Gender agreement in Italian compounds with capo-

2.1 Compounds with capo-

Compounds featuring capo- as their initial constituent have been productive in the Italian language since its earliest phases. The term capo originates from Latin căput, carrying different meanings, including ‘head’, in both physical and intellectual meanings, ‘extremity of something’, ‘initial part of a process or object’, and ‘essential or dominant element of a group or category’ (Micheli 2020:130–132).^[1]

In Medieval Latin, caput- appears within noun phrases with the second element in genitive case, emerging from a reanalysis of syntactic objects (Rainer 2021). They primarily denoted a person holding a leadership or authoritative position (e.g., caput ecclesiae. _GEN ‘church leader’). This semantic feature persisted in Old Italian, where compounds with capo- referred also to human referents possessing authority over a group (e.g., capofamiglia ‘head of household’) or a place (e.g., caporione, originally from capo del rione ‘district master’, nowadays meaning ‘ringleader’).^[2] As Micheli’s (2020:133–140) diachronic analysis shows, while originally coexisting with others meanings like ‘initial part of’ or ‘extremity of’ (e.g., capomese ‘first day of the month’) or ‘summit, top’ (e.g., caposcala ‘top of a stairway’), the denotation of a human referent became increasingly productive since the 16th-17th centuries, culminating in contemporary Italian, where this denotation is ‘firmly established’.

In contemporary Italian, we observe the structure capo-X identifying objects (as relics of Old Italian) and animate referents. Only the latter is a productive construction and therefore the one we focus on in our study.

Within this type of capo- compounds we can further classify between compounds where the second element is a noun referring to a group of animated referents both literally (e.g. capobranco ‘leader of the pack’) and metonymically (e.g., capostazione ‘station master’), and compounds where the second element is a professional noun (e.g., capocuoco ‘head chef’^[3]). The former group expresses a clear hierarchical relation between the constituents (i.e., subordinative compounds, where the elements composing the group are subordinated to the element capo), while the interpretation of the latter has undergone various semantic interpretations, placing these compounds into different classes:^[4]

1. subordinative compounds: capo- is interpreted as the syntactic/semantic head of the compound with the second element establishing a syntactic dependency with it (e.g., capo. _HEAD redattore. _COMPLEMENT ‘chief editor’ denotes the ‘chief of the editors’);

2. coordinative compounds: capo- and the second element belong to the same interpretative level (in the aforementioned example, capo. _HEAD redattore. _HEAD indicates a referent who is simultaneously capo ‘chief’ and redattore ‘editor’);

3. attributive compounds: capo is interpreted as the modifier of the head to the right of the compound (hence, capo. _MODIFIER redattore. _HEAD refers to a redattore ‘editor’ who is specifically the capo ‘chief’ among the others) (Micheli 2020).

As noted by Micheli (2016 and 2018, analyzing plural forms provides insights into the syntactic head of these structures. Regarding plural inflection, in fact, it is particularly interesting to notice that two constructions emerge: while compounds having capo-+group are clearly subordinate, and only inflect the first element in the plural^[5] (e.g., capo _SG.stazione ‘station master’ → capi _PL.stazione ‘station masters’), the same does not apply to the capo.redattore types, which appear both with external inflection (capo.redattori _PL) and double inflection (capi _PL.redattori _PL). Since exclusively internal inflection in the latter group is not attested^[6] (i.e., *capi _PL.redattore _SG), a subordinative interpretation appears not corroborated by number inflection, being the head the locus inflectionis. A coordinative reading would explain double inflection, but less satisfactorily external inflection.

More convincingly, both the external and double inflection can be explained by interpreting the second element as the head of the compound and capo- as a modifier, i.e., with an attributive reading. In both cases, the head bears the inflection. The lack of inflection on the modifier is common in Italian attributive compounds (e.g., parole._PL chiave._SG ‘key words’ vs stati._PL membri._PL ‘member states’, Radimský 2016: 197–198).

2.2 Grammatical gender in Italian

Italian morphology, as generally observed in Romance languages, features masculine and feminine gender distinctions for nouns denoting both animate and inanimate referents.^[7] While gender assignment for inanimate objects is arbitrary and not phonologically predictable (Thornton 2022: 19), it is based on biological factors for animate referents,^[8] with few exceptions to this pattern, such as soprano _M or sentinella _F ‘sentry’.

