Within formal linguistics, it is generally assumed that form-meaning relations in spoken language are arbitrary and not iconic. De Saussure (1916) argued that there is no logical or intrinsic relationship between the linguistic sign and the object or concept to which it refers. For example, there is no logical reason a chair is called chair; it could just as well be called table. Instead of iconic mappings, speech communities rely upon arbitrary, conventionalised forms in order to communicate. Linguistic forms that do employ iconicity, such as onomatopoeic words like bang or ideophones like splish splash, have been considered exceptions to this rule, with arbitrariness being considered a major principle in natural language. Hockett (1960) even named arbitrariness as a design feature of human language.

Recently, however, researchers have begun to examine iconicity in spoken language more closely. Blasi et al. (2016) found strong associations between basic vocabulary items and specific human speech sounds. Upon analysing word lists of 100 basic vocabulary items from almost two thirds of the world’s languages, Blasi et al. (2016) found that a large proportion of these languages show biases for carrying or avoiding specific sound segments. Biases were particularly common in items referring to body parts, with “tongue”, for example, often being associated with the lateral “l” and “nose” with the alveolar nasal “n”. This data raises questions about how arbitrary conventionalised linguistic items are, with iconicity seeming to play a larger role than previously thought. Furthermore, Blasi et al. (2016) argue that these associations emerge independently from one another, suggesting that they are not due to vocabulary items sharing linguistic origins, but rather that languages share preferences for sound-symbolic encodings.

Perniss et al. (2010) have also argued for the importance of iconicity in language after reviewing iconic mappings in signed and spoken language. It has often been claimed that sign languages are naturally more iconic than spoken language due to their occurrence in the visual modality, which, it is argued, allows for a greater exploitation of iconicity. Perniss et al. (2010), however, found that iconic mappings are used across both the spoken and signed modality and that both speakers and signers exploit iconicity in language processing and acquisition. For example, they cite studies which indicate that children may acquire iconic vocabulary more easily than arbitrary vocabulary. They argue that while arbitrariness still plays a crucial role in language, iconicity should also be considered as a general property of language. Flaksman (2017) has gone a step further and argued that language has a requirement for expressivity, which is satisfied by iconic words. She proposes the “iconic treadmill hypothesis”, where iconic words necessarily lose their iconicity over time, due to the pressure of the arbitrary system, meaning that they also lose their expressivity. This loss of expressivity then forces the introduction of new iconic coinages, which explains the continuing emergence of new iconic words across languages.

It has also been suggested that gestures should be taken into account when considering iconicity in spoken language, as they may allow speakers the same iconic expressiveness that is available to signers (cf. Goldin-Meadow & Brentari 2017; Schlenker 2018c). Indeed, in pioneering work conducted by Kendon (1980) and McNeill (1992), it was found that gestures can introduce information additional to that contributed by the accompanying speech. Nevertheless, while gestures have been of great interest in semiotics and other fields of linguistics, such as language acquisition, only recently has the nature of their meaning contributions become of interest to semanticists. Ebert et al. (2020), Schlenker (2018a) and Esipova (2019) have all proposed theories concerning the meaning contributions of iconic gestures, focusing on their pragmatic status and in particular on whether they are at-issue or not (see Section 2.3).

Similarly, there has been much research into the crosslinguistic typology of ideophones and their sound symbolism, but there has been little to no research into their pragmatic status and only little consideration given to their semantics (cf. Dingemanse 2012; 2013; Dingemanse et al. 2016; Henderson 2016; Kawahara 2020). In this research, we aim to give an initial overview of the evidence around the (non-)at-issueness of ideophones in the existing crosslinguistic literature before looking more closely at the at-issue status of ideophones in German and gathering initial experimental data using the framework established by Ebert et al. (2020) in their work on iconic co-speech gestures.

In the remainder of this section, we will provide an explanation of what is meant by at-issue status, as well as the theoretical background on at-issueness, before moving on to discuss the research conducted into the at-issue status of iconic gestures.

