Basque butterflies again (again)

[Larry Trask]

Roz Frank writes:

[on my objections to Lloyd Anderson's attempts at seeing certain Basque
words for 'butterfly' as ancient]

>  Well I guess that one can draw quite different inferences/conclusions from
>  the same data. The point I tried to make throughout my previous discussion
>  was precisely the opposite of "the lesson" that Larry drew from it, namely,
>  that the referentiality of the items discussed was, indeed, motivated.

>  Hence, if <bitxi/pitxi> has a clear etymology, it does not follow that to
>  use the same term in a compound to refer to a butterfly, a daisy and a
>  poppy would demonstrate "multiple unrelated senses" for an individual
>  formation, quite the opposite, for the three objects would be projected as
>  analogically similar; metaphorically the same, if you wish. To refer to a
>  colorful flower fluttering in the wind and a colorful butterfly with the
>  same term isn't any less motivated, in my opinion, than referring to the
>  front end of a rocket as a 'nose-cone', or to a kite as a 'cerf-volant'.
>  It's simply a demonstration of the capacity homo sapiens sapiens have for
>  analogical thinking (cf. Lakoff, Turner, Johnson, et. al.)

Oh, I agree that this *might* be so in particular cases.  But it does not
appear to be generally true in Basque.  A particular expressive formation
sometimes has such a diverse array of senses that a common semantic thread
is, at best, extremely elusive.

>  What is less clear, however, is how one ought to go about explaining the
>  etymology of the second element *<-leta>, assuming, of course, that it is,
>  indeed, derived from what was once a meaningful suffixing element in the
>  language, perhaps a compound one (as opposed to being merely an
>  "expressive" ending).

If you look at the full panoply of formations, I think you'll find that
all those final elements like <-leta>, <-lota>, <-papa>, <-dola>, and so
on are one-offs.  They do not recur in multiple formations, and hence they
are best regarded as meaningless (non-morphological) elements selected
purely for their agreeable sound.

>  To examine the question in depth, one would need a
>  listing of all words in Basque ending in *<-leta> and then, after examining
>  them, attempt to see whether any sort of a pattern of meaning could be
>  detected, particularly if one were to view *<-leta> as a compound suffix.

Are there any others?  I can't think of any.  I exclude, of course, cases
like the surname <Zabaleta>, in which the /l/ is clearly part of the stem
(here <zabal> 'wide'), and the final part is merely the familiar suffix
<-eta>.

>  Today there is no evidence in Basque for a productive suffix in *<-leta>,
>  as Larry has rightfully pointed out. However, if *<-leta> is viewed as a
>  compound suffix in *<-le-eta> things begin to look rather different. This
>  approach to the data would posit *<-le-eta> as a compound suffix that was
>  once productive in the language but no longer is and, hence, it is
>  encountered only as a fossilized suffix in compounds such as *<pitxileta>.

"Compounds such as"?  Can you think of any others?

>  In favor of this thesis/hypothesis one could muster the following facts.
>  First, it should be noted that <-eta> itself is not in any way an uncommon
>  suffix in Basque where it confers the notion of a "collective" or "abstract
>  extension" to the root-stem (e.g., <lapur> 'thief, to thieve' becomes
>  <lapurreta> 'theft; the act of thieving' [and, yes, <-eta> has a variant in
>  <(k)-eta>). It shows up in compounds that are a bit harder to translate
>  into English, e.g., <gogoeta> (sg.) '(processes involved in) thought,
>  thinking, desiring, remembering', from <gogo> 'memory, desire,
>  consciousness, thought'. In <gogoeta> the notion <gogo> is conceptualized
>  in terms of an "abstract extension" of <gogo>, i.e., an abstraction or
>  concept derived from the meaning of the root-stem.  At other times <-eta>
>  appears to refer to the place where X or an abundance of X is found,
>  <elorrieta> 'a place characterized by hawthorns, a hawthorn grove'.

Agreed.  But the evidence points pretty strongly to the conclusion that
the collective sense was the earliest sense in Basque.  And most specialists
believe, or suspect, that this <-eta> is borrowed from the Latin collective
suffix <-eta>, itself the direct source of the Spanish collective suffix
<-eda>, as in <castañeda> 'chestnut grove' and <alameda> 'poplar grove'.