Gender is evident through agreement with determiners and modifiers (e.g., il._M cielo._M azzurro._M ‘the blue sky’ vs la._F nostra._F casa._F ‘our house’), and through inflectional morpheme substitution (e.g., il._M maestro._M ‘the teacher._M’ vs la._F maestra._F ‘the teacher._F’). However other possibilities exist.

As illustrated by Thornton (2022), animated nouns can be classified based on their morphological behaviour:

1. Symmetrical/asymmetrical nouns: symmetrical nouns, exemplified by the previous example maestro versus maestra, involve simple substitution of the inflection morpheme; asymmetrical nouns are those where the feminine form is created through a derivational suffix (e.g. il. _M professore. _M ‘the._M professor._M’ vs la. _F professor-essa. _F ‘the._F professor._F’).

2. Nouns with distinct lexical morphemes for masculine and feminine (i.e., “eteronimi”) (e.g., padre ‘father’ vs madre ‘mother’).

3. Nouns where gender is only visible on agreement phenomena (i.e., “nomi di genere comune”) (e.g., il. _M cronista ‘the._M reporter’ vs la. _F cronista ‘the._F reporter’).

4. Epicene nouns, which always appear with a specific gender (e.g., la. _F vittima. _F ‘the victim’ vs il. _M personaggio. _M ‘the character’).

Thornton (2022) underlines that while this classification captures general tendencies, social biases of speakers can alter them. In cases where a professional title is typically associated with men, speakers may employ different strategies when referring to a woman. In addition to a regularly inflected feminine form of the professional title, the feminin form can manifest solely as an agreement phenomenon on determiners and modifiers (i.e., “semi-marked forms”, Cignarella et al. 2021; Formato 2014, 2016), as in (1):

Another option is not to express the feminine at all, even when referring to women, interpreting the professional title as an epicene noun, as in (2):

2.3 Feminine forms in occupational titles

As we have seen, while grammatical gender for inanimate objects is a semantically arbitrary category, using such opposition might raise issues when the referent is human.

Sabatini’s (1985) pioneering work on occupational titles in Italian, highlighted how these constitute a crucial aspect of sexist language use. She extensively analyzed masculine and feminine agentive nouns, particularly focusing on occupation titles that could easily be inflected according to Italian grammatical rules but are systematically left in the masculine form. The lack of a feminine inflection, is based purely on social reasons (see also Serianni 1997: 85: “grammatical uncertainties on [the point of feminine occupational titles] depend on extra-linguistic reasons, i.e., by the social transformative process accomplished during this century”, our translation).

To illustrate this extralinguistic bias, Sabatini (1985) listed several examples where the feminine inflection would not cause any phonological or morphological difficulty, yet the feminine version was systematically lacking or used in a derogatory sense: e.g. ingegnere ‘engineer._M’, for females engineer instead of ingegnera ‘engineer._F’ (contrasting to infermiere/infermiera ‘nurse._M/F’, cameriere/cameriera ‘waiter/waitress’, etc.); ministro ‘minister._M’ but not ministra ‘minister._F’; sindaco ‘mayor._M’ but not sindaca ‘mayor._F’; magistrato ‘magistrate._M’ and not magistrata ‘magistrate._F’, among others. She underlines that women were allowed to pursue many of these careers (e.g., diplomatic or magistrate) only in the 1960s, which is why Italian lacked the corresponding feminine inflection.

Resistance to social and linguistic changes has been so strong that occupational titles have remained in masculine forms, even when the job was performed by women, sometimes simply adding the word donna ‘woman’ before the masculine form (e.g. la donna medico ‘the woman physician._M’ instead for la medica, ‘the physician._F’). Sabatini mentions particular cases of agent nouns ending in the vowel -e, which can be either masculine or feminine. Interestingly, even in these cases, the feminine determiner was avoided, such as il giudice ‘the._M judge._M/F’ instead of la giudice ‘the._F judge._M/F’.^[9]

Several reasons have been proposed to explain the resistance to inflect professional names according to gender, from uncertainty about the correct form^[10] to a perceived cacophony of the feminine ending, the belief in the neutrality of a generic masculine^[11] (cf. Gheno 2019; Robustelli 2012), to even the “questionable belief […] that masculine may confer a higher degree of ‘seriousness’ compared to its feminine counterpart” (Zarra 2017: 26).

