Grammatical gender reversals: A morphosyntactic and sociopragmatic analysis

2.1 Positive evaluation

Across languages, grammatical gender reversals are commonly used for positive evaluation. For example, endearment, prestige, and solidarity can be thusly expressed.

2.2 Negative evaluation

Grammatical gender reversals can also be used productively for negative evaluation, for example, to show distress or derogation.

2.3 Data summary and questions

Below, I present a summary of the data with regard to the meanings and forms of grammatical gender reversals across languages.

3.1 Distributed morphology

The first work within the DM framework was a doctoral dissertation by Bonet (1991) on Catalan opaque clitics, which was followed by a well-known article by Halle and Marantz on DM and the pieces of inflection (1993). The DM framework was further developed in Halle et al. (1997), Marantz (1997, 2001), Borer (2005), Embick and Marantz (2006), Embick and Noyer (2007), Acquaviva (2009), Matushansky and Marantz (2013), Matushansky (2013), Embick (2012), Kramer (2012, 2015), and Bobaljik (2017), among many others. DM adopts the basic organization of generative grammar (Chomsky 1995, 2000, 2001), adding the level of Morphology as the interface between syntax and phonology, as diagrammed in (19).

(19)

The central claim of DM is that the relationships between morphemes are structurally identical to the relationships between words. In DM, there is no centralized Lexicon. The Lexicon in the traditional sense is “distributed” across the grammar in various lists (underlined in the structure in (19)): (i) the Formative List (bundles of features), (ii) the Exponent List (vocabulary items), and (iii) the Encyclopedia (a list of idioms). The Formative List comprises bundles of semantic and syntactic features that enter the syntactic computation. These bundles of features lack any morpho-phonological content. The exponent list contains vocabulary items that associate morphophonological content with bundles of features (in other words, they are “exponed”). Vocabulary Insertion occurs at spell out, only after all syntactic operations are over. The Encyclopedia contains a list of idioms in the language or “Encyclopedia entries” that relate vocabulary items to meanings.

DM distinguishes between word formation from √roots (the notation √ is from Pesetsky 1995) and that from syntactic categories (Josefsson 1995, 1997; Marantz 2001; Embick and Marantz 2006; Embick and Noyer 2007; Matushansky and Marantz 2013, among others). √Roots are category neutral, but they can never appear “bare,” as they must be categorized by combining with a category-defining head (a process called “Lexical Decomposition”). Examples include the “little” n, a, or v that form nouns, adjectives, or verbs, respectively, as shown in (20). The division between √roots and syntactic categories provides us with formal tools for handling the morphosyntactic processes that happen inside a word.

(20)

3.2 The location of gender

Grammatical gender is a system of nominal classification based on the agreement patterns to which the genders give rise (see, e.g., Aronoff 1994, Corbett 1991, Hockett 1958).

Previous works on gender in different languages have proposed that gender is a feature on n (Lecarme 2002 for Somali; Ferrari 2005 and Kihm 2005 for Bantu and Romance; Lowenstamm 2008 for French and Yiddish; Acquaviva 2009 for Italian; Kramer 2012 for Amharic).

I follow Kramer’s (2015) proposal within the framework of DM that gender features are located on n and come in two different types: (i) interpretable, i[+/−F], for natural gender, and (ii) uninterpretable, u[+/−F], for arbitrary gender, as listed in (21). In this system, the “plain” n, as in (21c), has no gender feature and results in morphological default. In the data described above, we are dealing with human nouns with natural gender; for this reason, the interpretable gender features i[+F] (for females) and i[−F] (for males) are most relevant for the current work.

In the framework of DM, roots are category neutral and have no grammatical features (Borer 2005; Acquaviva 2009; Embick and Noyer 2007; Embick 2012; Kramer 2015, among others). According to Kramer (2015), every language that has a natural gender interpretation must have the interpretable i[+F] and i[−F] features located on n, as schematized in (22).

