|Location point:||3°45′ S, 35°10′ E|
Hadza is a language isolate of Tanzania.
Location and Speakers
Hadza is spoken along the entire eastern shore of alkaline Lake Eyasi, which lies at the base of the Serengeti Plateau in central Tanzania, from Mount Oldeani in the north (just south of Ngorongoro Crater) to the Isanzu agricultural areas in the south. The Hadza do not live at altitudes much above 1500 meters, as they do not build shelters that will ward off the night chill at higher elevations.
There is a small population of Hadza to the west of the lake, in Dunduhina /duƞtuhina/ 'Sukumaland', but their number seems to be decreasing and many only speak Sukuma and Swahili.
There are approximately 1,000 speakers of Hadza, most now bilingual in Swahili. Other second languages include Bantu Isanzu in the south, Bantu Sukuma in the west, and to a lesser degree Nilotic Datooga in the center (e.g. near the Yaeda Valley to the southeast of the lake) and Cushitic Iraqw on the margins of the Iraqw highlands. The northern Hadza area, around the town of Mangola, was largely monolingual until the introduction of Standard Swahili after independence, though there are loanwords from colonial-era 'upcountry' Swahili.
As of 2005, about 40% of Hadza lived as full-time hunter-gatherers, and language transmission was robust in the areas east of the lake, but since then there have been reports of children speaking Swahili east of the lake, perhaps due to the influx of Dunduhinabee abandoning their territory in the west.
Hadza is a language isolate. Greenberg classified it as Khoisan due to its use of click consonants. If it were not for the clicks, it's likely that Hadza would have been classified as Cushitic. Hadza and Sandawe have very similar phonetic inventories and phonotactics, suggesting an old sprachbund. However, the only detected commonalities in vocabulary are old Cushitic loans or perhaps loans that passed from one to the other through Cushitic. Much of the borrowed Cushitic vocabulary is consistent with reconstructions of proto-West Rift.
There are no dialects of Hadza, presumably due to the mobility of its speakers. There are some regional differences of vocabulary, however, and speakers note that many more Bantu loans are used in the south. Some differences in pronunciation have been recorded from people from different areas, but it's not clear if these differences are regional accents or individual variation.
"Hadza" is the most common name in the literature. Haza /ɦad͜za/ simply means 'human being'. The derivative "Hadzane" (/ɦad͜zane/ 'in the manner of the people') is sometimes encountered as the name of the language. The feminine plural hazabee /ɦad͜zabeʔe/ may be used for the people, though it is usually spelled "Hadzabe" due to devoicing of the final vowel. (The feminine is the inclusive gender in the plural.) "Hadzapi" is the masculine-plural copular form hazaphii /ɦad͜zapʰiʔi/. "Hatza" and "Hatsa" are older German spellings.
"Tindiga" is from the Swahili name, watindiga 'people of the marsh grass' and its derivative kitindiga, presumably named for the large spring in Mangola. The forms wahadza, kihadza are found instead of watindiga, kitindiga in recent Swahili publications, but the Hadza do not consider watindiga to be pejorative. "Kindiga" may be a cognate from one of the local Bantu languages. "Kangeju" (pronounced Kangeyu) is an obsolete German name of unclear origin. The name "Wahi" (pronounced Vahi) found for the western Hadza in Kohl-Larsen is a Sukuma name for the Hadza.
Hadza does not have a written tradition. Two orthographies have seen recent use: the phonetic transcription of Tucker, Bryan & Woodburn (1977), used with some modification by Woodburn and other anthropologists, and the practical orthography of Anyawire et al. (2013). As it is based on the IPA, the Woodburn transcription is more accessible to linguists, while the Anyawire orthography is closer to Sandawe and other written languages of Tanzania. There is evidently some political controversy over the choice of orthography, with Hadza reporting they have been told that if they use the Anyawire orthography they will loose their land. There are two motivations for the Anyawire orthography: the perception that the IPA click letters of the Woodburn transcription portray the Hadza language as bizarre, when the Hadza people already have difficulties with being marginalized and a perception that they are not proper human beings, and the practical issue that the ubiquitous apostrophe of the Woodburn transcription makes it impractical for extended use. The apostrophes are generally omitted by those Hadza who have learned the Woodburn system, with the resulting texts often being unintelligible even to their authors; when the apostrophes are written, they greatly slow down the pace of writing. In the Anyawire orthography, doubled consonants are used for glottalized/ejective consonants and doubled vowels for glottal stop, with the apostrophe only being used for the Swahili convention of ⟨ngʼ⟩ for borrowed /ŋ/.
The Anyawire orthography is used alongside the IPA in this article for accessibility with the Anyawire dictionary. For clarity, an apostrophe will be retained in this article for glottal stop between unlike vowels, e.g. ⟨ae⟩ /ae/ vs ⟨aʼe⟩ /aʔe/. In practice, the Hadza are more likely to write an epenthetic glide between vowels in hiatus, e.g. ⟨aye⟩ for /ae/ vs ⟨ae⟩ for /aʔe/.
Hadza syllable structure is limited to CVN in native words. There are no vowel-initial roots, unless h is analyzed as an allophone of zero. It is possible that nasal codas are found primarily in loan words and ideophones. This N surfaces as a homorganic nasal when a following consonant has a place of articulation for it to assimilate to, and as nasalization of the vowel before a glottal consonant (glottal stop and h/zero) and pre-pausa. Coda N also appears allophonically before voiced nasal and glottalized nasal clicks. A moraic (syllabic?) nasal is also found word-initially in loans from Bantu and Datooga (and in a few Cushitic loans, some due to metathesis and some unexplained), in which case it may have a different pitch than the following vowel, or may be preceded by an epenthetic vowel; such nasals are analyzed here as NCV sequences rather than as a series of prenasalized consonants. Phonetically, initial #NCV may also occur in shortened forms when the initial hV syllable of a #hVC₂V-shaped word is elided and C₂ is a glottalized nasal click.
Stress and tone
It is not yet clear whether Hadza has phonemic tone. If it does, it is likely to have a minimal system of the kind often called pitch accent. Salient stress and pitch are not restricted to a particular syllable; their position may vary between elicitations of at least some words pronounced in isolation, but in some words pitch appears to be fixed. Heavy syllables tend to attract stress but again the pattern is not absolute. There are no known minimal pairs, grammatical or lexical, for stress or pitch. Minimal pairs claimed in the literature have turned out to either not be distinct in pronunciation, or to not be minimal pairs (e.g., one of the words may have an additional morpheme such as the habitual suffix -e, which frequently reduces to vowel length and attracts stress). It may be that some loanwords retain the tone of their source language.
There are five phonemic vowels, which are close to cardinal [a e i o u].
Phonetic nasal vowels, which are not common except before a consonant, are here analyzed as /VN/.
Phonetic long vowels are often sequences, as in kaha [kaː] ~ [kaɦa] 'to climb'. In some words, however, there is no allomorph with [ɦ], as in bôko [boːko] 'she' (forms a minimal pair with boko [boko] 'to be ill') and sêta [seːta] 'moon'. There is no known minimal set to establish long /Vː/ as distinct from a /VV/ sequence.
The mid vowels /e, o/ tend to rise when a high vowel, /i, u/, occurs in the following syllable, and /e, o/ may merge with /i, u/ in that position. However, such assimilation tends to be optional, even in suffixed pronouns such as onebii [ʔone̝biʔi] ~ unibii [ʔunibiʔi] 'we' (masculine, inclusive).
/u/ is relatively uncommon except in loans or due to vowel-height assimilation. /u/ does not occur at all in grammatical morphemes unless the morpheme has a second syllable with vowel /i/, and so is presumably a historical *o in such cases.