It has long been acknowledged that information contributed by conventional implicatures and presuppositions appears to differ from that of standard assertions; this information is not at-issue. Potts (2005) introduced the terms at-issue and not-at-issue to describe what Paul Grice (1975) termed “what is said” and “what is implicated”. At-issue information is that which the speaker primarily wants to focus on in conversation; it indicates the direction in which the speaker wishes to steer the conversation. Not-at-issue information, on the other hand, is that which the speaker provides as additional information, but that they do not want to pursue in the conversation. Simons et al. (2010) argue that at-issueness is determined by relevance to the Question Under Discussion (QUD). An assertion is relevant to the QUD if it contextually entails a complete or partial answer to the QUD, whereas a question is relevant if it has an answer which contextually entails a complete or partial answer to the QUD. Simons et al. (2010) define relevance to a QUD using the yes/no question associated with a proposition; ?p denotes the question whether p, or the partition of the set of worlds into p and ¬p. They then give the following formal definition of at-issueness (p.323):

According to this definition, at-issueness depends predominantly on the QUD; if information is a potential answer to the QUD, then it is at-issue. Simons et al. (2010) also highlight that some constructions and lexical items conventionally mark their content as not at-issue. In such cases, if the sentence is otherwise felicitous, speakers will assume the item linguistically marked as not at-issue does not address the QUD. Nevertheless, they also argue that such items are marked in a way that allows the items to be interpreted as at-issue, when necessary.

Expressives such as damn are common instances of lexical items with not-at-issue content, as can be seen in (4).

Here, what is at-issue is that Ed refuses to look after Sheila’s dog; the speaker asserts this and we can see intuitively that it is the main aim of their utterance. However, damn provides additional information about the speaker’s attitude to Sheila’s dog or the general situation at hand, for example, that the speaker does not like dogs or disapproves of Ed being asked to look after the dog. This meaning is not at-issue and is not put up for discussion by the speaker; they are not attempting to address the QUD via damn. As such, while it is perfectly possible to target the asserted, at-issue information with a direct denial, the not-at-issue contribution cannot be targeted in this manner, but must be addressed via a discourse interrupting interaction, as in (5) (see von Fintel 2004, who originally proposed the ‘Hey wait a minute’ as a way to test for presuppositional content; Potts 2015 who extended the test to all not-at-issue content; and Syrett & Koev 2014 who provide critical discussion of the limitations of this test).

Other key characteristics of not-at-issue information are that it cannot be interpreted in the scope of modal operators or negation, i.e. it projects, and it can be ignored under ellipsis (Potts 2005).

Alongside expressives, appositives are another oft-cited source of not-at-issue information. This can be seen in (6), where again it intuitively seems that the speaker does not want to discuss whether or not Peter is the best musician in town, but rather their aim is to communicate that Peter came for dinner the previous evening.

However, in both experimental research conducted by Syrett & Koev (2014) and theoretical work by Anderbois et al. (2013) and Nouwen (2007), it has been shown that while nominal appositives, such as in (6), and sentence-medial appositive relative clauses (i.e. who is the best musician in town) are predominantly not at-issue, sentence-final appositive relative clauses can be at-issue, as in (7), where the appositive can be targeted by direct denial. This indicates that the structural position of information can also impact upon its at-issue status.

Syrett & Koev (2014) argue that the reason why who is the best musician in town can be at-issue in (7), but not in (6) is because appositive relative clauses are able to compete with the main assertion for at-issue status. By varying their order relative to the main clause, they become more viable candidates for at-issue status. This explanation is compatible with a graded view of at-issueness (cf. Ebert 2017; Tonhauser et al. 2018), where information competes for at-issue status, so that the more standalone a piece of information is, the more likely it is to be at-issue; hence why appositive relative clauses occurring sentence finally would be better candidates for at-issue status.

Contrary to traditional literature on not-at-issue content, which claims that such items do not straightforwardly contribute to truth values (cf. Bach 1999; Dever 2001), Syrett & Koev (2014) argue that their experiment shows that false information presented in an appositive will impact on overall truth conditions, resulting in a truth value judgement of false for the entire sentence. However, experimental research by Kroll & Rysling (2019) investigated the impact of false information presented in appositives and main clauses on truth value judgements. Their findings indicate that information which is false, but also irrelevant to the QUD will, in fact, have less impact on truth values than false information which is relevant to the QUD. Assuming Simons et al. (2010)’s definition of at-issueness as relevance to the QUD, we can then surmise that information which is not relevant to the QUD and therefore not at-issue will have less of an impact on truth values. Interestingly, Kroll & Rysling (2019) argue that this impact is seen regardless of whether the false information is presented in an appositive or not. It is possible then that participants in Syrett & Koev (2014)’s experiment interpreted the false information in the appositives as relevant to the QUD and therefore at-issue, hence the greater impact that this had on truth value judgements. Whether this is due to their experimental design or a property of appositives themselves remains to be seen.

