Preposed function items are less likely to coalesce because speakers tend to pause before content items

Himmelmann (2014) makes a fresh attempt at explaining the suffixing preference in the world’s languages that was observed long ago by Sapir and Greenberg, and for which Hawkins & Cutler (1988) and Hall (1992) had proposed a processing explanation. But while these authors argued from word recognition, Himmelmann’s explanation starts from language production and combines research on spontaneous spoken language and clitic typology in a novel way. Specifically, he argues that preposed function words show a weaker tendency to become affixes than postposed function words because they are often separated from the lexical host by a prosodic boundary. Bybee et al. (1990: 29) had earlier observed that fusion between a grammatical element and the stem requires that they regularly occur in the same prosodic unit. Apparently, there is a greater tendency for a prosodic boundary to occur between a preposed function item and the host than between a postposed item and the host.

Himmelmann cites two specific pieces of evidence for this, one from typology and one from unplanned spoken language. From clitic typology, we know that ditropic clitics (i.e. items which phonologically lean on a form that is not their semantic host) are always phonologically linked to a preceding item and semantically linked to a following item (e.g. German zu-r Schule ‘to-the school’) – the opposite does not seem to occur. More surprisingly for a typologist reader, there is good evidence that in spontaneous spoken speech, hesitation pauses tend to occur between function item and content item, rather than before the function item (e.g. he wipes the [..] dirt off, rather than he wipes [..] the dirt off). With postposed function items, there is no such tendency for a pause to occur, as can be seen in auxiliary-mainverb combinations in German: When the order is auxiliary-mainverb (wird gekauft ‘is bought’), a hesitation pause may occur, but when it is mainverb-auxiliary in dependent clauses (gekauft wird), no pause occurs. Himmelmann cites extensive data from Tagalog, which has both preposed and postposed function items, and where speakers pause after the former, but not before the latter. There is a plausible reason for this: Content items are less frequent and automatized and require more processing time, and Himmelmann argues that speakers do not wait with producing the preposed function item because it signals that their turn is not over yet.

This seems to be a very important insight that will move the discussion forward, but it is still not quite clear how speakers’ behaviour in spontaneous speech relates to grammatical patterns. Himmelmann uses the term “prosodic boundary” in two different ways, to refer to grammatical patterns and to occurrences in speech. In the following, it seems to refer to speech (because of the reference to frequency):

“Prosodic word and phrase boundaries may occur after a clitic function word preceding its lexical host with sufficient frequency so as to impede the fusion required for affixhood.” (abstract, p. 927)

But in the discussion of clitic typology, a prosodic word is of course a conventional grammatical unit. Speech pauses may occur within grammatical units – one would not want to say that each time speakers pause because of a processing difficulty, they create new grammatical units.

However, I do find it plausible that due to the frequent pauses between a preposed function item and its host, there are fewer chances for phonological changes to occur that make the item appear “fused” with its host. By this I mean processes such as vowel reduction, consonant lenition, or tone loss that lead linguists to say that the preposed item is part of the same prosodic unit as the host. Perhaps surprisingly, Himmelmann does not talk about such changes at all, apparently taking them for granted.

However, for a complete understanding of the coalescence process, we need to understand the individual phonological changes, because we cannot take units such as “prosodic word” for granted. As Schiering et al. (2010) have pointed out, prosodic wordhood is not a monolithic uniform property, but is different in different languages and for different phonological processes. In other words, “prosodic word” is not a pre-established cross-linguistic category, but is rather a higher-level unit that linguists often find useful for describing phonological and morphological regularities. It’s the phonological changes that are primary, not the prosodic domains.

Since the kinds of phonological dependencies that lead linguists to speak of clitics and affixes rather than independent words are so diverse across languages, it is actually quite difficult to compare grammatical items with respect to affixhood across languages. Dryer’s (2005) WALS chapter on prefixing vs. suffixing does not say how affixes and clitics are distinguished, but in related work, Dryer admits that he often bases his coding decisions on the spelling choices of grammar authors. This raises the question whether there is a spelling bias of linguists (and missionaries), who typically have training in languages that write preposed items separately, but postposed items jointly (cf. French je travaill-er-ai ‘I’ll work’, rather than *je-travaill er ai). It seems very difficult or impossible to base one’s spelling on uniform principles, because languages are simply too diverse with respect to “affixhood/wordhood” criteria. As a result, linguists have not come up with a coherent cross-linguistically applicable notion of “grammatical word” (Haspelmath 2011), or “clitic” (Haspelmath 2015).