These issues are far from trivial. Language reflects the conceptualization of gender as social category, and gender representation in language not only serves a distinctive function, but also communicates and reinforces values and prejudices, in what Kite and Deaux (1987) label the “gender belief system”.

The need for terms denoting women in certain positions has become increasingly urgent, creating a “complex two-way dialectic process […] whereby language items both affirm and contest the status quo, and changes in social structures necessitate the development of new vocabulary” (Mills 2003:88).

A notable case is the postulated unmarkedness of the masculine gender (Jakobson 1939, 1960) in gendered languages like Russian. These considerations can be extended to Italian, where the feminine stresses that the noun cannot designate a masculine referent (with the interesting exception of when this is used in pejorative language), while the masculine can be used neutrally to signal a lack of sex reference when unknown or irrelevant in the context. However, experimental research has demonstrated that the supposed ‘masculine neutral’ gender is not truly neutral. Even in gendered languages like Italian, where gender is not a salient element of information, we still tend to interpret the referent as gendered. Thus, the use of generic masculine both mirrors and contributes to maintaining gender biases and prejudices (Cacciari and Padovani 2007; DeFranza et al. 2020; Formanowicz and Hansen 2021; Gygax et al. 2009, 2012, 2021; Moulton et al. 1978; Prewitt-Freilino et al. 2012; Stahlberg et al. 2007; Wasserman and Weseley 2009): a ‘generic masculine’ activates the mental representation of a male referent, aiding a circularity in which gender conceptualization and linguistic expressions conveying it concur to reinforce one another (see Violi 1986). Gender in fact is “not only a grammatical category that regulates purely mechanical issues of concordance; it is, on the contrary, a semantic category that conveys a profound symbolism within the language” (Violi 1986: 41, our translation).

Numerous studies have explored the connection between sexist language and gender inequality in the job market (Bem and Bem 1973; Gaucher et al. 2011; Mastromoro 2014). Additionally, research has shown that inclusive language incorporating feminine gender can positively impact recruitment and job expectations (Gabriel et al. 2008; Horvath 2015; Lindqvist et al. 2019; Lorenzon 2022; Vervecken et al. 2013).

In the case of the Italian language, guidelines advocating for a non-sexist language have been circulating since the 1980s, despite resistance from political institutions and public demands (Azzalini and Giusti 2019; Cardinaletti and Giusti 1991; Gheno 2019; Giusti 2009; Giusti and Iannàccaro 2020; Marazzini 2017; Robustelli 2018; Zarra 2017).^[12] However, the gender inflection of the word capo ‘boss’ has posed more challenges than other cases.

2.4 The feminine form of capo

As with the aforementioned ‘prestigious’ occupations and titles, the grammatical gender of the word capo did not present uncertainties when Italian society did not witness women in leading positions, thus capo almost exclusively referred to a man. This supposed unproblematic nature of capo gender was hence aided by a structural gender imbalance. However, contemporary society expects to see women too in leading roles, and, as we have seen in the previous section, many occupational titles have undergone progressive feminization.

Despite social changes that have impacted language and active propositions to combat sexist language (such as the pioneering work by Sabatini 1986, but also Robustelli 2012; Serianni 1997 and institutional guidelines like those from Piemonte Regional Council 2016), the word capo has consistently been referred to as uninflected according to gender. Capo is still considered in its original meaning of ‘head’, thus with the masculine gender^[13] and the use of feminine capa has been deemed as “playful”, “colloquial” and “substandard” (Bisetto 2004; D’Achille and Thornton 2008), or even “pejorative” (Sabatini 1987).^[14] Although etymologically masculine,^[15] capo’s contemporary meaning as ‘boss’ or ‘leader’ stimulates interesting reflections when the referent is a woman.

