Many linguists have the intuition that affixes are “attached” to their bases, and we often informally use metaphorical terms such as “fused”, “cohering” and “tightly linked”. There is nothing to be said against metaphors in technical terminology, but do these expressions have a precise meaning? Since my 2011 paper in which I expressed doubts about the “word” concept, I have been thinking about better ways of capturing key distinctions of morphosyntax terminologically. Even earlier, I questioned the well-known “agglutination vs. flection” distinction (in Haspelmath 2009; originally presented in 1999), which has its roots in racist ideas about the superiority of Indo-European (“Aryan”) languages that were popular in German Romanticism.
An intermediate result of my more recent thinking about “word”-related concepts is my 2021 paper on affixes (briefly summarized in this earlier post), where I build on the notion of (Bloomfieldian) boundness (= non-occurrence in isolation) and where I introduce the notion of welding (= segmental interaction between a bound morph and its host).
Now in actual practice in descriptive linguistics of the world’s languages and in typology, elements are often said to be “phonologically bound” or “phonologically dependent”, and something like this is often thought to be the basis for affix status (see Zwicky & Pullum 1983) – even though it is clear, upon further reflection, that clitics are generally phonologically dependent as well.
But what exactly is “phonological dependence/boundness/fusion/attachment”? The literature is not clear on this, and as a morphosyntactician who is not specialized in phonology, I do not have a proposal for making such a term precise. Evidently, it is not sufficient to say that X is phonologically dependent on Y if it occurs in the same phonological domain, because such domains can be very large (phonological words, phrases, intonation groups, etc.). We cannot say either that X must occur in the same “phonological word” as Y, because this is defined differently in different languages (Schiering et al. 2010), and comparative concepts must be defined uniformly across languages.
In this blogpost, I would like to highlight the fact that we do not really have a clear definition of phonological “dependence/attachment/fusion”, by discussing several passages from Bickel & Nichols (2007), an authoritative overview paper on inflectional morphology that gives examples from many of the world’s diverse languages and is thus potentially better informed than the usual textbooks (which tend to limit themselves to English examples). We will see that the terms used in this (by now fairly old) paper have multiple problems, by which I do not mean to play down its importance – I learned a lot from thinking hard about it, and my 2011 paper was partly inspired by it. The key distinction between isolating and concatenative formatives (= markers) is introduced as follows (Bickel & Nichols 2007: 180):
First of all, it should be noted that the term “fusion” has also been used both for cumulative expression (or synexpression), as in Sapir (1921), and for segmental interaction between adjacent forms, as in Bybee (1985) (e.g. Italian amic-o/amič-i ‘friend(s)’, or English buy/bough-t). Thus, the contrast between “agglutinative and flective languages” (mentioned above) has also been called “agglutinative vs. fusional” (e.g. Comrie 1989: 42). But of course, cumulative expression and segmental interaction are rather different concepts, so one wonders whether we are really dealing with a single scale of increasing fusion (“isolating < concatenative < nonlinear”). But in this post, I will mostly limit myself to what Bickel & Nichols (2007) describe as “isolating vs. concatenative markers” (see also their WALS chapter on fusion of inflectional formatives, where they use the same distinctions).
The next question is: Do we have good reasons to think that there is “a gradient of tightness of fusion”? B&N do not quantify the tightness numerically, and maybe by “gradient” they mostly refer to their three-term ordinal scale. Of course, many linguists have the intuition that “tightness of fusion” is a matter of degree, but we do not seem to have any way of measuring it numerically (yet?). Until someone proposes such a numerical measure, it is probably better to be careful with statements about degrees. Any numerical measure will ultimately have to rely on qualitative distinctions anyway, so let us try and get these straight first.