As well as being well attested in arbitrary linguistic items, research conducted by Ebert et al. (2020) and Schlenker (2018a) has shown that iconic co-speech gestures, similarly to expressives and appositives, can also contribute not-at-issue information in addition to the information contributed by the speech with which they co-occur. There has, up to now, been no experimental work conducted on ideophones. However, Dingemanse (2013) has argued that they do not describe, but rather depict, meaning that, through the use of an iconic performance, ideophones illustrate an event rather than using arbitrary linguistic signs to refer to concepts. Although this in itself does not necessarily indicate that ideophones are not at-issue, it does imply that they contribute information in a different manner to arbitrary items, which could suggest that they are not at-issue. A further indication of ideophones’ not-at-issue status is that they appear to share characteristics of typically not-at-issue items, namely that they appear to not be subject to negation (cf. Kita 1997; 2001; Dingemanse 2017; Toratani 2018) and similarly to supplements they tend to provide new information rather than old or backgrounded information (cf. Dingemanse 2017).

In the following section, we will provide background on the aforementioned studies that have been conducted on iconic gestures, focusing particularly on Ebert et al. (2020), as our research on ideophones uses an adapted version of their experimental design.

Gestures are for the most part non-conventionalised, non-combinatorial and non-recursive and are typically performed simultaneously with speech (co-speech gestures) (Kendon 1980; McNeill 1992). Co-speech gestures can, as previously discussed, add new information in addition to the speech (Kendon 1980; McNeill 1992), although all co-speech gestures typically rely on the context in which they occur for interpretation (cf. McNeill 1992; Lücking et al. 2013). There are also a variety of different gestures, including emblematic gestures, such as the thumbs up gesture, beat gestures and turn-taking gestures (McNeill 1992). In this paper, we will only be concerned with iconic gestures.

Several studies on iconic co-speech gestures have found evidence for their default not-at-issue status. Tieu et al. (2017; 2018), for example, performed experiments with truth-value judgements, picture-matching tasks and inferential judgements showing that co-speech gestures project in a way that would be unlikely if they were at-issue. Ebert et al. (2020) also provided evidence for the not-at-issue status of co-speech gestures in an experiment using a sentence-picture matching task.

The experiment used a 2 × 2 design with two mode conditions, iconic co-speech gestures and adjectives, and two match conditions, match and mismatch. In the experiment, participants were shown a video of a female speaker uttering a sentence which was either accompanied by a gesture, which was temporally aligned with an NP, or an adjectival description of an NP. The video was paired with a picture, which either matched the sentence or was a mismatch, as in Figure 1. Participants were then asked to rate how well the given description matched the image.

According to traditional literature on not-at-issue meaning, speakers will struggle to judge a sentence as true or false when it contains false not-at-issue information, but true at-issue information (cf. Bach 1999; Dever 2001). Per Bach (1999) speakers will find something both wrong and right with such a sentence. As such, this experiment employed an indirect test of at-issue status, where participants evaluated the truth of the target sentence relative to the situation given in the picture by rating how well the circumstances in the target sentence matched those in the picture. This use of a scale and meta-question, in place of directly asking if the sentence was true or false, allowed for subtleties of judgements, with participants still being able to indicate that they disliked sentences with false not-at-issue content. There was also an implicit, generic QUD such as ‘What is happening?’, which was assumed for the whole experiment. It was expected that speakers’ judgements concerning how well the description given in the video matched the picture they were shown should be more strongly impaired by mismatches induced by information deemed relevant to the QUD, i.e. at-issue material, than by mismatches induced by information not relevant to the QUD, i.e. not at-issue material. The specific hypothesis of the experiment was, therefore, an interaction of the factors mode and match such that, if iconic co-speech gestures are not at-issue, then the mismatch effect, i.e. the difference in ratings for matching and mismatching conditions, would be significantly larger for mismatches induced by adjectives, which clearly contribute at-issue information, than those induced by gestures.

The mismatch effect was significantly larger for sentences containing adjectival descriptions than those accompanied by iconic co-speech gestures, supporting the hypothesis of the experiment and therefore indicating that iconic co-speech gestures are not at-issue.

Although Schlenker (2018a) and Ebert et al. (2020) agree upon the default not-at-issue status of iconic co-speech gestures, the exact contribution of these gestures is still disputed. Schlenker (2018a) argues, for example, that they are a special type of presupposition, which he terms cosuppositions. Ebert et al. (2020), however, give an analysis of iconic co-speech gestures as Pottsian supplements, treating them as equivalent to appositives. These two theories make different predictions about the behaviour of the gestures in terms of their informativity and how they behave when they are embedded, for example under negation and negative quantifiers.