(22)

In this account, a language like Manambu, which has a natural gender interpretation but no overt gender marking on nouns, would be analyzed as nonetheless having the interpretable i[+F] and i[−F] features, as in (23).

(23)

3.3 The location of evaluation

Following Steriopolo (2008, 2014, 2015, 2016, 2017a, b, 2018a, b), evaluative suffixes across languages are specified for the grammatical feature [eval] “evaluation.”^[6] The suffixes differ in terms of their manner and place of syntactic attachment (see also Wiltschko and Steriopolo 2007, and Kramer 2015).

With regard to the manner of syntactic attachment, some evaluative suffixes merge as syntactic heads while others merge as syntactic modifiers. The distinction between heads and modifiers lies in the projection of category features (Schütze 1995; Bierwisch 2003; Bachrach and Wagner 2007, among others). Heads project, meaning that they can determine the syntactic category and/or grammatical features of the output (e.g., number, gender, noun class), as in (24a). In contrast, modifiers do not project. As such, they cannot determine the syntactic category or grammatical features of the output, as in (24b).

(24)

In regard to the place of syntactic attachment, evaluative suffixes can attach either to roots, as in (25a), or to various syntactic categories (already categorized roots), as in (25b), in order to form evaluative nouns (see also De Belder et al. 2014 on low and high diminutives in Italian and Hebrew, and Cinque 2015 on the ordering of evaluative heads in the extended nominal projection).

(25)

Consider, the Russian example (1), repeated in (26) for convenience. It contains an evaluative diminutive suffix -ok that attaches to a female first name, changing the grammatical gender of the noun from feminine to masculine: Liza “Liza (f)” – Liz-ok “little Lizzy (m).”

With respect to the manner of syntactic attachment, the suffix -ok behaves like a syntactic head as it can project. More specifically, it determines the grammatical gender of the output, which is masculine in Russian.

Regarding the place of syntactic attachment, I assume that the suffix is attached to a nominal category (an already nominalized root). The reason for this assumption is that the resulting derivation Liz-ok “little Lizzy (m),” although grammatically masculine, still has a female interpretation, because in Russian it is a female name and can only refer to a female, never to a male. Thus, following Kramer (2015), it must have an interpretable feature i[+F] on n, which triggers the female interpretation, as structured in (27).

(27)

In the structure in (27), the suffix -ok is a syntactic head which is merged above the noun Liza, thus creating the evaluative affectionate (affect) form Liz-ok “little Lizzy (m).” I propose that the suffix –ok is specified for the following morphosyntactic feature bundle, (28).

The suffix –ok productively forms evaluative derivations in Russian (Stankiewicz 1968: 109–13) and always triggers masculine agreement, no matter what the gender of the base is (it is the structurally highest feature that determines the grammatical gender of the whole nominalization; see Steriopolo 2008, Kramer 2009, Steriopolo and Wiltschko 2010, and Kramer 2015). For example, in (29), it attaches to male (i), female (ii), and unisex (iii) first names, creating masculine evaluative derivations with the meaning of endearment.

A structure for the evaluative form Vany-ok “little Vanya (m),” as in the data in (29i) above, is proposed in (30). Here the suffix –ok is attached to the male name Vanya “Vanya (m),” forming an evaluative noun Vany-ok “little Vanya (m) (affect).”

(30)

Thus, we notice in the Russian examples in (29) that the meaning of the evaluative suffix -ok remains constant, expressing endearment, and it does not depend on the gender of the base to which it attaches. There are, however, languages in which the evaluative meaning depends directly on the gender of the base. For example, as briefly described in Section 2.2.2, the language Tigre has the singular feminine evaluative suffix -ät and masculine evaluative suffix -ay. When a suffix attaches to a base whose gender coincides with the gender of the suffix, this results in a positive evaluation (diminutive (dim) meaning). However, when the gender of the suffix and that of the base do not coincide, the result is a negative evaluation (pejorative (pejor) meaning).

