The information status of iconic enrichments: modelling gradient at-issueness

3.1.1 Ebert et al. (2020)

Ebert et al. (2020) (hereafter EEH), spelling out and experimentally validating earlier work of Ebert and Ebert (2014), provide a supplemental analysis of iconic co-speech gestures, in which they contribute meaning in a similar manner to appositives, and argue, based on experimental data (see Section 3.2.1), that co-speech gestures are by default not at-issue. In order to provide a formal semantics for co-speech gestures, they adapt the approach to appositives developed by Anderbois et al. (2015) and employ the propositional variables p and p* to mark at-issue and non-at-issue content, respectively. Consider example (9) again, repeated below in (11).

Crucially, EEH argue that iconic gestures such as the BIG gesture introduce individuals or individual concepts, just like pointing gestures do. In other words, the gesture does not represent the size of the bottle, but the bottle itself.^[7] The gesture represents the bottle, but abstracted away from irrelevant properties and representing in particular the relevant one in the given context, which is the size property. Hence in this case, the iconic gesture BIG represents a big bottle. Formally, we argue that by producing the gesture, this act ‘lexically’ introduces this bottle as an individual concept, comparable to when uttering a name in spoken language. The gesture directly refers to an object or an individual, represented as the gesture referent g. By producing the gesture, an individual concept for this gesture referent is introduced. This is captured formally by introducing a novel discourse referent for a rigid designator I _g to the gesture referent g (see Ebert et al. 2020 for details).

Within this account the interpretation of the gesture is highly dependent on the temporal alignment of gesture and speech; EEH argue that the interpretation of an iconic co-speech gesture depends on how it aligns with the noun phrase and article of a given referent. They give the following analyses of co-speech gestures aligned with a noun phrase, an indefinite article and a definite article.

noun phrase. The relation between a gesture and a temporally aligned noun phrase is one of exemplification. This simply means that the individual gestural concept z must also have the property expressed by the noun phrase, i.e. N p * ( z ) .

indefinite article. When the gesture aligns with an indefinite article, the conveyed, non-at-issue meaning indicates similarity between the gestural and verbal concept. This is expressed via the two place predicate SIM,^[8] such that there is a non-at-issue predication SIM p * ( x , z ) , where x represents the verbal concept and z the gesture concept. SIM p * ( x , z ) is true in the case that the objects denoted by x and z are similar in the relevant dimension. What the relevant dimension is, is highly dependent on the context and determines the execution of the gesture itself, as mentioned above. In this example, the relevant property is the size property. Hence, the utterance in (11), where the gesture aligns with the indefinite article and the noun phrase, can be analysed as in (13).^[9]

Here the at-issue contribution of the utterance is that Cornelia brought a bottle, while the non-at-issue contribution is that this bottle is similar to the gesture referent in the relevant dimension (this is the non-at-issue contribution triggered by the temporal alignment of the gesture with the indefinite article) and that the gesture referent also has the property of being a bottle (this is the non-at-issue contribution triggered by the temporal alignment of the gesture with the noun phrase). The non-at-issue contribution hence is that what is gestured represents a bottle and that what is talked about, i.e. the referent that Cornelia brought, has to be similar to what is gestured. Since what is gestured is a bottle that is big in size, the contribution of the gesture (and its alignment with speech) eventually comes down to claiming that the bottle Cornelia brought is big.

definite article. The alignment of a gesture and a definite article conveys a strengthened relation between gestural and verbal referent, namely one of (relativised) identity or x = p * z , where the verbal concept x and the gestural concept z designate the same object in all the worlds in p*. Importantly, identity is only required on all p*-worlds, since the identity requirement, as the similarity requirement in the case of indefinites, is triggered by a co-speech gesture and its alignment with speech, which makes non-at-issue contributions. Hence, the utterance in (14) can be analysed as in (15).

In this case, alongside the at-issue and non-at-issue contributions, there is an additional presupposition which requires that there be a unique, contextually salient bottle. The at-issue contribution is that Cornelia brought this unique bottle. The non-at-issue meaning conveys that the gesture referent is the same object as this bottle and is itself also a bottle.^[10] Hence the non-at-issue inference is that the unique, contextually salient bottle that Cornelia brought is the same as the one indicated via the iconic gesture and hence big.

While EEH argue that co-speech gestures are non-at-issue by default, they also highlight that they can be shifted towards at-issueness status, when accompanied by a dimension shifter, such as a demonstrative, or particular types of focus, prosody and even facial expressions. They show that the German demonstrative SO,^[11] for example, shifts the non-at-issue content contributed by the gesture in (16) towards at-issueness status.