Co-speech gestures can also contribute at-issue information, although again, there are differing analyses as to when and why this is the case. Schlenker (2018a) argues that co-speech gestures, as cosuppositions, can be locally accommodated in the same manner as presuppositions. Ebert et al. (2020), on the other hand, propose that co-speech gestures can only be at-issue if they are accompanied by a dimension shifter. Dimension shifters can include particular prosody, focus or facial expressions, such as an eyebrow raise. Ebert et al. (2020) have also shown experimentally that demonstratives can shift co-speech gestures to at-issue status and have argued for demonstratives as dimension shifters. They repeated the previously described experiment, this time including a third MODE condition, where the iconic gesture was accompanied by the German demonstrative SO. All other aspects of the experiment were kept the same. The results for adjectives and iconic gestures without a demonstrative remained the same, however, in the mismatching condition, iconic gestures accompanied by a demonstrative showed a larger mismatch effect than those without a demonstrative, although not as large as that seen for the adjectives. This indicates that the demonstrative SO shifts the co-speech gesture towards at-issueness, but also suggests that at-issueness is not necessarily binary, but rather gradient. In contrast to Ebert et al. (2020) and Schlenker (2018a), Esipova (2019) presents a third view of when iconic gestures are at-issue or not, arguing that at-issueness is independent of modality and that at-issue and not-at-issue content should instead be considered in terms of restrictive and non-restrictive modifiers, respectively.

As noted by McNeill (1992), the temporal alignment of speech and gesture also plays an important role and it appears that it can also affect the gesture’s pragmatic status, just as the structural positioning of appositives affects their at-issue status (Syrett & Koev 2014). In a graded approach to at-issueness (cf. Section 2.2), co-speech gestures always face competition from the accompanying speech for at-issue status and therefore, unless they are shifted to at-issue status, for example by being accompanied by a demonstrative such as SO, they are generally not at-issue. However, in cases where gesture replaces speech completely, so-called pro-speech gestures (cf. Schlenker 2018b), there is no competition and so the gesture is at-issue. Ebert (2017) has also proposed that post-speech gestures are more likely to be at-issue due to the lack of competition with speech and their standalone temporal positioning, which would demonstrate another similarity between iconic gestures and appositives. It would then also be expected that pre-speech gestures are more at-issue, although these may be deemed more odd as they do not have a gesture referent given by previous speech.

In the following section, we will turn to ideophones and discuss previous research on these, focusing particularly on the predictions this makes for their at-issue status and the factors that may play a role in this, before looking at how this aligns with the evidence from ideophones in German.

Up to now, research on ideophones has predominantly focused on sound symbolism and iconicity, as well as their crosslinguistic typology, which includes exploring the semantic categories that can be expressed by ideophones. However, there has been little investigation of their semantics or pragmatics (see Henderson 2016; Kawahara 2020 for semantic analyses of ideophones in Tseltal and Japanese, respectively). In discussing their pragmatic properties, Nuckolls (1992) has argued that ideophones are an example of sound-symbolic involvement, where the speaker iconically simulates the salient features of an event and in doing so, encourages the listener to project themselves into the simulation and they, therefore, become involved in the narrative. Tolskaya (2011) has also given an account of ideophones, in which they are restricted to reports of events that the speaker themselves has witnessed, although the prevalent use of ideophones in narration and storytelling makes this theory seem unlikely. Dingemanse (2011) instead argues that due to their depiction of sensory imagery, ideophones index epistemic authority, where the speaker has relative authority over their interlocutor(s) with respect to what is said, which seems to encapsulate the core idea of Tolskaya (2011)’s analysis, without making such a strong claim. As far as we are aware, though, there is no research which directly investigates the at-issueness of ideophones. Nevertheless, the crosslinguistic research is rich enough that we are able to make reasonable predictions about the (not)at-issueness of ideophones and the factors that may impact upon it.

As discussed in Section 2.2, for us, the fact that ideophones depict rather than describe, indicates that they contribute information in a different manner to other more arbitrary items.