Vowels tend to devoice in pre-pausa position after a voiceless consonant. In the highly frequent pattern of word-final -V₁ʔV₁ (where the two vowels are the same, a pattern found in half of all grammatical suffixes and clitics though not frequently elsewhere), the vowel and preceding glottal stop tend to elide completely when pre-pausa, and either elide or reduce to a long vowel before a suffix/clitic or another word.
Hadza has at least 50 productive syllable onsets and several other marginal consonants.
Consonants in parentheses are thought to be loans, though the voiced obstruents at least are well-integrated and may be spreading to native roots.
Consonants with a medium grey background may follow a coda nasal in the preceding syllable (that is, they may be C₂ in a CVNCV pattern), whereas those on the lightest grey background may not. Alveolar /ǃ/ and lateral /ǁ/ (medium light grey) are not attested in this environment. However, it is suspected these gaps may be accidental, given that only two roots are known with dental /ǀ/ as C₂ after a nasal coda and that these sequences have not yet been investigated in the field. Moreover, the one known case of aspirated lateral /ƞǁʰ/, though its identity was clear when elicited, needs to be confirmed with additional speakers. A word with the same sequence reported in Tucker, Bryan & Woodburn could not be confirmed by Anyawire et al.
The voice-onset time of consonants is reduced after a nasal coda: tenuis consonants may be partially or completely voiced, and the aspiration of aspirated consonants may be reduced or lost. This is true of clicks as well as of plosives and affricates, and is reflected in both orthographies. There is only a two-way VOT distinction after a nasal coda. A /b/~/p/ distinction has not been found in post-nasal position despite /b/ being native.
Aspiration may only be distinct in the first several syllables of a word. Many speakers distinguish aspirated plosives in only a few words that contrast /k/ and /kʰ/, but maintain a robust aspiration distinction in affricates and clicks. The reason for this asymmetry is unknown.
The glottalized nasal clicks surface as [Ʞʔ] (where ⟨Ʞ⟩ stands for any click release) post-pausa and often as [ŋ͡nʔꞰ] intervocalically. Accounts differ as to whether there is voiceless nasal airflow preceding the click in post-pausa position.
Aspirated clicks may have a delayed and clearly audible posterior release, e.g. [Ʞʰ] ~ [Ʞkʰ]. The audible release has the same post-velar articulation as the k-series of plosives.
The alveolar clicks sound rather palatal with some speakers, and tend to be weakly articulated, though, so far as is known, not slapped as they can be in Sandawe. (Reports of slapped [ǃ¡] clicks in Hadza by a UCLA team in the 1990s could not be confirmed with the UCLA archives, and it would appear that weak/flapped alveolar clicks were erroneously identified with slapped Sandawe clicks that the team had recorded a few weeks earlier.) When intervocalic, nasal alveolar /ⁿǃ/ may surface as, or nearly as, a doubly articulated pulmonic nasal [ŋ͡n]. The lateral clicks, on the other hand, are typically quite salient, and such phonetic reduction has not be noted. They appear to have the same post-alveolar to palatal place of articulation as the lateral affricates. The 'dental' (denti-alveolar) clicks are unremarkable, but are less common than the others.
Among the ejectives, /pʼ/ is uncommon and may be ideophonic. Most other ejectives vary between affricate and fricative realizations. Fricative allophens [sʼ, ʃʼ, xʼ, xʷʼ] have been noted for /t͜sʼ, t͜ʃʼ, k͜xʼ, k͜xʷʼ/. Velar /k͜xʼ/ and /k͜xʷʼ/ also have plosive allophones, [kʼ, kʷʼ], and /k͜xʼ/ has a lateral allophone [k͜ʟ̝̊ʼ]. Ejective /t͜sʼ/ may correspond phonologically to both the plosive and affricate (/t/ and /t͜s/) alveolar series of pulmonic consonants.
The palatal lateral affricates may be pronounced with an alveolar onset, e.g. [t͜ʎ̥˔], or without involving the front of the tongue, e.g. [c͜ʎ̥˔]. Speakers express no preference when hearing these articulations. The relatively uncommon fricative /ɬ/, however, is alveolar; a palatal articulation is not accepted. Given that the fricative may be pronounced as an affricate in some prosodic situations, this means that there is a three-way phonetic place contrast among laterals: [t͜ɬ] vs [t͜ʎ̥˔ ~ c͜ʎ̥˔] and [t͜ʎ̥˔ʼ ~ c͜ʎ̥˔ʼ] vs [k͜ʟ̝̊ʼ].
/ɺ/ is typically realized as [l] post-pausa and [ɾ] elsewhere, though [ɺ] is not uncommon in either position. It tends toward [l] when it occurs as both C₁ and C₂, as in /ɺoɺa/ (sp. rabbit).
[ɦ] and zero may be allophones. If so, and given that vowel-initial suffixes and clitics may elide a short stem-final vowel, it would be simplest to posit that the phoneme is zero and that [ɦ] is an allophone of zero, rather than that /ɦ/ is elided under some conditions. A predictive conditioning environment for the presence or absence of [ɦ] has not been discovered, but prosody may play a role; even vowel-initial suffixes and clitics may be realized with an initial [ɦ] when given contrastive prosody. Assimilation or bilingualism may be involved in the case of loanwords. For instance, an Isanzu loan for 'snake' was recorded as ihato /ʔiɦato/ at the beginning of the 20th century (and perhaps in an Isanzu-biligual area) but as iyato /ʔiato/ at the beginning of the 21st century in Mangola.
The glides /j/ and /w/ are not distinctive next to high front /i/ and back /u/, respectively, and rarely (or uncertainly) distinctive next to mid front /e/ and back /o/. It is possible that medial y and w are phonemically /i/ and /u/ and that initial y and w are /ʔi/ and /ʔu/. When they occur word-initially, purported /jV, wV/ often appear as [ʔi(j)V] and [ʔu(w)V]. This is true even of a ubiquitous sequential auxiliary (yamo [jamo] ~ iamo [ʔiamo] in the 3msg posterior (past?) tense). [j] does occur in two object suffixes, and it is not clear if these can be explained away as /i/. In the anthropological and linguistic literature, glides y and w are often written between unlike vowels. For example, the 3msg non-past copula clitic -ha /ɦa/ is often transcribed as ya after a front vowel and as wa after a back vowel. (The copula sometimes contains a very clear [ɦ], and pre-pausa even [h] due to final devoicing, but often the /ɦ/ is realized as transitional murmur or possibly elided altogether.)
A labial click, variously reported as [ⁿʘʷ] or [ⁿʘˀ], is found in a single root, mcwa-, that is mimetic for a kiss and often made with an accompanying kiss to the hand. Mcwa- may be allomorphic with ncuwa- /ⁿǀua/ 'kiss', with a dental click.
A trill [r] is used by some speakers in some words, such as karro /karo/ (hunting/sighting name for a female giraffe), perhaps reflecting the pronunciation of the word in the language it was borrowed from. It is replaced with /ɺ/ (karo /kaɺo/) by other speakers.
An implosive [ɓ] is used by some speakers in some words, perhaps reflecting the pronunciation of the word in the source language, though in some instances it might co-occur with ejectives, which would be difficult to explain away as a loan (e.g. beggawuko /bekx’auko/ 'elephant', apparently [ɓekx’auko] in some recordings). It is replaced with /b/ by other speakers. Other implosives have not been noted.
A fricative [x] occurs lexicalized in an expletive, ahho [ʔaxo], where otherwise the root is akho /ʔakʰo/ 'genitals'. (/kʰ/ is commonly realized as [kx].)
Word-initial /mu/ and occasionally /mo/ are optionally realized as syllabic [m̩], for example in the common greeting mutana ~ mtana. This is a Bantu alternation.
