Parenthetical I’m telling you as a marker of private evidence

We present an analysis of parenthetical uses of the English expression I’m telling you as a discourse particle, i.e., an expression that conveys information about the epistemic states of discourse participants with respect to the propositional content of an utterance (Zimmermann 2011). The analysis connects I’m telling you to other discourse particles that mark the speaker’s assumptions about whether the (evidence for the) asserted proposition is shared knowledge between the speaker and addressee and whether or not the (evidence for the) proposition is “verifiable on the spot”, e.g. German ja (Kratzer 1999, 2004), Mandarin de (Soh 2018).

As Zimmermann (2011) notes, discourse particles convey "information concerning the epistemic states of the speaker, or her interlocutors, or both, with respect to the descriptive or propositional, content of an utterance" (see also McCready 2012).The analysis that we present for parenthetical I'm telling you (henceforth PITY) fits snugly into this characterization.We argue that it marks an epistemic difference between the evidence available to the speaker and the evidence available to the addressee.More specifically, PITY contributes use-conditional content to the effect that the speaker has private evidence for the proposition expressed by its sentential argument (henceforth the prejacent), private evidence being information available to the speaker but not the addressee.
Usage manuals describe PITY as a type of surprisal marker, the description in (2) being fairly typical. 1   (2) I'm telling you is used to emphasize that what the speaker is saying is true even though it may be surprising, or difficult for the addressee to believe.
We agree that PITY is often used in this way, but contend that (2) is unable to account for the full range of uses of PITY.Consider (3), for example.
(3) A and B are discussing B's new winter boots: a. A: Wow!Those are top-of-the-line boots.You must have paid a lot for them.b.B: I'm telling you, they're worth every penny during Minnesota winters.
Because A acknowledges that the boots are "top-of-the-line" in addition to their high cost, it should hardly be surprising to A that B believes they're worth their cost, nor should this be difficult for A to believe.Intuitively, rather than emphasize the truth of the proposition that the boots are worth their cost, B emphasizes the evidential basis of her assertion, namely her personal experience, which is information that B has privileged access to.Given these initial observations, it is the main goal of this paper to provide an account of the conventional semantic contribution of PITY that is able to account for the full range of its uses.Sections 2 to 4 develop our analysis of the semantic contribution of PITY.Section 2 explores the discourse restrictions on the appropriate use of PITY.Based on these observations, Section 3 provides a first attempt at an analysis of the semantic contribution of PITY.Section 4 discusses context specific uses of PITY and their relation to the analysis in Section 3. Section 5 discusses the broader implications of our analysis through a comparison with the German modal particle ja.Section 6 concludes the paper.
2. Contextual restrictions on the use of parenthetical I'm telling you.The appropriateness of parenthetical I'm telling you depends on certain features of the utterance context.The current section explores these contextual restrictions in detail, arguing that PITY is sensitive to the availability of evidence in the utterance context for the prejacent proposition and, furthermore, cares about which discourse participants have access to that evidence.

INFORMATION AVAILABLE TO BOTH THE SPEAKER AND ADDRESSEE.
If both the speaker and addressee have access to evidence in the utterance context, at utterance time, for the prejacent proposition, then the use of PITY is infelicitous.For example, in the context provided in (4), both A and B have access to sensory information about the weather and A's use of PITY is inappropriate (4a).The infelicity of (4a) is not a result of the fact that A is telling B something in this context that B likely knows already, for note that the plain assertion in (4b) is felicitous despite compelling contextual evidence for the prejacent proposition.While this evidence may be compelling enough that both A and B believe the prejacent is true and believes that each other believes it, the prejacent is not yet common ground.A's unadorned assertion in (4b) is understood as a proposal to make the prejacent common ground, the default dynamic effect of assertions (Farkas & Bruce 2010, Krifka 2015).
The difference in acceptability between (4a) and (4b) must therefore be traceable to the contribution of PITY, which appears, at least in part, to be sensitive to whether or not there is evidence in the utterance context for the prejacent.As a preliminary observation, we assume that PITY is infelicitous when there is contextual evidence available to both the speaker and addressee for the prejacent proposition.