Other characteristics of ideophones also appear to support a not-at-issue analysis, as they exhibit properties typical of not-at-issue content. Dingemanse (2017) notes that ideophones in Siwu are generally not subject to negation or used in questions and that they also provide new, rather than old or backgrounded information, which Potts (2005) has argued is characteristic of supplements (see Section 5.1 for further discussion of the exact not-at-issue contribution of ideophones). He also indicates that this is the case across many languages and indeed, it has been noted in Japanese by Kita (1997; 2001) and Toratani (2018). Kita (1997; 2001) in fact provides a two-dimensional analysis of Japanese ideophones that shares similarities with the multidimensional account of conventional implicatures given by Potts (2005). In this account, ideophones occur in the affecto-imagistic dimension, while other parts of speech occur in the analytic dimension, which, he argues, explains why ideophones in Japanese cannot be logically negated. Although more recent accounts of not-at-issue content have adopted a uni-dimensional approach (cf. Anderbois et al. 2013) to allow for anaphoric relations between at-issue and not-at-issue content, the analysis provided by Kita (1997; 2001) seems to be fundamentally compatible with a not-at-issue interpretation of ideophones.

There do appear however to be other additional factors which can impact on the at-issue status of ideophones. In particular, the crosslinguistic literature indicates three factors which influence the (not-)at-issue interpretation of ideophones, namely: how morphosyntactically integrated the ideophone is, whether it is accompanied by a demonstrative or quotative marker and its alignment and timing with respect to the main utterance. We will address now each of these factors individually.

As previously noted, German does not share the range of ideophones seen in prototypical ideophonic languages such as Japanese, however, there is sufficient evidence to suggest that ideophones do form part of its linguistic system. Ideophones in German meet the definition given by Dingemanse (2019: p.16) as “[…] an open lexical class of marked words that depict sensory imagery”:

open lexical class Dingemanse (2019) argues that the size of the ideophone class in languages such as Japanese are evidence for describing ideophones as an open lexical class, as they are comparable to other open classes in these languages. German does not have this size of ideophone class, however, corpus work conducted by Ćwiek (submitted) has shown that ideophones are both idiosyncratically manipulated and created in German, as well as being created ad hoc, which Dingemanse (2019) also argues is evidence for ideophones being an open class.6

marked Ideophones in German are marked through their unusual morphophonology, such as reduplication, as can be seen in examples such as husch-husch and zack-zack. This distinguishes them from regular German lexical items.

words German ideophones are conventionalised words, whose meanings can be specified and listed, with allowances for variations due to subjective speaker interpretations.

depict German ideophones also depict rather than describe, as they represent an event through some sort of structural similarity. As an example, plitsch-platsch in (12) iconically represents the sounds of a frog’s wet feet moving on the steps of the stairs.

sensory imagery It could be argued that German ideophones are predominantly sound-symbolic or onomatopoeic, which are categories of ideophones that, as Nuckolls (2019) notes, have been traditionally viewed as marginal and simplistic. However, Nuckolls argues against this view, highlighting that phonetic research shows producing sound-symbolic utterances involves effort in terms of vocal manipulation and that neurolinguistic research has shown the importance of cross-modal relationships between sound and other sensory modalities. Indeed, Nuckolls (2019) highlights that Quichua speakers rarely perceive sound-imitative ideophones as only encoding sound. Similarly, experimental work by Ćwiek (submitted) highlights that many ideophones in German, which could be described as onomatopoeic, actually encode multisensory information. A key example is holterdiepolter, which generally refers to a situation with loud, chaotic movement, roughly equivalent to English helterskelter and which speakers therefore perceive as encoding information about both movement and sound.

In terms of their at-issue behaviour, there appear to be clear parallels between the ideophones described in the crosslinguistic literature and those in German.

Firstly, sentence-medial adverbial ideophones in German demonstrate typical properties of ideophones as described by Dingemanse (2017), such as not being acceptable under negation, as in (9) and (10), which we argue are only acceptable as meta-linguistic utterances, and being restricted to contributing new information, as in (11), where it seems to be strange to use plitsch-platsch after already describing the wet, splashing sounds the frog makes.

As previously discussed, these properties are indicative of not-at-issue content and this is shown for German sentence-medial, adverbial ideophones in (12), where it only seems to be possible to target the information provided by the ideophone with a discourse interrupting interjection, as in (12d).

Furthermore, greater morphosyntactic integration of ideophones in German leads to a reversal of these properties and a more at-issue interpretation, which also fits with crosslinguistic descriptions of the differing behaviour of ideophones in differing syntactic categories (cf. Dingemanse 2017; Dingemanse & Akita 2016; Kita 1997; 2001; Toratani 2018). For example, it seems to be perfectly acceptable to target the ideophones in (13) and (14) with a direct denial.

Although the ideophones in (13) and (14) are clause final, graded at-issueness does not seem to be the cause of their at-issue interpretation.7 Instead, we argue that the verbs machen and gehen form predicates with the ideophones, resulting in a predicative ideophone construction, which causes the information they contribute to shift towards at-issue status and renders them an essential part of the main assertion.

Furthermore, ideophones in quotational-demonstration constructions also appear to be more at-issue in German. For example, ideophones accompanied by the demonstrative so also appear to be shifted towards at-issueness. For example, in (15), it seems possible to have an ideophone accompanied by a demonstrative in the focus of an answer, whereas this appears odd for ideophones not introduced by a demonstrative, as in (16).

The alignment of ideophones in German with respect to the main utterance also seems to play a role. For example, plitsch-platsch appears to be more at-issue in (18), where it occurs sentence finally, than when it occurs sentence-medially in (17) and hence plitsch-platsch in (18) is more easily targeted by a direct denial.

Finally, certain adverbial ideophones in German also seem to be more at-issue despite the lack of any factors outlined above. Although ratzfatz also occurs sentence-medially in (19), as plitsch-platsch does in (12), it seems to be easier to directly address the information provided by ratzfatz.

A second observation is that the meaning of ratzfatz in (19) appears to be more easily specified than plitsch-platsch in (12); ratzfatz clearly refers to the speed at which the bank earned the money. The ideophone plitsch-platsch, however, appears to evoke sensory images of both how wet the frog is and how this wetness affects the noise it makes as it goes up the stairs. This is perhaps an explanation for why ratzfatz seems to be more at-issue than plitsch-platsch, as the latter expresses a much more nuanced concept which is difficult to target with a direct denial. In general, ideophones such as ratzfatz and ruckzuck appear to have more specifiable meanings and also appear to be more at-issue than ideophones such as plitsch-platsch or holterdiepolter, while still being iconic. As an initial analysis for these ideophones, we propose the same mixed item analysis as that discussed in Section 3.1.4 for predicative and quoted or demonstrated ideophones. Alongside a core (more) at-issue meaning, ideophones such as ratzfatz and ruckzuck also have an iconic component, which we argue is not at-issue. The crucial difference between these ideophones and other adverbial ideophones is that the core meaning of such ideophones is more conventionalised and as such can be interpreted as more at-issue even in adverbial structures when they are not necessarily an integral part of the sentence.

The question then remains as to how ratzfatz and ruckzuck have developed a more specifiable or conventionalised meaning in comparison to other ideophones. We propose that this could be due to a process of deideophonisation. As described by Dingemanse (2017) for Siwu, more frequently used ideophones are more likely to be used as normal words. One potential explanation for ratzfatz and ruckzuck then could be that their core meaning has become more conventionalised due to frequent use and hence the ideophones themselves are becoming more at-issue.8

In summary, the previous examples indicate that ideophones in German are able to contribute at-issue and not-at-issue information. While we do plan to conduct further research on all of the cases discussed above, the experimental work presented in the following section focuses on non-predicative, sentence-medial adverbial ideophones in German, which we have claimed are not at-issue. Section 4 outlines the experiments we conducted on these ideophones, which provide evidence to support this claim.

The aim of this study was to gain information on how native speakers interpreted ideophones which were being considered for use in the main experiment. 18 ideophones were chosen and each embedded within a sentence. The pre-study was conducted using an informal questionnaire on Google Forms. A link to the questionnaire was circulated to native speakers among the staff and students within the Institute for Linguistics at Goethe-University, Frankfurt am Main. The stimulus sentences were presented one after another on a separate page. For each item, participants were first asked whether they understood the sentence. If they affirmed, the stimulus sentence was presented again with the ideophone replaced by a gap. Participants inserted a close paraphrase for the ideophone into a blank space by means of the keyboard. If participants did not affirm, they continued with the next sentence. 27 participants in total completed the questionnaire. The data were analysed qualitatively, with identical or closely similar paraphrases being grouped together and used to inform the second pre-study.

The second pre-study was a forced choice questionnaire, which aimed to establish the best adverbial paraphrases of the ideophones, which would then be used in the main experiment. 23 ideophone target sentences were each paired with three candidate paraphrases of the ideophone. 18 of the ideophones were used in Pre-study 1. The three candidates were (a) the paraphrase originally envisaged in preparation of the main experiment, (b) the paraphrase most frequently produced in Pre-study 1, and (c) one of the remaining paraphrases produced in Pre-study 1 that matched the context constructed for the main experiment especially well (see Section 4.3). The candidates for the additional five ideophones not tested in Pre-study 1 were paraphrases that were selected on the basis of dictionary definitions and native speaker intuitions. The questionnaire was created using Sosci.9 It was run online on the platform Prolific.10 Participants were compensated with £3.75. For each item, participants were first asked whether they understood the sentence. If they affirmed, they were shown the three alternatives and asked to select the candidate they considered the closest paraphrase of the ideophone. If they did not affirm, they continued with the next item. 12 German native speakers completed the questionnaire. For all ideophones except one, a majority of participants opted for one of the candidates, ranging from 100% (12:0:0) to 42% (5:4:3). This candidate was selected for inclusion in the corresponding adverbial phrase in the main experiment. For the remaining ideophone, there was a tie between two candidates (5:5:0). We selected the one that we considered to better match the context constructed for the main experiment.

This paper has, up to this point, focused on discussing the at-issue status of ideophones and has only hinted towards what the exact contribution of not-at-issue ideophones could be. This distinction is important, as the characteristics previously discussed for ideophones are typical of not-at-issue content, in particular the ability to project from under semantic operators such as negation, but are not a diagnostic for identifying exactly what kind of not-at-issue content an ideophone is, for example, a conventional implicature or a presupposition. This section then addresses this question, discussing evidence indicating the exact contribution of not-at-issue ideophones. We limit ourselves to two possible forms of not-at-issue content, which have previously been discussed in detail as analyses of iconic enrichments and co-speech gestures in particular. The first follows Ebert et al. (2020)’s analysis and treats not-at-issue ideophones as having the same meaning contribution as appositives, hence as supplements. The second approach would be to treat not-at-issue ideophones as cosuppositions following Schlenker (2018a) and his handling of iconic co-speech gestures. As the two approaches make different predictions concerning whether an ideophone should contribute old or new information and how they should behave when embedded under negation and negative quantifiers, we can make predictions about which approach better applies to ideophones by observing their behaviour in the given situations.

In terms of whether they contribute old or new information, cosuppositions as a form of presupposition should contribute old information, as they should already be established in the common ground and must be true according to the conversational context. Supplements, on the other hand, usually contribute new information (cf. Potts 2005). One potential way to determine whether not-at-issue ideophones behave as supplements or cosuppositions would be to see if they contribute old or new information. As we saw in (11), repeated in (22), ideophones in German appear to provide new information. The use of plitsch-platsch after the description of the sounds the frog makes appears odd.

As previously noted, this intuition is also supported by crosslinguistic literature, where ideophones most commonly contribute new information (cf. Dingemanse 2017). However, as Schlenker (2018a) argues, cosuppositions can easily be locally accommodated, therefore it is difficult to empirically test whether an item contributes new information. Further consideration must be given to how to experimentally evaluate this property.13

We can also consider the behaviour of ideophones under negative quantifiers and negation. Under a supplemental analysis, ideophones would be expected to be marked when embedded under these structures, whereas under a cosuppositional analysis they should project and therefore be acceptable within these structures. Again, as we saw in (9) and (10), repeated in (23) and (24), when ideophones are embedded in such structures, they seem to only be acceptable as meta-linguistic utterances, for example, used in response to something previously asserted.

The non-negatability of ideophones has also been observed in other languages such as Siwu (cf. Dingemanse 2017) and Japanese (cf. Kita 1997; 2001; Toratani 2018), which lends further support to a supplemental analysis of ideophones. Based on these initial observations, it does appear that a supplemental analysis of ideophones may be more applicable than a cosuppositional one, however, further research is most certainly needed to formally establish the exact contribution of not-at-issue ideophones.

One notable finding of research on ideophones is that they appear to frequently co-occur with gestures. Kita (1993; 1997), for example, examined a corpus of retellings of Tweety Cartoons by Japanese participants and found that 94% of ideophones occurred with a gesture. In re-examining Kita’s findings, Dingemanse (2013) argued that as the corpus mainly contained spoken narratives, where ideophones and gestures are more likely to co-occur, the frequency with which they co-occur in the corpus may not be fully representative of standard language use. He does, however, highlight that ideophones most commonly co-occur with iconic gestures and claims that ideophones and iconic gesture are two parts of a single depictive performance, where the speaker makes use of multimodality to give a full performance. For example, Dingemanse (2015) noted that speakers of Siwu tend to use iconic or depictive gestures when giving definitions of ideophones and highlighted that these gestures often helped to clarify aspects of the ideophone’s meaning that would have been difficult to express using ordinary words. In several cases, speakers also used very similar gestures with the same ideophones, which indicates that these ideophone-gesture pairs have become conventionalised within Siwu. Additionally, Nuckolls (2019) found that movement ideophones in Pastaza Quichua were often accompanied by iconic gestures that contributed additional information and which she argues served as a pragmatic embellishment of the ideophone. Dingemanse & Akita (2016) also found that ideophones in Japanese which are less morphosyntactically integrated and therefore more expressive were more likely to be accompanied by iconic gestures and that iconic gestures correlate with other expressive features of ideophones, such as expressive morphology and intonational foregrounding.

The experimental evidence gathered in this research has shown that, just as iconic co-speech gestures can contribute not-at-issue information, so too can ideophones. This is interesting as it shows that two of the most evident sources of iconicity in spoken language are both able to contribute not-at-issue information. However, ideophones do appear to be more at-issue than gestures. The size of the mismatch effect for ideophones in this experiment does seem to be smaller than that of co-speech gestures in Ebert et al. (2020).14 Therefore, although both sentence-medial ideophones in German and co-speech gestures are able to contribute not-at-issue information, how not at-issue this information is differs between the two iconic enrichments. One potential explanation for this difference is the different modalities that the two enrichments occur in; ideophones in spoken language occur in the same modality as the main linguistic signal, whereas gestures are in the visual modality and this may account for why ideophones are more at-issue than gestures. Equally, both also seem to be able to shift to being more at-issue when accompanied by a demonstrative, but Ebert et al. (2020) showed that the at-issue status of co-speech gestures accompanied by demonstratives was still not equivalent to standard German adjectives. We would expect then, that sentence-medial adverbial ideophones accompanied by a demonstrative are also more at-issue, but still not as at-issue as ordinary adverbials. This suggests a potential scale of at-issueness, such as in Figure 4, where iconic gestures without a demonstrative are less at-issue than sentence-medial adverbial ideophones, gestures accompanied by demonstratives are then more at-issue than sentence-medial adverbial ideophones, but sentence-medial adverbial ideophones accompanied by a demonstrative are again more at-issue than gestures accompanied by demonstratives, and so on.

This is by no means a complete scale and other iconic items could and should certainly be included. For example, pro-speech gestures would likely be more at-issue than sentence-medial adverbial ideophones as they are arguably more integral to the felicity of the sentence, whilst they may still be less at-issue than predicative ideophones, which are both integral to the felicity of the sentence and part of the spoken modality. The scale then provides an initial insight into the gradient nature of at-issueness and suggests further factors that may impact on at-issue status, such as modality. A further prediction then could be that iconic items signed on the hands in sign languages, such as movement and location features in classifier predicates, may pattern more closely to ideophones than gestures, as they also occur in the same modality as the main linguistic message.

Despite these potential differences between ideophones and gestures, they do appear to enjoy a symbiotic relationship; both employ iconicity and commonly co-occur when speakers exploit iconicity to enhance their speech, with evidence suggesting that iconic gestures can clarify or embellish ideophonic meanings (cf. Dingemanse 2015; Nuckolls 2019). Further research questions then include examining how exactly ideophones and iconic gestures interact with each other when they co-occur and whether this differs from how gestures interact with other speech, as well as exploring the combined meaning contribution and pragmatic status of ideophones and gestures.

Dingemanse (2012) also highlights that ideophones often occur with varying speech rates, loudness and phonation types, alongside gesture, all of which allows the speaker to make the most of the ‘depictive’ potential of the ideophone (p.660). Equally, recent research by Akita (2021) showed that varying phonation types influenced speakers’ ratings of the size and shape of novel words. For example, creaky voice was associated with larger, more pointed images. This then prompts further questions about the nature of multimodal iconic contributions in spoken language, or what Dingemanse (2013) has termed ‘iconic performances’. Outside of spoken language, multimodal performances have been observed in bilingual English/ASL children, who produce sound effects when code-blending English with ASL classifier predicates (cf. Davidson 2015).15 Furthermore, Sandler (2009) has argued that signers produce iconic gestures with their mouths while signing and that these gestures complement and add to the linguistic signal being given by the hands. Mouth gestures are commonly used to iconically modify sign language constructions such as classifier predicates and size and shape adjectives. Clearly, iconic performances across modalities often feature a range of different iconic enrichments, which occur in combination with each other. Any thorough investigation of iconicity in language will therefore need to explore how iconic enrichments interact with and complement each other.