Gradient at-issueness, minimum relevance, and propositional prominence

1 Introduction

It is great to see that with their target article, Barnes and Ebert try to crack down on the binary notion of (non-)at-issueness. That at-issueness is a binary notion is an idea that, even though not explicitly defended, is rooted deeply in the semantic and pragmatic literature ever since Potts’s introduction of the term in his 2005 book. In that work, Potts himself was not really interested in the pragmatics behind that notion and what it actually means for a proposition or some content to be »at issue«, and thus he built the distinction between at-issue and non-at-issue content (or to be precise, conventional implicated content) into the semantic system as deeply as the semantic type system. Hence, it is no wonder that he employed a binary distinction. However, even in the research that took up the notion of (non-)at-issueness and took a broader and more pragmatic view, the binarity of at-issueness was more or less carried over. And even though some people have hinted at the idea that at-issueness is a gradient rather than binary phenomenon (AnderBois et al. 2015; Gutzmann 2017; Syrett and Koev 2014; Tonhauser et al. 2018), it needed a paper like Barnes and Ebert’s to distill these hints to deconstruct the assumption of binarity and build the foundations for a gradient notion of at-issueness. For this reason alone, their target article will have a huge impact and certainly inspire a lot of research that pushes the field forward. Therefore, there are many aspects in their paper worthwhile to comment and expand on. However, in this comment, I will focus solely on the core ideas of gradient at-issueness and will concentrate on three aspects:

1. the application of gradient at-issueness to non-iconic expressions,

2. the minimum at-issueness requirement and some of its problems,

3. an extension of Barnes and Ebert’s ideas to a competition-based model of at-issueness as (propositional) prominence.

While the first point directly shows the strength of Barnes and Ebert’s approach, the second aspect will point out some of its weaknesses. The third point is then to be understood as a gentle suggestion to overcome these issues and connect their ideas about at-issueness to the more general idea of prominence. All this is not meant to take away from what I take to be an innovative and inspiring paper ,^[1] that surely will have a great impact not only on research on iconic enrichments, but also on our general understanding of (non-)at-issueness.

2 Gradient at-issueness beyond iconic enrichment

To start, it is worth to repeat the factors that Barnes and Ebert assume to influence the degree of at-issueness.

Barnes and Ebert are explicit that they »focus on iconic expressions and their information status« and »leave it to future work to incorporate other non-at-issue constructions such as appositives into this scale«. Let me directly pick that up briefly and sketch for some non-iconic expressions how the factors Barnes and Ebert establish can explain differences. Let’s start with the obvious case of appositives: As Barnes and Ebert already mention (following findings by Syrett and Koev 2014), the structural position of appositives has an influence on their at-issueness (at least, when tested by the common procedure).

It also seems that demonstrative reference to an appositive increases its at-issueness, even if it occurs in sentence-medial position. The differences may be subtle and surely have to be tested empirically, but they at least hint at demonstrative reference being a factor for non-iconic non-at-issue content as well.

For expressives, the difference between what Barnes and Ebert call internal versus external enrichment matters. So-called expletive expressives (Gutzmann 2013) can be left out without affecting the main content and are non-at-issue, whereas expressive nouns in predicative position cannot be omitted and are more at-issue.

These brief considerations show that the factors that Barnes and Ebert assume seem to be on the right track and may fruitfully be applied to non-iconic non-at-issue content as well.

3 Minimum relevance as a threshold for at-issueness

Besides the at-issueness influencing factors, the core idea of Barnes and Ebert’s approach to gradient at-issueness is their appropriateness condition for utterances that puts relevance thresholds on contents to model their gradient at-issueness (Barnes and Ebert: (58)).

The crucial ingredient to their approach is that every kind of »propositional construction« with content t _i comes with an at-issueness requirement of minimum relevance for that content, a(t _i ), so that the relevance of that content with respect to the QUD must exceed this threshold. Secondly, every propositional content of an utterance must fulfill this.

While this approach delivers many desired results, I see some problems with it. First, from an empirical point of view, it does not cover cases in which the content components on their own fail to meet their at-issueness requirement, but the overall utterance is still perfectly appropriate, because the utterance overall – with all its contents taken together – is relevant. The following example illustrates this:

In this example, neither the main clause content nor the content of the appositive provides a relevant answer to the question. Only if one takes the information of both components together, does the utterance provide a relevant answer to the question. And, as witnessed by (9B), this seems to be perfectly fine, whereas Barnes and Ebert’s condition given in 3 would predict (9B) to be infelicitous, because at least the main clause content does not cross the relevance threshold.

Besides this more data-driven reservation, I have a more conceptual one: The assignment of minimum relevance/at-issue values for every »propositional construction« is not really grounded in the factors that Barnes and Ebert themselves identify as being relevant for at-issueness. So, while, say, appositives and post-speech gestures are non-integrated (i.e. they are external enrichments; see (1a)) or while cospeech gestures occur in a different modality than the speech they accompany (see (1d)), this should make them less at-issue than the main content of the utterance they occur in/with. However, the system sketched by Barnes and Ebert only specifies a »minimum requirement« and it does not really provide a way to determine how this value is derived from the properties of the construction. As given, the a-function could assign rather arbitrary values without any connection to the at-issueness properties.^[2]

To be fair, I think it is rather clear that Barnes and Ebert think of the a-function as a formal implementation of the factors that they identify, so I think this alone does not really make the construction-specific minimum at-issueness requirement problematic. What I, however, think is a bigger issue is that the factors that Barnes and Ebert use clearly seem to be dependent on the »structural context«: whether a content component is of the same modality as its host utterance is not a property of the construction itself, but a relation between the construction and the linguistic context it occurs in. And the structural position in which an appositive or a gesture occurs in, is not a property of the appositive or gesture itself, but a property of, well, the structural context in which they occur. But, in its current form, the system can only put up minimal at-issueness requirements of say, appositives or gestures or ideophones, but not for cospeech gestures versus cosign gestures or sentence internal versus sentence final appositives. Therefore, I believe that the current implementation, as given in the article, cannot really capture the core ideas of what Barnes and Ebert’s are after.

For this reason, let me use the remainder of this comment to briefly put the sketch of an alternative proposal on the table which, I think, is better suited to capture the core insights that Barnes and Ebert are actually after.

4 Towards a competition-based model of gradient at-issueness as propositional prominence

I think the best way to capture the core insights of Barnes and Ebert’s inspiring paper while simultaneously circumventing the (maybe not so serious) problems sketched above and relating to other, independently developed ideas is i) to think about at-issueness in terms of propositional prominence (von Heusinger and Schumacher 2019) and ii) substitute the construction specific at-issueness requirement with a competition-based model.^[3]

The core idea of such an approach is to reconceptualize what is called at-issueness since Potts’s (2005) work as prominence at the propositional level. While the notion of prominence is well established when it comes to discourse referents (see von Heusinger and Schumacher 2019 for an overview), it has an analogous application for propositions. And just as the relative prominence of some discourse referents over others enables certain properties, like being able to be picked up anaphorically by simple pronouns, the relative prominence of some propositions over others makes them more at-issue and enables certain properties, like being directly targeted by yes or no (which in fact have been claimed to be propositional anaphors, Krifka 2013). In a sense, this means that the gradient notion of at-issueness is traced back to the gradient notion of propositional prominence.

Given these short elaborations, we can now link propositional prominence to the pragmatic notion of at-issueness (in the sense of addressing the QUD) by the following constraint:

This means that, in principle, the most prominent proposition expressed by an utterance is the most relevant answer to the QUD. But (10) is also compatible with no proposition (alone) really addressing the QUD. However, in this case, some combination of content has to address the QUD. This can be formulated by another constraint:

Now, what determines the prominence of a propositional component are so-called prominence-lending properties, which are similar to those discussed by Barnes and Ebert. Of course, the exact set of prominence determining features should be investigated empirically, but here is a list of what I take to be promising candidates:

Now, this only has an effect if there are competing content components in an utterance. If there is only one propositional component in an utterance, (10) is trivially fulfilled. But if there are two or more, they compete with each other in the context of each other. Moreover, some properties in (12) are gradient, which contributes even more to the relative prominence calculation of the propositions expressed by an utterance.

With such a list of factors, the difference in at-issueness of various content components can directly be derived (without building it into the system). Consider the following sentence:

The difference in prominence between the two contents in (13) obvious: The interjection is neither prosodically, syntactically, nor semantically integrated. It is syntactically very light and occurs at the beginning of the utterance. And while it is in the spoken modality and its content is conventionally expressed, its content is non-truth-conditional and not semantically composed (i.e. it expresses a content on its own) and neither is it asserted nor referred to by other parts of the utterance. In contrast, the main clause has many prominence-lending properties and clearly is the more prominent. Hence, there is no ambiguity as to which content addresses the QUD and, in virtue of that, is at-issue. This is similar to how certain discourse referents cannot be picked up by anaphors, because they are not prominent enough.

These are the basic ideas of how the gradient at-issueness of various propositional constructions can be derived from an interplay of prominence-lending factors. With a sophisticated enough empirical investigation, I think it would even be possible to assign different weights to the factors so that some factors (like, for instance, truth-conditionality or modality) have a stronger effect than others and some features are more prominence-lending the higher their value is, similar to how it has been done, for different purposes, in Linear Optimality Theory to model gradience in grammar (Keller 2000, 2006).

5 Summary

The core idea championed by Barnes and Ebert—that (non-)at-issueness is a gradient phenomenon—is both inspiring and thought-provoking and, I dare to say, absolutely correct. This will push the boundaries of our understanding of (non-)at-issue meaning and will shed new light on some phenomena that are puzzling under a binary perspective on at-issueness, as I briefly outlined in the first part of this comment. And this does not even touch on the implications for how we think about iconic enrichments. And even though I have raised some doubts about how gradient at-issueness is spelled out by Barnes and Ebert, I am sure the general idea is worthwhile to explore in more depth and can be pushed forward when connected to more general notions, as I have proposed here with respect to linguistic prominence. But these were just some rough ideas inspired by reading the target article, and I am sure that many better thought-out modifications, expansions and applications of the ideas developed by Barnes and Ebert will be put forward as soon as their paper properly hits the market.