/v/ may occasionally be found in loans, as in mimetic vûvuko ~ fûfuko /vuːvuko ~ fuːfuko/ 'propeller toy', though it is normally replaced with /f/. [v] (or perhaps [β]) has been reported for /b/, for example in Bleek's transcriptions of the plural suffixes -bee and -bii as 've' and 'vi'.
All clicks within a root must be the same, as in /ǀikiɺiƞǀa/ 'pinkie'. Although there is a tendency for click consonants and ejectives to occur as C₁, in a quarter of words with clicks the click occurs only as C₂, and in other words as both C₁ and C₂.
Clicks and ejectives may co-occur, though always in that order. When they co-occur, there appears to be some sibilant/lateral harmony, with a strong tendency for /t͜sʼ/ to occur after denti-alveolar clicks and for /c͜ʎ̥˔ʼ/ to occur after lateral clicks. (/k͜xʼ/ does not seem to be affected by the click, and is common after all places/types.)
When an aspirated consonant occurs more than once in a word, the first instance tends to deaspirate (as in Grassmann's/Katupha's Law, though generally optional in Hadza). A coda N may block deaspiration, as in the allomorphs /peǃeǃʰe/ ~ /peƞǃʰeƞǃʰe/ 'to rush'. Repeated NC sequences may loose their nasal segment as well, as in /ƞtʰaɺaƞtʰaɺaʔabiʔi/ ~ /taɺaƞtʰaɺaʔabiʔi/ '(arrow) barbs'. It is not clear to what extent these patterns hold if the aspirated consonants are different rather than repeated phonemes.
Only roots are given below. For example, the word for 'name', given here as the root akhana, is actually feminine plural akhanabee.
Wichmann & Holman basic 40-word list (ordered by stability)
- louse ccamazzi /ⁿǀˀamat͜sʼi/
- two piye /pie/, konxa /koⁿǁa/ (verb)
- water ati /ʔati/ (mpl; msg = 'rain, river')
- ear hajjapitchi /ɦat͜ʃ’apit͜ʃʰi/
- to die taxxi /taⁿǁˀi/, misi /misi/ (Datooga?)
- I *one /ʔone/, */-ko/
- liver xxe /ⁿǁˀe/
- eye akhwa /ʔakʷʰa/
- hand, arm ukhwa /ʔukʷʰa/
- hear nxaʼe /ⁿǁaʔe/
- tree zziti /t͜s’iti/
- fish ccama /ⁿǀˀama/, ccara /ⁿǀˀaɺa/
- name akhana /ʔakʰana/
- stone haqqa /ɦaⁿǃˀa/
- tooth aha /ʔaɦa/
- breast iriba /ʔiɺiba/ ('breasts, milk'. Cushitic); ggazza /k͜x’at͜s’a/ ('flesh at sternum')
- thou the /tʰe/, /-en-/ (m), /-n-/ (f)
- path yeke /jeke/
- bone midla /mic͜ʎ̥˔’a/ (cf. Dahalo)
- tongue ncata /ⁿǀata/
- skin ahu /ʔaɦu/
- night zzifi /t͜s’ifi/ (m; f = 'day')
- leaf hazzape /ɦat͜s’ape/
- rain = 'water' (msg = 'rain, river'; mpl = 'water')
- kill xô /ǁoː/
- blood athama /ʔatʰama/
- horn roo /ɺoʔo/
- person haza /ɦad͜za/ (human), unu /ʔunu/ (person)
- knee guringuri /ɡuɺiƞɡuɺi/ ~ gurunguri /ɡuɺuƞɡuɺi/ (Cushitic? pan-African)
- one itchâme /ʔit͜ʃʰaːme/
- nose intawe /ʔiƞtʰawe/
- full ggaʼe /k͜x’aʔe/, *cco- /ⁿǀˀo-/
- come za /d͜za/ (Bantu), botco /bot͜ʃo/ (stem of imperative)
- star ntsa /ƞt͜sʰa/, sa /sa/ (Cushitic?)
- mountain xxudle /ⁿǁˀuc͜ʎ̥˔’e/
- fire zzoko /t͜s’oko/
- we (exclusive) ô /ʔoː/, /-j-/
- to drink fa /fʷa/
- to see chî /ǀʰiː/ ~ cî /ǀiː/
- bark heggwa /ɦek͜xʷ’a/
- new zana /d͜zana/
- dog xhaano /ǁʰaʔano/, tîngi /tiːƞki/
- sun isho /ʔiʃo/
- This article uses the old IPA letter for a moraic nasal, ⟨ƞ⟩, which was retired from the IPA alphabet because it had no specific phonetic value. In Hadza, the moraic nasal is homorganic with a following consonant.
- cf. ccezze /ⁿǀˀet͜s’e/ 'tick'
- from hazzape /ɦat͜s’ape/ 'leaf'?
- If *-ko-a > /-kʷa/ rather than *-kʷa-ʔV > /-koʔo/.
- Anyawire et al. (2013) only found ntsa. Tucker et al. (1977) report sa 'star', and Woodburn confirms the form he knows does not have a nasal onset. /ƞ/ + /sa/ would generate /ƞt͜sʰa/. Indeed, speakers do not care if the latter is written ntsa or nsa. There are several Cushitic loanwords that have an onset that varies as C ~ NC in Hadza where the source language is only reconstructed with C.
- from za 'come'?
- Mariamu Anyawire, G.G. Bala, Kirk Miller (ed.) & Bonny Sands (2013). A Hadza Lexicon (ms).
- Roger Blench (2008). Hadza Animal Names. Paper presented at the 3rd International Symposium on Khoisan Languages and Linguistics, Riezlern, 7–9 July 2008.
- Niklas Edenmyr (2004). The semantics of Hadza gender assignment: a few notes from the field. Africa & Asia, no. 4, Department of Oriental and African languages, Göteborg University, 3–19.
- Edward Elderkin (1978). Loans in Hadza: internal evidence from consonants. Occasional Papers 3, Dar es Salaam.
- Peter Ladefoged, Ian Maddieson & Bonny Sands (1991). Hadza wordlist and sound files. UCLA Phonetics Lab Archive.
- ———— (1993). The Phonetic Structures of Hadza. In UCLA Working Papers in Phonetics No. 84: Fieldwork Studies in Targeted Languages, 67–87.
- Kirk Miller (2008). Hadza Grammar Notes. Paper presented at the 3rd International Symposium on Khoisan Languages and Linguistics, Riezlern, 7–9 July 2008.
- ———— (2009). Highlights of Hadza fieldwork. Paper presented at the 83rd Annual Meeting of the Linguistic Society of America, San Francisco, 8–11 January 2009.
- ———— (2016). Hadza Kinship Terms. In Shah & Brenzinger, eds., Khoisan Languages and Linguistics: Proceedings of the 5th International Symposium, July 13–17, 2014, Riezlern/Kleinwalsertal. Cologne: Rüdiger Köppe, 277–333.
- Bonny Sands (1998). The Linguistic Relationship between Hadza and Khoisan. In Schladt, Matthias (ed.) Language, Identity, and Conceptualization among the Khoisan (Quellen zur Khoisan-Forschung Vol. 15), Cologne: Rüdiger Köppe, 265–283.
- ———— (2013). Phonetics and phonology: Hadza; Tonology: Hadza; Morphology: Hadza; Syntax: Hadza. In Rainer Vossen, ed., The Khoesan Languages. Oxford: Routledge.
- Sergei Starostin. Hadza basic lexicon. Global Lexicostatistical Database. [much of the parsing is speculation, and Bleek's posthumous data is unreliable]
- A.N. Tucker, M.A. Bryan & James Woodburn as co-author for Hadza (1977). The East African Click Languages: A Phonetic Comparison. In J.G. Moehlig, Franz Rottland, Bernd Heine, eds, Zur Sprachgeschichte und Ethnohistorie in Afrika. Berlin: Dietrich Diener.