Feature set based on the work of Rose (2013) and Rose (2015).
Authors: Françoise Rose, Marine Vuillermet
Acknowledgements: Natalia Chousou-Polydouri, Oscar Cocaud-Degrève, Kellen Parker van Dam, David Inman, David Timothy Perrot, Diana Krasovskaya
Rose (2015: 531) defines gender indexicality as “a formal distinction depending on the gender of the speech-act participants[. It] is distinct from grammatical gender, which indicates the gender of a referent. Gender indexicality can be found in sentences where neither the speaker nor the addressee is involved as a participant in the event.” Languages displaying this phenomenon consequently have genderlects. The three examples below contrast how women vs men speakers say ‘tree’ in Karajá [kara1500] (Nuclear-Macro-Je; Brazil), ‘yes’ in Eastern Bolivian Guaraní [east2555] (Tupian; Argentina, Bolivia, Paraguay), and ‘his house’ in Mojeño Trinitario [trin1278] (Arawakan; Bolivia).
| ♀ speaker | ♂ speaker | ||
|---|---|---|---|
| (1) | Karajá (Ribeiro 2012: 131) | ||
| ‘tree’ | kɔwɔrʊ | ɔwɔrʊ | |
| (2) | Eastern Bolivian Guaraní (Giannecchini 1996 [1898]: 306, 310) | ||
| ‘yes’ | éé | tà | |
| (3) | Mojeño Trinitario (Rose 2013: 119) | ||
| ‘his house’ | ñipeno | mapeno | |
The symbols ♀/♂ stand for the two kinds of indexical gender, and are distinct from the abbreviations F(em)/M(asc), which stands for grammatical gender. Other terms used to refer to genderlects (the term we used in our survey) are gender indexicality (Fleming 2012), gender allocutive agreement or allocutivity (Oyharçabal 1993, Antonov 2015), or male/female speech. Although “male/female speech (act participants)” is frequent in the literature, previous works (e.g. Furfey 1944, Haas 1944) used the term “men/women's speech”. We intentionally chose the terms “men/women” over “male/female” speech act participants to refer to people’s gender rather than their biological sex.
Genderlects can vary according to the following parameters: 1. indexed speech act participant (speaker vs. addressee vs. both) 2. categorical vs. statistical genderlects 3. locus of gender indexicality (phonology, morphology, or lexicon) 4. (inter)subjective nature of gender-indexing elements 5. non-referential vs. referential gender indexicality.
While most parameters existed in the previous literature, the (inter)subjectivity parameter has been developed for the present study.
Haas (1944) distinguishes three types of genderlects on the basis of the speech act participant(s) indexed, and we have added a fourth one, for languages which have more than one type. Parameter 1 is surveyed with Gend-03 (phonology), Gend-05 (morphology), Gend-09 (lexicon).
The following examples show that in Garifuna [gari1256] (Arawakan; Belize, Guatemala, Honduras, Nicaragua; de Pury Toumi 2003), women speakers use the first word of each pair, while men use the second word.
| (4) | würinauga | gúñaru |
|---|---|---|
| ‘yesterday ♀’ | ‘yesterday ♂’ |
| (5) | hiñaru | würi |
|---|---|---|
| ‘woman ♀’ | ‘woman ♂’ |
In Basque [basq1248] (isolate; Spain, France), speakers use the first form in (6) to address a familiar woman, and the second a familiar man (Alberdi 1995: 276).
| (6) | diagon | diagok | |
|---|---|---|---|
| 3S.stay.ADD♀ | 3S.stay.ADD♂ | ||
| ‘he/she/it stays (woman vs man familiar addressee).’ | |||
This last case is much rarer. In Yana [yana1271] (isolate; United States; Sapir 1929 [1963]), the lexemes used for ‘quail’ and ‘acorn’ depend on the gender of both the speaker and the addressee: sika·ka and yuna in (7a) are only used by men speakers to men addressees, and in all other cases (i.e. men to women, women to men, women to women), including mixed audience, sika·khA and yuh in (7b) are used.
| (7) | a. | Men addressing men | |
|---|---|---|---|
| sika·ka | ‘quail’ | ||
| yuna | ‘acorn’ | ||
| b. | All other cases | ||
| sika·khA | ‘quail’ | ||
| yuh | ‘acorn’ | ||
Note that all attested Type III systems, also known as “relational gender” systems, are asymmetric: there is not a specific genderlect for each possible configuration (Rose 2015: 507-508).
