haspelmath at EVA.MPG.DE
Fri Sep 23 13:11:40 UTC 2005
Note that "forward gapping" in SOV constructions as defined by Ana
Carrera is not really a case of "gapping" in the strict sense because it
does not leave a "gap". While in SVO constructions we have "S[V]O+S[ ]O"
(where brackets identify the antedecent and the ellipsis site), in SOV
constructions we have "SO[V]+SO[ ]", where the ellipsis site is at the
margin. An example comes from German:
(1) ...dass Merkel die Grünen [liebt] und die SPD den Kanzler [ ].
'...that Merkel [loves] the Greens and the SPD [ ] the chancellor.'
Only the English translation has a medial gap here.
In SOV constructions, one might instead apply the term "gapping" in the
liteeral sense, for a case where the O is ellipted, as in (2):
(2) Turkish (Kornfilt 1997:120)
Hasan [istakoz-u] pisir-di, Ali de [ ] ye-di.
Hasan lobster-ACC cook-PAST(3SG) Ali and eat-PAST(3SG)
‘Hasan cooked, and Ali ate the lobster/Hasan cooked the lobster, and Ali
I would therefore propose that the term "gapping" should not be used in
a typological context. Instead, we can use the terms "forward ellipsis"
(or analipsis) and "backward ellipsis" (or catalipsis), and if we want
to focus specifically on verb ellipsis, as Ana Carrera does, then we can
talk about "verb analipsis" etc. (The term pair "analipsis/catalipsis"
is from my paper "Coordination" for the 2nd ed. of the Shopen volumes,
available on my website.)
hartmut at RUC.DK wrote:
>I'm trying to understand the question.
>If I remember the original discussion of gapping far back in the last century
>correctly, the possibilities in the directionality of gapping were strongly
>tied to SOV/SVO considerations; it was even discussed if one could decide the
>notorious issue whether German is ("underlying") SOV or SVO with reference to
>On the face of it, the idea that forward gapping is possible in SVO languages
>but not in SOV languages is plausible (statistically), and makes sense (which
>would provide an explanation).
>But Basque and Yoruba seem to be counterexamples.
>Basque is SOV and allows for forward gapping, Yoruba is SVO and does not allow
>Hence, a different hypothesis is needed. Here comes the new idea: forward
>gapping is only allowed in languages that have a "unified" coordinator.
>If a strong correlation can be shown, this would be extremely interesting. Among
>other reasons, because it would release os from the necessity to classify all
>languages into permutations of S, V, and O. (My favorite example is Modern
>Greek: usually considered SVO, while the statistically most frequentt word
>order is V(S)O. - (S) because of pro-drop. - But Greek has a unified
>coordinator kai, so it ought allow for forward gapping.)
>But a statistical correlation and an explanation are two quite different things.
>Can we say that the new hypothesis makes sense? And in which way?
>It could actually be that unified coordination and forward gapping are only
>indirectly related, because their cooccurrence is dependent on a third factor,
>not discussed so far.
>I would therefore be cautious with formulations like "therefore forward gapping
>is disallowed", if "therefore" is meant to refer to some explanatory link
>("accordingly" might be safer to write).
>Citat "Arnold M. Zwicky" <zwicky at CSLI.STANFORD.EDU>:
>>Begin forwarded message:
>>>From: ana carrera <gappingucl at yahoo.co.uk>
>>>Date: September 22, 2005 2:41:28 PM PDT
>>>To: "Arnold M. Zwicky" <zwicky at csli.stanford.edu>
>>>Subject: Re: coordination
>>>A few days ago prof. Zwicky sent a query on my behalf. From the
>>>responses which I have received it is possible that the query was
>>>ambiguous and therefore I would like to repost my query and I hope
>>>I have expressed myself more clearly this time.
>>>I am dealing with forward gapping in SOV, VSO and SOV languages as
>>>part of my dissertation. By forward gapping I understand the
>>>deletion of the verb in the second conjunct:
>>>(1) I study gapping and you VP-ellipsis.
>>>My research so far predicts that forward gapping is only possible
>>>in those languages which use one and the same coordinator for all
>>>the categories they conjoin, independently of whether they have the
>>>unmarked order SVO, SOV or VSO. For instance, in English (SVO) the
>>>coordinator "and" is used to conjoin NPs, Clauses, PPs, etc and
>>>forward gapping is possible. In Basque (SOV) the coordinator is
>>>"eta" both for NPs and clauses and therefore forward gapping is
>>>On the contrary, Yoruba (SVO) has a different coordinator for NPs
>>>(àti) and for clauses (sì) and therefore forward gapping is
>>>disallowed. The same is true in Japanese or Korean (SOV) which also
>>>show different coordinators for NPs and clauses and thus disallow
>>>My hypothesis is that coordinators which select for what they
>>>conjoin block the possibility of forward gapping.
>>>I would be very grateful if you could provide me with more data
>>>which confirm or disconfirm my results so far.
Martin Haspelmath (haspelmath at eva.mpg.de)
Max-Planck-Institut fuer evolutionaere Anthropologie, Deutscher Platz 6
Tel. (MPI) +49-341-3550 307, (priv.) +49-341-980 1616
More information about the Lingtyp