Now what is an isolating marker (or formative)? B&N also call such elements “particles”, a term that has been discussed (rather inconclusively) by Zwicky (1985). But what defines them? They are said to be “full-fledged free phonological words on their own”, but this does not work as a definition. First, “free” and “on their own” are really the same, if free is used in its Bloomfieldian sense (free form = a form that can occur on its own, i.e. in isolation). Second, “full-fledged” seems superfluous (as there are no “reduced” phonological words), and third, the notion of “phonological word” is not cross-linguistically generalizable (as I mentioned above). The Lai Chin ergative enclitic that is mentioned in the above passage is niʔ, which fulfills the minimality requirements for a phonological word in Lai Chin (as it has a CVC shape), but many languages have (non-affixal) adpositions or auxiliaries that are phonologically deficient and couldn’t occur on their own (e.g. the Russian prepositions s, k; the Arabic prepositions li, bi; the English future auxiliary ’ll). And even more importantly, many languages have affixes which could occur on their own phonologically, e.g. Spanish Imperfect -aba (cant-aba ‘sang’), Polish Plural Past -li (śpiewa-li ‘they sang’), or English –ing. Are these “isolating affixes”?
In any event, we surely wouldn’t want to say that these markers somehow make Spanish, Polish and English morphology more “isolating”. Maybe this problem could be solved by simply replacing “isolating” by “independent” in B&N’s fusion scale, but the criteria for “isolating” vs. “concatenative” markers are actually compatible, as we see when we read on:
Concatenative markers are said to trigger “(morpho)phonological adjustments” in the resulting form: vowel harmony in Turkish (ad-lar ‘names’ vs. el-ler ‘hands’), voicing assimilation in English (cat-[s] vs. dog-[z]), vowel elision, vowel epenthesis, consonant lenition, and so on. Such segmental effects are very salient, and they have been called welding in Haspelmath (2021: 15):
“definition: welded form
A welded form is a bound form which has shape variants phonologically conditioned by its host, or which phonologically conditions shape variants of its host.”
Now crucially, a welded form may of course be phonologically independent in the sense of being able to occur on its own. For example, the Russian superlative suffix -ejšij/-ajšij palatalizes the final consonant of the adjective (e.g. vysok(-ij) ‘high’, vysoč-ajšij ‘highest’), but at the same time there are no phonological reasons why it should not occur on its own. Thus, welding is different from “dependence” if the latter is understood as deriving from a minimality constraint that does not allow occurrence in isolation.
So when B&N say that “concatenative formatives are phonologically bound and need some other word for their realization”, they seem to state two different necessary conditions: On the one hand, a concatenative marker must be a (Bloomfieldian) bound form (which by definition needs some other form for its realization), and on the other hand, it must be welded (“phonologically bound”). This means that their typology does not have a place for non-welded bound forms, such as the English verbal suffix -ing, or for non-bound forms which do trigger (morpho)phonological adjustments in combination with adjacent forms (e.g. the English preposition/adverb before, which can occur on its own, but which has different shapes depending on the environment: bef[ɔ:] the war, bef[ɔr] another war, she came bef[ɔ:]).
The notion of “phonologically bound/fused/attached” is thus not an elementary notion, and it does not conform to Bickel’s (2007: §3) wise proposal that typologists should use fine-grained comparative concepts – talking about “phonological attachment/fusion” is simply too coarse-grained.
Note also that the reason why a form cannot occur on its own need not be its phonological properties (as the term “phonologically bound” suggests). While this is quite possibly the case with “deficient” markers which consist only of a single consonant or do not have a full vowel (e.g. English Genitive ’s, or German Past tense -t[ə]), there are of course many bound forms which are not phonologically deficient (English -ing, Spanish -aba, and so on). This also applies to many clitics, e.g. the Italian object person clitics la= ‘her’ and li= ‘them’ (cf. the free forms là ‘there’ and lì ‘there’). Thus, we cannot define clitics as “forms that are phonologically deficient” (as I noted in a 2022 conference presentation), even though one often reads this.