We added a fourth type, where different items participate in different types of genderlects. In Tupinambá [tupi1273] (Tupian; South America), several lexical items differ according to the gender of the speaker (Type I), and a couple of items are restricted to men speakers with men addressees (Type III).
| (8) | a. | Depending on the gender of the speaker | ||
|---|---|---|---|---|
| reí | reá | ‘affirmative particle ♀ / ♂’ | ||
| raré | rá | ‘truthfully ♀ / ♂’ | ||
| b. | Men speakers to men addressees only | ||
| hẽ | ‘olá, oh!’ | ||
| ahẽ | ‘he, this one, someone’ | ||
Genderlects are statistical in most world languages, but have been argued to be categorical in others (Fleming 2012). In categorical genderlects, each lect is exclusive to a type of speech act situation, depending on the gender of the speakers and/or addressees. The present survey exclusively deals with categorical genderlects.
Cases where lects do not depend solely on gender may look statistical. For instance, the formality of the social context plays a role in the Basque genderlects (example 6). However, the system is still categorical since, within the informal situations, the selection of one genderlect or the other is obligatory. A system is statistical if a speaker can select one of several varieties (intra-speaker variation). A system is categorical if a speaker must use one genderlect or the other.
Günthner (1996) lists the following major domains that can index gender: grammar (phonology, morphology, syntax, lexicon), pitch, choice of languages or varieties, communicative styles, discourse strategies, and discourse genre. Because we survey cases of categorical genderlects, the relevant domains are those of “grammar only”. The feature set is thus concerned with phonology (Gend-02-03), morphology (Gend-04-06, see example (6)), and lexicon (Gend-07-10, see examples (1)-(5), or (7)-(8)). Note that the present study does away with the additional locus called “illocutionary” in Fleming (2012) and “discourse markers” in Rose (2015), because the items previously categorized under these loci clearly belong to the morphology (i.e. epistemic or mood markers) or lexicon (i.e. interjections, ‘self’). The present study instead resorts to an orthogonal parameter of (inter)subjectivity.
The parameter of (inter)subjectivity is an original contribution of the present study, and is considered orthogonally to the morphological (Gend-06) and lexical domains (Gend-10). The term (inter)subjectivity forms a coherent concept covering both “subjectivity” and “intersubjectivity”, which we consider relevant to the study of genderlects.
Subjectivity is the linguistic expression of the speaker himself and their own attitudes and beliefs (Lyons 1982: 102). For example, Cocama [coca1259] (Tupian; Brazil, Colombia, Peru) has gender-indexing interjections for pain, frustration and self-encouragement (Vallejos 2015).
Intersubjectivity is the “explicit expression of the speaker/writer’s attention to the ‘self’ of addressee/reader” (Traugott 2003: 218). For example, Tiwi [tiwi1244] (Australia) has interjections which belong to the lexicon, and additionally, are intersubjective, since they are used to call people: aya (man addressee), agha (woman addressee), awi (more than one addressee) Lee 1987: 123, 377; Osborne 1974: 119.
Genderlects are further categorized into having referential or non-referential gender-inexing elements (for a detailed discussion, see Fleming 2012, Rose 2013). In all the cases illustrated above, gender-indexing elements were non-referential: the linguistic elements that index gender do not refer to the speech participant, the gender of which is indexed. In contrast, gender-indexing elements are referential when the linguistic elements refer to the speech participant the gender of which is indexed. This is the case with pronouns or deictics referring to 1st or 2nd person which have distinct forms for a woman or a man referent. In these items, gender is both indexical and referential (Fleming 2015, Rose 2013, Berg 2023). We did not use this parameter in our survey. It is further discussed in the How section, since it is relevant for the analysis of gender in the pronominal systems as grammatical or indexical gender.
In the first large study on categorical genderlects, Fleming (2012), 17 out of 20 languages exhibiting the phenomenon were in North and South America, as compared with the rest of the world. The subsequent study by Dunn (2013) also found a prevalence of genderlects in South America (8 of the 14 case studies), and was the first to suggest genderlects as a potentially Amazonian areal feature. Another survey, targeting South America only, attested the presence of 41 languages with genderlects across many different stocks in the continent, and suggested the phenomenon to be an areal trait within South America (Rose 2015: 503). Rose and Bakker (2016), based on an opportunistic survey of the literature, strongly supported genderlects as a specifically South American feature, with nearly half of the cases of known genderlects (50/102) occurring within South America.
Because the terms “genderlects” or “gender indexicality” are rarely present in grammars, we established a list of key terms (in French/Portuguese/Spanish/English, namely (fe)male, (wo)man~(wo)men, feminine/masculine, sex, gender, speech and dialect) to systematically search for in the existing documentation, primarily in grammars and dictionaries if available. This section discusses frequent diagnosis confusions, as well as infrequent edge cases.
If the categorical gender distinction is about the lexicon, even experienced linguists might get confused in the following cases, where the lexical distinction is based on gender of - an argument of the verb, e.g. for verbs expressing typical activities of women vs. men, as in (9); - the referent for animate nouns, as in the pair son/daughter, which contrasts the gender of the referent (one's offspring), not of the speech act participants; - the possessor (or ego) for kinship terms, as in (10), i.e. the ego from which the relationship is conceptualized.
In all these cases, there are no genderlects.
In Karajá (Nuclear-Macro-Je), two different lexemes translated as ‘cry’ refer to two different gendered activities: “The women cry or chant daily if a child is traveling or hurt. Women also cry for a period of a lunar month: the death chant. In contrast, men cry only during the death chant and even then less than the women” (Fortune and Fortune 1975). As an alternation based on the gender of the argument of a verb (rather than that of the speaker or addressee), these do not count as a case of genderlect.
| (9) | Karajá | |
|---|---|---|
| obu | ‘cry (feminine subject)’ | |
| hi | ‘cry (masculine subject)’ |
In (10), the three Ese Ejja [esee1248] (Pano-Tacanan; Bolivia, Peru) kin terms index the gender of the ego, i.e. of the person on which the kinship relationship is based, not that of the speaker or addressee.
| (10) | Ese Ejja kin terms for ‘sister’ (e-...-mese is a citation form circumfix for kin terms) | ||
|---|---|---|---|
| a. | e-a'i-mese | ‘older sister of a woman’ | |
| b. | e-shewe-mese | ‘younger sister of a woman’ | |
| c. | e-jjo'i-mese | ‘(older/younger) sister of a man’ | |
The status of kin terms can be misleading when possessed by a speech act participant (e.g. my sister), or in the vocative. All three terms illustrated in (10) can be used by women/men speakers, and to women/men addressees, as in ‘Eva's older/younger sister went to the field’. It is thus important to check whether speech act participants of both genders can use these terms with a 3rd person possessor.
Lexicalized expressions including grammatical gender like obrigada/obrigado ‘thank you’ in Portuguese are respectively uttered by women vs. men speakers. In fact, it comes from a participle that agrees in grammatical gender with the speaker, who is originally also the subject of a non-verbal predication (lit. ‘(I am) obliged.’). Because of this, obrigada/obrigado is considered as grammatical gender (i.e. not genderlect) in Portuguese. Had it been borrowed from Portuguese into a language with no grammatical gender, it would be considered as genderlect.
Still another source of complexity is address terms. For instance, Lady vs. Sir refer to the gender of their referent, and of the addressee, and could therefore be considered as indexing gender. In this case, because Lady and Sir also exist as referential terms, as in the lady is intelligent, it is more economical to consider the gender in their address functions as being primarily lexical. In other words, we only considered address term pairs if they did not exist in parallel as referential terms in the language lexicon.
