| EDITORS: Baerman, Matthew; Corbett, Greville G.; Brown, Dunstan; Hippisley, Andrew
TITLE: Deponency and Morphological Mismatches
SERIES: Proceedings of the British Academy, Volume 145
PUBLISHER: Oxford University Press
Michael Maxwell, Center for Advanced Study of Language, University of Maryland
Deponency is a term that has traditionally been used for an oddity in the
paradigm of Latin verbs, in which a subset of verbs lack forms bearing active
voice inflection, but use what appear to be passive voice affixes (for that is
their function on most other verbs) as if they were active. Extending this idea
beyond Latin (and what may be a similar situation in classical Greek), one might
consider ''extended deponency'' to involve an arbitrary subset of words for which
some set of affixes gets used in the ''wrong'' part of the paradigm (and not used
in the ''right'' part of the paradigm). That is, deponency in this extended sense
need not necessarily involve active and passive voice.
The questions addressed by the papers in this book include whether deponency (in
this extended sense) is found in other languages, and what this can tell us
about inflectional morphology.
Before going further, I want to address the question of who should be interested
in this book. One might think that this topic would be of interest to a narrow
readership, perhaps only to those who specialize in inflectional morphology. One
would be mistaken; for as in so many things in linguistics, pull on one piece of
string, and you will find that it is connected to the rest of this tangled knot
we call language. There is much of general interest, and I will highlight some
of these connections in the individual papers.
For each paper, I give a brief summary followed by specific comments on the
strength of the arguments and on connections to the larger field of linguistics.
Finally, in the Evaluation section, I make general comments about the book as a
In the introduction, the editors briefly summarize the problem, and give one- or
two- sentence summaries of each paper, which were originally given at a workshop
organized by the Surrey Morphology Group.
While Greville Corbett's paper (''Deponency, Syncretism, and What Lies Between'')
is not the first in this volume, it centers on the question of what ''canonical''
deponency is, a notion that is referred to in many of the other papers, and in
that sense is arguably the introductory paper. Corbett defines 'canonical
deponency' , contrasting this with the notion of syncretism, the situation where
one cell of a paradigm appears to ''use'' the form belonging to another cell. The
primary distinction is that in syncretism, the form is used in both the
''correct'' cell and the ''wrong'' (syncretic) cell; whereas in deponency, the form
is used only in the ''wrong'' cell (or more commonly, a set of ''wrong'' cells,
referred to informally as a ''slab'' of the paradigm), leaving a gap in the
paradigm where the forms would normally be used.
Corbett lists a number of other distinctions between syncretism and deponency,
but then finds that other languages exhibit morphological phenomena that
straddle the boundary between syncretism and deponency. In this way, Corbett
argues, deponency in Latin is more canonical than the similar phenomena in the
other languages, and certain instances of syncretism are more canonical than others.
While this is no doubt logical--the canonical instances of deponency are just
those instances where the term was originally defined--it seems dangerously
close to tautology. That is, it is surely an accident of history that some of
the earliest grammars were of Latin, and that deponency is a feature of Latin
verbal morphology (and of classical Greek morphology, where we again have very
early grammars). But suppose the earliest grammars had been of some other
language with a complex morphology, and some other phenomenon had been described
and labeled there--some phenomenon with a different cluster of properties from
those which define ''canonical deponency.'' Would we then see Latin deponent verbs
as being on the border between canonical instances of this other phenomenon, and
canonical instances of syncretism? What, if anything, makes deponency special,
so that ''canonical deponency'' should be relevant to linguistics?
Zoologists are in a somewhat similar situation: canonical mammals walk on four
legs, have a tail, and feed their young with milk; bats, whales and the platypus
are somehow less canonical. But for mammals, we have some justification for the
notion of a canonical mammal, since it seems likely that early mammals shared
these characteristics, while bats and whales branched off (and the platypus is a
special case). Furthermore, most branches of mammals fit the canonical picture.
So zoology has both phylogeny and numbers to justify the notion of ''canonical.''
Linguistics, on the other hand, has neither: it is not clear that languages
which exhibit deponency do so through any kind of common descent (or even
parallel evolution), nor are canonical instances of deponency necessarily more
common than the other phenomena Corbett discusses. Rather, ''canonical deponency''
runs the risk of being a category like ''animals with wings'', which includes
birds, bats, many insects, and (less canonically) flying squirrels and flying fish.