Despite the prior assumptions, it is worth noting that the term capa is extensively used by speakers, not only within informal settings or for playful purposes. Journalism, for example, frequently employs the term, even in publications not aligned with the most inclusive demands of society. A search in the archive of La Repubblica, one of Italy’s most widely read newspapers, deemed ‘generalist’ (Fortunati and Sarrica 2004) or ‘polycentric’ (Vaglio 2006) and with a wide distribution across the nation (Gabellieri and Scaglione 2021) reveals its usage in expressions such as capa di gabinetto del sindaco (‘chief of staff of the mayor’, 11.05.2022), capa dell’istruzione comunale (‘head of municipal education’, 02.04.2022), capa dell’esecutivo europeo (‘president of the EU executive’, 20.01.2022), capa della Task Force (‘head of the Task Force’, 24.04.2021), and also the compound capa dipartimento (‘head of department’, 15.04.2021). Such occurrences cannot be attributed solely to a recent trend, as examples from over 20 years ago also exist, such as capa della Federcasalinghe (‘head of the Federation of Housewives’, 29.03.2001).

Furthermore, the neutral use of the term capa has been observed by Cleis (2000) in the campaign “Internship. The advantage of practice”, promoted in 1999 by the Federal Office for Vocational Education and Technology in Switzerland. Billboards displayed images of young girls stating Anche la mia capa è stata apprendista (‘my boss._F too was an intern’). In this case, no particular connotation was attributed to the use of capa. Cleis (2000) indicates that, despite not being officially accepted, capa was widely used among young people at the time of the study.^[16] Moreover, recent analysis examining the use of gender stereotypes and linguistic sexism in children’s books confirms the use of capa in Italian translations (Veronese 2022).

It is intriguing that a term subject to such debate lacks a thorough analysis investigating its use, perception and history in contemporary Italian.

3.1 When capo is an element of a compound

In Section 2, we observed that internal inflection is the sole possibility for capo- _SUB compounds, occasionally accompanied by external inflection for capo- _ATTR. type. Italian dictionaries do not account for gender inflection based on the human referent of the compound, consistent with the assumptions related to the word capo discussed earlier. Consequently, a female station master is argued to be referred to as a capo.stazione.

However, capo-x _SUB compounds conventionally remain uninflected in both gender and number when referring to women, despite being inflected in number when the referents are men (as generally recommended by dictionaries, see also Dressler 2006; Gheno 2018; Padrosa-Trias 2010; Telve 2011). Hence forms such as (3) should be prescriptively express:

In the case of attributive compounds these challenges are arguably less visible, since these compounds have the option of external inflection ((4a) and (4b) are almost equally present in corpora, see Micheli 2016), and capo- can be treated more easily as a prefixoid:

A lack of gender inflection poses challenges for native speakers striving for a language free of sexist biases.

Our study aims to investigate the usage of feminine capa- as a compound element by Italian native speakers, discerning between social resistance, etymological reasons, and morphological obstacles. To achieve this, we will analyze gender inflection of capo- in comparison to traditionally feminine and masculine occupational titles, the isolated word capo, and two types of capo- compounds: capo-_SUB and capo-_ATTR.

4.1 Materials

The inflection of capo- as the first constituent of subordinate and attributive compounds was compared with that of i) words without capo firmly established in their feminine inflection because referring to occupational titles traditionally occupied by women; ii) words without capo firmly established in their masculine inflection because referring to occupational titles traditionally occupied by men; iii) the word capo in isolation to verify differences in its use as a free form versus within a compound. As said, the word capo (both in isolation and within compounds) refers to a leading position, and consequently it is traditionally associated to men. The five categories of nouns with examples are shown in Table 1.

Occupational titles traditionally associated to women were selected based on their well-established feminine inflection in the literature (Proudfoot and Cardo 2005; Cortelazzo 2017; Frabotta 2022; Giani 2019; Giusti and Iannàccaro 2020; Gheno 2019; Ricci 2021; Robustelli 2000; Sabatini 1987; Sulis and Gheno 2022).

Titles traditionally associated with men were selected for their resistance to accepting feminine inflection, as identified by Miglietta (2022) and Zarra (2017).

Professional titles with potential double endings (e.g., avvocatessa/avvocata ‘lawyer._F’) were excluded because of theoretical considerations (see footnote 11) and experimental findings suggesting gender bias is linked to specific suffixes (Merkel 2013; Merkel et al. 2012; Mucchi Faina and Baino 2006).^[17]

Into our sample, we included both subordinate and attributive compounds that appear among the 900 most frequent compounds with capo- as the first element in the itWac corpus. The second elements of the subordinative compounds were either masculine or feminine; the second elements of the attributive compounds were words whose feminine form is inflected both with or without a derivational suffix. Each category comprised eight words, as described below. The complete stimulus set is shown in Table 2.

