Some remarks on the fine structure of ideophones and the meaning of structure

1 The meaning of structure

Hierarchical structure, a core property of human language, is closely connected to the expression of meaning. A well-known illustration of this close bond between structure and meaning comes from syntactic ambiguity: the linear string John hit the dog with the stick has two different interpretations, where each reading corresponds to a different constituent structure: John [ _VP hit [ _NP the dog [ _PP with the stick]]]—meaning: ‘John hit the dog that was carrying the stick’—and John [ _VP hit [ _NP the dog] [ _PP with the stick]]—meaning: ‘John hit the dog and he did so with the stick’. A second illustration of the meaning of structure comes from the semantic roles associated with noun phrases. For example, when the noun phrase the dog occupies the complement (i.e., direct object) position of a transitive verb, as in John [ _VP hit the dog], the dog is interpreted as the Theme-argument of the verb hit. However, when it occupies the subject position of the clause, as in The dog bit John, the semantic role of the dog is completely different; it then acts as an Agent-argument. A third illustration of the close relationship between structure and meaning comes from the interpretation of nouns like dog. When dog combines with the indefinite article a, or the numeral one, as in The car hit a/one dog, it typically gets a count-reading, which feels like the default interpretation. Interestingly, dog gets a mass-reading when it is part of a different nominal structure, specifically one lacking an indefinite article, or one in which it co-occurs with the quantifiers much or some, as in Look, there is (some/much) dog on the bumper of your car! A fourth and final illustration of the close relationship between structure and meaning comes from the following minimal pair: This is [a good solution of the problem], and This is [a hell of a problem]. Even though the bracketed noun phrases look quite similar superficially, their meaning is very different. In the former bracketed noun phrase, (of) the problem acts as the complement of the noun solution; in the latter bracketed noun phrase, however, the noun hell has a (metaphorical)-evaluative meaning rather than a referential one. It is the speaker of the utterance that assigns negative valence (hell) to the referent of the noun problem. That these two nominal expressions have different underlying structures is clear from the fact that they display different sub-extraction behavior: It is impossible to say: a problem which this is [a hell of (a) which], but completely fine to say: a problem which this is a good solution of which]; see den Dikken (1998:186).

The search for the relationship between structure (including word order) and meaning also plays a role in Barnes and Ebert’s article on the information status of iconic expressions in spoken language, specifically German ideophones such as plitsch-platsch in (1a) and co-speech gestures such as ‘BIG’ in the English sentence in (1b):

They observe that the information status of such expressions can vary. Specifically, they can contribute more at-issue-content or less at-issue-content to the sentence in which they are embedded, where ‘at-issue content’ stands for information that is semantically, and also prosodically, (more) integrated into the clause, and accessible for denial by the interlocutor; ‘non-at-issue content’ stands for the opposite.

They argue that the extent to which ideophones and co-speech gestures have at-issue content is dependent on the distribution of these expressions within the clause (an external property), but also on the expression’s inner structure (an internal property). For example, the German ideophone plitsch-platsch receives a more at-issue reading in clause-final position, as in (2), than in clause-internal position, as in (1a). Interlocutor B’s denial, Nein, das stimmt nicht, etc., yields a more felicitous reading for (2A)—‘(?)’ according to Barnes and Ebert—than it does for (1a)—‘#?’ according to Barnes and Ebert.

When the demonstrative element so co-occurs with the ideophone, as in (3), the at-issue-reading also becomes more felicitous; that is, the information denoted by speaker A’s utterance so plitsch-platsch can be denied by speaker B.

In short, the empirical data discussed by Barnes and Ebert seem to confirm the generalization that structure matters for meaning. In this article, I aim to provide some further substance to this generalization by further exploring the internal syntax and external syntax of ideophones on the basis of data from Dutch.^[1] The article is organized as follows: In Section 2, I propose that ideophones are so-called roots. Section 3 examines the inner structure of ideophones, and the way in which they are integrated into larger syntactic structures. In Section 4, it is proposed that ideophones are deictic expressions. Section 5 discusses the syntactic distribution of ideophones, and the way in which structure (external syntax and internal syntax of ideophones) interacts with meaning. Section 6 concludes this article.

