[Lingtyp] words, bound forms, welded forms

Martin Haspelmath haspelmath at shh.mpg.de
Mon Jan 28 20:03:19 UTC 2019

Mark's examples from Eton (Bantu) illustrating different kinds of 
grammatical forms and rules remind us how complex languages can be -- 
thanks for this, Mark!

However, the reason I coined the new term "welded" was to have a term 
that does not come with any kind of baggage, and it would defeat the 
purpose to change the meaning that I proposed -- and note that my 
definition explicitly excludes non-segmental interactions 
(I know that this may sound Eurocentric, but I admit that I'm not a tone 
expert, and tonologists should feel free to coin whatever terms they need.)

We all have the intuition that there are degrees of "coalescence" or 
"tightness of combination", but we cannot measure such degrees yet, so 
most of the relevant literature remains vague and subjective.

In order to make headway here, we need to compare things that are 
objectively comparable, not merely appering comparable on an intuitive 
basis. Probably someone can eventually define a clear concept of 
"attachedness" that combines segmental interaction and tonal 
interaction, but my proposal was more modest. (I left out suprasegmental 
criteria for the reasons mentioned by Adam Tallman.)

I don't know how best to compare tonal patterns, but I would insist that 
we need a term (like "welded") for bound forms that show segmental 
interaction, and also a term "form" that is limited to segment 
sequences, because otherwise we cannot define "morph" and "root", some 
of the most basic terms of comparative grammar. (Thus, a non-segmental 
high tone pattern cannot be a "form", but must be something else.)

Sorry if such discussions sound boring, but I don't see an alternative 
to precise terminology that is used uniformly across the discipline.