EEH provide an analysis of demonstratives as dimension shifters for co-speech gestures, where the demonstrative simply shifts the proposition of evaluation from p* to p. Hence the utterance in (16) can be analysed as in (17).

Now what is at issue is that Cornelia brought a large bottle, as the previously non-at-issue contribution that this bottle is similar to the gesture referent in size has been shifted to the at-issue dimension. The utterance would be false in the case that Cornelia brought a small bottle. However, the additional contribution that the gesture referent also has the property of being a bottle remains non-at-issue.

While EEH do not directly discuss iconic gestures outside of co-speech gestures, Ebert (2017) has made predictions concerning post-speech gestures. She argues that as post-speech gestures are more standalone than co-speech gestures and do not occur simultaneously with speech, they may be more likely to be at-issue than co-speech gestures.

3.1.2 Schlenker (2018a)

Schlenker (2018a) presents an alternative analysis to Ebert et al. (2020), where co-speech gestures are a specialised form of presupposition. Schlenker highlights that the projection patterns of iconic co-speech gestures and presuppositions are highly similar; and argues that a co-speech gesture triggers a presupposition which requires its content to be entailed by the content of the expression it modifies. Hence presuppositions triggered by co-speech gestures are conditionalised on the at-issue content of the assertion. Schlenker gives the following definition of presuppositions triggered by co-speech gestures (p. 316–317):

Given the utterance in (19), Schlenker argues that the gesture triggers the cosupposition that if Johnny’s mother helped him, this involved some lifting.

Assuming that the asserted, at-issue proposition is true, i.e. that Johnny’s mother did indeed help him, and if the cosupposition holds, then Johnny’s mother must have helped him in a way that involved some lifting. Importantly Schlenker notes that, in order to prevent this analysis from overgenerating inferences, for example by positing that helping always entails lifting, the content of the co-speech gesture should be entailed by its local context and not the global context. We can therefore see that in a given situation, such as a gymnastics competition, for example, helping would entail some kind of lifting, but that this need not always be the case. Schlenker (2018a) further argues that co-speech gestures are comparable to weak presupposition triggers such as realise and can therefore be easily locally accommodated. In the case of cosuppositions triggered by co-speech gestures, local accommodation results in the content of the gesture being treated as an at-issue contribution, if the sentence is true and if the speaker is not in a context where the cosupposition is already known to hold.

Crucially, Schlenker furthermore argues that it is empirically adequate to state that the cosupposition also holds in cases of negation or subordination under a modal operator and discusses examples as the following ones.

Here, it is stated that Jonny’s mother did not help him. But additonally, the gesture triggers the (cosuppositional) inference that had she helped him it would have been via lifting.

Finally, Schlenker (2018b) has also argued that pro-speech gestures, as in (21) are by default at-issue.

This claim seems to be intuitive because the gesture stands in for crucial lexical material, which would be at-issue and therefore if the gesture were to be non-at-issue, then the sentence would be incomplete and infelicitous.

3.1.3 Esipova (2019)

Esipova (2019) provides an alternative approach to both EEH and Schlenker (2018a), arguing that co-speech gestures are neither inherently at-issue nor non-at-issue, and that their interpretation is not dependent on temporal alignment with speech. Her approach rests upon the idea that narrow syntax and semantics proper are both modality-blind, meaning that gestures compositionally integrate in the same manner as spoken items. Not all gestures can necessarily be compositionally integrated, however. Esipova argues that this is not a problem if we do not assume a uniform analysis for all gestures.

Giving the gesture LARGE as an example, Esipova claims that gestures are “non-lexicalized and [carry] little morphosyntactic information” (p. 92), hence they can either be property-like, in this case similar to the adjective large, and act as modifiers, or they can be nominal-like, in this case similar to the nominal phrase a large one, and be supplements. As modifiers, they can either be restricting and therefore at-issue or non-restricting and therefore non-at-issue. Supplements, per Potts (2005), cannot be restricting and therefore nominal-like gestures are non-at-issue. Supplemental gestures project as supplements, whereas only non-restricting modifier gestures project in the same manner as other non-restricting modifiers.

Esipova further points out that restricting modifier gestures tend to be dispreferred by speakers and argues that this is likely not due to constraints on restricting co-speech gestures in semantics or narrow syntax, but rather constraints in the phonology or pragmatics. She highlights that one potential explanation for this degradedness comes from Schlenker (2018a), who notes that there is a general preference for gestures to be semantically vacuous, i.e. making a redundant contribution, as when there are meaning contributions from two modalities within one utterance, the contribution in the secondary modality should be redundant.

In terms of pro-speech gestures, Esipova (2019) argues that these should be theoretically acceptable as restricting modifiers, but they are often degraded in English due to linearisation and prosodic grouping, which are to some extent language-specific. She does, however, provide examples of pro-speech gestures with restricting interpretations in French, as in (22), and Russian.

Overall, all three approaches observe that post-speech gestures and pro-speech gestures have an at-issueness status that is, at least by default, different from that of co-speech gestures. While Schlenker claims that co-speech gestures and post-speech gestures are non-at-issue, with pro-speech gestures being at-issue, he also argues that co-speech gestures give rise to cosuppositions, a special kind of presupposition, whereas post-speech gestures act as supplements. Esipova, on the other hand, argues that the at-issueness status depends on whether the gestures are restricting or not. However, likely due to the fact that they are performed at the same time as the verbal material they accompany, co-speech gestures are often not at-issue, which is not the case for pro- or post-speech gestures. Finally, EEH account for the default non-at-issueness status of co-speech gestures, with Ebert (2017) furthermore arguing that the more standalone a gesture is, the more at-issue it tends to be. This then makes the prediction that pro- and post-speech gestures are more at-issue than co-speech gestures.

3.2.1 Ebert et al. (2020)

In an experiment conducted with native speakers of German, Ebert et al. (2020) found empirical evidence for the non-at-issueness status of iconic co-speech gestures. The experiment consisted of a sentence-picture matching task, using a 2 × 2 design, which crossed two mode conditions, iconic co-speech gestures and adjectives, with two match conditions, match and mismatch, in a Latin square design. The two mode conditions were realised through 24 target sentences and the variation between gesture and adjective was realised as in (23) and (24).

The variation in the match condition was implemented with two pictures, where one picture matched the conditions in the target sentence, while the other did not. An example item can be seen in Figure 1.

Figure 1:

Sample item (Ebert et al. 2020).

Overall, there were 24 target sentences, each paired with one matching and one mismatching picture. The procedure for the experiment was as follows; participants were shown a video of a female speaker uttering a sentence. In the target sentences, the speaker also produced a gesture in the gesture condition. All videos were accompanied by a picture. Participants were asked to rate how well the description given by the speaker in the video matched the image using a scale from 1 to 5 with 5 being the sentence perfectly matches the circumstances in the picture and 1 being that it does not match at all.

The experiment was designed as an indirect test of at-issueness. The goal was to have participants evaluate the truth of the target sentence in relation to the situation given by the picture. These truth judgements were gathered via the participants’ response to the meta-question of how well the description in the video matched the picture. The use of a meta-question and rating scale also allowed for participants to express subtleties of judgements with respect to sentences where the information appears to be partly false and partly true. As per research conducted by Kroll and Rysling (2019), which showed that speakers’ truth value judgements are less impacted by information deemed irrelevant to the QUD, i.e. non-at-issue information, the specific hypothesis was that, if co-speech gestures were non-at-issue, there would be a significant interaction of the factors mode and match, such that the mismatch effect (the difference between ratings in the matching and mismatching conditions) would be significantly larger for target sentences containing adjectives than those with speech accompanying gestures. Since no overt QUD was given it was assumed that participants would construe a very generalised QUD matching their understanding of what is at-issue and what is not.

The results of the experiment supported the hypothesis, with the mismatch effect being significantly greater for adjectival target sentences than for co-speech gesture sentences (see Figure 2). EEH therefore argue that these findings corroborate the claim that adjectives contribute at-issue information, whereas co-speech gestures contribute non-at-issue content.

Figure 2:

Results of experiment 1 (Ebert et al. 2020, p. 173).

In a second variation of this experiment, EEH repeated the above experiment with the same materials and procedure, but an additional third mode condition, where iconic co-speech gestures were accompanied by the German demonstrative SO. The same target sentences were used as in the iconic co-speech gesture condition, except that the noun phrase was accompanied by the demonstrative SO and the condition was also realised with a video of a speaker uttering the sentence and simultaneously producing the gesture. The onset of the gesture was aligned with the onset of the German demonstrative SO. An example item can be seen in (25).

The hypothesis in this variant was that the demonstrative would shift the gesture towards at-issueness status and that the mismatch effect for demonstrative target items would therefore be closer to that of the adjective target sentences. This claim was supported by the results of the experiment, with the mismatch effect being larger for sentences with gestures accompanied by demonstratives than for sentences with gestures not accompanied by demonstratives (see Figure 3). However, this mismatch effect was still not as strong as for adjectives, indicating that while the demonstrative did shift the gesture towards at-issueness, it was still not as at-issue as the adjectives. This provides evidence for not viewing at-issueness as binary, but rather gradient, as was discussed in Section 2.2.

Figure 3:

Results of experiment 1 (Ebert et al. 2020, p. 174).

Having discussed gestures as one prominent type of iconic meaning in natural language, we will now turn to another class of iconic phenomena: ideophones. While gestures use the visual modality, ideophones use the same modality as other spoken language phenomena.