Consider first the data in (31) and (32) with the feminine suffix -ät (the data are from Palmer 1962, cited in Stump 1993: 11).

In (31b), the feminine suffix -ät is attached to the feminine base ’әssit “woman (f)” and the resulting evaluative form is diminutive: ’әssit-ät “woman (dim).” In (32b), the same suffix is attached to the masculine base ’әnas “man (m)” and the evaluative form is pejorative: ’әnes-ät “man (f) (pejor).” The proposed structures for (31b) and (32b) are presented in (33) and (34), respectively.

(33)

(34)

As indicated in the structures in (33) and (34), I propose that the evaluative suffix -ät is specified for the morphosyntactic feature bundle [eval][+F], as in (35) because it productively forms evaluative forms that trigger feminine gender agreement. However, the type of evaluative meaning (positive or negative) it expresses seems to directly depend on the gender of the base to which it attaches, a phenomenon that will be discussed a bit later in this section.

Consider now the data in (36) and (37) with the masculine evaluative suffix -ay. The way in which this suffix functions is a mirror image of the feminine suffix -ät, discussed above, namely, when the suffix -ay is attached to a feminine base, the resulting form has a pejorative meaning, as in (36b). However, when it attaches to a masculine base, the resulting form is diminutive, as in (37b).

Since the suffix -ay productively forms evaluative nouns which trigger masculine agreement, I propose that it is specified for the feature bundle given in (38).

The proposed structures for (36b) and (37b) are given in (39) and (40), respectively. In (39), the masculine evaluative suffix -ay is attached to the feminine base ’әssit “woman (f),” forming a pejorative noun ’әssit-ay “woman (m) (pejor).” In (40), it is attached to the masculine base ’әnas “man (m),” forming a diminutive noun ’әnes-ay “man (m) (dim).”

(39)

(40)

Thus, similar to the feminine suffix -ät, a match between the gender of the suffix -ay and that of the base results in a diminutive interpretation while a mismatch in gender results in a pejorative interpretation (Palmer 1962, cited in Stump 1993: 11). In this context, the following question arises: How can we account for the dependency of the evaluative meaning of the suffixes on the gender of the base to which they attach?

I follow Acquaviva’s (2009) approach to a constructional treatment of meaning, according to which “lexical” meaning arises in a syntactic construction. Thus, to understand how the same suffix can express fundamentally different (even opposite) evaluative meanings, the notion of syntactic context becomes relevant.

Kramer (2015) proposes the notion of “semantic licensing conditions” with respect to the treatment of roots. She states that

…the n that a root combines with has an impact on semantic interpretation (…) Because these licensing conditions affect interpretation, (…) they are encoded in the Encyclopedia as conditions on the semantic interpretation of a root in a context. (Kramer 2015: 51)

For example, a root like “mother” is interpretable in the Encyclopedia only in the context of a n [+F], as shown in (41) below. If this root is used in a different context, for example, in the context of a n [−F], the Encyclopedia won’t be able to interpret the derivation, which will cause it to crash.

I would like to suggest extending Kramer’s treatment of roots to account for evaluative affixes. The licensing conditions encoded in the Encyclopedia can affect semantic interpretations of evaluative affixes in a context. For example, the feminine evaluative suffix -ät in Tigre receives a negative interpretation (pejorative) in the context of a n [−F] and a positive interpretation (diminutive) in the context of a n [+F], as presented in (42). In a similar way, the masculine evaluative suffix -ay is interpreted as pejorative in the context of a n [+F] and as diminutive in the context of a n [−F], as in (43).

Furthermore, I propose that the structure as presented in (44) is universal for evaluative gender reversals across languages. In (44), an evaluative gender reversal projects an evaluative head, n2 [eval], above n1. The evaluative head [eval] is specified for the gender features [+F] or [−F], and thus it is capable of changing the grammatical gender of the base to which it attaches.

(44)

Languages, however, differ with respect to whether the head n [eval] is spelled out morphologically or not. Consider, for example, the data from Manambu. The language has no overt gender morphology on nouns, but grammatical gender agreement is visible on words agreeing with nouns (Aikhenvald 2012), as shown in (45). In (45a), the noun du “man” has no overt gender morpheme, but it triggers masculine agreement with the agreeing adjective numa “big” and the demonstrative ke “this.” In (45b), the noun ta:kw “woman” also has no overt gender morphology, but it triggers feminine gender agreement.

As already described in Section 2.2.2, gender reversals in Manambu express a pejorative attitude of the speaker, as shown in (46). To the best of my knowledge, the language does not apply gender reversals to express a positive attitude (unlike Russian and Tigre, discussed above).

I propose that similar to the discussed data from Russian and Tigre, Manambu reversals project an evaluative head, n [eval]. However, unlike in Russian and Tigre, the evaluative head in Manambu is not spelled out morphologically. The morphologically null head n [eval] is nonetheless specified for the gender features [+F] or [−F] and, thus, it can change the gender of the base to which it attaches, as evidenced by the grammatical gender agreement with the adjective and demonstrative in the data in (46). Syntactic structures for the gender reversals in (46a, b) are proposed in (47a, b) below.

In (47a), the noun du “man” has the interpretable gender feature i[−F], as it has a male interpretation. The morphologically null head n [eval] is projected above that noun and is specified for the feature [+F]. The resulting evaluative noun du “man (f); (pejor)” triggers feminine gender agreement, as in (46a) above. A parallel structure for the female noun ta:kw “woman” is given in (47b). The noun is specified for the interpretable gender feature i[+F], as it has a female interpretation. The evaluative head n [eval] is projected above the noun and contains the feature [−F]. As a result, the evaluative noun ta:kw “woman (m)(pejor)” triggers masculine gender agreement, as in (46b) above.

(47)

This analysis is to some extent similar to that in Mathieu (2012) who analyzes the singulative in the Algonquian languages Fox and Ojibwe. According to Mathieu (2012: 653), “the singulative is a process by which a collective or a mass noun[…] is turned into a unit.” Some languages can use derivations from a mass noun to a singulative noun via a gender shift. For example, in Fox and Ojibwe, a gender shift (from inanimate to animate) occurs, targeting mass and collective nouns and turning them into units. Mathieu (2012) proposes that the singulative is encoded in the head Div “division,” which is projected above an nP, as shown in (48).

(48)

The difference between the languages Fox and Ojibwe is that in Fox, the head Div contains an overt gender suffix (the animate suffix -a), while in Ojibwe, this suffix is morphologically null.

The structures in (49a, b) below illustrate how a mass noun is turned into a unit of measure in Fox, according to Mathieu (2012). In (49a), the mass noun “money” is undivided (the head Div is not projected) and the noun has an inanimate suffix -i. In (49b), the head Div is projected, and the gender shift occurs from inanimate to animate, as evidenced by the animate suffix -a. The resulting noun zhooniyaah-a “coin” with the meaning “a unit of measure”. Thus, the mass noun “money” becomes a unit of measure “coin.”

(49)

As Mathieu (2012: 670) claims, the same derivation takes place in Ojibwe. The language also uses a gender shift from inanimate to animate to mark singulativization. However, unlike in Fox, the singulative in Ojibwe is not morphologically visible. In (50a), the head Div is not projected, and the mass noun “money” is undivided. In (50b), the head Div is projected, and a gender shift occurs from inanimate to animate represented by a morphologically null morpheme -ø.

(50)

3.4 Summary

I have proposed that three different forms of evaluative gender reversals across languages, as shown in Table 3 above, have the same syntactic structure, as presented in (51), in which the syntactic head n [eval] is projected above a noun with interpretable gender features.

