Clitics and Cliticization in Hebrew, Part 2

This post is the second part of my brief discussion of clitics in Biblical Hebrew. In my first post on this topic, I ended with a question about the nature of the construct/סמיכות form of nouns: are they clitics or not? If the answer is yes, then how do we deal with the non-use of the maqqef and the presence of an independent accent? If the answer is no, then what do we call these forms?

To review, I summarized the facts in the following way:

The challenge, as indicated above, is that many such clear cases of cliticization are not marked by a maqqef, which is the normal Masoretic indicator of a clitic. The phrase עַל־פְּנֵ֥י הַמָּֽיִם ‘upon the surface of the water’ in Gen 1.2 is illustrative: the maqqef signals the clitic status of the preposition עַל, but the bound word פְּנֵי is not connected to its clitic host הַמָּיִם by a maqqef. Yet the clitic status of construct/bound forms is not only suggested by the examples that do appear with a maqqef (e.g., אַדְמַת־קֹ֖דֶשׁ ‘land of holiness [= holy land]’, Exod. 3.5) but also by the vocalization differences between the free and bound forms: assuming an underlying /dabar/ for ‘word’, the free form, which has a primary word stress, דָּבָר exhibits pretonic and tonic backing and raising ([a] to [å̄], IPA [ɔ], whereas the bound form דְּבַר exhibits no pretonic change and tonic reduction to schwa, suggesting that originally the form did not carry primary word stress.

The solution I suggest below is heavily dependent on a recent conversation with Elan Dresher, Professor and Chair of the Linguistics Department at the University of Toronto:

It may be that the explanation for the absence of the maqqef in many construct forms is historical. At some point construct forms began to be reinterpreted syntactically (or morphologically), rather than prosodically. That is, the Masoretic accents do not govern the appearance of construct forms. For a bound form like דְּבַר that has an independent accent and no maqqef, there must have been some lexical mark that indicated it was a clitic, and the phonology must have been sensitive to that, rather than to its prosodic status under the accents. Put differently, construct forms are partially fossilized prosodic clitics that are no longer necessarily real prosodic clitics. This suggests that the Masoretes themselves were not quite sure what to do with the bound forms — that the old system had broken down but no new generalizations had emerged to replace it. Or perhaps the Masoretes inherited a system that resulted from the breakdown of the older one, where individual lexical items go their own way in the absence of a new generalization that would govern them.

So, to bring my post on the clitics to an end, using the principles and criteria deduced by those engaged in the typological study of clitics (e.g., Zwicky, Zwicky and Pullum, Klavans, Anderson) allows us to classify numerous function words in Biblical Hebrew as simple clitics, either attached directly to their host or connected with a maqqef (see also Holmstedt 2010 [RBL review of Alan Kaye’s Morphologies of Africa and Asia]). However, it requires a bit more fluidity and perhaps (as described above) a historical perspective to account accurately both for the status of bound words and for those words that are unexpectedly cliticized (to avoid a stress crash). I agree fully with Dresher when he writes,

“Whether an orthographic word is cliticized or not depends on a complex set of prosodic, phonological, and syntactic conditions, some of which are reviewed in the following sections. It turns out that cliticization is tightly tied in with the entire Tiberian prosodic system, and cannot be understood without taking into account the principles of phrasing.” (2009:99)

15 Responses to “Clitics and Cliticization in Hebrew, Part 2”

  1. ed cook Says:

    It seems odd to think of construct (lexical) nouns as clitics; I tend to think of small words, functional words, quantifiers and determiners as “cliticizable,” but not contentive words. Does this appear cross-linguistically? I’m not challenging, I really want to know. Couldn’t the prosodic phenomena be a by product of a compounding like process (e.g. Eng. blackbird vs black bird)?

    Love the blog, keep up the good work!

    • robertholmstedt Says:


      That’s a very good question.