4.2 Procedure

The test was carried out through a web-hosted survey platform. At the onset, respondents received information that they participated in a study on feminine occupational titles, and that they would be presented with sentences in which a male referent would be mentioned which they needed to replace with the corresponding word for the female referent. An example of a sentence is given in (5):^[18]

Each category’s target words were equally split between plural and singular forms, with their syntactic roles balanced between subject and direct object. All target words were preceded by a definite article.

In Italian compounds with capo- are sometimes written juxtaposed (e.g., capo redattore), univerbated (e.g., caporedattore) or hyphenated (e.g., capo-redattore) (Micheli 2016: 21). This graphic variation may introduce a potential confounding factor (because the element capo is sometimes isolated and other times not (Juhasz et al. 2005; Marelli et al. 2015), potentially affecting its inflection. To mitigate this, sentences were presented auditorily using a natural-sounding text-to-speech synthesizer.

Participants could listen to the sentences as many times as needed, with no time constraints on their responses. They were instructed to rely on their intuitions, with responses not judged as right or wrong. After listening to a sentence like (5), they completed a written sentence such as (6) with the corresponding noun in its feminine form, preferably including the preceding definite article:

After receiving instructions, participants provided information about their occupation, language background and proficiency, educational attainment, gender, and age. Subsequently, the sentences were presented in a randomized order. Each participant listened to sentences containing two out of the eight targets within each category, totaling 10 sentences. Following this, participants indicated on a five-point scale: i) their awareness of the ongoing debate on inclusive language, ii) the frequency of their own use of inclusive language, and iii) the importance they attributed to inclusive language (see Appendix 2).

4.3 Analysis

The analysis aimed to provide a description of the forms that the participants used to refer to the occupational roles in feminine form. Participants had various options for marking grammatical gender: they could omit marking altogether, mark it on the target noun, the definite article, or both. For plural nouns, they could use either a plural or singular form. Furthermore, there was an asymmetry between marking in subordinate and attributive compounds. In subordinate compounds, participants had limited choice, primarily marking gender or number on the first constituent (capo-), while in attributive compounds, they had the flexibility to mark on either or both constituents. This diversity of possible responses resulted in varied participant choices. The grammatical gender (masculine/feminine) and number (singular/plural) of the responses were manually categorised. This was performed for both the determiner and the noun. Additionally, the gender and number of the first and second constituents of subordinate and attributive compounds were categorized separately. Consequently, for these compounds there was a categorisation for the first constituent, another one for the second, and a third one for the two together (i.e., feminine when at least one of the constituents was inflected according to feminine gender).

The analysis focuses on the distribution of the gender and number variables. For gender, the frequency of gender marking on the noun and on the article within each of the five noun types was examined. Furthermore, the effects of syntactic function in the sentence (subject/direct object), number (singular/plural), and three participant background variables (attitude towards gender-inclusive language, age, and gender) were tested.

The analysis of number was limited to items presented in plural in the test, particularly focusing on the two constituents of the compounds. Differences between the five noun types and any potential effects of syntactic function (subject or direct object) and participant background variables were considered, but not for number.

The statistical analysis is a mixed effects logistic regression model, with gender of the noun and that of the article as the outcome variables and the explaining variables described above as predictors. The predictor awareness of gender-inclusive language was estimated as the sum of the three ratings of the three statements at the end of the questionnaire. We performed this analysis once for the gender of the articles and the nouns, and a second time for the element capo in isolation and as the first constituent of the compounds. Additionally, we asked whether the participants marked gender of the attributive compounds on the first element (e.g., capacuoco), the second (e.g., capocuoca) or on both (e.g., capacuoca).