I dwell on this issue of whether canonicity is important because it is a
question that appears explicitly or implicitly in many of the papers in this
volume. For example, in the paper ''Morphological Typology of Deponency'', Matthew
Baerman lists a number of properties of Latin deponent verbs, then turns his
attention to other languages where some words exhibits a subset of these
properties. The outcome of Baerman's analysis is that each of the properties of
deponent verbs seems to be independent, so that deponency becomes a name for an
arbitrary cluster of properties.
Andrew Spencer (''Extending Deponency: Implications for Morphological
Mismatches'') examines morphological mismatches involving a subset of words in
one category, which take the inflection of another category--a subset of nouns
which inflect as if they were adjectives, for example, while still behaving in
other ways as if they were nouns. Spencer erects a typology of morphological
mismatches--the third such typology in the first three papers. Like Baerman's
typology, Spencer's is based on a number of individual properties. One might
hope that by grounding the typology in well-chosen features, such a typology
would distinguish between the possible and the impossible, and the typology
would thus turn out to be a theory. As it turns out, very little is ruled
out--apart from two of Spencer's combinations which are logically incompatible,
nearly every combination seems to be instantiated in some language, although
some combinations are less clearly attested than others. Spencer concludes that
the mismatches are the result of historical accident, and that ''the old labels
[including deponency] have outlived their usefulness.''
By this argument, then, 'deponency' may be a useful term for linguistic
discussions, but has no more (or less) claim to being theoretically important
than does any other combination of features.
I turn next to Gregory Stump's article, ''A non-canonical pattern of deponency.''
Stump points out that there are two ways that deponent verbs might work: the
first possibility is that the affixes such verbs choose might be out of synch
with their morphosyntactic features, while the second possibility is that the
semantic interpretation of such affixed verbs might be out of synch with their
morphosyntactic features. ''Canonical'' deponent verbs cannot help us distinguish
between these two possibilities, but Stump constructs an argument from a kind of
deponency in Sanskrit in favor of the second of these possibilities.
The argument requires close attention by the reader (particularly if, like me,
you don't happen to know Sanskrit), but seems convincing. Assuming Stump is
right about the behavior of deponent verbs, there are interesting implications
for linguistic theory in general. For example, the simplest theory would be that
morphosyntactic features (at least those of the sort involved in deponency) have
a direct interpretation in semantics. Person features, for example, would seem
to have a direct connection to meaning. Yet even here there are exceptions; the
use of third person singular indefinite ''one'' for first person plural in French,
for example. Stump's argument based on deponency seems to be another nail in
this simple theory's coffin. Perhaps it is time to question what morphosyntactic
features are, in the same way that phonologists such as Jeff Mielke (2008) have
recently questioned the universal nature and phonetic groundedness of
phonological features. If it turned out that morphosyntactic features are
neither universal or innate, then there is a significant learning problem; there
would also be implications for theories of the structuring of morphosyntactic
If Stump is right, another area which would merit re-investigation, is the
connection between inflectional morphology and syntax. In particular, if
deponent verbs with their passive morphology select direct objects in the same
way as active non-deponent verbs do, then there are implications for theories of
subcategorization. Stump does not discuss this for Sanskrit, but the
subcategorization properties of deponent verbs do come up in Lavidas and
Papangeli's article (see below). However, there are differences between
deponency in Sanskrit (Stump) and in Greek (Lavidas and Papangeli), so the usual
caveat that ''further investigation is necessary'' applies. The interpretation of
morphosyntactic features in syntax should be questioned--indeed, the term
''morphosyntactic feature'' may turn out to be a misnomer. This is not of course
the first indication that syntax does not interpret features expressed in the
morphology in any simple way. Other well-known cases include agreement (or the
lack of agreement) between a verb and a coordinate NP subject in English; the
use of agreement affixes only in the absence of explicit subjects (Irish,
McCloskey and Hale 1984; the Bora languages of Peru and Colombia, Walton and
Walton 1975, Thiesen 1996, Walton, Hensarling and Maxwell 2000, Thiesen and
Weber ms.); and first person plural verbal agreement with NP subjects (which
would appear to be third person) in Spanish (''Las mujeres somos...'' '(We) women
are...', cf. Corbett 2006: 132).
