Adjacent nasals across word boundaries

KL note that a common context for the omission of orthographic “n” is when two nasals are adjacent. They also note that syllable-final nasals are not missing within a phonological phrase when the following syllable is vowel-initial. And in fact, this is a context in which they note that nasal intrusion also occurs in the written form (note, with monosyllabic words). The examples cited in KL are summarized below in 5):

They attribute this pattern to resyllabification in which the final nasal is carried over to the syllable initial position. The implied argument is that because resyllabification happens across word boundaries, the tlacuilo could be unclear as to which syllable the nasal belonged to. That is, in the first case, the sequence of two nasals in in nehuatl would potentially sound like a single sound in connected speech, and therefore be represented with a single orthographic “n”. By the same token, resyllabificaion of a word-final nasal to the onset of the following syllable would make it possible, in connected speech, to not be clear which syllable the final nasal belonged to. One can almost imagine the colonial-era tlacuilo saying “tlen” in isolation, and then the phrase “tle.no.tech.tla.o.co.li” (periods represent syllable boundaries). This could be the source of the intrusive “n” in words like tlennotechtlaocoli in (5b-c) above. This parallels the synchronic word cannot in English. Here, can and not are represented together as a single word in the orthography, including both ns, but is typically syllabified [kæ.nɑt]~[kɛ.nɑt].

The phonological patterns observed in ChN, seem to support this conclusion. In ChN, resyllabification occurs both word-internally and across word boundaries following the maximal onset principle (^{Kahn 1976}; ^{Selkirk 1982}; ^{Clements and Keyser 1983}). Cross-linguistically, there is a pattern of maximizing the number of consonants in the onset that is of course to the degree allowed by the phonotactics of the particular language. In the case of ChN, like many varieties of Nahuatl, syllables can maximally have a single consonant in the onset and coda, with the possibility of off-glides. The data below in 6) show syllabification across morpheme boundaries word-internally.

In example 6a) above, the [h] that forms part of the reduplicant template is syllabified as the onset of the vowel initial syllable of the stem. Similarly, in 6b) the final [ʃ] of the morpheme [alaʃoʃ], ‘orange’, forms the onset of the following syllable comprised of the morphemes making up the word [aː͡tɬ], ‘water’, in this compound form such that the final syllable is [ʃaː͡tɬ].

The maximal onset principle can also apply across word boundaries in ChN. However, prosody does condition when this occurs. Specifically, resyllabification can occur in connected speech, but will not occur at word boundaries that straddle a pause. This echoes Karttunen and Lockhart’s assertion that this occurs within a phonological phrase pauses occur at a phrasal boundary. In ChN, word-final nasals are realized as [h]. I will describe this pattern in much more detail in the following section, “Lenition of final nasal sounds in Nahuatl”. For now, we can look at the syllabification of a phrase naman ni tonatiuh “today/ hoy día”. The underlying form is /naman ni tonatiw/. In careful speech it is syllabified as in 7a), where the final nasal in /naman/ is realized as [h]. However, in normal to rapid speech the final nasal in naman is lost and the string of sounds is syllabified as in 7b).

Word-internally, sequences of nasal stops across morpheme boundaries are reduced to the onset of a syllable during syllabification rather than as a long or geminate nasal. The examples in 8) show this pattern.

This pattern in ChN, both word-internally and across word boundaries, accounts for the form cited in 5a) in the classical text. We can assume that such patterns were present in the Nahuatl spoken by the tlacuilo of the colonial manuscript. Patterns of resyllabification in ChN, can account for both omission and intrusion of orthographic “n” in these contexts. In the next section I will address the patterns in (1b-d) above and discuss the patterns for coda nasal sounds in ChN.

Lenition of final nasal sounds in Nahuatl

A wide range of diachronic and synchronic processes in which there is a weakening effect on a sound, such as debuccalization and devoicing have been described as lenition (^{Bauer 2008}; ^{O’Brien 2012}; ^{Kirchner 2013}; ^{Katz 2016}; inter alia). Certain types of lenition results in the loss of contrasts in perceptually weak positions (^{Katz 2016}). The omitted and intrusive nasals described by KL occur specifically at a perceptually weak position, the coda. By looking at modern Nahuatl, KL note that in some varieties, syllable-final nasals are weakened. They cite a process in which phrase-final nasals are devoiced in Veracruz Nahuatl (^{Wolgemuth 1969}), Puebla Nahuatl (^{Robinson 1969}) and Milpa Alta (^{Whorf 1946}). They go on to report that coda nasals following a long vowel are realized as nasalization of the vowel rather than as a nasal consonant for the dialects spoken in Veracruz, in which the vowel length distinction is maintained (^{KL 1976}, 381). Finally, they describe that both nasalization and voiceless nasals often disappear in rapid speech. They conclude that the weakening of underlying nasals in the phrase-final position could lead to the omission of the letter n in the manuscript because it is so weakened in the spoken form.