INFORMATION AVAILABLE TO THE SPEAKER BUT NOT THE ADDRESSEE.
Our second observation is that the use of PITY is appropriate whenever the speaker's assertion is based on information that she believes is available to her but not to her addressee.In other words, the speaker's basis for the assertion is private.In the context established in (5), for example, A has information that she believes B lacks but which she believes is relevant to B in some way.In this case, the information is relevant to B's decision of whether or not to approach the woman.Both PITY and the plain assertion are felicitous in this context, as shown in (5a) and (5b) respectively.
(5) A knows that the woman across the room, C, is married and believes that B doesn't know this.
B is clearly thinking about approaching her.a.A (to B): I'm telling you, she's married.b.A (to B): She's married.
According to our intuitions, (5a) offers a more forceful warning against approaching C than does the plain assertion in (5b).There is a potential implication associated with (5a) that B may be inclined to doubt the truth of A's claim that C is married or may not recognize its relevance.The use of PITY conveys that A presumes to know something that B does not.
Along similar lines, the prejacent of PITY often expresses a proposition that relates to the speaker's own actions (6), beliefs (7), physical or emotional states (8a), preferences (8b) or intentions (8c).2Speakers stand in a privileged epistemic relation to such propositions, having direct access to the content of their own physical and mental states, access that addressees lack.The overall effect of using PITY in these cases is minimal, perhaps simply serving to emphasize the prejacent, as suggested by (2).For example, in (6a Propositions relating to personal tastes are matters of opinion and since opinion-holders once again stand in a privileged epistemic position to their opinions, and to the basis of those opinions, relative to addressees, we expect the use of PITY to be appropriate with statements containing predicates of personal taste, like tasty or fun.This is, in fact, what we observe in (9).( 9) Looking up from a hearty bowl of soup at the dinner table : I'm telling you, this soup is tasty!
In this context, the other diners may have tasted the soup and formed their own opinions about its tastiness on the basis of their personal experience and preferences.PITY is still felicitous because the speaker's evidence for the prejacent proposition is HIS personal experience and preferences, which the other diners lack access to.Thus the discourse effect of PITY in ( 9) is to emphasize the speaker's opinion that the soup is tasty.
When matters of taste are debated, statements about taste often take on a more objective feel and what information is available to speaker and addressee becomes important.In (10) B uses PITY to attempt to convince A that the soup is (or will be) tasty to A (or to people like A and B).Importantly, in this context A cannot have already tried the soup.We argue that this is because B is not in a privileged epistemic position with respect to the prejacent proposition that the soup is tasty to both A and B. Since this proposition depends at least in part on the personal opinions of A, B cannot have evidence for it that A lack, especially since A has indicated that they believe it to be false in (11b).3

WHAT INFORMATION IS "AVAILABLE" IN AN UTTERANCE CONTEXT?
The discussion in the previous two sections attempts to establish that the use of parenthetical I'm telling you is felicitous when there is evidence in the utterance context for the prejacent proposition that is available to the speaker but not to the addressee.But what does it mean for evidence to be available in an utterance context and how do we know who this information is available to?In particular, what information in the immediate context of utterance is available to the speaker but not to the addressee?Is information conveyed by extra-linguistic stimuli in the immediate discourse context available to every agent in that context, whether or not they are attending to that information, or must an agent be consciously attending to contextual information for it to count as available?We offer a preliminary discussion of these questions in this section, returning to them again in section 5 when we relate our analysis of PITY to related phenomena.As noted in section 2.1, PITY is inappropriate when there is sensory evidence for the prejacent proposition in the immediate utterance context (see example 4, which is repeated below).The infelicity of PITY in this context suggests that certain sensory information, for example McCready argues that sentence final man is a strong assertion operator that forces update of the addressee's information state with the prejacent proposition ϕ.In case ϕ is inconsistent with the addressee's beliefs, this update is preceded by appropriate belief revision.The analysis captures the insistent nature of (ic).information about the ambient temperature, is perceivable, and thus available, to every agent in an utterance context.
(4) A and B are visiting a tropical island for the first time and it is extremely hot when they arrive.
a. A (to B): #I'm telling you, it's hot here.
However, contextual evidence need not be actually perceived in order to count as available.In ( 12) and ( 13), adapted from Lindner (1991), the addressee is not attending to the available visual evidence for the prejacent proposition.Nonetheless, PITY is infelicitous.Presumably, the addressee could attend to the relevant external stimulus, and thereby gain access to it, with minimal effort.The use of PITY is once again felicitous, however, if the addressee demonstrates an inability, or refusal, to attend to the relevant stimulus ( 14).Such denials explicitly mark information as unavailable to the denier, thereby licensing the use of PITY.
(14) A is climbing the stairs in front of B: a. B: You've got a hole in your sleeve.b.A: Yeah right.This is a brand new shirt.c.B: I'm telling you, you've got a hole in your sleeve.
Other kinds of sensory information, for example, smells or sounds, are less obviously readily available in utterance contexts.Note the felicity of PITY in (15).
(15) I'm telling you, something is rotting around here.
Perhaps external stimuli that do not occupy a definite location in the utterance context do not count as readily available in that context.We leave further exploration of these issues for future research.
3. The semantic contribution of parenthetical I'm telling you.Based on the preceding discussion, we propose the analysis of the conventional semantic contribution of PITY in (16).We assume that PITY conveys that the speaker has private evidence for the prejacent proposition.The contribution is modeled formally as a use-condition, following previous work on discourse particles (Kaplan 1999, Kratzer 1999, 2004, Gutzmann 2015).PITY otherwise makes no truth-conditional contribution: a sentence I'm telling you, S is true at a possible world w just in case S is true at w.The notion of private evidence is elaborated in ( 17) and ( 18).
(16) I'm telling you, S c,w a. is true in context c at world w just in case the proposition p expressed by S in c is true at w; and b. is appropriate in c just in case the speaker's evidence for p in c is PRIVATE.
(17) An agent x has PRIVATE EVIDENCE for a proposition p iff that evidence is ACCESSIBLE to x and to no other salient agent y.
(18) ACCESSIBILITY: Individuals have access a. to their own knowledge bases; and b. to readily available perceivable evidence in the utterance context at utterance time.
The semantic proposal in ( 16) together with the definitions in ( 17) and ( 18) predict that PITY is inappropriate if the speaker's evidence for the prejacent proposition consists of readily available sensory evidence in the utterance context at utterance time or the speaker's evidence is part of the addressee's private knowledge base at utterance time.Importantly for our analysis, these restrictions are about the speaker and addressee's EVIDENCE for the prejacent proposition and not about whether this proposition is among their individual beliefs.Thus the definitions do not rule out the use of PITY when the speaker and addressee both BELIEVE the prejacent proposition, as long as the evidence that the speaker has for it is private and thus not available to the addressee.The use-conditions assigned by PITY therefore emphasize the reasons the addressee should believe the prejacent rather than the belief itself.These observations are relevant to the discussion of the discourse specific uses of PITY discussed in section 4.