One may wonder where deponency comes from: why should such a non-iconic system
arise? While the ultimate origin of deponency in Latin and Classical Greek is
not known, clues might be found in the historical processes affecting deponency
in the descendent languages. Deponent verbs are not found in the Romance
languages, but Nikolaos Lavidas and Dimitra Papangeli show that far from
disappearing in Greek, deponent verbs are found at all later stages of the
language, down through the present. From their article, ''Deponency in the
Diachrony of Greek'', it appears that some verbs which were not deponent in
ancient Greek have become deponent, while some formerly deponent verbs have
become non-deponent--and in at least one case, a verb changes from deponent to
non-deponent (displaying both active and passive morphology), and then back
again to being deponent.
It is not however clear why this has happened, and one wonders whether the
traditional response of historical linguists looking for regularity where none
is apparent--that the irregularity comes from borrowing from unattested
dialects, or that certain forms are simply unattested in the corpora because
they are too rare--might not hold here. What is needed to rule out the latter
explanation (forms are unattested because they are rare) is a statistical
analysis. Doubtless some linguist somewhere is already planning such a study; it
should not be difficult, given the tools now available. (It is also possible
that deponency in Greek is an entirely different phenomenon than deponency in
Latin, perhaps one more tied to verbal semantics.)
While Lavidas and Papangeli's article is important for the data (it is replete
with tables of deponent verbs at various stages of Greek), it does suffer from
inconsistent terminology and unclear explanations. For example, the authors make
it clear that there were at least three kinds of deponency (or phenomena similar
to deponency) in Ancient Greek: deponent verbs with passive morphology; deponent
verbs with middle morphology; and verbs which were always active, never bearing
passive or middle morphology--in addition of course to verbs which could be
found in active, middle and passive forms. Unfortunately, Lavidas and Papangeli
later use the term 'active verb' with an indeterminate meaning; it could refer
to the class of verbs found only in the active voice, but it seems to be used to
refer instead (or perhaps also) to verbs which can be active or passive (there
was by the Hellenistic times no distinct middle morphology or meaning). The
discussion becomes particularly confusing in their summary of the data (p. 116),
where categories of change are mixed together in the text, only to be
distinguished in a table (table 12, p. 117).
In Evolutionary Phonology (Blevins 2004), one must explain the existence of
''unnatural'' but (relatively) common phonological patterns by explaining their
origin out of some combination of ''natural'' patterns and events, and their
persistence despite their unnaturalness. In morphology, while the origin of
deponency in Latin is obscure, one might ask the second question: why such an
apparently non-iconic pattern should have persisted for a long time in Latin.
The paper by Zheng Xu, Mark Aronoff, and Frank Anshen explains this persistence
by saying that Latin speakers in effect found a use for deponency: they made it
(partially) iconic. The presentation is based on Levin's verb classes, here used
for Latin rather than English.
However, the explanation seems to have little to do with those classes; rather,
one generalization (not perfect, to be sure) is that verbs whose objects are
physically affected are almost never deponent. (The converse of this, which is
the generalization that one would like to make, seems to be less true: not all
transitive verbs whose objects are not physically affected are deponent.)
Several other generalizations, such as the origin of some deponent verbs in
derivational morphology, take care of many of the remaining cases of deponent
Xu, Aronoff and Anshen conclude that many Latin deponent verbs are verbs which
were low on Hopper and Thompson's (1980) scale of transitivity; that is, that
deponency came to be used by Latin speakers as marking the class of verbs which
were low on this scale. This is an interesting conclusion, one which runs
counter to many of the papers in this volume, which assume that deponency (or
similar phenomena) are largely arbitrary. It would be good to see such a study
extended, by looking at how verbs entered or left the class of deponents through
the history of Latin; or perhaps whether, as Latin gradually turned into the
Romance languages (where deponent verbs are no longer found), the first verbs to
lose their deponent status were those which were higher on the transitivity
scale. Again, we now have the tools for such studies (although the early history
of Romance languages is regrettably under-attested).
Andrew Hippisley's paper, ''Declarative Deponency: A Network Morphology Account
of Morphological Mismatches'', describes a computational implementation of the
notions of canonical behavior that Corbett's paper describes.