When gender-indexing elements are referential, both an analysis in terms of indexical gender and in terms of grammatical gender may be possible: - if two forms are available in the 1st person, and if they refer to the gender of the speaker, they could be either analyzed as 1st person feminine / masculine or 1st person speaker ♀/ speaker ♂; - if two forms are available in the 2nd person, and if they refer to the gender of the addressee, they could be either analyzed as 2nd person feminine / masculine or 2nd person addressee ♀/ addressee ♂.
In such cases, we favor an analysis which is consistent with the analysis of the rest of the paradigm, i.e. the other grammatical persons (if available) (Rose 2013, Rose 2018). For instance, in Cubeo [cube1242] (Tucanoan; Brazil, Colombia, Venezuela), where there is a gender distinction in 1st and 3rd persons (Chacon 2012: 270), the 3rd person distinction is about the referent and must be analyzed as grammatical gender, so the 1st person distinction is also considered as grammatical gender, and not indexical gender.
By contrast, in Cocama-Cocamilla [coca1259] (Tupian; Brazil, Colombia, Peru), there is a gender distinction in both the 1st and 3rd persons (Vallejos 2015). The 3rd person distinction depends on the speaker's gender (and not on the referent's gender), and so the 1st person distinction is also considered indexical gender. To summarize, when both an analysis in terms of indexical vs grammatical gender is possible for speech act participant pronouns, we have consistently adopted the analysis coherent with the rest of the pronoun paradigm (Rose 2013).
{ yes | no }
Two types of edge cases, coded as <?>, occurred. The first one accounts for contradicting speaker perspectives in a language with only a handful of speakers. The second accounts for our doubt whether the lect difference depended solely on gender or on some other (i.e. social, register) parameter.
{ NA | yes | no }
All the phonetic/phonological phenomena must occur below the level of utterance. Pitch differences such as exaggerated pitch distinctions between men and women in Japanese (van Bezooijen 1995) are not relevant here.
Guarayo displays genderlect distinction at the phonetic level: the same phoneme is realized differently in the different lects (Hoeller 1932: 2).
| (11) | otso ~ odso | oso | ‘(s)he went away ♂ / ♀’ |
The Karajá genderlects have different phonological inventories: the velar stop phoneme is absent from the men speech (Ribeiro 2012:131).
| (12) | a. | kɔwɔrʊ | ɔwɔrʊ | ‘tree, wood ♀ / ♂’ |
|---|---|---|---|---|
| b. | hãlɔkɔɛ | hãlɔɛ | ‘jaguar ♀ / ♂’ |
Archaic Creek forms, preserved in tales where a female character is talking, are a case where a phonological rule applies in a prosodically-defined position within the phrase. “Women's forms end in a long vowel with a falling pitch-stress while the corresponding men’s forms shift the stress to the penultimate syllable, altering it to a high pitch-stress, and in addition shorten the long vowel and add /s/. If the shortened vowel is /i/, it is often dropped altogether” (Haas 1944: 145-146).
| (13) | okikâ | okíkas | ‘he was ♀ / ♂’ |
|---|---|---|---|
| ó·kickî· | ó·kíckis | ‘you are ♀ / ♂’ | |
| o·kakaŋkî | o·kakáŋks | ‘I meant ♀ / ♂’ (short i dropped) | |
| apo·kiphoykâ· | apo·kiphóykas | ‘let him stay ♀ / ♂’ |
In Kadiwéu (Sandalo 2011, non-noble men use binary moraic feet (14a), noble men ternary moraic feet (14b), and noble women binary syllabic feet (copying vowels to fill the syllabic pattern, 14c)).
| (14) | a. | Gokidi ['Go.'ki.di] |
|---|---|---|
| ‘in the afternoon (non-noble ♂)’ | ||
| b. | Gokidi ['Go.ki.di] | |
| ‘in the afternoon (noble ♂)’ | ||
| c. | necodi ['ne.e.'co.di] | |
| ‘man ♀’ |
{ NA | speaker | addressee | both | mixed }
See discussion above on Parameter 1 above, and examples (11) and (12) for <speaker> in Guarayo (Tupian, Bolivia) and Karajá (Nuclear-Macro-Je; Brazil).