4.1.1 Kawahara (2020)

Kawahara (2020) argues that Japanese predicative ideophones are subjective predicates and claims that they have both a core at-issue meaning and a subjective meaning, which encodes their sound-symbolic nature, adapting the counterstance analysis of subjective attitude verbs given by Kennedy and Willer (2016) in order to account for this subjective nature.

Kennedy and Willer (2016) argue that subjective predicates differ from objective ones based on the pragmatic distinction speakers make between objective facts and arbitrary matters of linguistic practice. They give an example where two speakers agree on the objective fact that Lee eats oysters, but not meat, but disagree on whether eating oysters makes him a non-vegetarian or not and hence whether Lee is in the extension of vegetarian. In order to model this formally, they introduce information states, s ⊆ W, which are sets of possible worlds and the set of all information states, S = ℘(W). They furthermore assume that the context c gives a context set of s _c of what is in the common ground, as well as two functions: κ _c and κ c * . κ _c then derives a set of counterstances for an information state. The formal definition is given in (26).

These counterstances agree with the information state with respect to all objective facts, but disagree on how to resolve matters of linguistic practices. For example, in all counterstances Lee does not eat meat, but does eat oysters, however, in these counterstances vegetarian is a property of individuals who don’t eat meat, but either a) includes or b) does not include those who eat oysters. A proposition such as Lee is vegetarian is therefore subjective, or counterstance contingent,^[14] if there is at least one counterstance within the set of generated counterstances κ _c (s) where the proposition is not true, i.e. if there is at least one counterstance where Lee is not within the extension of vegetarian. Counterstance contingency can be defined formally as in (27).

Kennedy and Willer (2016) argue that speakers can make certain stipulative discourse moves which act to fix certain contextual parameters and therefore help to resolve uncertainties in meaning in discourse. For example by determining the kinds of eating habits that should be considered when deciding if someone is a vegetarian. They term this COORDINATION BY STIPULATION. However not all contextual parameters can be naturally fixed, for example what makes something tasty or what characteristics make someone fascinating. They highlight this distinction with the examples in (28) and (29). According to Kennedy and Willer (2016) the examples in (28) seem natural, whereas those in (29) appear odd.

Formally, this difference is modelled using the second function provided by the context set, κ c * . κ c * partitions κ _c (s) into “a set of equivalence classes of counterstances that agree on parameters that support coordination by stipulation” (Kennedy and Willer 2016, p. 919). κ c * is formally defined as follows:

As such, propositions with predicates such as tasty are said to be radically counterstance contingent meaning that no matter how the counterstance space is partitioned, within each partition there will always be one counterstance where the proposition is false. The definition for radically counterstance contingent propositions is given in (31).

Adopting this approach, Kawahara (2020) then argues that propositions with predicative ideophones in Japanese such as karikari, sakusaku, paripari (Engl: crispy) are radically counterstance contingent. However, according to Kawahara (2019, 2020 these ideophones are gradable and related to a subjective scale, meaning that, although they cannot combine with measure phrases, they can be compared when related to the same subjective scale. This means that these ideophones differ from other subjective predicates, as they can be sorted into sets based on the shared standard for their scales. Speakers can therefore select different ideophones from among this set based on their stance. Formally, this can be expressed as in (32).

Kawahara argues therefore, that for a sentence such as (33-a), a speaker may respond as in (33-b). Here the speakers both agree that the pie has the general property of being crispy, but disagree on the exact properties of the pie that make it crispy.

Kawahara claims that this is where the sound symbolic nature of ideophones comes into play; the speaker of (33-b) may feel that sakusaku better depicts the crispiness of the pie, for example, due to its form better representing the manner in which the layers of the pie break. The approach proposes then that the ideophones encoding the core meaning of crispiness in Japanese form a set and that speakers may choose from the ideophones in this set based on their subjective view of which ideophone best depicts a given referent.

This analysis is intuitive, as it is likely that speakers interpret the iconicity of ideophones in idiosyncratic and subjective ways and may therefore have a preference for one ideophone over another. The examples Kawahara uses are, though, all predicative ideophones, which we argue necessarily have an at-issue contribution.^[15] Nevertheless, we believe that Kawahara’s account provides a good analysis of the dual nature of predicative ideophones where part of the meaning is at-issue,^[16] but the iconic component of the ideophone remains non-at-issue (see Section 4.1.3 for further discussion of this point).

4.1.2 Henderson (2016)

Henderson (2016) provides an account of ideophones as demonstrations based on data from Tseltal. To do this, he adapts and formalises the analysis of quotation as demonstration given by Davidson (2015). The basic ideophone construction in Tseltal is formed by combining the verbal stem of the ideophone with the reportative speech marker chi. However, Henderson argues that ideophones are not simply quoted in Tseltal, but that the basic ideophone construction represents an ideophone-specific form of demonstration. He analyses ideophone stems (without the reportative marker) as predicates of events, as in (34) and introduces an operator ideo-demo,^[17] which is included in the basic ideophone construction. Just as Davidson (2015) argues that the demonstration argument in spoken quotations in English is introduced by the be like construction, ideo-demo is introduced through the use of the ideo + chi construction in Tseltal.

This operator selects for ideophone stems in the syntax and in the semantics gives an expression that can be embedded under chi; in other words, ideo-demo “[…] takes a linguistic expression […] and derives a relation between demonstrations and events” (p. 673). Henderson furthermore argues that the ideophone satisfies the theme argument of the basic ideophone construction with the reported speech marker chi, similarly to how the quoted utterance would satisfy the theme argument in a be like construction. In order to become the theme of the basic ideophone construction, the ideophone must therefore be of a particular type, namely a linguistic entity. Henderson (2016) simplifies Potts (2007a) and takes linguistic entities as pairs, ⟨string, semantic representation⟩, where the string is the orthographic/phonological representation of the natural language expression and the semantic representation is the lambda term denoting the appropriate function. For example, the unquoted natural language expression woman is translated to the lambda term as in (36), whereas the quoted natural language expression “woman” is translated as the constant of type μ^[18] whose denotation is the pair of the unquoted string and its denotation, as in (37).

The ideo-demo operator therefore requires that the utterance of an ideophone is itself a demonstration and that there be a structural similarity between the ideophone utterance event and the event it depicts. Davidson (2015) intentionally underspecifies the relation between demonstrated and demonstration event, requiring only that the demonstration reproduce salient aspects of the demonstrated event. However, Henderson (2016) argues that the event depicted by the ideophone must satisfy the relevant aspects of the ideophone’s lexical content. For example, only events including an inhaling sound can be depicted by the ideophone jik’. The struc-sim _⌞u⌟ relation provides this condition. struc-sim _⌞u⌟ ensures that the “utterance of an ideophone as a linguistic object” stands for an event that satisfies the ideophone predicate. Formally, this is the case if the demonstrated event can be partitioned so that:

1. all subevents satisfy the relevant aspects of the ideophone predicate’s lexical content;

2. the cardinality of the partition is equal to or greater than the number of atomic parts of the demonstration;

3. there is a temporal similarity between the partition and the atomic parts of the demonstration.

Henderson gives an example using the Tseltal ideophone tsok’, which encodes the sound of something frying in oil, seen in (38).

Henderson (2016) gives the truth conditions of (38) as:

1. there is an event e that takes place in the lard and the agent is x₁ (an individual given by the context or variable assignment).

2. the demonstration event has the linguistic object tsok’ as its theme.

3. the demonstration event is structurally similar to e:

   1. As d₁₃ is an atomic event, e must also be partionable into an atomic event (trivial partition).

   2. e must satisfy the predicate λe[TSOK′(e)], i.e. it must be an event of frying sound emission.

This is shown formally in (39).

This approach directly integrates the ideophone into the truth conditions of the sentence, and while Henderson (2016) does not make explicit reference to the at-issueness status of these ideophones, we argue that in these cases the ideophone is predicative and is therefore necessary for both the well-formedness of the sentence and its semantics and is hence used for the computation of the at-issue semantics. Similarly, the combination of the ideophone with the reported speech particle clearly marks the ideophone as part of a quotational or demonstration structure, which as we argue in the next section, also renders the ideophone necessarily at-issue.

4.1.3 A supplemental approach

Here we present an initial, supplemental analysis for adverbial and predicative ideophones which draws upon parts of the approaches in Kawahara (2020) and Henderson (2016) and is based upon data from ideophones in German. However, before we discuss the formal details of the approach, we would like to highlight three factors which appear to impact on the at-issueness status of ideophones (see Barnes et al. 2022 for a more detailed discussion of these factors).

morphosyntactic integration. Crosslinguistically, it has been noted that ideophones demonstrate different behaviour depending on their syntactic category (cf. Kita 1997, 2001; Toratani 2016). For example, Dingemanse (2017) argues that more morphosyntactically integrated ideophones in Siwu, such as predicative ideophones, are, in contrast to adverbial ideophones, able to be negated, used in questions and can contribute old information. These properties appear to indicate more at-issue behaviour and indeed, we argue that predicative ideophones must be partly at-issue; they are essential to the integrity of the sentence and must make some at-issue contribution in order for the sentence to be interpretable. However, we claim that predicative ideophones are mixed items (cf. Gutzmann 2011; McCready 2010), similar to Pottsian expressives such as Köter ‘mutt’, which has the at-issue contribution of denoting a dog and the non-at-issue contribution that the speaker has a negative attitude towards said dog.

Dingemanse (2017) furthermore argues that there may be a typological scale of morphosyntactic integration of ideophones, where in languages with more integrated ideophones, such as Somali, the ideophones tend to be less expressive, while in languages where ideophones are less integrated, as in Semai, the ideophones are more expressive. Seeing as we predict that more integrated ideophones are also more at-issue, we could then also predict, based on this scale, that languages with more integrated ideophones will be more likely to have more at-issue ideophones than those with a lesser degree of morphosyntactic integration.

quotation and demonstration with ideophones. Drawing upon the analysis of ideophone demonstration in Henderson (2016), we propose that ideophones which are quoted or used in demonstrations are also shifted towards at-issueness status. When an ideophone is accompanied by a quotation or demonstration marker, it is directly referred to in the speech and must therefore be partly at-issue in order for the sentence to be felicitous. We furthermore argue, in line with Davidson (2015) and Ebert and Hinterwimmer (2022), that demonstratives such as like in English or so in German can act as quotation/demonstration markers and therefore an ideophone accompanied by a demonstrative will also be shifted towards at-issueness status.

alignment and timing. Dingemanse (2013) notes that ideophones at clause edges generally have more prosodic foregrounding and expressive morphology than those embedded within other constructions, which would likely make them more prominent prosodically and semantically and therefore could indicate that they are more at-issue.

Recall that we find the same tendency as discussed here for ideophones also with gestures: both tend to be more at-issue towards the end of a sentence or the more standalone they appear to be. Both also appear to be shifted towards at-issueness by way of demonstratives. And both are interpreted as at-issue when they form an integral component of an otherwise incomplete utterance.

In general, the crosslinguistic literature indicates several properties of ideophones which are similar to those of non-at-issue content and in particular, co-speech gestures and Pottsian supplements. This includes the fact that ideophones are generally not negated, nor used in questions and that they tend to provide background information (cf. Dingemanse 2017; Kita 1997, 2001; Toratani 2016). These properties are mostly true of adverbial ideophones, which appear to be the most common realisation of ideophones crosslingusitically (cf. Akita 2009).

We propose, therefore, that adverbial ideophones have a similar meaning contribution to co-speech gestures as per the supplemental approach outlined in EEH. While there has been little semantic work on ideophones, other linguistic research provides some evidence to support an analysis of adverbial ideophones as supplements. Kita (1997, 2001 in fact provides an analysis of ideophones in Japanese that closely resembles the multidimensional approach to CIs given by Potts (2005), where he argues that standard linguistic information occurs in the ‘analytic’ dimension and ideophones in the ‘affecto-imagistic’ dimension. The following examples illustrate how the four key properties of supplements as described by Potts (2005) seem to apply to most cases of adverbial ideophones.

nondeniability. Similarly to appositives and other non-at-issue content, it appears that adverbial ideophones cannot be directly denied, but should rather be addressed via a discourse interrupting interjection, such as in the Hey, wait a minute! test. This is shown in (40).

scopelessness. Adverbial ideophones, like appositives, also appear to be odd when they appear in the scope of negation, as in (41) and (42), where we argue that they are only acceptable as part of meta-linguistic utterances, used in response to a previous utterance.

Other researchers have also noted that adverbial ideophones cannot appear in the scope of logical negation, including Dingemanse and Akita (2016) for Siwu and Kita (1997, 2001) and Toratani (2016) for Japanese.

antibackgrounding. Supplements are said to generally contribute new information, as was pointed out above, where we also critically discussed this claim (see footnote 5). Likewise, it has often been noted that adverbial ideophones tend to only contribute new information (see Dingemanse 2017).

However, as in the case of supplements, it is at least debatable whether this holds in general. Consider the following example.

Although the continuation in example (43) seems a bit odd with the redundant ideophonic information about how the frog went up the stairs, it is certainly not illicit.^[19] We leave the discussion about the antibackgrounding condition for supplements and ideophones for future research.

nonrestricting. Supplements cannot be used to restrict the at-issue content of the utterance in which they occur. (44) appears to make little sense unless a very specific context is set up, for example one where boxes are on the stairs in such a way that Peter would need to zig-zag between them to get down the stairs. However, in this case we argue that the ideophones would be somewhat meta-linguistic and this would therefore not be a true example of restriction.

Given these properties of ideophones and based on the evidence gathered in experimental work by Barnes et al. (2022) (see Section 4.2.1), we present a formal analysis of adverbial ideophones building upon both the analysis that EEH give for iconic co-speech gestures and the analysis of gestural-quotational demonstration by Ebert and Hinterwimmer (2022). Here, we still adopt a binary view on at-issueness, which we will refine in Section 6 below.

We argue that ideophones have two meaning components, like mixed items such as Köter ‘mutt’ do. The first is the ideophone’s conventionalised meaning. In contrast to ordinary mixed items, however, both meaning components of ideophones contribute non-at-issue information, while for Köter it is assumed that the meaning that the individual in question is a dog is at-issue.^[20] Hence, the conventionalised meaning contribution of an adverbial ideophone is the same as that of a standard adverbial, namely an event modifier. For example, the use of plitsch-platsch in (45) indicates that the frog went up the stairs in a splashing manner. The second meaning contribution is the iconic meaning of the ideophone. Dingemanse (2013) argues that varying aspects of the ideophone’s utterance contribute to its iconicity, including phonology, prosody and gesture. In order to capture the iconic nature of the ideophone, we follow Henderson (2016) and model the utterance of an ideophone as a demonstration, per Davidson (2015). This enables us to capture all elements of the utterance which contribute to the iconic mapping of the ideophone to the event. Unlike Henderson (2016), however, we do not attempt to specify the exact iconic relation between the event and the ideophone utterance or to integrate this into the truth conditions. Instead, we argue that the utterance of an ideophone alongside a report of an event results in the default non-at-issue inference that the ideophone utterance iconically depicts the event described by the main assertion. We choose to model this using the SIM predicate, which allows for these varying aspects to contribute to the iconic depiction in differing manners depending on the individual utterance and the context in which it occurs.

Applying this to the previous examples, we can then give an analysis of the form of (46) for an utterance such as (45). We again adopt the propositional variables p and p* to mark at-issue and non-at-issue content respectively.^[21]

The at-issue contribution of the utterance then says that there is an event of a frog going up the stairs. The non-at-issue contribution is that this event has the property of being a splashing event and that there is a demonstration, namely the utterance of plitsch-platsch, which is similar in the relevant dimensions to the event of the frog going up the stairs.

As in the cases of co-speech gestures, adverbial ideophones accompanied by a demonstrative also appear to shift towards at-issueness, as in (47). While a plain ideophone seems to be odd as an answer to a wh-question as in (47-b) (see (47-c)) due to its non-at-issueness status, shifting at-issueness via a demonstrative improves the ideophone as an answer (see (47-d)).

We therefore analyse (48) as in (49).

The at-issue contribution of the utterance then says that there is an event of the frog going up the stairs, which has the property of being a splashing event and that there is a demonstration, namely the utterance of plitsch-platsch, which is similar in the relevant dimensions to the event of the frog going up the stairs. The meaning contributions of the ideophone have been shifted to at-issueness status by the demonstrative.

Adverbial ideophones appear to also behave like co-speech gestures, in that they seem to receive a more at-issue interpretation when shifted to the end of the sentence they occur in, as in (50), where it seems that the ideophone can be directly denied. At least such a response seems to be an improvement when compared to (40-c) as an answer to (40-a).

As previously noted, this analysis applies only to adverbial uses of ideophones. In German, it appears to be possible to make a predicative structure with an ideophone, although in these cases there seems to be a requirement for an auxiliary verb to form the predicate. As can be seen in (51), these predicative uses can be directly denied, in contrast to adverbial uses, suggesting that these ideophones may be partly at-issue.

We propose the analysis in (52) for an utterance such as (51-a).

Here, the predicative ideophone takes the event as an argument giving the at-issue contribution that some sort of event of splish-splashing occurred, in which the frog was the agent. However, there is still the non-at-issue contribution that there is a demonstration, namely the ideophone utterance, which is similar to the main event, thus encoding the iconic component of the predicative ideophone. This demonstrates the mixed item analysis we mentioned for predicative ideophones earlier in this section. Predicative ideophones, like ordinary mixed items, make some sort of at-issue contribution, but still maintain a non-at-issue iconic contribution, just as Köter provides at-issue reference to the property of being a dog, while also providing the non-at-issue implication that the speaker has a negative opinion of dogs.

There also appear to be some ideophones in German, such as ratzfatz and ruckzuck, that, even when used adverbially, have this mixed interpretation. In (53), it seems to be possible to directly deny the ideophone contribution.

The question then arises as to why these ideophones can receive a mixed interpretation, even when used adverbially, whereas other ideophones in German, such as plitsch-platsch, cannot. One potential explanation comes from Dingemanse (2017), who highlights that in Siwu, more frequently used ideophones are more likely to be able to be used predicatively and hence behave more like ordinary words and are more prone to deideophonisation. It could be then that ratzfatz and ruckzuck have become more conventionalised due to frequent use and therefore the conventionalised part of their meaning has shifted to at-issueness status. We would expect more at-issue ideophones undergoing a process of deideophonisation to occur more frequently overall, but particularly in written language. An informal survey of German corpora seems to support this claim; the DWDS-Kernkorpus (1900–1999); the Metakorpus WebXL (01/1995-03/2021); the DWDS-Zeitungskorpus (from 1945) and the ZDL-Regionalkorpus (from 1993), give the following combined frequency values for ratzfatz, ruckzuck, holterdiepolter and plitsch-platsch:

1. ruckzuck:

   1. 5,530 entries = Frequency Level 1

2. ratzfatz:

   1. 10,200 entries = Frequency Level 1

3. holterdiepolter:

   1. 1,239 entries = Frequency Level 1

4. plitsch-platsch:

   1. 48 entries = Frequency Level 0

The DWDS lists 7 frequency levels (0–6), with frequency level 0 covering items with 5–1,029 entries and frequency level 1 items with 1,030–10,295 entries. While we have to leave further investigation of this question to future research, we believe it is worth pursuing and we would like to also compare the frequency of these ideophones in written and spoken language, looking at when specific ideophones first appeared and seeing if there are particular peaks in their frequency and how their usage has developed over time.

Our approach is not incompatible with either Kawahara (2020) or Henderson (2016) and in fact incorporates elements of both approaches. The main difference, however, is that our analysis makes explicit reference to and can account for the at-issueness status of ideophones in both adverbial and predicative structures. The next section will build upon this theoretical basis by discussing experimental work on the at-issueness status of ideophones.

4.2.1 Barnes et al. (2022)

The research goal of Barnes et al. (2022) was to establish whether sentence-medial adverbial ideophones in German contribute non-at-issue content. It adapted the experimental design of EEH, with two variants of the experiment, one using visual stimuli, the other audio stimuli.

The first variant of the experiment used a 2 × 2 design, which crossed the two category conditions, ideophone and adverbial, with the two match conditions, match and mismatch in a Latin square design. The two category conditions were realised through 20 target sentences and the variation between ideophone and adverbial was realised as in (54). The adverbials were selected based on two pre-studies, which were used to identify the best paraphrases or equivalents for the ideophones.

The variation in the match condition was implemented through fairytale contexts, which were chosen as it was hoped that ideophones would appear more natural and less marked in such scenarios than in other, more everyday ones. An example of the contexts can be seen in (55).

The procedure for the experiment was as follows; participants were shown the context and asked to read it carefully before also reading the target sentence. They were then asked to rate how well the target sentence matched the circumstances described in the context on a scale from 1 to 5, where 5 indicated a perfect match and 1 that they did not match at all.

The experiment assumed the same theoretical background as EEH, and predicted that if ideophones are non-at-issue and hence irrelevant to a constructed generalized QUD, then they should have less impact on participants’ judgements of how well a sentence matches a given context than adverbials, which are taken to be at-issue. Or in other words, the mismatch effect should be stronger for adverbials than for ideophones.

This hypothesis was supported by the results of the experiment, which showed that the mismatch effect was larger for the adverbials than for the ideophones, indicating that ideophones are less at-issue than adverbials. However, ideophones were rated more poorly than adverbials in the matching conditions, which were taken to be indicative of the general markedness of ideophones and the fact that speakers will generally disprefer them compared to ordinary words.^[22] (See Figure 4 for results).

Figure 4:

Results of audio and visual experiments (Barnes et al. 2022).

The second variant of the experiment was conducted in order to establish whether ideophones would be less marked when spoken than when written. The experiment used the same design and materials as the first variant, except that the target sentences were spoken by a native speaker instead of written.

The procedure was the same as the first variant, except that after reading the context, speakers clicked on a play-icon to hear the spoken sentence and then rated how well the sentence matched the circumstances in the context.

The hypothesis was that the experiment would replicate the results of the first experiment, but that ideophones would be rated more favourably in the match conditions due to being spoken. While the results did replicate the first experiment in showing a larger mismatch effect for the adverbials than for the ideophones, they did not show significantly more favourable ratings for ideophones in the match conditions.

Notably, this experiment indicates that while sentence-medial adverbial ideophones in German are not at-issue, they appear to be more at-issue than co-speech gestures. The differences in the mismatch effects for ideophones and adverbials seem to be smaller than the difference between gestures and adjectives in EEH.

Having seen the evidence for the at-issueness status of iconic enrichments such as gestures and ideophones and how this status can be affected by various factors, in the next section we will return to the concept of binary at-issueness and discuss how, with respect to iconic enrichments, a categorical distinction between at-issue and non-at-issue is inadequate.