Elaborating on the formal details, we adopt Gutzmann's (2015) multidimensional hybrid semantics in which sentences encode both truth-conditional and use-conditional content. 4ruth-conditional content corresponds to standard sets of possible worlds propositions (t-propositions), while use-conditional content corresponds to the set of contexts in which the sentence is appropriately used (u-propositions) (Kaplan 1999).
PITY corresponds to Gutzmann's category of functional expletive use-conditional items (UCIs) since it takes the proposition expressed by the prejacent as an argument, expressing an attitude towards that proposition and contributes only use-conditional content.In the use-conditional dimension, PITY denotes the type s,t , u function in (19), where u corresponds to the type of use-conditional propositions.In the truth-conditional dimension PITY denotes the identity function I s,t on t-propositions.
(19) PITY = λ p s,t .{c : c S has private evidence that p is true in c W } PITY semantically composes with the proposition expressed by the prejacent via Gutzmann's multidimensional application rule.
The conventional semantic content of the sentence in (20) thus consists of the truth-conditions in (20a) and the use-conditions in (20b).The sentence is true just in case Jones resigned and is appropriate just in case the context of utterance is one in which the speaker has private evidence for the proposition that Jones resigned.Evidence for the use-conditional status of PITY's evidential contribution comes from the observation that rejections-which typically target truth-conditional content-only target the t-proposition in (20a) and not the u-proposition in (20b).
(21) a. A: I'm telling you, Jones resigned.b.B: That's not true.He denied the allegations and said he plans to stay on.c.B': #That's not true.I have the same evidence that he resigned that you do.
4. The uses of parenthetical I'm telling you in discourse.In this section, we discuss two ways that speakers use PITY in discourse and the relation of these uses to the conventional use-conditional contribution proposed in section 3. We focus specifically on the use of PITY to mark disagreement (and sometimes agreement) and its use in different kinds of directive speech acts.
4.1.AGREEMENT AND DISAGREEMENT.Many speakers have a strong intuition that PITY marks disagreement, as in ( 22).An anonymous reviewer suggested that PITY is only licensed in explicit or implicit disagreements, for example.Furthermore, unlike many of the previous examples, the plain assertion in ( 22c) is relatively unacceptable compared to the corresponding PITY-marked assertion.
While the proposal in ( 16) is consistent with this use of PITY, we argue that it does not conventionally encode disagreement.
For the sake of discussion, assume the definitions in ( 23), which are adapted from Coppock 2018.5 (23) a. Agents a and b DISAGREE with respect to a proposition p in w just in case a accepts p in w and b rejects p in w. b.An agent a ACCEPTS p in w just in case p is true in every world consistent with a's evidence in w. c.An agent a REJECTS p in w just in case p is false in every world consistent with a's evidence in w.From these definitions it follows that if a and b disagree about a proposition p, then a and b have inconsistent, and therefore different, evidence; a has access to information that b does not have access to and thus has PRIVATE EVIDENCE for p.In other words, whenever a and b disagree, the conditions for the appropriate use of PITY are met.The use of PITY then highlights that the speaker has information that the addressee does not and suggests that the disagreement should be resolved in favor of p on the basis of this privileged information.The addressee should "take the speaker's word for it" that there is good reason to accept p.
The use-conditions in ( 16) therefore nicely align with contexts in which the speaker and hearer disagree.However, they do not require disagreement.There are any number of reasons for a speaker to highlight that they have private evidence for the prejacent proposition.In (3), repeated below, A and B have at least some shared evidence for the value of the boots given A's acknowledgment that the boots are "top-of-the-line".B's use of PITY highlights evidence gained from B's PERSONAL EXPERIENCE to which A does not have access.In ( 24), B might use PITY to suggest that her evidence licenses a stronger statement than one A is willing to make.
(24) a. A: I saw John driving around town in a fancy new car yesterday.He must be doing alright.b.B: I'm telling you, he's stinking rich.
There is no disagreement in either (3) or ( 24), at least not as defined in ( 23), which requires that the addressee REJECT the prejacent proposition.In (3), there is at least tacit agreement.In ( 24), A might be avoiding making the stronger statement, either for reasons of politeness or because their information state does not support it.In either case, A does not explicitly reject it.Rather, the speaker uses PITY to highlight their personal experience or indicate that they do have evidence for the stronger statement.