>  Furthermore as Larry and others have pointed out, the same suffix of <-eta>
>  is used as the marker of grammatical plurality in the oblique cases.

An identical marker.  I myself believe it is the same suffix, and at least
some of my colleagues agree, though I don't know if all do.

>  Indeed, this along with other aspects of <-eta> suggest that it existed in
>  the language before the system acquired the concept of singular/plural
>  contrast which is now has.

This was my own suggestion a few years ago, though more specifically I
suggested that this <-eta> was used to construct plural forms for the
local cases, after a plural had been created otherwise for the grammatical
cases.

>  The evidence suggests that previously this
>  suffix had a slightly different function in the noun phrase (or lexical
>  chain) than it does today. More work needs to be done on Basque along the
>  lines of what Lucy (1992) did for Yucatec Mayan since in Basque the marking
>  for number  (as singular and plural) appears to be a relatively recent

I agree, though I don't know how many of my colleagues agree.

>  and
>  not fully consolidated phenomenon as demonstrated by certain aspects of the
>  morpho-syntactic structure of the language, e.g., <-eta> as a suffixing
>  element still crops up with its older meaning and it has even been
>  suggested that <eta> in its modern meaning of 'and' is etymologically
>  linked to the same entity.

Possibly, but I find this idea a bit hard to swallow.  Not impossible,
though.

>  For example, today it is not particularly
>  unusual to find a sentence in a novel or book of essays that begins with
>  <Mikel'eta> (or 'Mikel-eta') and this expression is understood to refer to
>  'Michael and (the rest)' or it might be glossed as 'the collection of
>  Michael'; as 'Michael in his extended form'. It's not all that easy to
>  render the Basque meaning into English. Stated differently, there is every
>  reason to believe that the suffix <-eta> shouldn't be considered the new
>  kid on the block, rather it would seem that it dates back to
>  morpho-syntactic structures found in Pre-Basque.

Not if it's borrowed from Latin, which it very likely is.  After all,
Basque has borrowed lots of word-forming suffixes from Latin and Romance.

>  And in the case of <-le>, it, too, is quite common in Basque being an
>  agentive suffix (does it have another name?),

'Agentive' is correct: it derives an agent noun from a verb-stem -- though
normally only from verb-stems of a certain class: those containing the
prefix *<e-> in their non-finite forms.

>  regularly used with verbal
>  stems to refer to 'actors', e.g., from the non-finite verbal form <i-kus-i
>  'to see', one constructs <ikus-le> 'spectator'; it can also be added to
>  non-verbal stems where the compound takes on the same meaning, i.e., of an
>  'agent' or 'actor', even when the compound refers to a non-animate entity.

Sorry; I don't follow.  I don't think <-le> is ever added to anything but
a verbal stem, except in a couple of ill-formed neologisms.

>  For instance, from the same root-stem, i.e., <bitz/pitz-> based in turn on
>  a palatalized form of <bizi>, we have <phiz-le> 'that which lights,
>  animates, illuminates, enlivens, brings to life; brings about conception'
>  (Azkue II, 174) where <phizle> demonstrates a totally normal compounding
>  process.

Well, no.  First, this isn't a compound, but a derivative.  Second, it is
far from being totally regular, since a <-tu> class verb like <piztu>
cannot normally take the suffix <-le>: it "should" take the other agent
suffix.

This <p(h)izle> is recorded by Azkue only for Lapurdian and High Navarrese,
where it competes with the regular derivative <p(h)izta(i)le>.  I don't
have a date of first attestation, so I don't know how old it is, but
I suspect not very.

>  Also, it is clearly related to <bizi> 'to live; to be alive',

Well, no; I can't agree.  The word <bizi> is not a verb, but only a noun
meaning 'life' or an adjective meaning 'alive'.  You can only obtain a verb
from it by applying a suitable derivational process.  One derivative is the
compound intransitive verb <bizi izan> 'live', 'be alive', 'dwell', with the
auxiliary <izan> 'be'.  Another is the transitive verb <biztu> ~ <piztu>
'light, kindle, ignite', 'animate', with the verb-forming suffix <-tu>.

>  e.g., we have examples of <biztu> and even one document, Leizarraga's
>  translation of the New Testament, in which <viztu> appears (cf. Agud &
>  Tovar III, 147). Again, there is no reason to assume that <-le> is a recent
>  addition to the language.

