Martin Haspelmath (2011) has argued that the notion ‘word’ is incoherent, and therefore is not useful as a cross-linguistically valid comparative concept or as a language-internal descriptive category. As a result, in his view, the distinction between bound formatives, e.g., affixes and clitics, is untenable. For example, in Lithuanian, the bound formative –si is traditionally analyzed as a suffix when it is in one position (after the verb when the verb has no prefixes) and as a prefix when it is in another position (between the prefix and the verb in prefixed verbs):
(1) Lithuanian (Indo-European, Haspelmath 2011: 42)
1SG skutuo-si ‘I shave’ ne-si-skutu ‘I don’t shave’
2SG skutie-si ‘You shave’ ne-si-skuti ‘You don’t shave’
3SG skuta-si ‘He shaves’ ne-si-skuta ‘He doesn’t shave’
Haspelmath points out that due to its mobility, –si could also be analyzed as a clitic.
Haspelmath’s entire article is devoted to demonstrating that no single criterion or battery of criteria can pick out morphosyntactic words. Interestingly, all of his examples are glossed as though they comprise affixes or clitics. This is because the current glossing conventions used by descriptive linguists and typologists (mainly the Leipzig Glossing Rules) do not have a way of glossing bound formatives without committing to a particular analysis.
Without entering into a discussion of the merits of any particular analysis of any particular construction in any particular language, it seems that a way of glossing bound formatives in a non-committal way is desirable. This is especially important because, e.g., a typologist collecting data on bound formatives might be misled by a particular gloss into thinking that a particular bound formative is an affix, when in fact it could also be analyzed as a clitic. For example, the Modern Hebrew definite marker is typically glossed as a prefix (and so separated by a hyphen), while it has recently been shown to have behavior typical of clitics (Doron & Meir 2015+). Interestingly, the authors of this very article argued for its clitic status while glossing as though the definite marker is a prefix.
(2) Modern Hebrew (Afroasiatic, Doron & Meir 2015+)
ma ze ha-davar ha-mamaš taʕim ha-ze
what DEM DEF-thing DEF-really tasty DEF-this
Ideally, the glossing convention adopted would be one easily accessible on any keyboard, and since the plus sign (+) is not used to indicate any other type of morphosyntactic relationship, it would be possible to use it in order to indicate a bound formative without committing to its status as an affix or a clitic. (Admittedly, the plus sign might not be everyone’s cup of tea, aesthetically speaking.) When the precise nature of the bound formatives is unimportant, or if a linguist follows Haspelmath in jettisoning the principled distinction between affixes and clitics, this convention could be used.
For example, there is some disagreement about the proper analysis of bound formatives in the Coptic verb, with some linguists considering them to be affixes and others to be clitics. The proposed convention would allow linguists to gloss such examples without committing to a particular analysis.
(3) Coptic (Afroasiatic, Grossman 2014: 205)
‘He created him.’
Similarly, the Lithuanian examples could be glossed skutuo+si (‘I shave’) and the Hebrew examples could be glossed as ha+davar (‘the thing’).
It is important to stress that this proposal is simply to add another optional convention for linguists to use, in the face of doubts regarding the cross-linguistic validity of the criteria for distinguishing between affixes and clitics.
Doron, Edit & Irit Meir. 2015+. The Impact of Contact Languages on the Degrammaticalization of the Hebrew Definite Article. Forthcoming in Journal of Jewish Languages.
Grossman, Eitan. 2014. No case before the verb, obligatory case after the verb in Coptic. In: Eitan Grossman, Martin Haspelmath & Tonio Sebastian Richter (eds.) Egyptian-Coptic Linguistics in Typological Perspective, 203-225. Berlin: De Gruyter Mouton.
Hasplmath, Martin. 2011. The indeterminacy of word segmentation and the nature of morphology and syntax. Folia Linguistica 45(1): 31-80.
Je vais aller explorer le site Web que vous présentez.
The introduction of a more general use of the plus sign might solve a problem that hasn’t been discussed before as far as I know. Hyphens are sometimes part of standard orthographies, and this makes them ambiguous when they are used in glossed examples, since the reader will not be able to tell if the hyphen was there in the original source or if it has been added by the author of the paper. Thus, in an English example, egg-beater could represent both egg-beater and eggbeater. Even if this may be a marginal problem, I find it is wrong on principled grounds to introduce changes in the examples that are not retrievable.
While broadly sympathetic to Eitan’s proposal (as also to the arguments by Martin that underlie it), it seems to me that the problem is actually more serious than is acknowledged by the above discussion. Specifically, in many of the languages that I work on, the problem isn’t so much distinguishing between affixes and clitics, as it is between distinguishing affixes/clitics on the one hand and separate words on the other. Or in other words, between “bound” and “free” forms. If, following Eitan, we adopt the “+” symbol for a boundary involving either an affix or a clitic, then we should also adopt some other symbol for a boundary that may involve either a bound form or separate words. Perhaps “_”?
It’s an interesting proposal, but if one reads carefully the Leizpig Glossing Rules (revision of February 2008), one sees it stems from a misunderstanding. Rule 2 state: “Segmentable morphemes are separated by hyphens, both in the example and in the
gloss.” Since the term “segmentable morphemes” covers both affixes and clitics, the hyphen is already non-comitative with respect to this question.
It is only the equal sign that unequivocally signals clitic boundary, and thus one may assume that an author that uses the equal sign for clitics, uses the hyphen only for affixes, but this is not necessarily the case.
So in fact what is needed is to add a sign for unequivocal affixes, and for this the + symbol could work very well.
That’s true, but I think that pragmatic enrichment operates here: because of the extensive use of the equal sign for clitics, linguists often interpret the hyphen as meaning ‘affix’ (“If it were a clitic, the author would have used the equal sign; since s/he didn’t, it’s probably an affix”.) Moreover, the “+” is pretty ugly, and affixes are common, so I don’t think that linguists would adopt it for very frequent use. Relatedly, the use of the hyphen for affixes is so established that no one would make such a big change in their practices.
However, in the end, the idea is to have practices that can be clearly and unambiguously interpreted, and the suggestion here is just for the probably rare cases where (a) linguists care about this problem and (b) want to remain explicitly non-committal about it. This could also be done verbally in books and papers, as people often do when they have unusual gloss labels or deviate from IPA in some way, but that’s true across the board; in general, it’s helpful to have conventions that don’t require special comments.
You’re right about the pragmatic enrichment, but this is enrichment is on the responsibility of the reader. Browsing through linguistic literature, I get the impression that many linguists use the hyphen as a default segmentation marker, without paying any attention to the clitic/affix question at all, even when occasionally they may use the equal sign to signal out some particular clitic. In fact, the sharp distinction between clitics and affixes is only done in some frameworks. So, if at all, the committed signalization of something as a “definitely affix and not clitic” should be specially marked, and I think that marking it by + would respect an already existing tradition in some schools of linguistics to mark morpheme boundaries by + (in contrast to # symbolizing word boundary, which probably includes clitic boundaries as well).
I’m just afraid that if your proposal gets accepted, since now it is in vogue not to distinguish clitics/affixes, we will see an affluence of + signs cluttering our examples 🙂
There are typos in the Lithuanian examples: the negated 1s is ne-si-skutu, not -skuto, and the nonnegated 2s is skutie-si, not skutue-si (as correctly given in Haspelmath 2011).
Fixed, thanks a lot!
This is without doubt a much needed addition to the LGR. It will be especially useful to write articles discussing whether a particular morpheme or set of morphemes in a particular language should be analyzed as clitics or affixes, allowing the writer to remain non-committal before expounding the arguments for and against each possible analysis.
As you point out, the symbol + is not particularly nice-looking amidst a string of letters, but we don’t have much choice if we want to choose an easily typeable character, since the chevron, underscore and tilde are already taken for other purposes.