DIRECTIVE SPEECH ACTS
. PITY is natural in utterances that offer advice, warnings or suggestions (see 5, repeated below, and 25) and with prejacent sentences in the imperative mood (25).
(5) A knows that the woman across the room, C, is married and believes that B doesn't know this.
B is clearly thinking about approaching her.a.A (to B): I'm telling you, she's married.
(25) I'm telling you, order the lasagna.
The speaker in these examples uses PITY to indicate that she has private evidence for p, where p is relevant to a contextually-salient decision problem faced by the addressee.In (5), for example, B is deciding whether or not to approach C. A offers the information that C is married, implicating that B shouldn't approach C, and also indicating that they have access to evidence for this claim that B does not.A's utterance may also implicate that A believes B might have reason to doubt that C is married.
PITY is degraded in commands, as shown by ( 26).Commands differ from requests and suggestions in their MODE OF ACHIEVEMENT.The illocutionary point of a command is achieved via the speaker's capacity as a person with authority over the addressee (Searle & Vanderveken 1985, Green 2017).
(26) Drill Sargent to recruit: #I'm telling you, drop and give me twenty.
We suggest that the speaker's authority is sufficient for the addressee to perform the commanded action in these cases, so there is no reason for the speaker to further "justify" the directive by indicating they have private evidence that it is in the addressee's best interest to perform the commanded action.
The corpus example in ( 27) is a possible counter-example to the infelicity of PITY in commands.
(27) Cut your stake in half, all right?You've been doing a great job, everyone knows that, but I'm still your boss and I'm telling you: sell those contracts.
(COCA) However, it is not clear whether the addressee should sell the contracts because their boss, as someone with authority over them, commands it, or because their boss, as someone who knows better and has more relevant experience, advises it.The distinction is a subtle one, but given the presence of PITY, it is the intuition of the first author that the latter reading is more salient.Admittedly, the intuition is a subtle one.At the very least, the presence of I'm telling you in ( 27) serves to soften the command.
5. Comparison with German modal particle ja.In this section, we compare English PITY and the German modal particle ja (Jacobs 1991, Lindner 1991, Kratzer 1999, 2004, Gutzmann 2008: cf.).We argue that these expressions are sensitive to similar aspects of utterance contexts. 6ection 5.1 summarizes the previous research on ja and section 5.2 makes connections between this literature and the analysis of the semantics of PITY presented in section 3.
5.1.BACKGROUND ON GERMAN ja.The literature on the unstressed modal particle ja argues that it is appropriately used when the prejacent proposition is shared knowledge between the speaker and addressee.So, for example, ( 28) is appropriate if, and only if, it is common knowledge that Max is at sea.Ja is licensed in (29) because it is pragmatically presupposed that the writer and the reader know German.
( können.know 'We understand the meaning of this sentence, since we know German.' (Kratzer 2004: 127) However, it is not always the case that the proposition expressed by the prejacent of ja is shared knowledge between speaker and addressee.It is often noted, for example, that ja is felicitous when the prejacent's truth can be "easily ascertained" (Lindner 1991) from information "in the immediate surroundings" (Kaufmann & Kaufmann 2012), or is "verifiable on the spot, given extra-linguistic evidence" (Kratzer 2004) or "perceivable contextual evidence" (Zimmermann 2011).So for example, ja is used in (30) to inform the addressee that there is a hole in their sleeve and marks that information as, in some sense, uncontroversial.Ärmel.sleeve 'There is a hole in your sleeve.' (Kratzer 2004: 127, modified from Lindner (1991: 171)) When neither the shared knowledge nor non-controversiality condition is met, the use of ja is inappropriate.For example, ja cannot be used as an answer to an information seeking question (31) or in a statement conveying "breaking news" (32).
5.2.COMPARISON WITH PITY.Summarizing the conclusions of sections 2 and 3, the use of English PITY is appropriate just in case the speaker has private evidence for the prejacent proposition.Private evidence is evidence to which the speaker, but not the addressee, has access, where an agent has access to information that is part of her knowledge base or that is readily available in the utterance context.The use of the German modal particle ja, on the other hand, is appropriate just in case the prejacent proposition is shared knowledge between speaker and addressee or if it is verifiable from readily available contextual evidence.PITY and ja are thus sensitive to very similar aspects of utterance contexts.( 34) restates the use-conditions of ja using the terminology of the present paper.
(34) ja(ϕ) is expressively correct if there is PUBLIC EVIDENCE for the prejacent in the utterance context, where public evidence is evidence that is accessible to both speaker and addressee.
If information is accessible to both speaker and addressee, then it is either part of the knowledge base of both (and is therefore at least on its way to becoming common ground) or it is readily available in the utterance context.In the latter case, the prejacent would be "verifiable on the spot", to use Kratzer's (2004) words.On the reformulation in (34), German ja and English PITY are sensitive to the same features of utterance contexts, namely the evidence available to the speaker and the evidence available to the addressee.They differ only in the information they highlight.German ja highlights the public nature of the evidence for the prejacent in the utterance context, as summarized in Table 1.English PITY, on the other hand, highlights the private nature of the speaker's evidence for the prejacent in the utterance context, as summarized in Table 2  One question that needs to be addressed is whether there is a substantial difference between the use-conditions in (33) and (34), or is (34) just a paraphrase of (33) in terms of private and public evidence.We believe there is a real difference.(33) refers the status of the PREJACENT itself as shared knowledge and to possible contextual evidence for it.(34), on the other hand, refers to the status of the EVIDENCE for the prejacent.While it is easy to extend the analysis of PITY in (16) to German ja, as in (34), it is not so easy to extend the standard analysis of ja to PITY, for example as in ( 35).
(35) PITY(ϕ ) is expressively correct iff either the speaker believes that ϕ is neither shared knowledge between speaker and hearer nor verifiable on the spot.
One problem with the reformulation in (35) in our estimation is that the prejacent of PITY can be shared knowledge, as long as the speaker has evidence for that proposition that the addressee does not.( 24), repeated below, is a possible example of this kind of scenario, especially when A's assertion that John is doing "alright" is understood as a sort of rhetorical understatement.B's use of PITY then conveys that B has evidence other than the fancy car that John is doing well financially.
(24) a. A: I saw John driving around town in a fancy new car yesterday.He must be doing alright.b.B: I'm telling you, he's stinking rich.
Admittedly, the proposal in ( 34) is in need of more thorough investigation.There are examples for which (33) and (34) make different predictions.For instance, if the prejacent proposition ϕ is shared knowledge between speaker and hearer but each holds their belief on the basis of distinct evidence, is the use of ja appropriate-a point for (33)-or not-a point for (34)?Another question to be addressed is whether the analysis can be extended to related modal particles, for example doch, whose semantic contribution has been argued to over-lapping with that of ja (Kaufmann & Kaufmann 2012, Grosz 2014).
It might be that PITY and ja really do carve out distinct conceptual niches and the reformulation of ja's use-conditions in (34) is on the wrong track.What purpose would the use-conditions in (34) serve?There is an intuitive asymmetry between "shared" versus "not shared" evidence for a proposition.While it is natural to emphasize one's private evidence for a particular proposition in order to convince or persuade another to adopt a certain belief, or to provide additional evidence for a certain proposition, it is unclear why one would emphasize one's shared evidence for a proposition.Emphasizing that the proposition itself is shared knowledge, on the other hand, can function as a reminder, among other things.We leave these questions and further comparison for future research.