Expanding on their survey shows that lenition processes targeting final nasals seem to be a common pattern across the Nahuatl dialectological landscape. Typically, descriptions make reference to the word-level domain rather than the phrase referenced by KL in 4b) above. These patterns can be grouped into four general types:

Nasal absorption [Ṽ] (^{Trigo Ferre 1988}, ^{O’Brien 2012}) 9a) is a form of debuccalization in which oral articulations are lost, and nasality is realized on the vowel. This has been documented some varieties of Nahuatl, including ChN (^{Aguilar 2020}). Some varieties seem to have a velar nasal [ŋ]-realization of word-final nasals 9b). While in this case there is not a loss of oral articulation, this realization of final nasals can also be analyzed as the realization of a placeless nasal. Placeless nasals have been argued to surface as [ŋ] or an [ɴ]-like sound (^{Harris 1984}; ^{Trigo Ferre 1988}; ^{De Lacey 2002}). The realization of velar nasals has been shown to overlap acoustically and visually with vowel nasalization (^{Ohala 1975}; ^{Johnson, DiCanio and MacKenzie 2007}). This has been argued to account for excrescent nasal [ŋ] consonants developing from nasalized vowels such as in some varieties of Brazilian Portuguese where [ŋ] occurs where final nasalized vowels are expected (^{Shosted 2011}). Sound changes in either direction from [Ṽ] ⇆ [ŋ] can be attributed to shared acoustic properties and visual cues. This accounts for the types of sound correspondences [Ṽ] ⇆ [ŋ] in French varieties (^{Johnson, DiCanio and MacKenzie 2007}), as well as the nasal absorption [Ṽ] and velar nasals [ŋ] correspondence in contemporary Nahuatl.

Devoicing 9c) is reported utterance-finally in Naupan and Mecayapan. This process seems to be related to the final pattern identified in this survey, voiceless breathy debuccalization (9d), which can be seen as a further lenited voiceless nasal in 9c), diachronically [n̥]→[h]. This pattern is found in ChN (and will be demonstrated more fully below), and can be extrapolated as at least a possible surface form from examples given by ^{Pittman (1954}) for Tetelcingo.

Word-final nasal lenition seems to be widespread in Nahuatl phonologies. For example, ^{Boas and Beck ([1917] 2017)}, contrast Pochutec voiced word-final nasals and /l/ with voiceless counterparts in Nahuatl varieties from the Valley of Mexico. The effect of these processes is evident when surveying contemporary placenames of Nahuatl origin. The variation in the presence or absence of final nasals, being loanwords into Spanish not-withstanding, suggest that word-final nasals were quite variable across varieties when the place names’ orthographic representation were formalized in the colonial period. The locative suffixes “-tlan”, “-can” and “-pan” occur without final “n” in contemporary place names throughout Mexico and Central America.

These patterns for the realization of nasals across varieties provide insight into why the colonial tlacuilos might have omitted nasals in some contexts (e.g. ipilhua from expected ipilhuan, ^{KL 1976}, 380). An additional piece to this puzzle is the way in which nasal and laryngeal sounds are coextensive in ChN, providing further insights as to what might have been the source for the orthographic “n” in the colonial era texts. I will first describe the distribution of nasals in ChN and then discuss interactions with laryngeal sounds in the section titled “Intrusive nasals” in order to discuss the intrusive patterns identified by KL.

Chicontepec Nasal Sounds

In ChN, there are two contrastive nasals in the prevocalic/onset, position: /m/ and /n/. In the coda position, there are three patterns associated with position in the word and stress. The examples below in 11) show this contrast with minimal pairs.

Word-medial nasals undergo place assimilation to the following consonant in clusters as shown in 12).

In the word-final position, there are two realizations depending on stress. Since stress is predictably on the penultimate syllable, most word-final nasals follow this pattern: /n, m/ alternates with the glottal fricative [h]. The contrast with /h/ is neutralized. Thus, the stem cited in 12) above is [ikah] /i-kaN/ ‘her/his/its cheek’ in the possessed form. The following data show that both contrastive nasals alternate with [h] in word-final contexts.

Unpossessed nouns with the absolutive suffix 13a) and 13c) provide evidence that there is an underlying nasal that is present in the possessed construction. In this construction, however, the nasal sound is word-final and realized as [h]. In sum, word-medially, coda nasals, which necessarily occur in consonant clusters undergo place assimilation. Word-finally placeless nasals are realized as [h] when unstressed.

There are three exceptions to this pattern. The first one is a sequence of /n-m/where the first nasal does not assimilate: [inmoitah], ‘you all see each other.’ Second, stressed word-final nasals surface as [n]~[Ṽ]. This seems to be free variation between nasal absorption and a default alveolar nasal (e.g., ^{Harris 1984}). Stressed final nasals occur in monosyllables 14) and loanwords.

Spanish loanwords in ChN also demonstrate the effect of stress on the realization of word-final nasals. Words like asadón and violín have variable forms across speakers: [asˈaðoh]~[asaˈðon], [fiˈolih]~[βioˈlin], where the realization of the nasal is dependent on stress. The realization of nasals in ChN provides evidence for the pattern KL identified in 4b), phrase-final omission. It is likely that a phonological distribution of nasals similar to that of ChN was present in the Nahuatl spoken by the colonial tlacuilo. Nasal absorption ([Ṽ]) and breathy realization ([h]) with their loss of oral features, could certainly be a source of the omission of an orthographic “n”. This is especially likely in the case of [h] given that many colonial texts did not represent glottal sounds with any regularity. This realization of final nasals as [h] will also be relevant to the discussion of intrusive nasals in the “Intrusive nasal” section.

Omission before consonants

KL identified patterns in which nasal sounds are omitted before consonants: before nonnasal sonorants 4) and less frequently before consonants generally 4d). Some insight can be gained from the patterns found in ChN for nasals before /j/ and /w/. The other sonorant in ChN is /l/ is rare morpheme-initially such that no examples of /n-l/ or /m-l/ sequences are found in the data. There is considerable variation in how the nasal is realized in these contexts, however in both, there are examples in which the nasal sound is deleted. The data in 15) below show examples of morphologically complex words in ChN in which the nasal stop /n/ precedes the labiovelar glide in the underlying form.

The data in 15) show that /n/ in the context of /w/ has variable realization with a velar nasal, deletion, or nasal absorption as possible surface forms. In 15a) we see an n-final word with the glottal fricative realization. In 15b) the nasal sound is deleted in this same morpheme in the context of a /w/. In 15c) and 15d) we have a similar pair of morphologically related words, however in the context of a /w/ there is debuccalization with nasal absorption resulting in a nasalized vowel. In 15e)-15f), /m/ and /n/ surface as [ŋ]. The data in 15a)-15d) mirror the pattern in the colonial text suggesting that oral articulation of the /n/ was deleted for the colonial author as well. It would be easy to imagine the tlacuilo omitting an orthographic “n” if the oral articulation of the nasal sound were absent. ChN nasals in the context of /j/ exhibit similar variation: it surfaces as the palatal nasal [ɲ] or it is deleted. As was the case for sequences of /n-w/, there is variation across speakers in my field recordings. The two contexts represented in the data below are specifically the adverbial suffixes /-joh/, ‘covered in/full of’, and /-ja/, ‘already’. For some speakers, sequences of /n-j/ result in a surface form [ɲ]. The nasal takes on place features from the /j/ -16b) and 16f). For other speakers, the nasal sound is lost completely in this context -16c) and 16g). The diminutive form is provided in 16e) as a point of comparison. No examples of /m-j/ clusters are captured in my field recordings.

The patterns for the omission of orthographic nasals before non-nasal resonants 4c) in the colonial texts is not surprising given the variable realization of nasals in at least the context of /w/ and /j/ in modern varieties like ChN. A similar pattern has been documented in Mecayapan Nahuatl, where /n/ is reported to delete before /j/ and /w/ (^{Wolgemuth 1969}).

ChN, however, does not shed light on 4d), omission before consonants in general -in ChN, nasal sounds undergo place assimilation in consonant clusters. I turn to Karttunen and Lockhart’s account which points to Nahuatl varieties such as Mecayapan Nahuatl, where ^{Wolgemuth (1969}) describes coda nasals, which word medially necessarily occur in consonant clusters in Nahuatl (maximal onset principle), as reducing to nasalization on the vowel without oral articulation after long vowels (nasal absorption).

In 17) below, I summarize the insights from ChN and other Nahuatl varieties that can help account for patterns of nasal omission identified by KL (in 4) above).

Additional sources of breathy sounds

So far, I have established that /h/ and /n, m/ are contrastive sounds in ChN, but undergo neutralization to [h] in the word-final position, e.g., [nosih] /no-sin/ ‘my corn, mi maíz’. There is an additional neutralization pattern in ChN relevant here: /w/ is also realized as [h] in the word-final position. This is in fact part of a larger pattern in ChN where all non-nasal sonorants in undergo lenition in the coda position, summarized in 19).

In ChN, all non-nasal sonorants are voiceless and spirantized (and therefore breathier) in the coda position. /w/ is fully debuccalized. Discussed previously in the “Omitted nasals” section, the [j]~[ʃ] alternation in 19) is a long-standing feature of Nahuatl and is reflected in the orthographic conventions of colonial era texts. With respect to the other sonorants in colonial Nahuatl, the orthography does not tell us if there were the types of alternations seen in ChN. The graphemes “l” and “uh”/“u”/“hu” are not as revealing in colonial orthographies as the alternation between “y/i” and “x”. Alternations between voiced [l] and voiceless and/or spirantized [l̥]~[ɬ] have been documented in Hauhtla (^{Beller and Beller 1979}), Ixquihuacan (^{Sasaki 2014}), Matlapa (^{Croft 1951}), Mecayapan (^{Wolgemuth 1969}), Orizaba (^{Goller, Goller and Waterhouse 1974}), and Sierra Puebla (^{Robinson 1969}). Similarly, alternations between voiced [w] and voiceless and/or debuccalized [w̥]/[h] have been documented in Ixquihuacan (^{Sasaki 2014}), Mecayapan (^{Wolgemuth 1969}), Naupan (^{Brockway 1963}), Pomaro (^{Sischo 1979}), Sierra Nahuat (^{Key and Key 1953}), Sierra Puebla (^{Robinson 1969}), and Tuxpan (^{Cuauhtli 2022}). It is possible that patterns of lenition targeting coda /l/ and /w/ were also present in the colonial era, but were not represented in the orthography. The example in 20) shows this alternation for the labiovelar glide in ChN for the verb ‘to buy’.

These spirantized alternants found in ChN and other varieties are of interest here because of the laryngeal component. Both [ɬ] and [h] realizations of /l/ and /w/ involve a higher air turbulence than their voiced counterparts, and perceptually overlap with other sounds with distinct underlying forms. A ChN speaker who hears an [h] in the speech stream might map the surface form to underlying /h/, /n, m/, or /w/. The spirantized lateral [ɬ] also introduces its own ambiguity. Onset sonorants tend to devoice after an [h] in ChN: a sequence such as /hl/ is realized as [hl̥ ]~[hɬ] such as [ihɬia] /ihlia/ ‘to say’. In addition, sequences of /l͡tɬ/ are realized as [ɬː] e.g., [tʃiɬːakʷa͡tsih] /͡tʃil-͡tɬakʷa-͡tsin/, ‘food with a lot of chile’. Both patterns result in a voiceless lateral onset. It is not difficult to imagine possible reanalysis of any of these /l/ realizations encountered in the speech stream as corresponding to an /hl/ phonological representation. In fact, I have seen contemporary speakers write “hl” in informal writing (i.e. on social media) for an “l” in a coda position.

KL reference the generalized pattern of syllable-final sonorants weakening in Nahuatl, but contend that only the glottal stop provides the phonological motivation for hypercorrection with a nasal. However, it seems laryngeals broadly correspond to intrusive nasals. The patterns for coda sounds directly relate to the process for nasal intrusion as KL note that “intrusion of l and uh” (^{KL 1976}, 382) also occurs in the contexts for nasal intrusion. It seems clear then that it is the relationship between breathiness and the phonological alternations with nasal sounds, /l/, and /w/ that are the source for a number for the intrusive nasals. The fact that “uh/hu” and ‘l’ intrusion occurs in the same contexts as nasal intrusion in KL’s analysis points to phonological distributions like that of ChN where [h] from /n,m,w/ and [ɬ] from /l/ all share being breathy sounds. Let us now turn to a final piece to this complex laryngeal landscape is glottalization.

Prosodic glottalization in ChN

In many varieties of Nahuatl, there is typically one contrastive laryngeal sound. In the central varieties like Milpa Alta (^{Whorf, Campbell, and Karttunen 1993}), Nauhpan (^{Brockway 1963} and ¹⁹⁷⁹) and reconstructed for classical Nahuatl, it is glottal stop /ʔ/. In a great number of varieties -ChN, Río Balsas (^{Guion et al. 2010}), Huauhtla (^{Kimball 1990}), Hueyapan (^{Campbell 1976}), Ixquihuacan (^{Sasaki 2014}), Matlapa (^{Croft 1951}), Orizaba (^{Goller, Goller and Waterhouse 1974}; ^{Campbell 2011}), Pómaro, Michoacán (^{Sischo 1979}), San Agustín Buenaventura (^{Canger 2000}), Tetelcingo (^{Pittman 1954}, ^{Tuggy 1979}), Tuxpan (^{Valiñas 2013}), Zacapoaxtla (^{Key and Key 1953}, ^{Robinson 1969}), among others-, this sound corresponds to the glottal fricative /h/ such that there are correspondences across varieties between words like [paʔ͡tɬi] ~[pah͡tɬi], ‘medicine’. There are at least two varieties reported to have contrastive /h/ and /ʔ/ including Cuentepec (^{Velázquez Patiño 2014}) and Mecayapan (^{Wolgemuth 1969}, ²⁰⁰⁷). Even in varieties where there is no evidence that [ʔ] is a contrastive sound, there are a number of predictable prosodic contexts in which glottalization occurs. In ChN, glottal stop is not a contrastive segment. Nonetheless glottalization occurs at the prefix-stem boundary, stemstem boundaries, hiatus contexts and in pre-pausal contexts (^{Aguilar 2020}). This laryngeal articulation can range from realization as a canonical glottal stop to a period of aperiodicity (irregular pulsing) during phonation. This is expected given that glottalization has variable realization cross-linguistically (^{Pierrehumbert and Talkin 1992}; ^{Garellek, Ritchart and Kuang 2021}; ^{Garellek 2022}). For clarity, glottalization is transcribed here using [ʔ].

In ChN, prosodic glottalization occurs with variable frequency within the word. Pre-pausal glottalization is a more obligatory pattern. I will briefly demonstrate these distinct types of glottalization, beginning with glottalization sensitive to morphological structure. Glottalization at morphological junctures in ChN occurs at the prefix-stem boundary. It never occurs at the stem-suffix boundary -e.g., 21) below.

Similarly, glottalization occurs in hiatus contexts at morphological junctures 22). Note that glottalization occurs most frequently with homorganic vowels, but also occurs regularly in other hiatus contexts.

Glottalization also occurs in hiatus contexts across word boundaries 23).

As 21)-23) show, within a single utterance in ChN, there will be many possible laryngeal articulations, both contrastive (i.e. [h]) and prosodic (glottalization).

In ChN, glottalization also occurs at pauses in natural speech -pauses may align with prosodic boundaries. Utterance-medially, glottalization is typically realized as a canonical glottal stop. Utterance-finally, this pre-pausal glottalization is realized as a canonical glottal stop after vowel-final words with an audible release. The examples in 24) show that glottalization is not a lexical sound associated with the word [tsik͡tɬi], ‘gum’, or [kʷaː], ‘chew’, but rather an articulation associated with the end of the utterance.

When the utterance-final word is consonant-final, glottalization is realized as aperiodicity during phonation on the final vowel(s) -the context determines the realization as aperiodicity (consonant-final) or glottal stop (vowel-final). Vowel-final utterance-final context glottalization seems to be obligatory, while variable in its overt realization in the consonant-final utterance-final context (^{Aguilar 2023}). Huauhtla (^{Kimball 1990}), Hueyapan (^{Campbell 1976}), Ixquihuacan (^{Sasaki 2014}), Orizaba (^{Goller, Goller and Waterhouse 1974}), Sierra Nahuat (^{Key and Key 1953}), Sierra Puebla (^{Robinson 1969}), Tuxpan (^{Cuauhtli 2022}), and Matlapa (^{Croft 1951}) are all described as having both a contrastive [h] and some form of phrase/word-final glottalization. Descriptions generally focus on the vowel-final context which makes sense since glottalization is most salient there.

There are two other issues to note here regarding utterance-final glottalization in ChN. First, this glottalization seems to behave like a suprasegmental feature, moving leftward to the penultimate and even antepenultimate syllable in cases where the final syllable(s) is devoiced (^{Aguilar 2020}, ²⁰²³). Thus, glottalization, can occur in multiple places in an utterance-final word -glottalization at prosodic boundaries as well as from pre-pausal glottalization on the ultimate, penultimate and antepenultimate syllables.

A key context is the utterance-final position where both pre-pausal glottalization and glottal fricatives can be specified to cooccur. In ChN, an utterance like /ki-ita-h/, ‘they see it’ (quiittah), is produced as [kiʔita̰h], [kiʔitaː], [kiʔita̰], or [kiʔitah (versus /ki-ita/, ‘s/he/it sees it’ [kiʔitaʔ] with a canonical glottal stop). In addition, ChN data show that the hiatus resolving glottalization described above in 23) can also overwrite glottal fricatives utterance medially: [noˈtata ʔiˈmilah] /no-tatah# i-milah/ 1POS-father# 3POS-field ‘my father’s field’. The expected form [notatah] is produced as [notata] in the context of the glottalization. It is likely that some of the omitted nasals observed by KL reflect the interaction between lexical glottal fricatives and prosodic glottalization.

Laryngeals in ChN are complex and present an interesting piece to the puzzle of intrusive nasals. It seems likely that most varieties of Nahuatl exhibit some amount of laryngeal complexity beyond the presence of a contrastive laryngeal sound. Let us look at one example, Milpa Alta Nahuatl, which gives insight to the relationship between glottal stops and nasality. Whorf describes 1930s Milpa Alta Nahuatl of Mexico City as having only one laryngeal, transcribed as ’h, which mostly occurs in the coda position (^{Whorf, Campbell and Karttunen 1993}). Whorf describes a strong or aspirated release. He also describes the glottal sound as pre-aspirating the following consonant; i.e. [notaʔhtsin̥] ‘my father’ (p. 175). In a recent paper, ^{Hejna (2023}) showed that the fortis contrast in Welsh English was not only marked with preaspiration, but also preglottalization in some cases, as well as with both in other cases demonstrating that glottalization and aspiration are not necessarily exclusive. Perhaps what Whorf observed in Milpa Alta was inverse to Welsh English where the specification for a laryngeal sound could be realized as glottalization as well as glottalization with a period of aspiration, and perhaps solely aspiration (i.e. [h]) as in other varieties of Nahuatl. In both the Nahuatl and Welsh English cases, the contrast is maintained by something laryngeal. Whorf contrasts the ’h articulation with what seems like preglottalization of geminates; i.e. [iʔta] or [i^ʔtːa], ‘to see’ (p. 176) (cf. ChN [ita]). This point of comparison in Whorf’s description suggests that varieties with the phonemic /ʔ/ (rather than /h/ like ChN) most likely also have additional laryngeal articulations associated with prosody or the phonetic implementation of other contrasts such as the geminates above. Thus, forms like 18f), yntonca, where the first intrusive nasal corresponds to the site of prosodic glottalization found in ChN (at morpheme boundaries), might have been produced with a breathy release as described for Milpa Alta and be the source of the intrusive nasal.

ChN and other contemporary varieties of Nahuatl feature both glottalization and aspiration in some form or another. Let us now explore the relationship between laryngeal sounds and intrusive nasals.

Laryngeals and nasals: rhinoglottophilia

In the first part of this section, I demonstrated how the phonological alternations between a glottal fricative and nasal stops in ChN gives insight into one of the types of nasal intrusions. Here, I explore KL’s observations and discuss the relationship between laryngeal sounds and nasal sounds.

KL hypothesize that there were two types of nasality in Nahuatl: one corresponing to the contrastive sounds in the language, and the other a “constant suprasegmental feature of all vowels stemming from nondistinctive leakage of air through the velum” (^{KL 1976}, 382) -in other words, KL hypothesized that vowels were generally nasalized in Nahuatl. In their explanation, the fact that there were weakened syllable-final nasals, combined with a general nasal quality to segments not specified to be nasal lead to instability of nasals in Nahuatl, creating the context for the colonial tlacuilo to be unsure when to include an n in the written form. While this is a possibility, it alone cannot account for the fact that intrusive nasals, in their own description, seem systematic. Were this the only factor responsible for intrusive nasals, then one would expect to find them anywhere and everywhere there was a somewhat nasalized vowel. Their list of contexts for nasal intrusion is long, but still constrained. Instead of a suprasegmental velar leakage, I argue that the relationship between laryngeal sounds and nasality cross-linguistically, is a more likely explanation for intrusive nasals.

I begin with the phonologized pattern in ChN in which the phonetic [h] has multiple phonological representations (/h,n,m,w/) with an additional source of breathiness from [ɬ]. These patterns specifically point to the relationship between nasality and breathiness in the acoustic signal. The diachronic and synchronic relationship between nasality and breathiness has been called “rhinoglottophilia” (^{Matisoff 1975}; ^{Blevins and Garrett 1993}; ^{Igartua 2008}; inter alia), applied to cases in which nasal segments developed from aspirated ones, or conversely, where glottal fricative segments developed from nasal segments. Oft cited examples include British English [hãːvəd] ‘Harvard’ as a synchronic example, where nasalization is associated with the breathiness of [h]; and Hindi [sãp] ‘snake’ from Sanskrit sarpa as a diachronic example where nasality is associated with the breathiness of [s] (^{Ohala, 1980}). With respect to nasality and breathiness, their acoustic correlates are quite similar, despite arising from distinct articulations. Both, for example, have been shown to have a weaker first formant (lower amplitude) and therefore increased spectral tilt (H1*-H2*) relative to oral vowels/modal voicing: from of velopharyngeal coupling in the case of nasality (^{Chen 1997}; ^{Maeda 1993}; ^{Stevens 2000}; ^{Simpson 2012}), and a larger open quotient of vocal fold vibrations in breathy phonation (^{Bickley 1982}; ^{Klatt and Klatt 1990}; ^{Garellek and Keating 2011}; inter alia). In addition, both nasal and breathy vowels share an increased first formant bandwidth and a stronger first harmonic (^{Chen 1997}; ^{Simpson 2012}; ^{Styler 2015}), and are characterized by an increase in amplitude of lower frequency energy (^{Stevens 2000}; ^{Gordon and Ladefoged 2001}; ^{Garellek and Keating 2011}). Moreover, breathiness has been shown to be able to function as an enhancement strategy for nasality in both nasal vowels and consonants (^{Garellek, Ritchart, and Kuang 2016}) -that is, breathy phonation during a nasal vowel or consonant can heighten the perception of nasality for the listener. Warekena (Maipuran) provides an interesting example of how nasality and breathiness cooccur. Utterance-medial and final pauses are marked with a final-h segment along with a nasalized copy vowel: /-hṼ/. Following ^{Garellek, Ritchart, and Kuang (2016)}, we can posit that the nasalization on the copy vowel enhances the breathiness of the [h] marking the pause in the utterance.

The relationship between nasality and glottalization is less straightfor-ward. Breathiness and glottalization have been modeled as being on opposite ends of Ladefoged’s Continuum model of glottal states (^{Gordon and Ladefoged 2001}). Evidence such spectral tilt (H1*-H2*) seem to support this difference between glottalization and breathiness. Breathiness has been shown to have a higher spectral tilt (H1*-H2*) and glottalization lower as compared to modal phonation (^{Garellek 2015}). Also, glottalization is associated with aperiodicity in f0, while breathiness is not. Nonetheless, there are cases in which there seems to be a relationship between nasality and laryngeal stricture. For example, in Thai, vowels have been shown to nasalize without a nasal antecedent after glottal fricatives and to a lesser degree after glottal stops (^{Matisoff 1975}, ^{Johnson et al. 2019}). Similarly, ^{Ulrich (1993}, 439) notes that glottal stops have an affinity for nasalized vowels in Chickasaw, occurring “disproportionately” after nasalized vowels. As a Nahuatl-specific phenomenon, glottalization and breathiness are connected. In ChN, utterance-final glottalization is realized as a canonical glottal stop, followed by a breathy release. In the acoustic signal, an utterance-final phrase like [ne ͡tʃi͡tʃi] ‘that dog’, could be narrowly transcribed as [ne ͡tʃi͡tʃiʔh]. This parallels Whorf’s description of Milpa Alta Nahuatl /ʔ/ which is transcribed as ’h to express the aspirated portion after the glottal stricture (^{Whorf, Campbell, and Karttunen 1993}). Whorf includes glottal stop as part of the class of sounds along with /p, t, k/, which are strongly aspirated. In ChN, /p, t, k/ are also aspirated consonants and would be narrowly transcribed as [pʰ, tʰ, kʰ] -this especially salient in the word-final position. Interestingly, in Milpa Alta, Whorf describes expected /ʔ/ in the word-final position to be weakened or surface as h or ʰ (which perhaps indicates a weak [h]) that attaches itself to the following vowel-initial word (onset maximalization), suggesting that in Milpa Alta, at least, glottalization and breathiness can occupy the same place. Perhaps specification is not for glottal adduction nor abduction, but rather, broadly laryngeality that is generally realized as [ʔh] but sometimes [ʔ], [h], or weaker forms of either, along the lines of the Welsh English pre-glottalization/aspiration example (^{Hejna 2023}).

We have already seen a context noted by KL, 18c), where nasal intrusion corresponds to an expected laryngeal sound. The relationship between breathiness, glottalization, and nasality in Nahuatl suggests that the presence of prosodic glottalization as well as various sources of aspiration (e.g., neutralization/ debuccalization of consonants, spirantization of sonorants, and complex glottal consonants [ʔh] such as in Milpa Alta Nahuatl, and aspirated voiceless stops), may together help account for the multitude of nasal intrusion contexts. Let us now return to the unaddressed contexts for nasal intrusion observed by KL summarized in 18) above.

One possibility is that the increased number aspirated/laryngealized articulations in a single word might have compounding effect on the acoustic properties of breathiness such that the tlacuilo perceived nasality and thus included an orthographic n. For example, 18a) the expected form quimatican, ‘let them know’, is instead quimatincan. In phonology like ChN, a narrow transcription would be [kʰimatʰikʰah]. It is possible that nasal intrusion occurs here in a context sandwiched between an aspirated consonant and a very breathy syllable. The overall effect of this sequence of aspirated articulations could work together to misguide the perception of nasality.

An alternative account is that intrusive nasals in the orthographic representation do not necessarily reflect a mistaken nasality but rather, simply the perception of breathiness. Since [h] in a variety like ChN often corresponds to two orthographic representations in colonial orthographies: n in the case of underlying nasals and zero graphemic representation in the case of underlying /h/. In this account we can imagine the tlacuilo, interpreting the aspiration/breathiness in the speech stream and having to decide if it should be represented orthographically with an n (or even l or uh) or not represented at all.

We have already addressed the final intrusive n in sintiyon, ‘site’ 18f), as possibly coming from the common pattern of adding an [h] to vowel-final loanwords. The first intrusive n here seems to occur in a similar context as the [sãp] Hindi example. Here the aspiration of the [s] combined with the overall breathiness of rest of the word [sitʰʲoh] might have also resulted in minimally the perception of breathiness if not the perception of nasality, either of which to be represented by an orthographic n. Ytoca, ‘his name’, occurring as yntonca, also in 18f), provides a similar context. In ChN, this word is [itʰokʰah]. The final [h] in ChN [itʰokʰah], ‘her/his name’, like other words such as [imah], ‘her/his hand’, correspond to forms in Classical Nahuatl with an i such as tocaitl, ‘name,’ and maitl, ‘hand’. While not represented in the orthography, it is likely that the form in classical Nahuatl also had a final laryngeal sound, moreover, vowel-initial words are candidates for prosodic glottalization in ChN. Again here, the multiple sources of laryngeal sounds could have been the source of the intrusive nasals in the written form: [ʔitʰokʰah] represented as yntonca. The fuller phrasal context of the tokens described by KL would be useful here.

The second example in 18a) palpan tlanca from the expected palpan tlaca, ‘Palpan people’, also likely had a glottal fricative that was not systematically represented in the orthography. In some varieties, the plural of ͡tɬaka͡tɬ man/person is ͡tɬakah. Here, we may have an example of metathesis as identified in 18b), only in this case metathesis of a glottal fricative, represented as an n. Were this historical token pronounced in the ChN phonology, it would be [pʰaɬpʰah ͡tɬakʰah] (cf. ChN [tɬakameh]). In this case not only are there a number of aspirated consonants in the word, there is also a final glottal fricative that the tlacuilo would have had to decide whether to represent it in some way or another. Following this theme of metathesis, KL suggest that the forms ynpa from ypan, ‘on it’, and Teohuanca from Teohuacan (place name) reflect methathesis of a nasal segment from an adjacent syllable. Metathesis here may also have been motivated aspiration on stops combined with some of the nasal realizations described above: [n̥]~[h]~[ Ṽ]. All would potentially have the perceptual effect of leftward movement of the acoustic properties associated with an n in the orthography. The combined breathiness of the aspirated release of the stop combined with either [n̥]~[h] (e.g., [ipʰan̥]~[ipʰah], [tʰeowakʰan̥]~[tʰeowakʰah]) could have either inspired the tlacuilo to represent the breathiness with an n earlier in the word, or alternitively have resulted in the percepts of nasality earlier in the word to be represented by an n. Similarly, the loss of a final segment through nasal absorption would have moved the nasal articulation closer to the stop (e.g., [ipã], [tʰeowakʰã].

Two additional examples in 18d) seem to be the result of a different process than what I have described above. KL described intrusion as occurring before “nonnasal resonants”. The unexpected nasal in manyor may reflect nasality spreading across the syllable from the initial m. On the other hand, onpohuanlli from onpohualli, ‘40’, seems to be accounted for by the pattern of preaspiration found in some dialects such that /l͡tɬ/ sequences surface as [hlː]. In these cases, the preaspiration might have been interpreted as a sound to represent with the letter n or created the percept of nasality. In the case of chicueponhualli from chicuepohualli, ‘160’, there is no straightforward account. Perhaps in the context of both preaspiration of the ll sequence and the aspirated release of the /p/, the percepts of nasality were created, or at least the precepts of breathiness. For these examples, a possible variable is vowel height. In Thai, spontaneous nasalization following both glottal fricatives and glottal stops occurs more often with mid and low vowels, and may be the result of lower resting position of the velum in the production of these vowel qualities (^{Johnson et al. 2019}).

In addition, glottalization emerging from phrasal structure might also have contributed to some of the intrusive nasals, especially since some glottal stops (i.e. those of Milpa Alta) are described as being accompanied by aspiration. Taken with the pattern observed in ChN, where glottalization can move leftward to the nearest voiced syllable, it is possible that some of the harder to explain cases reflect prosodic glottalization and something about the phrase structure, especially in the cases described as occurring less frequently. However, without knowing the phrase structure in which the examples given by KL were in, it is impossible to know if this had an effect. One example suggests that this nonetheless could have been a factor in some of the intrusive nasals. In 18e) KL note that a common site of nasal intrusion is “finally.” The example given is niquitohuan from the expected niquitohua, ‘I say’. I suspect that the intrusive nasal here corresponds to a laryngeal articulation associated with the pre-pausal context. For example, in the Testaments of Culhuacan (^{Cline and León-Portilla 1984}), both forms, niquitohua (293 instances) and niquitohuan (36 instances), ‘I say’, are found. In most cases they occur in the same context: at the beginning of a declaration with the formulaic phrase Ihuan niquitohua(n), ‘and I declare’, followed most often by a subordinate clause introduced with yn. It is plausible that in the phrasal structure of the text analyzed by KL, that a pause in the spoken form could align and be accompanied by some form of glottalization as in ChN -perhaps realized like the Milpa Alta [ʔh]. In the example below, the unexpected n at the end of niquitohua in the sentence below from the Testaments of Culhuacan might align with pre-pausal glottalization that the tlacuilo perceived as a segment to be represented with n: “-yhuan niquitohuan yn nocaltzi yn onpa ycac yn tiyanquiztenco tonatiuh yquiçayanpa ytzticac monamacaz. ‘And I declare that there is a house of mine in Tianquiztenco which faces east’” (^{Cline and Portilla 1984}, 23).

Finally, a possibility is that some of the intrusive nasals reflect local variants not widely attested in other documents or varieties. One such example is the observation of nehuantl, ‘I’ (KL p. 380), from the expected nehuatl. While most attestations of the first-person singular pronoun do not feature an additional nasal, there is at least one that does. From a 1598 document from Temascalapan (^{Rojas Rabiela, Rea López and Medina Lima 1999}), the form “nehuantli notoca don Francisco Ximenes” is given, very similar in form to nehuantl. Similarly, ^{Melton-Villanueva (2016}) notes that Nehuatli was the preferred form used in the valley of Toluca. Here we can posit the presence of a laryngeal consonant as the -tli absolutive is used but no consonant is represented in the consonant cluster (similar to cases like patli [pah͡tɬi]~[paʔ͡tɬi]). In these two variants, nehuantli and nehuatli, KL’s observed intrusive nasal in nehuantl corresponds to either a laryngeal or nasal sound. Interestingly in ChN, one of the forms of the first-person singular pronoun, [nahwa], also contains a laryngeal sound.

In 25) below, I summarize the insights from ChN and other Nahuatl varieties that can help account for patterns of nasal intrusion identified by KL (in 18) above).