4.4 Results

The questionnaire was completed by 192 respondents, including 134 women, 57 men, and one participant who did not identify as either male or female. All were native speakers of Italian, with six reporting a second native language (i.e., Croatian, Sardinian, English, Spanish, Turkish or German). Their age ranged from 19 to 74 years, with an average of approximately 44 years. 168 respondents lived in Italy at the time of the survey, the remaining 24 reported living in another country. The highest educational level obtained by 32 participants was middle school or high-school, while the remaining 160 had obtained some form of university degree. The participants’ responses towards gender-fair language were generally favourable. Over 97 % reported having at least some awareness of gender-fair language, more than 90 % used it at least every now and then, and more than 85 % had a neutral or a positive attitude towards it. The attitude towards gender-fair language correlated weakly with age (younger participants had a somewhat more favourable attitude towards gender-fair language), with participant gender (female participants had a somewhat more favourable attitude towards gender-fair language) and academic degree (attitude towards gender-fair language was somewhat more favourable in participants with a university degree).

The total number of responses was 1920. However, not all responses, could be used for the analysis. These included 327 responses where the determiner was missing, and 190 responses with nouns other than the target nouns (e.g. responsabile instead for capo/capa, or multiple possibilities like “capa or capo”) in the stimulus sentences.

Table 3 shows the proportions of determiners and nouns inflected for number and gender. There was considerably more variation in the gender of the nouns compared to determiners, which were predominantly feminine. As expected, the results for morphologically simple words revealed a scalar inflection (occupations traditionally linked to women → occupations traditionally linked to men → word capo in isolation).

In the case of compounds, there was a substantially higher percentage of feminine inflection for attributive compounds compared to subordinate compounds. This is due to the exclusive consideration of the lemma: In attributive compounds, respondents had the option to inflect only the second element, leading to a higher percentage of feminine inflection. The difference between subordinate and attributive compounds stems from the tendency for inflection to affect the rightmost element in attributive compounds, while subordinate compounds typically undergo internal inflection on the capo element, which is more resistant to inflection (as seen in capo in isolation and the type capocuoca in Table 7).

We tested the effects of syntactic function and number together with respondent age, gender, and attitude towards gender-fair language on the gender of determiners and nouns. The results are shown in Table 4. The values in this table represent the predicted proportions (on a logarithmic scale) of feminine gender for the five noun categories and the effects of the five predictors on these proportions. The only predictor with a significant effect was respondents’ attitude towards gender-fair language. Respondents who scored higher on this variable, were more likely to use the feminine form of the determiner or of the noun.

The impact of attitude towards gender-fair language is illustrated in Figure 1, showing the predicted proportions of feminine responses for the five stimulus categories across the range of responses to the gender-fair questions.

Figure 1:

Proportions feminine nouns predicted by attitude towards gender-fair language.

Table 5 shows the proportions of feminine forms for capo in isolation and as the first constituent of subordinate and attributive compounds. The proportion is obviously highest when capo was used in isolation and considerably lower in the two compounds. Interestingly, the proportion of feminine responses appears to be affected by the stimulus number, that is the proportions of feminine responses were higher in plural. Table 6 presents the estimates of the effects of the same predictors used in the previous analysis, indicating that the effects of number and attitude towards gender-fair language were both significant.

Table 7 shows how often the first and the second constituents of attributive compounds were marked for gender. As expected, the table shows that feminine gender was more commonly marked on the second constituent. Specifically, it was predominantly marked on only the second constituent, less frequently on both the first and the second constituent, and least frequently on only the first constituent. In singular forms, the first constituent (capo-) was marked as feminine in 12 out of 186 cases, while in plural forms, it was marked as feminine in 55 out of 177 cases. Noun number, therefore, appears to have an effect on the gender marking of capo- in these compounds. For the second constituent, there was a less pronounced number effect. This constituent was marked as feminine in 132 out 186 singular cases and in 133 out of 177 cases.

In sum, the results reveal significant differences in the gender usage across the five noun categories examined. Participants consistently employed feminine forms for occupations traditionally performed by women, while only roughly 15 % of the time when that occupation was expressed by a subordinate compound with the word capo- as its first constituent. Respondents with a more favourable attitude towards gender-fair language demonstrated a higher likelihood of employing feminine forms.

The feminine form capo was relatively frequent when used in isolation,^[19] but less when it constituted the first element of a compound. Interestingly, participants exhibited a higher propensity to use feminine forms for plural items compared to singular ones, and again, respondents with a favourable view on gender-fair language were more likely to use the feminine form than those who had a less favourable view. Regarding attributive compounds, respondents preferably marked feminine gender on the second element and less often on the first. When marking feminine gender on the first element, they were more likely to do so for plural than for singular compounds.