2 Ideophones as roots

With Dingemanse (2019:16), I assume that ideophones form “[…] an open lexical class of marked words that depict sensory imagery”. Following Borer’s exo-skeletal model (2005) and work on distributed morphology (Halle and Marantz 1993; Harley and Noyer 1999), I assume that (Dutch) ideophones such as pats ‘bang’, boem ‘bang’, and tsjoek ‘choo/chuff’ (sound of a train/engine) are roots, that is, lexical vocabulary items that are not specified for categorial information or any other formal-syntactic features; see also Corver (2014; 2015. According to these theories, the categorial status of a “word” is determined by the structure on top of the root. In DM, for example, it is a categorial marker (e.g., n, v) that merges with the root and determines the categorial nature of the projected structure. For example, the English root kiss becomes a noun by merging with the nominalizer n, as in n+√kiss, and it becomes a verb by merging with the verbalizer v, as in v+√kiss. An important consequence of this analysis is that lexical categories are syntactic objects with a composite structure.

Along similar lines, an ideophonic root atom like tsjoek can be ‘assigned’ categorial status on the basis of locally available functional material. For example, the numeral twee and the plural morphology –s in (4a) define the nominal nature (and count interpretation) of the phrasal expression twee tsjoeks. The past tense morphology –te (i.e., T) in (4b) defines the verbal character of tsjoekte.

It is clear that, in these examples, tsjoek is highly integrated, not only syntactically—being a finite verb, it undergoes the so-called Verb Second operation in the Dutch main clause; Koster (1975)—but also semantically—it forms a verbal predicate that takes de trein as its external argument—and phonologically (the voiceless dental /t/ of the past tense morpheme -te results from the preceding voiceless consonant /k/ at the end of the root).^[2]

3 On the syntactic integration of ideophones

The question arises as to how ideophones are integrated into the syntactic structure, when they have a superficially bare form, as in the examples in (5):

In current generative grammar, it is assumed that there is one major computational device for building structure, namely Merge (Chomsky 1995). Merge constructs syntactic objects from lexical material and from the syntactic objects that it has already constructed. For example, merge of the lexical items the and dog yields the syntactic object (noun phrase) the dog, which, in turn, can be combined with the preposition at, yielding the prepositional phrase at the dog, as in He looked at the dog.

If Merge is the computational device for building syntactic structure, then arguably it is also involved in syntactically integrating ideophones such as tsjoek-tsjoek and boem in larger syntactic structures. Evidence in support of the syntactic integration of ideophones and, more generally, what are traditionally called ‘interjections’, comes from the structure-dependent nature of the relationship between the interjection/ideophone and other material within the clause. As shown in James (1973: chapter 4), interjections “refer” to a string of words that form a constituent. They do not “refer” to a linear sequence of words that forms a non-constituent. James gives the following examples to illustrate this structure dependence:

The natural interpretation of sentence (6a), which contains the hesitation marker oh, is that the speaker is selecting Wendy over other people, e.g. Mary and Joanna. In other words, oh and Wendy go together, as expressed by ‘boldface’ in (6b). A reading in which the speaker is choosing between ‘Wendy that she shouldn’t come’ and ‘Joanna that the party would start at 8 pm’ is infelicitous; see (6c). We see the same contrast in (7): in saying oh, the speaker is selecting one of the things that Princess Anne did, as in (7b), where oh goes together with the VP launched a ship. As indicated by (7b), the speaker cannot be selecting whether to say “launched a ship and Prince Charles went to the reception” or “opened an agricultural show and the Queen Mother reviewed the Twelfth Hussars”; see James (1973:115).

Similar illustrations of structure dependence can be given for ideophones. Consider, for example, the Dutch example in (8a):

The sound information depicted by boem only applies to the (eventive) VP tegen een boom aangereden, as represented in boldface in (8b), but not to the linear sequence and non-constituent boem tegen een boom aangereden en Marie is tegen een muur aangereden. For (8c) to become felicitous, boem must be repeated in the second conjunct: …en Marie is boem tegen een muur aangereden.

Now that we have shown that ideophones, and interjections more in general, display structure-dependent behavior, let us examine in more detail how they get integrated into the larger syntactic structure. As my starting point, I take the proposal that bare roots (e.g. √car) cannot participate in syntax “on their own”; see Chomsky (2013). They must have a minimal amount of functional structure (e.g., nP, as in [ _nP n+√car]) to be visible for syntactic computation. If so, also bare ideophones like boem, as in (8), and hesitation markers like oh, as in (6)–(7), must have functional structure on top of the root. The question then arises as to what exactly this functional structure is. I tentatively propose that these superficially bare elements are “small” noun phrases, that is nPs, and possibly sometimes DPs. Evidence in support of their nominal nature comes from a number of phenomena that deserve further investigation. First of all, it turns out that many interjections (including ideophones and exclamations) display the element -s at the end. Consider the following examples:

In line with Corver (2022), I take this element -s to be a minimal phonological realization of the categorial head n ^o . It is the same -s that we find in nominal forms such as iet-s mooi-s ‘something beautiful’, and forms such as langzaam-pje-s (slow-DIMINUTIVE-s, ‘slowly’) and op-een-s (at-one-s, ‘at once, suddenly’), which are, traditionally, taken to be adverbs but should be analyzed as hidden nominal expressions. Under this reinterpretation, opeens, for example, has the following structure: [ _PP op [ _QP een [ _nP n ^o (= -s) [√TIME]]]], where TIME is a silent “noun” (see Kayne 2003). Following this line of analysis, floeps has the derivation in (10a). I assume that the superficially bare form floep has the same nominal structure, i.e. nP, but that the root remains in situ and that n does not surface; see (10b). Thus, the categorial head n only surfaces when there is a phonological host for the suffixal element -s.

Notice by the way that this -s-pattern is also found at the end of other types of interjective expressions:^[3]

A second reason for saying ideophones such as those in (9) are nominal forms comes from diminutive formation. Some of these interjections can be combined with -ie, which is a more informal variant of the diminutive morpheme -je:

Following De Belder (2011), I assume that there is a functional layer within the noun phrase, encoding ‘diminutive meaning’ (De Belder’s ‘SizeP’); see also Corver (2022). This functional layer is located on top of nP, as in (13):

The form floep-s-ie is derived by moving the root to n ^o , yielding floeps, and subsequently raising floeps to -ie, yielding floepsie. In short, the ideophone floepsie has a composite inner structure.

A third potential argument in support of the nominal status of (apparently) bare ideophones, comes from reduplicative patterns of the following type:

The ideophones in (14) feature the “linking element” de, which is homophonous with the non-neuter definite article de, as in de zolder ‘the attic’ and de bloem ‘the flower’. Interestingly, similar reduplicative patterns can be found as proper names in children’s verses and songs (15a), and also as a kind of replacement name if you have forgotten someone’s real name (15b).

The reduplicative proper names in (15) are reminiscent of complex proper names consisting of a first name and a second common noun that designates the individual’s profession, as in Jan de tuinman (Jan the gardener), and Jan de professor (Jan the professor). I tentatively propose that all these X de Y-patterns instantiate a Determiner Phrase (i.e. DP), with de being a definite article (i.e. D), and the pre-article phrase and the post-article phrase occupying D’s specifier-position and complement-position, respectively. Thus:^[5]

Having given some empirical support for a nominal analysis of ideophones such as floep, floeps and floepsie, let us return to the question of how ideophones, and other interjective material, become part of the (clausal) syntactic structure. I propose that it is the structure building operation Merge which combines the ideophone, a phrasal expression, with its “host”, that is, the phrasal constituent whose content is “specified” by the ideophone. Thus, the sequence boem tegen een boom aangereden in (17) has the following representation:^[6]

Being part of the VP, boem can move along with tegen een boom aangereden in so-called VP-topicalization constructions:

That boem can move along with the fronted VP shows again that it forms a structural unit (a constituent) with its structural “host” (i.e., the VP).

4 The deictic nature of ideophones

In Wilkins (1992:131ff), it is argued that interjections are indexical in the sense of being context-bound (see also C.S. Peirce (1955:119)). Just like the indexical pronouns I, you, this, that, and the indexical adverbs here and now, they must be tied to the actual speech moment, that is, the situation of utterance. For example, the ideophone zjoef ‘whoosh’ in (19) references some relation between the falcon’s wing movements and the flying event at the moment of utterance.

The question arises as to whether this indexical meaning has consequences for the syntactic structure of ideophones, and also other types of interjections. As noted in Barnes and Ebert’s article, German ideophones such as plitsch-platsch can co-occur with the demonstrative element so, as exemplified in (3), repeated here as (20):

Similar examples can be found in Dutch; see Corver (2015):

I take the demonstrative element zo to be the referencing element which establishes a relation with the contextual situation. The element zjoef designates the sound contents of the deictic element zo. More specifically, I assume that zo zjoef represents a small-clause-like structure, with zo, which I take to be a (pro)nominal expression (Corver 2023), as the small clause subject, and the nP zjoef as the small clause predicate. Following den Dikken (2006), I call this small clause representation ‘Relator Phrase’:

I take the (optional) element van in (21b, c) to be a functional preposition that functions as an exponent of the Relator-head:

Interestingly, patterns of the type van + ideophone are also possible, as exemplified in (24):

I assume that the representations of the ideophones in (24) contain a silent demonstrative ZO, as in (25):^[7]

The silence of demonstrative ZO in (25) is not unexpected. In colloquial spoken language, other deictic elements can also be absent at the sound surface, if their contents (meaning) is contextually recoverable. Some illustrations are given in (26):

In summary: ideophones that appear as “satellite elements” (e.g., zjoef in (19)) in a clause have inner structure. Specifically, the root (e.g., √zjoef) is embedded within a (nominal) functional layer, which minimally equals nP but possibly includes DP (see the linking element de). Furthermore, the “functionally dressed” ideophone acts as a predicative element within a small-clause representation (RelP), that encodes a predicative relationship between an overt (zo) or silent (ZO) demonstrative element which establishes a (deictic) connection with the utterance context.