I have mixed feelings about this article. On the one hand, I am all for
computational implementations of theories and language descriptions. In the
languages where I have had the opportunity to build and test computational
descriptions of syntax, morphology and phonology, I have found without exception
that existing descriptions of languages leave gaps, ambiguities, and
inaccuracies, which only become apparent when the grammar is implemented and
tested. This is true regardless of the linguistic theory employed in the
description, the depth of coverage, the number of linguists who have worked on
the language, etc. In sum, human minds are not as good at testing grammars as
computers are. On the other hand, Hippisley's formalism (Network Morphology) and
its concrete implementation as a DATR program will, I believe, border on the
undecipherable for most linguists. To take just one example, there is a
distinction among stems in angle brackets (such as <stem1>), stems in quoted
angle brackets (''<stem1>''), and stems in bracketed quoted brackets
(<''<stem1>''>). (Programmers used to programming languages like C may be reminded
of indirection and double indirection, a source of many software bugs.) I may be
wrong, but it seems that there must be clearer ways of representing inheritance
in linguistics; possibly there could be a linguistic notation layer on top of
DATR, where the linguistic notation would capture the facts in a way that
linguists are used to seeing, and the notation would then be translated into the
DATR notation. This might for example be based on the representation of multiple
inheritance used by Koenig (1999).
If I understand Hippisley's analysis of Latin deponent verbs correctly, it does
not appear to be capable of capturing the distinction made by Stump between form
deponency and property deponency. I may however be wrong about this, since DATR
seems to be capable of expressing a wide range of facts, and a DATR programmer
might come up with a way to do it. The distinction is at any rate not discussed
in Hippisley's paper.
As I suggested above, it is open to question whether the term 'deponency' has
any linguistic status, other than as a label for a language-particular
phenomenon in Latin and (maybe) Greek. In the article ''The Limits of Deponency:
A Chukotko-centric Perspective'', Jonathan David Bobaljik approaches this
question from the viewpoint of generative linguistics, and concludes that
deponency is an emergent phenomenon, resulting from a variety of unrelated
causes. The specific case concerns an anti-passive construction in Chukchi. (I
note in passing that once again, crucial data for the theory of linguistics
comes from an endangered language; Chukchi has ten or fifteen thousand
speakers.) In the true anti-passive, the subject (which would be marked in the
ergative case in an active sentence) is marked with the absolutive case, while
the object (which would be marked by the absolutive case) is marked in an
oblique case. The verb in an anti-passive clause shows agreement with just the
absolutive case NP. Thus, the anti- passive clause looks like an intransitive
clause, with respect to both case marking and verbal agreement. In the 'spurious
anti-passive' (SAP) construction, on the other hand, while the transitive verb's
agreement marking is like that of an intransitive verb--i.e. like an
anti-passive--the case marking on the subject and object follows the transitive
model, i.e. it is like an active. The SAP is thus the ergative language analog
of a deponent. However, the 'use' of the SAP construction in Chukchi differs
from the use of deponents in Latin: whereas in Latin, deponents are a lexical
matter (some verbs are deponent, others are not), in Chukchi the SAP is used to
avoid 'inverse' agreement, i.e. a verb whose subject is lower on the person
hierarchy scale than its object (for example, a third person subject acting on a
first person object). Bobaljik gives a theoretical account for this, and argues
that because of the difference in how the SAP arises in Chukchi, it cannot be at
heart the same thing as deponency.
It is striking that Xu, Aronoff and Anshen (in their article discussed above)
argue that Latin speakers were ''trying'' to find a meaning for deponency, a
meaning that centered on a notion of deponency being linked to verbs which were
low on the scale of canonical transitivity--and canonical transitivity involves
exactly the notion of an agent high on the person hierarchy acting on a patient
low on the person hierarchy. From this perspective, there is a remarkable
similarity between Latin deponents and the Chukchi SAP--almost an argument that
deponency is, if not a theoretical notion, at least a common response to a
certain kind of cognitive pressure.
This paper is slightly marred by missing or mis-identified citations. (p. 192:
Johns et al 2006 does not appear as such in the references, but can be
identified from the Bobaljik 2006 reference; Xu et al 2006 is listed as a
conference paper, but in fact it is in this very same volume.)
Jeff Good's ''Slouching Towards Deponency'' discusses a variety of phenomena in
Bantu languages which at first glance resemble deponency. (The title refers to
the historical changes that gave rise to these phenomena; this and the Xu,
Aronoff and Anshen article will thus be of interest to historical linguists.)
The problem, as Good is careful to point out, is that it is not always easy to
know whether a certain verb is actually deponent in these agglutinating
languages, because it is hard to know whether a particular verb bears a passive
suffix (or another valency changing suffix), or if a verb simply happens to end
(or begin) in the same sequence of phonemes that a passive suffix would contain.
In fact, it appears that in some languages, verb roots which historically bore a
chance resemblance to passive forms were re-analyzed by the language's speakers
as actually being passives--and since there was no active counterpart, these
verbs were then deponent.
Once again, the data comes from small languages; the weakness is that the
descriptions on which the analysis is based are not always as clear as one might
wish. In some cases, the only evidence that a verb is intransitive (and
therefore potentially deponent) comes from its gloss; in others, the lack of a
non-passive form has to be inferred from the absence of such a form in
dictionaries or word lists. This has implications for the documentation and
description of languages: for some purposes, it is not sufficient to collect and
gloss texts. Directed elicitation, based on theories or at least on specific
questions that need to be answered, will therefore be necessary.
However, in the next paper, ''Spanish Pseudoplurals: Phonological Cues in the
Acquisition of a Morphology Mismatch'', Ricardo Bermudez-Otero takes exactly the
opposite stance: individual linguists' intuitions cannot be trusted, and the
best data is large corpora. Fortunately for his study, Spanish corpora--unlike
Bantu corpora--are voluminous.
Bermudez-Otero describes noun plurals, which in Spanish usually consist of an -s
suffix; a few singular nouns also end in /s/, and the question is how native
speakers analyze such forms as stems which happen to end in /s/, or as stems
which take the plural -s suffix, but are treated as singulars--and are therefore
deponent, in the extended sense of this term. Bermudez-Otero argues that
speakers in fact do both, analyzing some words in one way, and some in another.
The data needed to motivate these claims is subtle, and Bermudez-Otero argues
that first language learners generally do not have access to the crucial forms
in their primary data; so the question is how they come to their analysis.
Bermudez-Otero argues that they use heuristics, based in part on avoiding
analyses of marginal cases which violate tendencies in clear cases. While this
superficially resembles explanations in Optimality Theory, it differs in that
the heuristics are applied in analyzing forms to be included in the lexicon,
i.e. during learning. This explanation is also, as Bermudez-Otero points out, in
contrast to the hypothesis of Xu, Aronoff and Anshen (in the article discussed
above) that speakers of Latin tried to find a grammatical or semantic ''reason''
for deponency, although the two hypotheses are not necessarily mutually exclusive.
Rather than appealing to heuristics of avoiding certain configurations, a
different explanation for how learners decide between analyzing singulars ending
in /s/ as being stems which happen to end in /s/, or deponent words, might be
that analogy plays a role: that is, the analysis of a particular word is modeled
on the single word which it most closely resembles. This is of course a very old
idea, but it is not obvious that it is wrong, and Spanish pseudo-plurals might
provide a useful test case.
A fair portion of this article is devoted to motivating the source of the data
used to support these claims. There has in fact been disagreement among
linguists studying Spanish morphology over what forms are acceptable (or
grammatical) in precisely the crucial cases. Previous linguists have largely
relied on their intuition to provide forms; Bermudez-Otero instead uses corpus
counts as evidence (readers will recognize that this is an old argument).
Bermudez-Otero makes several claims: first, the data on the crucial forms is
vanishingly small; only by the use of a large corpus--in particular, the web--is
it possible to amass sufficient data to demonstrate which forms are used.
Second, learners do not have access to sufficient forms to have actually learned
them, therefore they must be basing their usage on something else (such as the
heuristics he proposes).
I find this last point less than persuasive. It seems that many Spanish speakers
do not have any reliable intuitions about the forms in question; indeed, that is
why the previous studies by other linguists have reached differing conclusions:
the linguists either had a contrary intuition about the 'correct' forms, or had
no intuitions at all. The question, then, is why are the forms Bermudez-Otero
bases his analysis on dominant in the corpus? One possible answer is that the
forms are found in a particular sub-language, and that those speakers (or
writers!) who use them have picked them up only because they frequently read
texts in that sub-language. That is, they are familiar with--they have
learned—''virusito'' 'little virus' for the same reason that readers of LINGUIST
List are familiar with the name ''Helen Aristar-Dry''. (Judging by the examples in
this paper, 'virusito' is most common in the domain of on-line forums about
Nicholas Evans' paper is ''Pseudo-Argument Affixes in Iwaidja and Ilgar: A Case
of Deponent Subject and Object Agreement.'' These languages are, again,
endangered languages, this time of Northern Australia. The form which deponency
takes here is different; certain verbs take subject or object agreement affixes,
but there is no corresponding real-world argument. For some verbs, there is at
least the hint of such an argument, in a sort of metaphorical sense; for other
verbs, even this is absent. Evans draws a parallel with expletive subjects in
English, and with expletive objects in Australian English and Italian; in these
languages , the ''extra'' arguments are pronominal, not affixal, leading Evans to
propose a still more extended notion of deponency, extending into the syntactic
realm. This is perhaps related to yet another issue on the boundary between
morphology and syntax, that of paradigms which are partly analytic and partly
Finally, there is a summary article by P.H. Matthews, ''How Safe Are Our
Analyses?'' Rather than reading this as a summary, it might be better to read it
as an introduction, for its clear description of Latin deponency, and for the
enlightening parallel Matthews draws between deponency and phonological
In my introduction to this review, I suggested that this book should be of much
wider interest than its title might suggest. In the descriptions of individual
articles, I have mentioned some of the related linguistic issues that are
raised: the historical sources of deponency, ranging from the phonological
(Good, Evans) to the semantic (Xu, Aronoff and Anshen); the meaning of
morphosyntactic features (Stump, Matthews); language acquisition and
learnability (Bermudez-Otero); elicitation vs. introspection (Good,
Bermudez-Otero); and of course phonology, semantics, and syntax, not to mention
Unlike in many collections of conference papers, the authors of the individual
papers collected here have listened to each other, and at least some of the
papers have been revised to show these insights.
The book is indexed by subject, language and author. Surprisingly, the word
''deponent'' is not to be found in the subject index.
In sum, I found the book to be more thought provoking than I expected, and
Blevins, Juliette. 2004. _Evolutionary Phonology: The Emergence of Sound
Patterns_. Cambridge: Cambridge University Press.
Corbett, Greville G. 2006. _Agreement_. Cambridge Textbooks in Linguistics.
Cambridge: Cambridge University Press.
Hopper, Paul, and Sandra A. Thompson. 1980. ''Transitivity in grammar and
discourse.'' _Language_: 56:251-299.
Koenig, Jean-Pierre. 1999. _Lexical Relations. Stanford Monographs in
Linguistics_. Stanford: CSLI Publications.
McCloskey, James and Kenneth Hale. 1984. ''On the Syntax of Person-Number
Inflection in Modern Irish.'' _Natural Language and Linguistic Theory_ 1: 487-534.
Mielke, Jeff. 2008. _The Emergence of Distinctive Features_. Oxford: Oxford
Ross, John R. 1972. ''Double -ing.'' _Linguistic Inquiry_ 3:61-86.
Thiesen, Wesley. 1996. _Gramática del idioma bora_. _Serie Lingüística
Peruana_, 38. Pucallpa, Peru: Ministerio de Educación and Instituto Lingüístico
Thiesen, Wesley, and David Weber. ms. A Grammar of Bora.
Walton, James, Grace Hensarling and Michael Maxwell. 2000. ''El Muinane.'' In
María Stella González de Pérez and María Luisa Rodríguez de Montes (eds.),
_Lenguas indígenas de Colombia: una visión descriptive_. Pp. 255-73. Bogotá:
Instituto Caro y Cuervo.
Walton, James W. and Janice P. Walton. 1975. _Una gramática de la lengua
muinane_. Bogotá: Ministerio de Gobierno.
Yip, Moira. 1998. ''Identity Avoidance in Phonology and Morphology.'' In Steven
Lapointe, Diane Brentari, and Patrick Farrell, _Morphology and its relation to
phonology and syntax_. Pp. 216-246. Stanford, Calif.: CSLI Publications.
ABOUT THE REVIEWER
Dr. Maxwell is a researcher in computational morphology and other computational
technologies and resources for low density languages, at the Center for Advanced
Study of Language at the University of Maryland. He has also worked on
endangered languages of Ecuador and Colombia, with the Summer Institute of