June 8, 2022 at 7:52 pm #137sraKeymaster
A recent exchange of emails with Martin Haspelmath provided an opportunity for me to clarify various aspects of my views on some basic issues in Linguistics. This began with a FaceBook query by Martin about what was meant by saying clitics are “prosodically deficient”; I responded with references to my 2005 book on clitics, but the discussion went on well beyond that point. I quote below my most recent reply to Martin, which goes into more basic things.
We are clearly coming at these matters with very different world views, and it is doubtful that discussion or citation of facts will persuade either of us that the other has the right end of the stick.
On 7 Jun, 2022, at 8:52 AM, Martin Haspelmath
Thanks, Steve, for these comments!
My research programme is to find generalizations independently of assumptions about innate building blocks,
Here we definitely part company. My sense is that you (Martin) want to define categories within a language on the basis of whatever language-particular properties you find associated with a pre-systematic linguistic notion. This Nominalism reminds me of the kind of symptomatology that prevailed in our field in the 1970s and other times: defining e.g. “subject” on the basis of a bunch of behavioral properties, some (but not all) of which might be found in particular languages, and stopping there. As far as I’m concerned, this is where the project starts: you have a bunch of properties, and you want to deduce that behavior from the way grammars are organized, the language-particular implementation of a form provided by the language faculty. For example, I am a great fan of the typology of clitics offered by Pullum and Zwicky, and their demonstration that English “n’t” is an inflectional affix, not a clitic, but I am unhappy that they considered their work done when they had provided a useful and insightful set of behavioral criteria for calling something a clitic (of one or the other sort). I tried to improve on that in my book, by providing a view of grammatical architecture from which their behavioral properties followed. But that architecture necessarily involves a set of things that you would call innate building blocks. I’m an unrepentant Idealist, here.
so I cannot work with notions such as “p-word” that are defined differently in different languages:
(Martin:) The problem I’m trying to solve is how to characterize clitics in general terms without presupposing abstract categories such as “phonological word” that are defined differently in different languages.
(Steve:) Agreed. “PWord” is a universal category of prosodic structure, though the phonological effects associated with it differ from language to language.
I’m wondering if you would agree that one needs to assume that “p-word” is an innate (pre-established) category in order to use different criteria in different languages. This is a very strong assumption (which some would say is incompatible with Darwin’s Problem), but one can of course make it. (However, one cannot at the same time claim that one doesn’t assume innateness beyond Merge; because of this contradictory attitude, I wrote this paper: https://ling.auf.net/lingbuzz/005158)
As far as I’m concerned, “PWord” is a category in a universal hierarchy of prosodic structure, from segments through syllables, feet, PWords up to (at least) Intonational phrases. The phonological correlates of e.g. syllabic structure will vary from language to language, and the same for other, including categories, such as PWords. I don’t see this as a problem. As you say, it does imply that I have a rather richer (some would say out of date) view of the content of the human language faculty. I certainly don’t adhere to the “Merge is all there is” school of thought.
In Anderson (2005: 23), you give the following definition:
Phonological Clitic: A linguistic element whose phonological form is deficient in that it lacks prosodic structure at the level of the (Prosodic) Word.
But this sort of deficiency applies to most affixes – would you want to say that most affixes are phonological clitics?
What I’d say is that (most) clitics are affixes. Morphosyntactic (“special”) clitics are affixes introduced at the phrasal level, and most of these are prosodically deficient (though some, such as maybe Italian loro, are not). Virtually all word level affixes, introduced by rules of the morphology (and thus not lexically listed, contrary to the “morphology is just the syntax of small domains” view), are prosodically deficient.
So what you seem to be saying is that an *affix* is a segment sequence that is not “lexically listed” but “introduced by rules”.
Now this seems to presuppose that there is a difference between “rules” and “lexicon”, something that I don’t find obvious at all. I don’t want to make this strong assumption, so I cannot define “affix” in this way.
“Rules” in this sense are a formal expression of language-particular systematic relationships. Such relationships commonly have formal markers. For instance, the relation between declaratives and interrogatives in e.g. English is marked, in part, by a difference in word order. The relationship between singulars and plurals, where this is not lexicalized as an idiosyncratic difference, is marked by an additional /z/ at the right edge of the noun. And so on. An “affix” on this view is some phonological alteration (including not only the addition of material, but also changes such as Ablaut, Umlaut, subtraction, metathesis, etc.) in the shape to one term of the relation to yield the other term. For me, virtually all morphological markers are such correlates of relationships, a notion that can be expressed formally by rules (but perhaps in other ways as well). That is distinct from contentful bases. The difference between a notion of morphological relations as mediated by the addition of lexically specified affixes, such that all complex morphology is basically compounding, and the way I prefer to see it goes back at least to the difference between the views of René de Saussure, on the one hand, and his brother Ferdinand on the other: see my essay in Louis de Saussure’s and my little (2018) book on René.
(Steve) Purely phonological (“simple”) clitics are prosodically deficient lexically listed items manipulated by the syntax and later by the phonology.
It seems that the idea of “simple clitics” comes from English reduced auxiliaries, but such items aren’t all that common, don’t you think?
On the contrary: most of the things that get called “particles” in many languages are in fact phonological (né “simple”) clitics, though of course many of them also occupy unusual structural positions, so they are (also) morphosyntactic (né “special”) clitics. Sorting out these two independent dimensions of what are indiscriminately called clitics in the literature was a major point of my clitics book.
I’m wondering about elements such as English “if”, “at”, “or”, “only”, “too”, which do not occur on their own and are not like ordinary verbs, nouns and adjectives – aren’t these clitics, too? They are almost always unstressed (“phonologically deficient”?), and they combine with hosts of different word classes. These kinds of elements are very common and found in all languages. Why aren’t we treating them as clitics most of the time? Is it because they don’t present orthographic problems in English?
Interesting question, and one I haven’t thought about. Of course stressability (or the lack thereof) is not, for me, a diagnostic property for clitic status, and in most of these cases the lack of stress is surely grounded in information structure: note the all of your examples have segmental structure that is incompatible with consistent relegation to fully unstressed status. All of them also seem to me to be distributed by principles of the syntax, and not as phrasal markers (the status of morphosyntactic clitics, on my account). So I don’t see any reason to call any of these things clitics of any sort, though more thought might persuade me that there’s more to be said here.
I think you’ll agree that our differences of opinion are profound enough that it’s unlikely they will be resolved.
Anderson, Stephen R. (2005). Aspects of the Theory of Clitics. Oxford: Oxford University Press.
Anderson, Stephen R., and Louis de Saussure [eds]. (2018). René de Saussure and the Theory of Word Formation. (Classics in Linguistics 6). Berlin: Language Science Press.]
- You must be logged in to reply to this topic.