5 The distribution of ideophones

Having a more refined picture of the inner structure of “satellite” ideophones, let us now have a look at their distributional behavior (i.e., external syntax). As noted in Barnes and Ebert, German ideophones can occupy different positions within the clause; see, for example, (1a) and (2A). As exemplified in (27), also Dutch ideophones can occur in different positions:

In (27a), zjoef occupies a clause-internal position. In (27b) and (27c), on the contrary, zjoef occupies a left-peripheral position and a right-peripheral position, respectively. The question arises as to whether zjoef in (27b) occupies the left-peripheral position as a result of a displacement operation (so-called I(nternal)-Merge). Specifically, could the ideophone zjoef have been moved from a clause-internal position, as in (27a), to a left-peripheral position, as in (27b)? Schematically:

As represented in (28), zjoef occupies the [Spec,CP]-position after displacement, and the finite verb is located in the C(omplementizer)-position as a result of the so-called Verb Second operation (Koster 1975).

The existence of the minimally different pattern in (29), however, raises the question as to whether the pattern in (27b) really involves displacement of the ideophone zjoef to [Spec,CP]. In (29), the locative d(emonstrative)-word daar ‘there’ immediately precedes the finite verb, and thus seems to occupy [Spec,CP], the position immediately preceding the finite verb. If daar occupies [Spec,CP], the left peripheral ideophone arguably does not occupy the left-peripheral position as a result of movement (i.e., I-Merge). As an alternative analysis, one might propose then that zjoef is base-generated,^[8] that is E(xternal)-merged, with the clausal structure as depicted in (29):^[9]

That the ideophone zjoef is not input to I-merge, is also suggested by the fact that long distance dependencies are ill-formed:

In this example, one cannot interpret the ideophone zjoef as providing information about the sound associated with the falcon’s flying over my head. Under a movement analysis, in which zjoef orginates in the embedded clause and moves upward in a successive-cyclic fashion, such long-distance readings are (incorrectly) predicted to be possible. Under an E-merge analysis, the impossibility of the long-distance reading may be accounted for in terms of some sort of locality constraint: an ideophone (i.e., the RelP representing the ideophone) can only stand in a meaningful relationship with the phrase (e.g. CP or vP) with which it has merged directly.

So far, I have argued that clause-internal zjoef (27a) and left-peripheral zjoef are base-generated in their surface position. What about (27c), in which zjoef occurs clause-finally? For the analysis of this pattern, I base myself on Kayne’s (1994:78) analysis of the English Right-Dislocation pattern in (31a):

Since, right-adjunction is not possible in his antisymmetry theory, Kayne proposes that John is not a right-adjoined constituent. Rather, he proposes that John (is) in (31a) is a reduced clause that has he’s real smart left-adjoined to it, as in (31b). As indicated, Kayne assumes that there is an empty functional head X^o mediating that adjunction.

In the spirit of Kayne’s analysis, I propose that the pattern in (27c), in which zjoef occurs in final position, has the base structure in (32a), where I abstract away from the mediating functional head. A slightly more refined representation is the one in (32b), where zjoef is represented as a relator phrase:

Importantly, zjoef (i.e. the relator phrase) is the “matrix expression” and the preceding clause (De valk vloog over mijn hoofd heen) the subordinate or embedded expression.

Having more refined structural representations of the sentences in (27) may help us in further exploring the subtle information-structural meaning differences between the various patterns discussed in Barnes and Ebert’s article. I will leave this quest for the exact relationship between the structural placement of ideophones and their informational (i.e., meaning-related) contribution at the clausal level for future research.

6 Conclusions

Hierarchical structure is a core property of human language. It is a property that is closely connected to the expression of meaning. The main aim of this article was to show that Dutch ideophones, even though often “simple” at the (sound) surface, have a complex inner structural organization. For discovering the subtle meaning contributions of ideophones, I think it is important to lay bare the hidden inner structure of ideophones. The same holds for the clausal structure in which the ideophone functions (or appears to function) as a “satellite constituent”. The various distributional patterns of ideophones correspond to specific positions in the hierarchical structure of the clause. These syntactic positions arguably matter for the informational contribution made by ideophones.