{ NA | yes | no }
In our survey, we consider pronouns and demonstratives as grammatical categories rather than lexical ones.
{ NA | speaker | addressee | both | mixed }
{ NA | exclusively | non exclusively | no }
{ NA | yes | no }
{ NA | 1-5 | 6-20 | 21 and more }
{ NA | speaker | addressee | both | mixed }
{ NA | exclusively | non exclusively | no }
{ NA | exclusively | non exclusively | no }
{ NA | speaker | addressee | both | mixed }
Our systematic survey in a genetically diverse worldwide sample of 325 languages, with a focus in the Americas, shows that categorical genderlects are not a rarity, with 58 cases (18%). The 58 languages with genderlects are distributed over 13 isolates and 33 families. We confirm their very high prevalence in South America (28% of the South American languages of the sample), and their relatively still high prevalence in North America (15%) compared to outside the Americas (10%).
With regard to the domains targeted in the grammar, genderlects primarily target the lexicon (79%), then morphology (36%), and lastly phonology (12%).1 Expectedly, among the 17 languages which target two domains, the great majority (76%) have gender indexicality in both the lexicon and the morphology. It is noteworthy that none of the surveyed genderlects consist of two very distinct lects. If phonology is gender-indexed, then a few phonemes only are concerned, or limited suprasegmental features like phrase-final modification. If morphology is gender-indexed, the system maximally targets a handful of morphemes. If lexicon is, then almost three quarters of the languages have only 1 to 5 gender-indexed lexemes, a quarter have 6 to 20, and only one has more than 21.
A little over half of the genderlects surveyed exclusively encode the speaker gender (52%). Those exclusively encoding the addressee gender are the second most frequent (19%). Genderlects encoding jointly the speaker and addressee gender ('both' systems, Haas Type III) or mixed (depending on the gender indexing element) are less frequent (14 and 15% each, respectively), but still represent together almost a third of all types. Interestingly, our sample shows drastic differences inside vs. outside the Americas: systems based on the speaker gender are much more frequent in the Americas, while systems based on the addressee genders or on both the speaker and addressee genders are much more frequent outside the Americas. These figures need to be confirmed with a larger sample outside the Americas.
Gender indexicality seems to be highly associated with (inter)subjectivity: 88% of languages with categorical genderlects display gender indexicality exclusively in (inter)subjective elements. This figure is particularly high, especially if we take into account the fact that the (inter)subjectivity parameter is not relevant for genderlects in the phonological domain.
Conceptualization: Françoise Rose, Marine Vuillermet
Data collection: Natalia Chousou-Polydouri, Kellen Parker van Dam, Oscar Cocaud-Degrève, Anna Graff, David Inman, Diana Krasovskaya, David Timothy Perrot, Françoise Rose, Marine Vuillermet, Diana Krasovskaya
Supervision of data collection: Françoise Rose, Marine Vuillermet
Alberdi, Jabier. 1995. The development of the Basque system of terms of address and the allocutive conjugation. Amsterdam Studies in the Theory and History of Linguistic Science Series 4. 275–294. John Benjamins BV.
Antonov, Anton. 2015. Verbal allocutivity in a crosslinguistic perspective. Linguistic Typology 19(1). 55–85. De Gruyter Mouton.
Berg, Thomas. 2023. Gender marking in the first-person singular: A case of paradigm (in) consistency. Journal of Linguistics. 1–35. Cambridge University Press.
Chacon, Thiago Costa. 2012. The phonology and morphology of Kubeo: The documentation, theory, and description of an Amazonian language. University of Hawai'i at Mānoa dissertation. (xxi+376pp.)
Pury Toumi, Sybille de. 2003. Vice versa. Le genre en garifuna. In Landaburu, J. and Queixalos, F. (eds.), Méso-Amérique, Caraïbes, Amazonie Volume 2, 155–162.