(51)

The evaluative head can be specified for the gender features [+F] or [−F]. When a mismatch in gender features occurs between the gender of the evaluative head and that of the noun to which it attaches, we observe a change in the grammatical gender of the base. I have also shown that languages differ with respect to whether the head n [eval] is spelled out morphologically.

4.1 A sociopragmatic framework

As the crosslinguistic data in Section 2.1 show, the evaluative meaning cannot be an entailment of the interpretable gender features i[+F] or i[−F]. For example, in the Manambu examples (45a, b) above, we observe the nonevaluative sentences in which the nouns “man” and “woman” trigger masculine and feminine agreement, respectively. In (46a, b), there are gender reversals with evaluative meanings. Thus, it is not the gender features themselves, but their reversed use that produces an evaluative effect. The question arises: How can we account for this effect?

I suggest that the evaluative effect of gender reversals can be accounted for by the sociopragmatic principles developed in Acton (2014).

The author proposes a sociopragmatic framework which interrelates semantics, pragmatics, and sociolinguistics. Consider some basic principles of this framework. First, the Violations of Expectations (VE) principle, as in (52).

The principle suggests that when the speaker violates conversational expectations, this should have a special significance for the hearer. The VE principle is a broadening of Horn’s (1984) neo-Gricean framework. Horn (1984) analyzes exclamativity of demonstratives and proposes a principle of Division of Pragmatic Labor (DPL) based on the notion of markedness (marked forms convey marked meanings), as in (53).

It is important to note that while the VE principle signals that an utterance has a special significance, it does not tell us what it is, which brings us to the principle of Full Significance (FS), as in (54).

First, the FS principle is intended to capture the role of context, where the context includes not only details of a conversational situation, but also expectations regarding this situation, the broader ideologies and beliefs of discourse participants, etc.

Second, FS suggests that the full significance of an utterance is dependent on its relation to relevant alternatives (or “pragmatic alternatives”). In other words, it is a comparison of the actual utterance the speaker used with those which the speaker could have used in a given situation.

The next principle, Differential Importance of Different Alternatives (DI), states that the various alternatives have different degrees of importance, which is determined in relation to conversational expectations, as stated in (55).

4.2 An analysis

Consider the Manambu data with gender reversals in (46) (the data are repeated in (56) for convenience). These data have been selected for an analysis here because of the detailed description of the context available in Aikhenvald (2012: 53–4).

The author observes that such data can only be uttered in a casual conversation (not at a formal village meeting). The feminine gender in (56a) can be used to refer to a small or fat man, but it can never be uttered directly to that man’s face, as it is considered extremely insulting and demeaning. Aikhenvald (2012: 54) describes a situation in which a man was referred to as “woman,” because he displayed “inappropriate” social behavior. The man remained in his wife’s village after getting married instead of taking her to his village, which is the normal practice in the Manambu culture. He breached that tradition and demonstrated a social behavior which is associated with women. Thus, as Aikhenvald (2008, 2012) observes, in Manambu, a mismatch in gender can reflect a “culturally inappropriate situation.”

Consider now the data (56b). These data can be used to refer to a woman who is too boisterous, large in size, or “too big for her boots” (Aikhenvald 2012: 54). It expresses a derogatory attitude toward the woman and would not be said directly to that woman. For example, Aikhenvald (2012: 54) describes a situation in which the utterance in (56b) was used to refer to a woman who sported knowledge of totemic names, which is traditionally the province of men. Another situation occurred in the Iatmul village of Palimbei, when a girl accidently saw men blowing long flutes in a fenced-off enclosure, which a woman is not allowed to see. As a result, “[…]she was subjected to scarification and a shortened version of male initiation” (Aikhenvald 2012: 54). Aikhenvald (2012: 55) also observes that “[…]the girl felt degraded and shamed, despite the fact that she had gained what was considered important ritual knowledge.”

According to the principles of the Division of Pragmatic Labor (Horn 1984: 22) and the Violations of Expectations (Acton 2014: 38), the data in (56a) and (56b) can be interpreted as having a special significance (or conveying a “marked message” in Horn’s 1984 terms). Here the gender markings do not correspond to the natural genders of the referents. The mismatch between the grammatical and natural genders violates the hearer’s expectations as to what a normal or appropriate utterance should sound like. As a result, such an utterance is interpreted as having a special significance.

The Full Significance principle (Acton 2014: 35) captures the role of context, whereby the notion of a context includes not only a certain conversational situation but also cultural ideologies and beliefs. As we have seen from the descriptions above (Aikhenvald 2008, 2012), in the Manambu society, men and women are expected to maintain a strict division of social roles. Thus, breaching these roles is looked down upon by other members of the society and can be a reason for mockery and offense. As a result, referring to a man as if he were a woman and vice versa expresses a highly negative attitude.

Additionally, the Full Significance principle states that the full significance of an utterance depends on its relation to relevant alternatives. The relevant alternatives of the Manambu data in (56a) and (56b) are the sentences in (45a) and (45b). In these sentences, a woman is referred to with the feminine gender and a man with the masculine gender. There is no mismatch in gender, and the meanings are nonevaluative. Thus, if we compare the derogatory utterances in (56a, b) with their nonevaluative (neutral) alternatives in (45a, b), the observed difference in meaning stems from the speaker’s evaluation vs the lack thereof (corresponding to Potts’ 2007 notions of expressive vs descriptive content).

The next principle discussed above, the Differential Importance of Different Alternatives principle (Acton 2014: 36), states that different alternatives have different degrees of importance. This means that not all possible alternatives will receive equal consideration. The importance of a given alternative varies directly with how well it squares with conversational expectations both conceptually and in terms of form. The more aligned with expectations and similar to the observed form the alternative is, the more likely it is to receive consideration. In the case of gender reversals (evaluative forms), for example, as in (46a, b) above, the alternative that is most aligned with expectations is the one without a gender reversal (the neutral form), as in (45a, b). The neutral alternative would be highly relevant to interpretation because it would be both (i) generally expected and (ii) similar to the evaluative utterance.

The importance of the evaluative alternative with gender reversal can be observed in the social behavior of the speakers of Manambu, described in detail in Aikhenvald (2012, 2016). The evaluative expressions with gender reversals as in (46a, b) are so insulting that the speakers tend to avoid using them in the presence of the referents. Compared these with the neutral utterances in (45a, b), which have no derogatory effect and can be uttered in the presence of the referents.

4.3 A summary

Three principles developed in Acton (2014) allow us to decode the non-entailed evaluative meanings of grammatical gender reversals as follows: (i) They violate the hearer’s expectations and thus, they are interpreted as having a special significance (the VE principle), (ii) they are uttered in a sociocultural context in which breaching the social gender roles is highly inappropriate (the FS principle), and (iii) they are compared to neutral forms without gender reversals as the ones that are most aligned with conversational expectations both conceptually and in terms of form (the FS and DI principles).

5.1 Data

A change in the grammatical gender of nonhuman animate and inanimate nouns is crosslinguistically used for two main purposes: first, to express the speaker’s attitudes and emotions, and second, to indicate unusual or unexpected properties of the referent (e.g., an unusual size or shape).

First, consider data that express the speaker’s emotions. For example, in Oromo (an East Cushitic language, spoken in Ethiopia and Kenya), a change in gender can indicate the speaker’s attitude toward the referent (Clamons 1995: 392). In (57a), the noun sareé “dog (f)” triggers feminine grammatical agreement. In (57b), a masculine suffix is used, creating an evaluative form sareé-n “dog (m),” which expresses the speaker’s negative attitude toward the dog (“nasty dog”).

In Benchnon (or Bench; a dialect of an Omotic language, spoken in southwestern Ethiopia), the speaker changes the usual masculine gender of the noun tʃ’āmá-ī “shoe (m),” as in (58a), to feminine tʃ’āmá-ā “shoe (f)” in order to express a derogatory attitude (the speaker does not care that the shoe was lost), as in (58b).

Now consider data that indicate unusual or unexpected properties of the referent. In Halkomelem (a Central Salish language spoken in southwestern British Columbia, Canada), the word šeł “road” is masculine (the default gender in the language), because roads are perceived as long, thin, and rigid (Gerdts 2013: 423–4), as in (59a). However, this noun can also be marked as feminine when a road is perceived as unusually curvy, as in (59b).

In Alamblak (the Sepik Hill family, spoken in the Angoram District of East Sepik Province, Papua New Guinea), inanimate references that are short, squat, and wide are usually feminine, while long, slender, and narrow referents are masculine (Bruce 1984: 97). A change in gender is used to indicate an unusual size of the referent. For example, the noun for “house” is normally used with the feminine suffix -t, as in (60a), but the masculine suffix -r can also be applied if the house is perceived as unusually long, as given in (60b).

An example from Manambu is presented in (61). Here a change in grammatical gender indicates the increasing size of a pregnant woman’s belly (Aikhenvald 2008: 118). The noun ya:l “stomach, womb” is usually feminine in Manambu because of its round shape. However, a very large size can be indicated by using the masculine gender. Thus, at the beginning of the sentence in (61) the noun ya:l “belly” triggers feminine agreement with the adjective numa-ø “big (f).” The agreement later changes to masculine, numa-də “big (m),” in order to indicate the remarkably large size of the growing belly.

5.2 A syntactic analysis

I will show how the syntactic structure proposed in (44) above can be applied to account for grammatical gender reversals in nonhuman animate and inanimate nouns. As diagrammed in (62), the evaluative head n2 [eval] is projected above the noun. The head n2 [eval] can be specified for the gender features [+F] or [−F], and thus, it is capable of changing the grammatical gender of the base to which it attaches.

(62)

The only difference between the structure in (62) and the one proposed earlier in (44) lies in the type of gender feature (interpretable or uninterpretable) of the noun to which the evaluative head is attached. As I briefly described in Section 3.2, Kramer (2015) suggests that there are different types of n across languages: (i) n with interpretable gender features i[+/−F], (ii) n with uninterpretable features u[+/−F], and (iii) n with no gender features (“plain” n). Across languages, sex-differentiable nouns that denote humans and some higher animals (which may vary in different languages) can have interpretable gender features i[+/−F] or a “plain” n with no gender features (e.g., same-root human nouns). Nouns that denote nonhuman animates (in which biological sex is unimportant to humans) as well as those referring to inanimate referents can have either uninterpretable gender features u[+/−F] or a “plain” n .

Consider, for example, Halkomelem. In this language, masculine is the unmarked gender (Gerdts 2013). I propose that the noun šeł “road” in the data in (59a) above is masculine by morphological default and corresponds to the “plain” n in Kramer’s (2015) terms, as structured in (63). The evaluative head n [eval] is specified for the feature [+F]. Thus, the resulting evaluative form triggers feminine grammatical agreement with the determiner: θə šeł “the (f) road (unusually curvy),” as in the data in (59b).

(63)

In Benchnon (the data in 58), the noun tʃ’āmá-ī “shoe (m)” is marked for the uninterpretable feature u[−F]. In the structure in (64), the evaluative head n2 [eval] is projected above this noun and contains the feature [+F]. The resulting noun tʃ’āmá-ā “shoe (f) (derogatory)” is evaluative “adding a derogative sense to the utterance” (Rapold 2006: 184).

(64)

In example (60) above from Alamblak, the noun kuñ- t “house (f)” is feminine. I propose that it is specified for the uninterpretable feature u[+F], as shown in (65). When the head n2 [eval] with the feature [−F] is projected, the resulting evaluative form kuñ- r “unusually long house (m)” has masculine gender.

