided. Earlier in this section we established that with jānā, the agency of the Agent-like argument is reduced from the point of view of the speaker during the performance of the action itself, either because he/she loses control (cf. example in 6) or because he/she is forced (cf. example in 7). However, in the case of baiṭhnā, the lack in volitionality with respect to the result of the action is sometimes ascribed after the fact, in particular, after looking at the result of the action in its broader context -this is the case in sentence (16) where the narrator thinks that being in onesided love is foolish and he reinterprets the action of falling in love in that light and concludes that it was a mistake and that the result was not what he intended. Hence, many times it is translated into English using adverbs like "by mistake" or "foolishly" (Shapiro 1989;Snell 2016, 225) emphasising the fact that the Agent-like argument was not fully aware of the result of his actions. Let us look at this final example that illustrates this difference:

17.
rām ne is bār uskā mārg rok liyā aur bolā -"vākya purā kījie na -yūṁ kisī bhī bāt ko adhūrā choṛna ṭhīk nahīṁ hai". "rahne dījie. āp kaheṁge -gaṁgā bahut adhik boltī hai". "viśvās kījie -mujhe āpka adhik bolnā bhī acchā lagegā. kahie na". This example is interesting because the Agent-like argument is deliberating whether he has made a mistake by speaking out his heart. From the context we can see that he was in control of the situation to a great extent and he knew that he wanted Ganga or himself to express certain feelings. However, having done that and after looking at Ganga's reaction, he is now wondering if that was appropriate or not. Thus, the use of kah baiṭhnā shows that Ram is wondering if what he just said to Ganga was a mistake. In conclusion we can say that with atypical constructions, where transitive polar verbs are used along with intransitive vector verbs, there could be a reduction in the overall transitivity of the construction due to: 32 The use of the present perfect, as per Montaut's analysis (2006), also confirms the fact that the present perfect can have experiential and resultant connotation. Ram in this case has integrated the possibility that he made a mistake into his information set and is not astonished by it.  Drocco (2018, 276-82) concluded that in many instances similar to (18) even if there is volitionality on the part of the Agent-like argument and he/she in no way is affected by the action, jānā can still be used. Indeed, according to the same author, atypical constructions with the vector verb jānā are often used when the speaker, as a consequence of the fact that the Agent-like argument 'went beyond', expresses a negative reaction (astonishment, shock, etc.) towards a particular action or event. This 'negative reaction' can be explained as a 'negative surprise' or also as a 'shock'. 36 Therefore, it is not always clear if 'volitionality' and 'affectedness' are the only factors correlated with the use of jānā in an atypical construction.
In general, and also to anticipate some of our conclusions, we can say that very often atypical constructions are used for talking about 'unexpected events'. By 'unexpected events' we mean events that should not have happened according to the speaker. In (18) for instance, cannibalism is seen as something not morally correct and desirable. One of the ways in which the speaker/writer (and also the reader) can make sense of an event like this is by reducing the agency of the Agent-like argument while reporting it. Even though the Agentlike argument acted consciously with respect to the goal he wanted to achieve (by taking the consent of the person he ate), in a broader sense, when his actions are judged in contrast to another 'normal' or 'mentally sound' agent, he is classified as being 'out of his mind' by the speaker.
Thus, it seems that there is some other parameter associated with this use of jānā which impacts the transitivity category in H/U and which is apparently not included in Hopper and Thompson's transitivity parameters. This parameter seems to be normative in nature, where the speaker, after looking at an action, asks himself whether a 'normal' agent would act in this way or not. This helps the speaker to somehow rationalize an extreme event which perhaps was too shocking to understand. Let us look at another example from Drocco (2018, 282) (21) is the informational context. Hook (forthcoming), Jagannathan (1981) and Drocco, Tiwari (forthcoming), have talked about how the use of typical compound verb constructions is linked to prior information. 42 While in (20), it is hard to say without the context whether there was something expected about the action, 43 prime facie, the use of a typical compound construction has the inverse effect of making the event sound like it was somehow expected. In (21) however, the simple verb does not have any such implication and hence increases the element of surprise over (20). However, compared to (19), (21) expresses less astonishment regarding the event. This is because, as we have seen in sections 2 and 3 above, when jānā is used as a vector verb with transitive polar verbs, it often shows that the speaker regards the event as 'unexpected' because it goes against a particular habit or norm or it is seen as excessive. Thus, khā jānā already carries this flavour of being 'unexpected' and this implication can be carried over to cases of unexpected events in general even when they may have been volitional and transitive and the basic transitivity parameters were not in question. 44 Thus, based on the analysis above, we can say that the element of surprise/astonishment increases in the order below: 1. khā liyā (cf. 20) 2. khāyā (cf. 21) 3. khā gayā (cf. 19) This does not mean that the three constructions are interchangeable in the sense that the speaker can choose one over the other purely based on how much surprise he wants to express. On the contrary, the informational context in which the event takes place and who he is speaking to, also impacts this choice.
42 In the context of any sentence (or utterance), 'prior information' is that state of the world in which the speaker and the listener both already have some information about the event that is being spoken of in the sentence and they both know this to be true for the other person as well: i.e., the speaker knows that the listener is aware of the event and vice versa. In H/U, 'prior information' is encoded by the use of compound verb constructions with vector verbs jānā, denā and lenā restricted as follows: 1. jānā is used with intransitive verbs; 2. denā is used with transitive verbs; 3. lenā is used with transitive verbs and some intransitive verbs. Therefore, the concept of 'prior information' is applicable for 'typical' compound verb constructions and not for 'atypical' constructions.
43 Interestingly, the article from which (19) is taken, opens with (19) as the headline and reiterates it with khā lenā in (20), which is similar to (19), but, as already pointed out, with a typical compound verb construction. This could be because once the reader has read the headline, what follows is expected by him/her.
44 But perhaps should not be so from the point of view of the speaker.