Agreed.  It shows every sign of being ancient.

>  Compounds, such as <biztu> are of interest for another reason since they
>  show that non-finite verbs such as <bizi> can be utilized to form new verbs
>  by the addition of the verbalizing suffix <-tu>.

No; I'm sorry, but <bizi> is not a verb.  You can't add <-tu> to a verb
to obtain another verb.  You can add <-tu> to almost anything else -- noun,
adjective, adverb -- but not to a verb.

>  In the case of <biztu>,
>  the final /i/ is lost in the compound.

Derivative.  And, yes, this loss of /i/ is regular.

>  And as we have seen, the palatalized
>  form of <bizi> went on to become a free-standing form, i.e., <bitxi/pitxi>,
>  at least that is a relatively standard interpretation of events.

Yes, but <bizi> was a free form to begin with.

>  That means
>  a non-finite verb in <i> produced a free-standing stem.

Nope.  Sorry; I can't agree.

>  I mention this
>  since Larry has argued that this never happens,

This is not quite what I said.  What I said was that a verbal root --
by which I mean a root that takes the prefix *<e-> in non-finite verb-forms
-- never appears bare (unprefixed) in any derivative.  The *stem* (prefix
*<e-> plus verbal root) can occur as the first element in word-formation,
though not as the last.

>  i.e., with reference to
>  whether the stem <bil-> in <bildu> could be related to the verbal radical
>  or stem <-bil-> in <ibili>. However, I must say that I agree with Larry in
>  that (at least today) non-finite verbal stems (such as <bizi, ibili>) do
>  not tend to produce free-standing root-stems nor, for that matter,
>  non-finite verbs in <-tu>.

Indeed.  No verbal stem ever takes <-tu>, nor is there any evidence that
such a process has ever been possible in Basque.  I except here the process
of borrowing Romance verbs by replacing the Romance infinitive ending with
<-tu>, as in <erreibindikatu> 'claim', from Spanish <reivindicar>.

>  When speaking of the way that verbs can be constructed in Basque using
>  <-tu>, the following is one of the more curious examples of Basque's
>  morpho-syntactic ingenuity. The verb is <zendu> which Mikel Morris
>  translates in his _Euskera/Ingelesa/Englis/Basque Dictionary_ (1998) as 'to
>  pass away, to give up the ghost; to disappear.' If one were to try to
>  unravel the etymology of this word following the normal discovery
>  procedures one would fail miserably. I mean that the normal strategy
>  involves looking first at the other lexical items demonstrating what
>  appears to be the same or a highly similar root-stem, i.e., phonologically
>  similar items.

Er -- Roz, who says it is "normal" to look at phonologically similar but
semantically unrelated items?  Especially when the etymology of <zendu>
is transparent?

>  In this case, we would find dozens of examples of compounds
>  in <zen-> and it is well known that in the case of these other examples the
>  root-stem <zen> has a phonological variant in <zein> and that that variant
>  derives in turn from <zeren> 'of what (indeterminate)'. So one's first
>  inclination would be to assume that the etymology of <zendu> should be
>  traced back somehow to, say, <zenbat> 'how many'.

*Whose* "first inclination"?  ;-)

>  But that would be wrong
>  for <zendu> is a non-finite verb that has been constructed from a finite
>  verb form of the verb <izan> 'to be', concretely from the conjugated form
>  of the third person singular past tense <zen> 's/he/it was'.

>  Actually one might argue that <zendu> is based on a relative clause
>  's/he/it that was'. For instance, it is commonplace in Basque to speak with
>  respect of the deceased. So when talking about one's mother who is
>  deceased, one might say, <Ama zenak horrela egin zuen> '(My) deceased
>  mother did it this way [the way you/the interlocutor are doing it]' where
>  <zenak> is <zen + n (relative claus marker) + determiner + erg.> which
>  converts the relative clause into an ergative subject. Hence, a root-stem
>  of <zendu> derives from a relative clause that in turn is based on a third
>  person singular past tense of a verb. I must admit that the English
>  translation '(My) deceased mother' fails to capture the affectionate and
>  respectful tone of the Basque phrase.

Almost, but not quite.  The interesting verb <zendu> has the following origin.