(65)

If we compare the examples above in terms of their evaluative meanings, we observe that they have different evaluative interpretations. For example, by using an evaluative form in Benchnon, the speaker expresses a negative attitude (derogation) while in Halkomelem and Alamblak, the speaker indicates an unusual physical form of the referent (it is the unusual property of a physical form that makes it evaluative). A sociopragmatic analysis will allow us to account for the differences in these evaluative interpretations.

5.3 A sociopragmatic analysis

The Violations of Expectations principle (Acton 2014) and the Division of Pragmatic Labor principle (Horn 1984) address cases in which a violation of conversational expectations (or a “marked expression” in Horn’s terms) has occurred. For example, when a noun which normally triggers feminine agreement unexpectedly triggers masculine agreement, it violates the hearer’s expectations. Thus, such an utterance is likely to be interpreted as having a special significance (or conveying a marked message).

The Full Significance principle captures the role of the sociocultural context, which differs in Benchnon and Alamblak. Consider how grammatical gender is assigned in these languages.

In Alamblak, the gender of nonhuman nouns can be assigned on the basis of the size and shape of the referent. Inanimate objects that are tall, long, slender, and narrow are usually treated as masculine, while objects that are short, squat, or wide are usually treated as feminine (Bruce 1984: 97). For example, the terms for “house” and a “short/squat tree” are normally feminine, while the terms for “arrows,” a “signal trumpet,” and a “tall slender tree” are normally masculine (Bruce 1984: 97). Thus, a change in grammatical gender is perceived as a sign of something unusual in the object’s physical form. For example, the conversational expectation for the noun “house” is feminine, because of its wide shape. According to the Full Significance principle, the observed form is compared to relevant alternatives. In this example, the feminine gender is expected on the noun “house,” but what we see is the masculine gender. Since the masculine gender in this language correlates with long and narrow inanimate objects, it stands to reason that the speaker is suggesting that not only is the particular house of an unusual form, but more specifically that it is unusually long, which is consistent with the masculine gender marking in the language.

In Benchnon, gender is assigned according to the principles that differ from those in Alamblak. Thus, nouns denoting animates (human and nonhuman) usually have sex-based gender – masculine for males and feminine for females, while nouns denoting inanimates usually trigger masculine agreement (masculine is the default gender in Benchnon). In contrast to Alamblak, gender in Benchnon is not assigned on the basis of the size and shape of the referent; it is either sex based (for animates) or masculine default (for inanimates). As Rapold (2006: 184) points out, a change in gender is perceived as a sign of the speaker’s attitude. For example, in the data in (58b) above, the gender of the inanimate noun “shoe,” which is normally masculine by default, is changed to feminine, expressing derogation. In this language, deviation from the expected default gender seems to indicate a negative attitude of the speaker. Observe in addition the data (66) below, in which a man is referred to with the feminine gender, also expressing a negative attitude. In this example, too, we see deviation from the expected masculine gender with the resulting derogative attitude on the part of the speaker.^[7]

According to the Full Significance and Differential Importance principles, an utterance is interpreted based on its relation to relevant alternatives which the speaker could have used in a particular discourse situation. The most obvious alternatives for the evaluative expressions in Benchnon (58b) and Alamblak (60b) are their nonevaluative counterparts, as in (58a) and (60a) above. The nonevaluative counterparts carry the normally used gender markings, which would correspond to the hearers’ conversational expectations. A violation of these expectations in (58b) and (60b) results in the evaluative interpretation of the utterances.

5.4 Summary

I have proposed that grammatical gender reversals in nonhuman animate and inanimate nouns across languages can be accounted for by using the same syntactic structure and sociopragmatic principles which were applied to account for human nouns, as described in Sections 3 and 4. The only difference between the accounts for human and nonhuman nouns concerns the interpretability of gender features on the head n to which n [eval] is attached. In human nouns, we deal with the interpretable gender features i[+/−F], while in nonhuman animates and inanimate nouns, there are either the uninterpretable features u[+/−F] or the “plain” n with no gender features, as in (67).

(67)