Where does this leave us? It seems to me that the notions of (Bloomfieldian) boundness and welding are sufficiently clear to be used in a precise way as technical terms, and I have also proposed definitions of “affix” (and even “word”) which (ultimately) rely on boundness. Intuitively, phonological minimality requirements and welding (as well as various prosodic restrictions, e.g. on stress and tone) are of course also important for the “tightness” of grammatical elements, but there do not seem to be any clear proposals yet for defining “phonological fusion” in such a way that it applies to all languages.
Bickel, Balthasar. 2007. Typology in the 21st century: Major current developments. Linguistic Typology 11(1). 239–251.
Bickel, Balthasar & Nichols, Johanna. 2007. Inflectional morphology. In Shopen, Timothy (ed.), Language typology and syntactic description, vol. III: Grammatical categories and the lexicon, 169–240. Cambridge: Cambridge University Press.
Bybee, Joan L. 1985. Morphology: A study of the relation between meaning and form. Amsterdam: Benjamins.
Comrie, Bernard. 1989. Language universals and linguistic typology: Syntax and morphology. 2nd ed. Oxford: Blackwell.
Haspelmath, Martin. 2009. An empirical test of the Agglutination Hypothesis. In Scalise, Sergio & Magni, Elisabetta & Bisetto, Antonietta (eds.), Universals of language today, 13–29. Dordrecht: Springer. (Accessed May 27, 2016.)
Haspelmath, Martin. 2011. The indeterminacy of word segmentation and the nature of morphology and syntax. Folia Linguistica 45(1). 31–80. (doi:10.1515/flin-2017-1005)
Haspelmath, Martin. 2021. Bound forms, welded forms, and affixes: Basic concepts for morphological comparison. Voprosy Jazykoznanija 2021(1). 7–28. (doi:10.31857/0373-658X.2021.1.7-28)
Sapir, Edward. 1921. Language: An introduction to the study of speech. New York: Harcourt, Brace & Co.
Zwicky, Arnold M. 1985. Clitics and particles. Language 61(2). 283–305.
Interesting article, as always.
One question: you use the definition: A welded form is a bound form which has shape variants phonologically conditioned by its host, or which phonologically conditions shape variants of its host.
What does it mean for a bound form to have “shape variants”?
Since “A form is a recurrent pairing of content and segmental shape”, wouldn’t different shape “variants” technically be different forms altogether?
Naively, I would say that [/z/, PLURAL] (as in dogs, buds, cogs) is a different form from [/s/, plural] (as in cats, rats, puffs) because their segmental shape is different.
I feel like if you want to talk about “shape variants”, [/z/, PLURAL] and [/s/, PLURAL] should count as variants of a single form. And [/en/, PLURAL] (as in ox-oxen) should count as a separate form from them. But can we find a way to make this rigorous across languages?
Relatedly, how is “phonological conditioning” defined? I feel like the relationship between [/s/, PLURAL] and [/z/, PLURAL] “looks” phonological, whole the relationship between these and [/en/, PLURAL] does not “look” phonological, but a clear definition of “phonological conditioning” would be useful.
Yes, I thought about the definition of “form” and “shape variant” a lot, and I have some discussion in §5 of my 2021 paper on weldd forms. I start out from the presupposition that there is wide agreement that the English “-s”-plural and the “-en”-plural represent different affixes, and that [s] and [z] (as in cat-s and dog-s) are shape variants of th “-s”-affix. This seems to be uncontroversial, and it’s also uncontroversial that there are many clear cases like these, in many different languages. The same applies to stems: While “good” and “bett(-er)” are different stems (or forms), “ho[t]” and “ho[ɾ](-er)” are variants of the same stem (or form). I think the problem is that few linguists have tried to give general definitions of terms like “affix” or “root”, so they have not felt the need to say things like “an affix is a kind of form…”. So there is no clear sense of what exactly a “form” is – the term is used more narrowly in actual practice. (I say that a form is the same as an expression, but these terms do not have the same ranges of actual use; though I think that they have the same denotation.)