On 25.01.19 13:23, Mark Van de Velde wrote:
> Dear colleagues,
> The tone system of the Cameroonian Bantu language Eton may be useful 
> for illustrating the usefulness of distinguishing between what Martin 
> calls "bound" versus "welded" (if I understood it correctly), and to 
> show that the definition of welded form should better not be 
> restricted to *segmental* phonological interaction.
> Eton has a rule that copies high tones. The copied high tone 
> subsequently attaches to its right according to a set of rules that 
> take into account the type of morphological boundary the tone has to 
> cross. (The prosodic status of the syllable to which it attaches plays 
> a role as well, which is not relevant here.)
> Three types of boundary have to be distinguished: a strong (word) 
> boundary, a weak (affix) boundary and an intermediate boundary that I 
> called "clitic" due to its intermediate status. A high tone that 
> crosses a weak or intermediate boundary deletes any following low 
> tone, and then further spreads (if it crossed an affix boundary) or 
> stays put (if it crossed a clitic boundary). A high tone that crosses 
> a strong boundary delinks a following low, which becomes floating. 
> Therefore, we need three degrees of "welding" in an analysis of Eton 
> tonology.
> These three degrees are defined by tone rules, but they correspond to 
> typical morphological distinctions. The identification of the strong 
> boundary as a word boundary and that of the weak boundary as an affix 
> boundary are unproblematic. The intermediate boundary is found between 
> a genitive marker and a noun. Genitive markers are generally phrasal 
> affixes with an intermediate degree of welding in the Bantu languages, 
> so "clitic" seems right here too.
> Crucially, high tone forms - linguistic forms that lack a segmental 
> shape and consist of a high tone only - come in three types according 
> to the way in which they attach to the right. By analogy and for 
> descriptive ease, I called them tonal affix, tonal clitic and tonal 
> word. Clearly, all of these are clear examples of bound forms, but 
> they show three different degrees of welding. In the following 
> examples H stands for a floating high tone, # is a word boundary, = a 
> clitic boundary and - an affix boundary, ? downstep (caused by a 
> floating low tone in between high tones).
> ??-???? 'clan', ??-???? 'hole', H# LOCATIVE, H= GENITIVE, H- CONSTRUCT 
> H# ??-???? --> ?????? 'in the clan',  H# ??-???? --> ??????? 'in the hole'
> H= ??-???? --> ?????? 'of the clan', H= ??-???? --> ?????? 'of the hole'
> H-??-???? di? --> ?????? di? 'this clan', H= ??-???? di? --> ?????? 
> di? 'this hole'
> (For more details: Van de Velde, Mark (2008) /A Grammar of Eton./ 
> Berlin: Mouton de Gruyter, p 52-59)
> If I understood Martin right, welding is relative, gradual, and 
> language/construction/domain specific, whereas bounding is discreet 
> and crosslinguistically definable.
> Best wishes,
> Mark
> On 25/01/2019 11:56, Martin Haspelmath wrote:
>> On 25.01.19 11:29, Sebastian Nordhoff wrote:
>>> On 1/25/19 10:27 AM, Martin Haspelmath wrote:
>>>> *a welded form is one that shows segmental phonological interaction *with its neighbour.
>>> I can't help wondering whether "phonologically conditioned allomorphy"
>>> would not be an existing concept which has identical properties to the
>>> concept of "welding":
>>> - there are two or more forms to choose from
>>> - the choice of form depends on the shape of the "neighbour"
>>> Best wishes
>>> Sebastian
>> Yes, a welded bound form would always exhibit "phonologically 
>> conditioned allomorphy", but the latter notion is much broader and 
>> not well-defined. For one thing, the "allomorph" concept depends on 
>> the "morpheme" concept, on which there is no agreement, and much 
>> confusion. For example, are German /-er/ and /-en/ allomorphs of a 
>> single {PL} morpheme? Are English /-ness/ and /-ity/ allomorphs of a 
>> single {ABSTR.NOUN} morpheme? Are French /tomb(-er)/ 'fall' and 
>> /chute/ 'fall' allomorphs of a single {FALL} morpheme?
>> If one includes such suppletive alternants under "allomorphy" (as 
>> almost everyone does), then one also has phonologically conditioned 
>> allomorphs that are not welded forms according to the proposed 
>> definition (because these must be variants of the same form) (cf. 
>> Carstairs 1988 on "phonologically conditioned suppletion").
>> For example, Dutch has the plural forms /-s/ and /-en/, which are 
>> distributed according to phonological conditions, and the division of 
>> labour between the English suffixes /-ize/ and /-ify/ (/computer-ize, 
>> French-ify/) is also more phonological than anything else, it seems. 
>> There is no weldedness in Dutch /-s/ or English /-ize/.
>> I find the notions of "form" and "form variant" (= a form with a 
>> somewhat different shape due to a phonetically natural sound 
>> alternation) much more viable than the traditional "morpheme" and 
>> "allomorph" notions (the confusion of "morpheme" has not improved 
>> since Mugdan's (1986) paper, where he traced the diverse uses of this 
>> term since the 1880s).
>> Best,
>> Martin
>> -- 
>> Martin Haspelmath (haspelmath at shh.mpg.de)
>> Max Planck Institute for the Science of Human History
>> Kahlaische Strasse 10	
>> D-07745 Jena
>> &
>> Leipzig University
>> Institut fuer Anglistik
>> IPF 141199
>> D-04081 Leipzig
>> _______________________________________________
>> Lingtyp mailing list
>> Lingtyp at listserv.linguistlist.org
>> http://listserv.linguistlist.org/mailman/listinfo/lingtyp
> -- 
> Mark Van de Velde
> director of LLACAN (CNRS - Inalco), Paris
> http://llacan.vjf.cnrs.fr/pers/vandevelde/
> Please note my new address:mark.vandevelde at cnrs.fr

Martin Haspelmath (haspelmath at shh.mpg.de)
Max Planck Institute for the Science of Human History
Kahlaische Strasse 10	
D-07745 Jena
Leipzig University
Institut fuer Anglistik
IPF 141199
D-04081 Leipzig

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/lingtyp/attachments/20190128/902d6e10/attachment.htm>

More information about the Lingtyp mailing list