The Morphological Limitations of Navajo Sibilant Harmony
Blair Jackson
1 Introduction
Coronal harmony is a type of consonant harmony found in most languages of the Na-
Dene family. Navajo, however, is unique among its relatives as this process of assimilation
does not affect the outer (leftmost) verb domain, also known as the disjunct. Thus far, there
has been no attempt to explain why Navajo consonant harmony is limited to the two
rightmost domains of conjunct and stem. Using this inquiry as a base, I would like to
explore how the phonological process of coronal harmony may or may not be
morphologically conditioned.
In Navajo, unlike in most languages with consonant harmony, the process of sibilant
harmony may not be conditioned solely by distance from the trigger. The evidence for this
comes from the fact that the trigger for consonant harmony can be found in the conjunct
verb domain, and harmony would still not spread to the disjunct domain to its left, despite
the two being adjacent. This might suggest to an observer that the phonological process is
interacting directly with verb morphology. However, conventionally, phonology cannot
seemorphology to such an extent as would be happening here, making this explanation
unlikely. I would like to suggest an alternative hypothesis based on conservatism regarding
phonology and morphology interfacing, while still explaining this quirk.
I will suggest that before the development of Proto-Athabaskan, what would
eventually become the disjunct and the conjunct of the verb were understood to be
different words. At a certain later point, partial morphological fusion occurred. However,
emphatic separation was retained between the disjunct and the conjunct resulting from
these being separate domains containing different information. As a result, in Navajo,
sibilant harmony never spread to the disjunct, because even though it had already
undergone morphological fusion, it was considered by speakers as separate enough from
the stem and conjunct to be exempt from the phonological process.
2 Background
The Navajo language is a member of the Southern Dene branch of the Na-Dene
language family, spoken primarily by the Navajo People located within the southwestern
United States. Most closely related to Navajo are the Southern Dene or Apachean languages
such as Chiricahua and Jicarilla Apache. The Na-Dene language family is known to be one of
the immediate sub-Arctic at the northernmost, all the way to just below the United States-
Mexico border at the southernmost. These languages are best known for sharing a similar
verb morphology, known for its complexity.
Any discussion relating to Navajo would be incomplete without a mention of the
documentation that much of the research and analysis of the language relies on. Young and
dictionary and grammar of the Navajo language is a massive work, meticulously
documenting the -and- It is so
dense that there is more than one separate work aiming to guide readers through solely the
process of using The Navajo Language.  its various editions,
has been an essential cornerstone in describing patterns within the language and its sheer
size has made Navajo one of the most well-documented languages in the world, providing a
wealth of data on the most widely-spoken Indigenous language in the United States.
2.1 The Navajo Verb
 our scope closer to the subject of this investigation: the Navajo verb.
Notoriously complex, Na-Dene verbs are tough to tackle and thus commonly modeled using
a slot-and-filler type template. Navajo is no exception to this trend. Simply put, figures
such as Table 1 below, created by Young and Morgan, demonstrate the rich verbal
placed, and describe the ordering of the morphemes that make up the verb. Note that not
all the time when analyzing a specific example, and in fact, the Navajo
verb is minimally disyllabic (Young and Morgan 1987).
three different sections known as the verb domains: namely the conjunct domain, disjunct
domain, and stem (Sapir and Hoijer 1967). The conjunct domain is located closest to the
stem, whereas the disjunct domain lies on the outer edge of the verb. The rightmost
morphemes, positions 7, 8, and 9 in the conjunct domain, and 10, the stem, are all
necessary for a minimal fully inflected verb (McDonough 2003). No positions in the
disjunct morpheme group are required.
Table 1: The Structure of the Navajo Verb According to Young and Morgan (1987)
0: Object of Postposition
1a: Null Postposition
1b: Postpositional: Adverbial
Thematic: Nominal
1c: Reflexive
1d: Reversionary
2: Iterative mode
3: Distributive Plural
4: Direct Object Pronominal
5. Subject Pronoun: Non-local
6a: Thematic/Adverbial elements
6b: Thematic/Adverbial elements
6c: Adverbial-Thematic
7: Modal Conjugation Marker/Peg
8: Subject Pronoun: Local
9: Classifier
Note: the slots in this model are conventionally numbered using Roman numerals, but considering these are
falling out of use with each passing generation, for ease of understanding, I have preserved the original value for
each slot but translated it into the corresponding Arabic numeral.
2.2 The Conjunct and Disjunct Domains
The conjunct domain contains the required information for the verb, as well as the
qualifier and agreement markers located further away from the stem. Not only are they
bound tightly to the stem, but they are also known to combine to the point where
portmanteau is realized in some cases (Rice 2000),
The disjunct domain is described by Young and Morgan as a group of morphemes
less tightly bound to the verb than those in the conjunct. Similarly, McDonough describes
the disjunct as a group of morphemes with clitic-like properties. The disjunct domain is
morpheme. Young, Morgan, and McDonough suggest that the conjunct attaches first to the
stem, and then the disjunct to the left edge of the resulting compound.
3 Consonant Harmony
Harmony, as a phonological process, is a type of long-distance assimilation. Where
harmony occurs, specific segments will be required to agree regarding a specific
phonological feature. Vowel harmony systems are more often attested than consonant
harmony systems and are relatively well-documented consequently. However, consonant
harmony is present throughout most of the Na-Dene family (McDonough 2003).
This 
to investigation by a number of linguists, including Rose and Walker. This long-distance
action can be illustrated with an example from Kikongo, a Bantu language, where nasal
assimilation occurs.
Example 1: An Instance of Consonant Harmony in Kikongo from Rose and Walker,
I hit
I washed
we planted
we ground
In Example 1 above, the consonant in the suffix agrees in nasality with a voiced nasal if one
appears earlier in the stem (but not outside of the stem).
Because of the curious way consonant harmony can operate at long distances, it is a
particularly interesting case in the field and there could be much to learn in the process of
3.1 Coronal harmony in Na-Dene languages
Returning to the realm of the Na-Dene language family, we focus on a specific type of
consonant harmony: coronal harmony. Involving consonants with a coronal place of
articulation, coronal harmony is the most well-documented variation of consonant
harmony.  typically includes consonants with a dental, alveolar, palato-
alveolar, and retroflex place of articulation. In fact, the phonemic inventories of Na-Dene
languages are known to be coronal-heavy. Below are some examples of coronal harmony in
Tahltan, a Na-Dene language spoken in Northern British Columbia, given in Margaret
Example 2: Examples of Tahltan Coronal Harmony from Hartwick 1984
nedenesdit󰕏 󰕏dit󰕏
The above example shows the process of the first-person singular subject pronoun,
/s/. This pronoun typically surfaces as [s] with a range of stem consonants, but when
followed by an instance of /󰕏/ or // in the stem, it will be realized as [󰕏] or [],
respectively. In the case of the first two examples, nedene󰕏dit󰕏 and di󰕏t󰕏a, the
phonological feature in the stem that seeks to agree is [󰡹󰡃󰢔󰢀󰕎󰠼󰢈󰕎]. The harmony process
thus spreads from a trigger in the stem to the participating segment or segments in the
leftward direction.
3.2 Sibilant harmony in Navajo
The coronal harmony process found in other Na-Dene languages is indeed also
found in Navajo. In the Navajo-specific consonant harmony phenomenon, sibilants are
required to agree in [󰡹󰡃󰢔󰢀󰕎󰠼󰢈󰕎]. This results in an affected word having either alveolar or
postalveolar sibilants but not both. Below is an overview of the sibilants in the Navajo
consonant inventory, organized by whether they exhibit [󰡹󰡃󰢔󰢀󰕎󰠼󰢈󰕎].
Table 2: the Navajo Sibilants
/s/ /z/ /
s/ /
/󰕏/ /󰕖/ /
󰕏/ /
󰕏󰤹/ /
Example 3: Examples of Navajo Sibilant Harmony from Sapir and Hoijer 1967
a. si-󰍧
b. 󰕏i-󰋦 󰕖aa
he (4
Navajo sibilant harmony ensures that if the process of morpheme concatenation
creates a disagreement in the [󰡹󰡃󰢔󰢀󰕎󰠼󰢈󰕎] feature, sibilants within the prefixes assimilate to
the anteriority of the sibilant in the root. Therefore, like the process in Tahltan shown
previously, harmony spreads right-to-left from its trigger, which is always located in the
stem or the conjunct domain. This process is considered mandatory in many environments
(McDonough 2003). However, it also seems to weaken with distance (Berkson 2010).
Above are examples of Navajo sibilant harmony from Sapir and Hoijer 1967.
Example (3.1) involves the stative perfective si-. In 3.1a, it is realized faithfully since
no postalveolar sibilant is present in the stem. However, in 3.1b, this same stative
perfective prefix is realized as 󰕏i-, assimilating to the anteriority of the 󰋦 󰕖 within the root.
Example (3.2) is one with more intervening segments, yet harmony still occurs. The
underlying [-󰡹󰡃󰢔󰢀󰕎󰠼󰢈󰕎] sibilant in
sii- assimilates to [+󰡹󰡃󰢔󰢀󰕎󰠼󰢈󰕎], triggered by the
[+󰡹󰡃󰢔󰢀󰕎󰠼󰢈󰕎] sibilant in the stem.
solely triggered by sibilants in the stem. The process can also
originate from somewhere within the conjunct domain. However, it has never been shown
to affect the disjunct domain and is thus believed to be limited to the rightmost groups of
stem and conjunct only. Below is an example taken from Oberly 2014, showing a verb with
morphemes inhabiting the disjunct domain. Oberly separates the disjunct with a # for ease
of understanding the domain separations.
Example 4: Disjunct Domain Unaffected by Sibilant Harmony
t󰕏#ni󰕏- t󰕏#nis-
Even though the process occurs elsewhere in the verb, triggered in the stem with ts
and then spreading into the underlying ni󰕏 in the conjunct, it notably does not need to
spread to the disjunct and would be considered incorrect if it did.
Berkson 2013 indicates that consonant harmony in Navajo is, of course, a non-local
process, distance does play a significant role in whether or not it is brought about.
Puzzlingly, the sibilant harmony phenomenon can not only be triggered in the stem but
also in the conjunct. Even when this is the case, however, the process will never occur in
the disjunct domain, despite it being even closer to the trigger in this instance. This
suggests that the reason for the inability of sibilant harmony to reach the disjunct may not
be simply distance-related, and could, in fact, be evidence that some other process is to
4 Discussion
This behavior, the spreading of sibilant harmony assimilation everywhere in the
verb except for the disjunct domain, raises some questions. How does this phonological
morphological realm? Are the phonological and morphological realms separate at all in this
I initially sought to attribute this behavior to some kind of interaction or bleed-
through of the phonological and morphological realms of the Navajo verb, but I am not an
Athabaskanist, so seeking to do too much innovation is not my aim at this time. However, a
typical informal rule of thumb for the whole field is the expression, 
 This is an oversimplification aimed at someone with an underdeveloped
understanding of the field, but it spurred me to start considering explanations outside of
this area, to break out of the thought cycle, and ironically, to be more conventional about
the hypothesis I offer.
Lately, I have been reading research on Proto-Indo-European. It is there that I
encountered the phenomenon within documented developments of Indo-European
branches wherein new nominal case forms are sometimes created through the
univerbation of adpositional phrases (Milizia 2020).
4.1 Univerbation
"Today's morphology is yesterday's syntax." Talmy Givón on univerbation
Univerbation is a process whereby an expression containing a specific group of
morphemes thought once to be separate words becomes a single word over time (Brinton
and Traugott 2005). It is known to have occurred in numerous languages and has been
documented in many of these. A well-known example is from French, where par ce que (lit.
) became parce que . In some Indo-European languages, pronouns or
auxiliaries following the verb eventually synthesized with the verb itself. Below is an
example from Lithuanian, courtesy of Milizia 2020.
Example 5: Univerbation in Lithuanian
In this example, the 1SG pronoun 
form a single word.
4.2 Proposal
What if a process during the development of Proto-Athabaskan related to
univerbation is behind the lack of sibilant harmony in the disjunct domain? There is
already plenty of evidence pointing to the disjunct domain being less tightly bound to the
verb than the conjunct (Young and Morgan 1987). Furthermore, the disjunct morphemes
are less likely to undergo any phonological alternations than the conjunct morphemes,
according to Oberly 2014. This could suggest that these morphemes were not always
bound to the verb structure and could in fact have been considered different words at a
time. Unfortunately, documentation from the time period where this hypothetically
happened does not exist so there is no available proof of this theory.
At a certain point, partial morphological fusion related to our understanding of
univerbation occurred. However, emphatic separation was retained between the left edge
of the conjunct and the newly added disjunct, as a relic of its past as a totally separate
word. As a result, Navajo sibilant harmony never spread to the disjunct, because although it
had already fused with the rest of the verb, it was seen as separate enough from the stem
and conjunct to be exempt from the phonological process, possibly as a way of marking it.
4.3 Further Questions
The caveat of such a hypothesis rooted in historical linguistics, is, of course, how to
support it beyond speculation. The ideas I presented herein certainly spawn further
questions, the most glaring in my opinion is, 
rightmost domains in Navajo and not in other Na-Dene languages exhibiting the same
 Thinking of possible explanations only immediately served me another
happens because variation happens. Another question regarding coronal harmony in
Navajo has already been asked: why would a language with an already slim number of
places of articulation and where much of the consonant inventory is coronal consonants,
further reduce its phonemic inventory by having a process that disallows contrasts among
the coronal consonants? I encountered a question myself while flipping through various
examples of sibilant harmony in my sources. It seemed that there are examples of verb
constructions that would undergo sibilant harmony but where apparently either
instancea version with harmony taking place or one with no harmonycould be
preferred by speakers. This paper relies on the context that sibilant harmony should
happen when there is a trigger. The presence of examples where the phenomenon is
optional  and is an
encouragement to be innovative in the systems we use to describe and analyze both this
specific instance and phonology at a distance in general. Berkson 2013 undertakes
something related to the optionality of Navajo Sibilant harmony.
To end on a more radical note, I will bring back my invocation of Talmy Givón.
"Today's morphology is yesterday's syntax." Some researchers, like Oberly, suggest that
syntax, morphology, and phonology should be taken together as components of the
grammar of languages like Navajo. This is certainly an exciting proposition.
Berkson, Kelly Harper. 2013. Optionality and Locality: Evidence from Navajo Sibilant
Harmony. Laboratory Phonology, vol. 4, no. 2.
Brinton, Laurel J., & Elizabeth Closs Traugott. 2005. Lexicalization and Language Change.
Cambridge: Cambridge University Press, p. 48.
McDonough, Joyce. 2003. The Navajo Sound System. Dordrecht, Kluwer.
Milizia, Paolo. 2020. Morphology in Indo-European Languages. Oxford Research
Encyclopedia of Linguistics.
Oberly, Stacey. 2008. An Optimality-Theoretic Analysis of Navajo Sibilant Harmony. Coyote
Papers, vol. 16.
Rice, Keren. 2000. Morpheme order and semantic scope. Cambridge UK: CUP.
Rose & Walker. 2004. A Typology of Consonant Agreement as Correspondence. Language,
vol. 80, no. 3, 475531.
Sapir, Edward and Hoijer, Harry. 1967. The Phonology and Morphology of the Navaho
Language. (University of California Publications in Linguistics, 50.) Berkeley:
University of California Press.
Young & Morgan. 1987. The Navajo Language. Albuquerque, University of New Mexico