Himmelmann leaves these issues aside, apparently simply assuming that clitics and affixes can generally be distinguished, and that the phenomenon in question can be sufficiently characterized as “the further grammaticization of (clitic) function words as affixes on their lexical hosts” (p. 955). I think we’ll need a more rigorous approach, but it also seems clear to me that Himmelmann’s paper constitutes significant progress: The connection between (i) the prefixing dispreference, (ii) the tendency for preposed function words to lean on preceding items (as in zu-r Schule), and (iii) the tendency for hesitation pauses to occur before content items seems very plausible, and future work can build on this important insight.


Bybee, Joan L., William Pagliuca & Revere D. Perkins. 1990. On the asymmetries in the affixation of grammatical material. In William Croft, Keith Denning & Suzanne Kemmer (eds.), Studies in typology and diachrony: Papers presented to Joseph H. Greenberg on his 75th birthday, 1–42. Amsterdam: John Benjamins.

Dryer, Matthew S. 2005. Prefixing vs. suffixing in inflectional morphology. In Martin Haspelmath, Matthew S. Dryer, David Gil & Bernard Comrie (eds.), The world atlas of language structures, 110–113. Oxford: Oxford University Press. (See

Hall, Christopher J. 1988. Morphology and mind: A unified approach to explanation in linguistics. London: Routledge.

Hawkins, John A. & Anne Cutler. 1988. Psycholinguistic factors in morphological asymmetry. In Hawkins, John A. (ed.) Explaining language universals, 280–317. Oxford: Blackwell.

Haspelmath, Martin. 2011. The indeterminacy of word segmentation and the nature of morphology and syntax. Folia Linguistica 45(1). 31–80.

Haspelmath, Martin. 2015. Defining vs. diagnosing linguistics categories: A case study of clitic phenomena. In Joanna Błaszczak, Dorota Klimek-Jankowska & Krzysztof Migdalski (eds.), How categorical are categories? New approaches to the old questions of noun, verb and adjective. Berlin: De Gruyter Mouton (to appear)

Himmelmann, Nikolaus P. 2014. Asymmetries in the prosodic phrasing of function words: Another look at the suffixing preference. Language 90(4). 927–960.

Schiering, René, Balthasar Bickel & Kristine A. Hildebrandt. 2010. The prosodic word is not universal, but emergent. Journal of Linguistics 46(3). 657–709.

OpenEdition suggests that you cite this post as follows:
Martin Haspelmath (April 21, 2015). Preposed function items are less likely to coalesce because speakers tend to pause before content items. Diversity Linguistics Comment. Retrieved July 19, 2024 from

5 thoughts on “Preposed function items are less likely to coalesce because speakers tend to pause before content items

  1. Pingback: What is an affix? A fresh attempt | Diversity Linguistics Comment

  2. Just reporting a new study by Reinhöl and Casaretto. Based on comparative Indo-European data, the authors build on Himmelmann’s (2014) approach and claim that prosody-syntax mismatches prevent grammaticalization.

    Reinöhl, Uta & Antje Casaretto. 2018. When grammaticalization does not occur – Prosody-syntax mismatches in Indo-Aryan. Diachronica 35(2). 238-276.

  3. I do believe that a simple rephrasing can save the argument, and that the prosodic word as a universal category is not a necessary condition for the argument to work. Regardless of whether a given language distinguishes 2 phonological domains or eleven, there are smaller phonological domains (e.g. the Syllable) and larger ones (e.g. the Phonological Word). Note that both terms are capitalized in order to signal their language particular status. Those domains have boundaries. If a boundary occurs between an affix and the host, some phonological processes will be blocked (note that the boundary on the other side has no effect, suggesting that it is not the unit that is important, but really rather the boundary.) The higher up in the hierarchy, the more “solid” the boundary, the more likely that phonological processes will be blocked.

    The universal category which has to be assumed is “boundary” not phonological word. Furthermore, an ordering of boundaries as to their importance has to be possible. Those are assumptions which I do not find problematic.

    Thus we can rephrase

    “Prosodic word and phrase boundaries may occur after a clitic function word preceding its lexical host with sufficient frequency so as to impede the fusion required for affixhood.”


    “Phrase boundaries and boundaries of phonological domains of a higher order may occur after a clitic function word preceding its lexical host with sufficient frequency so as to impede the fusion required for affixhood.”

    and the argument works without relying on the pre-established category of phonological word.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.