Conclusion.
We have argued in this paper that I'm telling you should be analyzed as a marker of private evidence, contributing a use-conditional proposition to the effect that the speaker has access to evidence for the prejacent proposition that the addressee lacks.We showed how this analysis is consistent with certain discourse functions of PITY, for example, its use in disagreements and in certain kinds of directive speech acts.Finally, we compared our analysis to analyses of the German modal particle ja, showing that both expressions appear to be sensitive to the same kinds of information, though in complementary ways.

( 4 )
A and B are visiting a tropical island for the first time and it is extremely hot when they arrive.a.A (to B): #I'm telling you, it's hot here.b.A (to B): It's hot here.

(
10) A has not tasted the soup: a. B: Try some of this soup.b.A: No thank you. c.B: I'm telling you, it's tasty.However, if both A and B have tried the soup and disagree about its tastiness, B cannot use PITY to try to convince A that the soup is tasty, as in 11c.(11) A and B have both tasted the soup: a. B: This soup is tasty.b.A: No, it isn't.c.B: #I'm telling you, it's tasty.
(i) a. A: John came to the party.b.B: No, he didn't.c.A: John came to the party, #(man).
(12) A and B are talking to each other at a party; A is facing B, who has his back to the door.A notices Jeff enter through the door.a. B: It's too bad Jeff couldn't be here tonight.b.A: (#I'm telling you,) Jeff is here.(13) A is climbing the stairs in front of B: a. B: (#I'm telling you,) you've got a hole in your sleeve.

(
20) I'm telling you, Jones resigned.a. Truth conditions: {w : Jones resigned in w} b.Use conditions: {c : c S has private evidence that Jones resigned in c W }

( 3 )
A and B are discussing B's new winter boots: a. A: Wow!Those are top-of-the-line boots.You must have paid a lot for them.b.B: I'm telling you, they're worth every penny during Minnesota winters.

( 30 )
The addressee is walking up the stairs in front of the speaker: ), A indicates that he knows better than anyone what he did or did not do, and that B should therefore trust him.

Table 1 :
on page 11.The use-conditions contributed by each expression are thus complementary.Discourse conditions governing the use of German ja.

Table 2 :
Discourse conditions governing the use of English PITY.