      First, I suggest that the compounding analysis doesn’t work so well, though, since the result in a compound is something different than simply the sum of the two words. That is, the referent of “blackbird” is different (more restrictive) than “black bird”, the semantic overlap notwithstanding. To my knowledge (and according to Zwicky and Pullum’s descriptive criterion #4, listed in my first post), such semantic changes are not characteristic of cliticization. And that’s the problem: we have lexical words that are obviously bound and do not reflect a lexical-semantic change, both of which are characteristics of clitics.

      So, second, while some linguists (see, e.g., Caink 2006 in my blibiography) may limit clitics to grammatical words, it is definitely worthwhile to note that it is *extremely* difficult to find any claim about clitics only coming from a certain lexical class (grammatical words vs. content words) in a wide range of works, from Zwicky 1977 (and 1994) to Anderson 2005. Similarly, Borer (2003, also in my bib) makes room for groups of words in various languages that only fit some of the prototypical characteristics (which is why, for example, there is the very good question about whether “clitic” is word class like “noun, verb, particle”).

      Finally, in light of more prosodic-syntactic (versus word-class or syntax-dominant) approaches to cliticization, I think that we can say that the understanding of cliticization simply needs to be flexible enough to include the Semitic bound/’construct’ strategy.

      (Again, good question — I should incorporate some of this in my essay.)


  2. ed cook Says:

    That makes a lot of sense. I’ll have to give it some more thought. Also in favor of the construct=clitic view is that lexical constructs often grammaticalize as prepositions, the latter being the prototype, almost, of proclitic forms in Hebrew, yet without phonological change. (Although tell me — again, a sincere question due to ignorance — aren’t prepositions considered as “lexical” in the generative tradition, but “functional” from a grammaticalization standpoint? Or is it a question of degree?)

    • robertholmstedt Says:


      Hmmm … I don’t think prepositions are considered lexical (=content words) in generative grammar. Typically, they are classified as grammatical words (=function words = non-content words = non-refering words, that is words that don’t reference either a concrete entity or class of entities but rather signal spatial, temporal, etc. relationships between referring words) with selectional features (e.g., requiring a noun complement).

      Sorry for all the alternate definitions. I felt that not doing so left too much ambiguity (and since it’s all still off the top of my head, I’m sure there’s still some ambiguity in there somewhere).


  3. ed cook Says:

    I get that prepositions are non-referring, etc., but they are lexical heads of their phrases (like e.g, Noun), not functional heads (like e.g, Tense). I guess I’ve been thinking that lexical : functional = content : grammatical. But it looks like the first opposition is syntactic, the second semantic. Thanks for the clarification.

    • robertholmstedt Says:


      Sorry if I misread you — it’s hard to know from what direction someone’s coming at such issues. Obviously, you sorted it out yourself more clearly than anything I wrote contributed!

      But to elaborate a bit: You’re last statement is how I understand the distinction — whether or not something is a syntactic head of its phrase is a different issue than what word class that same item belongs to.

      And in generative phrase structure, the distinction between functional head of a phrase (e.g., T(ense) in the TP, Foc(us) if the FocP) and lexical head (e.g., N(oun) in the NP) has to do strictly with the architecture of the phrase.

      Although I’ve not thought of it before, probably because I cut my teeth on generative grammar, that the overlap in terms is annoyingly confusing.

      Great discussion — keep stopping by!


  4. Chip Hardy Says:

    Unless one uses a very broad definition for cliticization, that is, something like “the loss of primary stress”, and then extrapolates that a clitic is anything which exhibits cliticization, I don’t think the bound forms of a construct phrase should be understood as clitics.

    Many words have reduced forms and even loss of accent when placed in certain constructions determined by the phonotactics of their individual language, but this reduction per se does not make them clitics. Consider, for example, the common English expression, “What is up?” which undergoes loss of stress and phonological erosion SUBSEQUENT TO the copula becoming the enclitic, ‘s: /wot iz up/ > /woz up/ (cliticization) > /wəzup/ (loss of stress) > /sup/ (reduction).

    Another problem for the Hebrew bound form as clitic proposition is that bound forms themselves may be modified as if they are independent words, whereas clitics do not typically take modification. For example one finds Hebrew phrases such as, מלכי הארץ הצדיקים ‘The righteous kings of the land’, where the adjective clearly modifies the nomen regens (i.e. the bound element of the construct phrase).

    Just a few scattered thoughts, but I’ll have to side with Ed this time around.

    Thanks for the great blog, Rob!

    • robertholmstedt Says:


      Perhaps we should both keep keep reading on clitics and cliticization. I, too, started with some of the narrower notions and the broader reading on the issues forced me to modify my views. The typological literature indicates clearly that the story is not so cut and dried as many brief treatments, which is what my essay (following Dresher’s work) makes clear. Anderson’s book (cited in the biblio to part 1 of the post series) is a good one to start with, since he pulls in diverse data and deals expertly with the phonological and morphological complexities.

      Your third paragraph illustrates precisely what Dresher has called for — a modification of the standard view of clitics to include Semitic bound words. Simply put, the definition of cliticization must be flexible enough to incorporate the Semitic data. Moreover, it seems to me you’re still working with the “word class” notion of clitics, which has shown to be seriously inadequate.

      Keep pushing, though. ;-)

      • Chip Hardy Says:


        The typology approach is good and helpful when describing a linguistic category like clitics that occurs cross-linguistically (and I readily admit I have not read the Anderson study); however, in such studies there is necessarily a tendency to smooth out some of the rough edges. My intention was merely to point out one particular rough edge in Ancient Hebrew, that is, a clitic such as על is functionally different than the bound member of a construct phrase though phonologically they share certain properties, and further it may be possible that the phonological similarities are explainable by other means like prosodic conditions of the individual language without an appeal for a broader definition of clitics.

        Thanks for your response and keep pushing back!

      • robertholmstedt Says:


        We certainly agree on the limits of typology — you’re talking to a generativist who uses typology, after all. (I find Neymeyer’s criticism of typology spot on.)

        The tension of having a preposition like על as well as a bound noun like אדמת both fall under what we call cliticization is precisely what Dresher is talking about, though. You should read his article (cited in my biblio) — he’s a phonologist, after all, and is fully aware of the alternate possibilities you’re suggesting. And to be honest, this is exactly what he and I discussed earlier this summer when I was writing this essay and trying to figure out how to analyze the construct state. The problem with this appeal to other prosodic conditions, though, is that instead of modifying the description of cliticization — which is still in flux anyway! — is the lack of elegance and economy. Why create a whole separate phenomenon when we can simply nuance an existing one?

        Pushing, pushing, pushing, keep those objections coming … rawhide! (Sorry, I got carried away.)

      • Chip Hardy Says:

        Yes, but as you well know, parsimony cuts two ways! Are bound forms clitics and therefore follow certain phonological conditions thereof OR are clitics in Hebrew, historically speaking, bound forms and follow certain preconditions of Semitic phonology? I will leave that gordian knot for someone else; it is way above my pay grade… In the mean time, I will take a look at Dresher’s paper.

        Onward and, hopefully, upward!

      • robertholmstedt Says:


        Not so! I decided long ago that parsimony only cuts my way. ;-)

        Email me for a pdf of Dresher’s paper.


  5. ed cook Says:

    FWIW, John Huehnergard agrees that “Bound (construct) forms are proclitic” (Historical Grammar, p. 8). But what do you do with construct forms that don’t change form in construct, e.g., /sefer ha-ish/? Do we say that /sefer/ here is somehow prosodically different than absolute /ha-sefer/? Is it a clitic, without changing form or shifting stress?

    • robertholmstedt Says:


      That’s interesting and encouraging.

      The lack of morphophonological (endings and/or just vowels) change is secondary from a prosodic perspective. It’s the lack of primary word stress that is determinative. But notice also the historical reconstruction that is part of my explanation — that the Masoretes has reanalyzed some of these bound nouns, which is why some have no separate ta’am and some do.

      Thanks for the input and Huehnergard reference. I wonder if that comment is John’s addition or if it goes back to Lambdin. I’ll have to check my old copy… I wouldn’t be surprised if Lambdin was ahead of the field on this, too.

Comments are closed.

%d bloggers like this: