A flag is a cover term for an adposition or a case-marker, as I explain in my recent 2019 paper on flagging and indexing (in the journal Te Reo, run by the Linguistic Society of New Zealand). All comparative linguists know that in many cases, it is not quite clear whether we should treat an element (such as the Japanese accusative marker o, or the Arabic dative marker li) as an affix and thus a case-marker, or as an adposition (i.e. a form that is not an affix). The term flag serves as a convenient cover term for comparative linguists in situations where it does not matter.
I first used this term in my 2005 paper on ditransitive alignment, and it has been adopted by quite a few people in the meantime, so there is obviously a demand for it. The term comes originally from Relational Grammar (I found it, for example, in Judith Aissen’s work of the 1980s).
Accusative flags, dative flags, instrumental flags and so on behave rather similarly, regardless of whether they are written separately or written as affixes. Authors who want to distinguish strictly between adpositions and case-markers run into all kinds of problems. For example, as I noted in my 2018 book review and in my recent 2020 paper on “Ergativity and depth of analysis”, Baker’s (2015) dependent case theory is weird in that it posits not only ordinary segmental case-markers, but also accusatives and ergatives with zero exponents, zero adpositions that assign visibile cases, and visible adpositions that look like case affixes. And for Hungarian, there are two prominent papers that argue that the case suffixes are actually postpositions, i.e. not really different (other than with repect to vowel harmony) from the flags that are written separately (Creissels 2006, Trommer 2008). Clearly, we need a general term for case affixes and adpositions, because these kinds of elements share many properties in common.
Analogous to the terms “case marking” and “adpositional marking”, we can use the term “flagging” in two senses which are differentiated by the context: (i) the process of providing a nominal with a flag; (ii) the set of flags of a language. And we can say, quite elegantly, that a nominal is unflagged when it lacks a flag (compare this with the awkwardness of “uncasemarked”, or “not marked adpositionally”).
In my Te Reo paper, I also note that flagging is what people mean most of the time when they talk about “dependent marking”, except that the term flagging is more precise and somewhat narrower. Dependent marking presupposes the notion of “dependent”, which is not generally well-defined, and according to the original definition, it includes concord marking on modifiers (e.g. gender agreement on adnominal adjectives). But we want a special term for dependent markers that are relators, i.e. that provide information about the kind of relation that holds between a noun/verb and its arguments/modifiers.
Now interestingly, Te Reo published a commentary to my paper alongside it, by my distinguished colleague (and friend) Bill Croft, who agrees with me on many points – not suprisingly, because I have been strongly influenced by him ever since his 1990 book “Typology and universals” (2nd edition 2003). Croft’s piece in Te Reo was originally a reviewer’s report but was expanded because Bill felt he had quite a few things to say. One of his points concerns my definition of “flag” which he is not fully happy with:
“A flag is a bound form that occurs on a nominal and that indicates the semantic or syntactic role of the nominal with respect to a verb (in a clause) or with respect to a possessed noun (in a complex nominal).” (Haspelmath 2019: 96)
Croft has problems with the part “bound form”. He thinks that (i) it is not always clear whether a form is bound or free, and (ii) the difference is really gradual and boils down to degrees of grammaticalization.
But my inclusion of “bound form” was quite conscious, because not all relation-denoting forms are flags. When I talk about an event “in the middle of the winter”, or a paper “on the topic of grammatical marking”, or a dispute “that concerns the nature of linguistics”, I use relation-denoting words: “middle”, “topic”, and “concern”. But these are not bound forms that “occur on a nominal” – they are nouns and verbs that are not dependent in the sense in which grammatical markers are dependent on their hosts. And this dependency is best characterized as boundness, i.e. as inability to occur in isolation. The term bound was introduced precisely in this sense by Bloomfield (1993) and is routinely used in this sense by morphologists (e.g. Bauer 2019: 10). Linguists don’t worry much about this term (for example, Bauer does not even include it in his subject index, let alone his glossary of technical terms), but whenever it is defined clearly, a bound form is defined as a form that cannot occur in isolation (some linguists talk about forms being “phonologically bound”, but the term “welded” is more suitable for this sense).
But what about Croft’s second point? Is the difference really gradual, so that instead of making a binary distinction between bound and non-bound forms, we should try to measure degrees of grammaticalization? Maybe, but how would we measure degrees of grammaticalization? How can we tell whether languages tend to have a binary contrast or whether such distinctions are gradual? It is clear that underlying every measurement, there must be a concept that is defined in a non-gradual way. For example, to measure temperature, we need to make a binary distinction between temperature and weight. If there were only a gradual distinction between the two (maybe some stuff, such as smoke, is not readily measured in terms of either temperature or weight, but should be measured by some intermediate scale?), it is unclear how measurement would work.
So I think that it is the other way round: Grammaticalization (and its degrees) must be assessed in terms of basic binary notions such as boundness. There may well be degrees of grammaticalization, but we will know only once we have measured these degrees via nonscalar basic concepts. I do not insist that boundness (i.e. non-occurrence in isolation) must be the best concept, but it has worked very well for many purposes so far, and I do not know of an alternative for defining flags.
Another interesting terminological question concerns the term “case”: Croft makes a side remark (in footnote 4) that he does not find the term “case marker” (as a cover term for case affixes and adpositions) confusing:
“This use of terms is transparent: marker is commonly used for terms regardless of boundness or wordhood, and case expresses the grammatical semantic category.” (Croft 2019: 127)
Now it is true that since Fillmore’s (1968) paper “The case for case”, the term case has also been used for semantic roles, so that any instrumental marker (including instrumental prepositions like with) is a “case marker” in this sense. But the term case was originally used for systems of role marking that involve affixes. When we talk about a language having a “case system”, we do not think of adpositions. Thus, I think it is confusing to use the term “case” in a sense in which it refers to anything other than a type of affix (or a feature value) in a system that crucially involves role-marking affixes. Fillmore created a kind of confusion that we should leave behind us, and the term flag allows us to do that.
(There is also another respect in which I disagree with Croft’s terminology: While it is true that the term “marker” is used regardless of “wordhood”, it is not used regardless of boundness: a grammatical marker is defined as a type of bound form, as I noted in an earlier blogpost.)
In his commentary text on my 2019 paper, Croft also discusses the difference between term pair indexing and flagging on the one hand, and his distinction between indexical and relational marking on the other. This is an interesting discussion that I may come back to on another occasion. Bill and I agree that the well-known distinction between head-marking and dependent-marking is not ideal in a number of ways, and that comparativists should spend more time thinking about these core aspects of grammar, rather than simply assume that we have already found the right conceptual distinctions.
Baker, Mark C. 2015. Case: Its principles and parameters. Cambridge: Cambridge University Press.
Bauer, Laurie. 2019. Rethinking morphology. Edinburgh: Edinburgh University Press.
Creissels, Denis. 2006. Suffixes casuels et postpositions en hongrois. Bulletin de la Société de Linguistique de Paris 101(1). 225–272.
Croft, William. 1990. Typology and universals. Cambridge: Cambridge University Press.
Croft, William. 2019. Comparative concepts and practicing typology: On Haspelmath’s proposal for “flagging” and “(person) indexing.” Te Reo 62(1). 116–129.
Fillmore, Charles J. 1968. The case for case. In Bach, Emmon & Harms, R.T. (eds.), Universals in linguistic theory, 1–88. New York: Holt, Rinehart and Winston.
Haspelmath, Martin. 2005. Argument marking in ditransitive alignment types. Linguistic Discovery 3(1). 1–21.
Haspelmath, Martin. 2018. Review of “Baker, Mark. 2015. Case. Cambridge: Cambridge University Press.” Studies in Language 42(2). 474–486.
Haspelmath, Martin. 2019. Indexing and flagging, and head and dependent marking. Te Reo 62(1). 93–115. (doi:10.17617/2.3168042)
Haspelmath, Martin. 2020. Ergativity and depth of analysis. Rhema (to appear).
Trommer, Jochen. 2008. “Case suffixes”, postpositions, and the phonological word in Hungarian. Linguistics 46(2). 403–437.