Dunn, Michael. 2013. Gender determined dialect variation. In Corbett, Greville (ed.), The expression of gender, 39–68. Berlin: De Gruyter Mouton.
Fleming, Luke. 2012. Gender indexicality in the Native Americas: Contributions to the typology of social indexicality. Language in Society 41(3). 295–320. Cambridge University Press.
Fleming, Luke. 2015. Research Note: Speaker-referent gender indexicality. Language in Society 44(3). 425–434. Cambridge University Press.
Fortune, David & Gretchen Fortune. 1975. Karajá men's-women's speech differences with social correlates. Arquivos de Anatomia e Antropologia 1. 111–124.
Furfey, Paul Hanly. 1944. Men's and women's language. The American Catholic Sociological Review 5(4). 218–223.
Günthner, Susanne. 1996. Male–female speaking practices across cultures. In Hellinger, Marlis and Ammon, Ulrich (eds.), Contrastive Sociolinguistics, 447–73. Berlin: Mouton de Gruyter.
Haas, Mary R. 1944. Men's and women's speech in Koasati. Language 20. 142–149.
Hoeller, Alfredo. 1932. Grammatik der Guarayo-Sprache. Verlag d. Missionsprokura d. PP Franziskaner.
Lee, Jennifer. 1987. Tiwi Today: A Study of Language Change in a Contact Situation. (Pacific Linguistics: Series C, 96.) Canberra: Research School of Pacific and Asian Studies, Australian National University. vii+449pp.
Lyons, John. 1982. Deixis and subjectivity: Loquor, ergo sum?. In Jarvella, R. J. and Klein, Wolfgang (eds.), Speech, place, and action: Studies of deixis and related topics, 101–24. New York: Wiley.
Osborne, C. R. 1974. The Tiwi language: grammar, myths and dictionary of the Tiwi language spoken on Melville and Bathurst Islands, northern Australia. (Australian Aboriginal Studies, 55.) Canberra: Australian Inst. of Aboriginal Studies. xii+170pp.
Oyharçabal, Bernard. 1993. Verb agreement with non arguments: On allocutive agreement. Amsterdam Studies in the Theory and History of Linguistic Science Series 4. 89–89. John Benjamins.
Ribeiro, Eduardo Rivail. 2012. A grammar of Karajá. University of Chicago dissertation. (x+291pp.)
Rose, Françoise. 2013. Le genre du locuteur et de l'allocutaire dans les systèmes pronominaux: Genre grammatical et indexicalité du genre. Bulletin de la Société de Linguistique de Paris 108(1). 381–417.
Rose, Françoise. 2015. On male and female speech and more: categorical gender indexicality in indigenous South American languages. International Journal of American Linguistics 81(4). 495–537. University of Chicago Press Chicago, IL.
Rose, Françoise. 2018. A typology of languages with both grammatical gender and genderlects. In Fedden, Sebastian and Audring, Jenny and Corbett, Greville G. (eds.), Non-canonical gender systems, 211–246. Oxford: Oxford University Press.
Rose, Françoise & Peter Bakker. 2016. A typological survey of genderlects.
Sandalo, Filomena. 2011. Estratificação social e dialetos prosódicos no kadiwéu. In Silva, Giovani J. (ed.), Kadiwéu: senhoras da arte, senhores da guerra. Cuiabá: Editora CVR.
Traugott, Elizabeth Closs. 2003. From subjectification to intersubjectification. In Hickey, Raymond (ed.), Motives for language change, 124–139. Cambridge ; New York: Cambridge University Press.
Vallejos, Rosa. 2015. La indexicalidad de género en kukama-kukamiria desde una perspectiva tipológica. In Fernandez, Ana and Alvarez, Albert and Estrada, Zarina (eds.), Estudios de Lenguas Amerindias 3: contribuciones al estudio de las lenguas originarias de América, 199–225. Hermosillo (Mexico): Universidad de Sonora.
Bezooijen, Renée van. 1995. Sociocultural aspects of pitch differences between Japanese and Dutch women. Language & Speech 38(3). 253–265.
The percentages add up to over 100%, since genderlects may target more than one domain in a given language. ↩