In Basque, <zen> is a finite verb-form meaning 'was', a typically irregular
inflection of <izan> 'be'.  Like any finite form, this one can take the
relative suffix <-n>, producing here the regular relative form <zen> 'who
was', still normal in the language today.  But this relative form has become
specialized as a lexical adjective meaning 'late', as in <Mitxelena zena>
'the late Michelena'.  And it is the *adjective* <zen> which takes the
verb-forming suffix <-tu>, quite regularly, to yield the derived verb <zendu>
'pass away, die'.

Note also that both <zen> in the sense of 'late' and its derivative <zendu>
are recorded only from the 1850s.  They do not appear to be ancient.

The only parallel case I can think of is <den> ~ <dan> 'all, every', derived
from a relative form of <da> 'is'.  The original sense of <den> was therefore
'which is'.  But this is recorded no earlier than 1761, and it too is not
ancient.

>  Which other languages do this sort of thing? I know that in Slavic
>  languages there are some pretty nifty ways of creating verbal compounds in
>  noun phrases. But I don't know of any thing that would correspond very
>  closely to what happens in the Basque example. Any ideas?

This process is not productive in Basque, and the two examples just cited
are the only two I know of.

>  In conclusion, a much more rigorous analysis of the data concerning the
>  suffixing element *<-leta> would be needed before alleging that 1) it is a
>  compound suffix composed of *<-le-eta> and/or 2) that <-leta> in
>  <pitxeleta> (*<pitxi/bitxi-le-eta>) is actually derived from that suffix
>  and not from a totally unmotived expressive formation.

Not possible, I'm afraid.  The agent suffix <-le> is added to verbal stems,
and only to verbal stems.  But <bitxi> (and variants) is not a verbal stem.

>  However, given that
>  1) the old collective suffix in <-eta> gaves rise to the plural marker in
>  oblique cases in Basque

Possibly.

>  and 2) it is found as a semi-fossilezed form in
>  toponyms,

Indeed.  It is common in toponyms and in surnames, but it is unproductive
and rare to nonexistent in the ordinary lexicon.

>  it follows that previously formations in <-eta> were more common

Well, quite likely, though I'm nervous about that "it follows".

>  and that consequently if *<-le-eta> was once a producive suffixing element,

It couldn't have been.  First, <-le> is only ever attached to verbal stems.
Second, <-le> is never followed in any known case by another suffix --
though it can *follow* another suffix.  Third, Agent-Collective makes
no semantic sense in a word for 'butterfly'.  Fourth, there is no trace
in Basque of any such word-forming suffix as *<-le(e)ta>, in any function
at all.

>  a formation like *<pitxileta/bitxileta> could be considered to date back to
>  Pre-Basque. It's all in how one looks at the data.

No, sorry; it isn't.  We can look at the data any way we like, but the
results are always the same: <bitxileta> contains no recognizable suffixes,
it cannot contain agentive <-le>, and it cannot be ancient.

>  I would close by saying that Azkue lists:

>  <pitxeleta> 'butterfly';
>  <pitxilota> 'butterfly' (var. in <mitxilota>);
>  <pitxilote> 'daisy, poppy, butterfly'
>   <pitxoleta> 'daisy, poppy, butterfly'.

>  Was the protoypic form *<pitxileta>?  I don't really know. The argument
>  outlined above is simply one way of looking at the data. Indeed, until
>  additional examples in *<-leta>

*What* additional examples?  There don't appear to be any.

>  are subjected to rigorous analysis, the
>  case for *<le-eta> must remain a highly tentative one.

I'd say it can be rejected out of hand.  Sorry to be such an old grouch,
as usual, but the evidence is uniformly against any such analysis.

>  It could be that
>  *<pitxileta/bitxileta> dates back to an earlier stage and integrated what
>  was at that time a productive suffixing element made up of <-le> and
>  <-eta>.  Over time the compound suffix fell into disuse and ceased being
>  productive in the language. At that point the suffix's phonology would have
>  become unstable, as often happens when a once meaningful element in a
>  compound can no longer be disambiguated. But at the same time we need to
>  remember, as has been mentioned previous on this list, there is a
>  possibility that the Spanish suffixes in <-ota/-ote> may have played some
>  role here.

>  So it would appear that once again Larry and I will need to agree to
>  disagree, at least on some of these points.

OK.  I agree to disagree. ;-)

Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk