Optimizing Polish Gender: A Preliminary Analysis
Master’s ThesisName: Agnieszka Gimżewska
Student number: 5627893
Programme: Linguistics (Research Master)
Department: Department of Language and Literature Faculty: Faculty of Humanities
University: University of Amsterdam Thesis supervisor: René Genis
Second reader: Arjen Versloot
Date: September 2019
Abstract
This thesis is about the grounds on which singular nouns of Contemporary Polish may be considered to belong to a particular gender category with the perspective of an adult native speaker in mind. The variety of the language investigated is Standard Polish spoken in most of the Polish territory. The vantage point is a three-way gender division into masculine, feminine, and neuter reflected largely in an underlying three-way formal distinction instantiated by the nominative singular (of nouns) (even where currently Polish morphophonology-phonetics makes a four-way contrast). In Stefańczyk (2007) more or less loosely stated rules and tendencies which generate grammatical gender of nominative singular nouns (native and loaned) found in Słownik języka polskiego (Doroszewski, 1958-1969) and in Uniwersalny słownik języka polskiego (Dubisz, 2003) amount to twenty nine pages and relate the grammatical gender of nouns to their semantics as well as to their phonological and morphological conditioning (involving derivation and compounding) in the nominative singular case form. In this thesis the rule set which generates pre-syntactic gender of nouns listed in Słownik języka polskiego PWN (Drabik et al. 2014) amounts to one page only. It was possible to show that nearly 99% of GA in my sample can be accounted for by a theory which recognizes (1) a markedness hierachy of masculine over feminine and neuter, and feminine over neuter and (2) a set on non-ranked (Steinmetz, 1986, 2006) morphophonological-phonetic and semantic gender assignment rules. A preliminary Optimality Theory analysis of gender assignment in Contemporary Standard Polish is presented using Steinmetz’ (1986) fundamental work and Rice’s continuation (2003, 2004, 2005, 2006).
2
Keywords: Grammatical gender; Gender assignment; Phonological gender assignment rules; Morphological gender assignment rules; Semantic gender assignment rules; Optimality Theory; Standard Polish
The following abbreviations, symbols and other conventions are used throughout: Acc.=Accusative, aff.=affectionate, ANIM=animate noun, [+ANIM]=the referent is marked for animacy, arch.=archaic, coll.=colloquial, CSP=Contemporary Standard Polish, dim.=diminutive, E.=English, f/F=feminine gender, [+FEMALE]=noun referring to a biological female, G.=German, GA=gender assignment, Gen.=Genitive, I.=Italian, L.=Latin, m/M= masculine gender, [+MALE]=noun referring to a biological male, [-MALE, -FEMALE]=the referent (human and non-human) has a sexless connotation, n/N = neuter gender, Nom.=Nominative, obs.=obsolete, pej.=pejorative, Pl.=Plural, Sg.=Singular.
Throughout the thesis, the noun entries are in Nom.Sg. case, unless otherwise stated. Also note the use of bracket conventions when citing Polish nouns <>, {}, //, and []: <o> is a grapheme, {-uni-} is morphological (derivational), /o/ is phonemic/phonological, and [ɔ] is phonetic (how the sound is pronounced). For the morphophonological-phonetic specification of noun entries in Nom.Sg. case form, the following order will be applied (e.g. dziadunio dziad{-uni-}/+o/ [ʥad-uɲ+ɔ] “grandpa, (aff.)”).
1. Main claims and plan
This is a study into the grounds on which singular nouns of CSP may be considered to belong to a particular gender category with the perspective of an adult native speaker in mind.1 Research has established that nouns are not arbitrarily assigned to their grammatical gender categories (e.g. Corbett, 1991; Comrie, 1999). In Russian, traditionally recognized as having three grammatical genders, such nouns as for exampleдом “house” and Фрукt “fruit” have a fixed masculine gender in the lexicon. In German, which also has a three-gender system, both haus “house”, and obst “fruit” are neuter. These are just a few examples, but the explanation of what mechanism is responsible for GA to nouns in a given language had not yet been given, until the late 1980’s when Steinmetz, a prominent researcher in the area of GA, took a stand on this issue.
1 The GA of plural nouns in the nominative form in CSP is left out of consideration in the present study. See Rice (2005, p. 2) for a comment on Optimality Theory based plural GA.
3
Building on the foundations laid by Steinmetz (1986), Rice (2006) writes as follows:
In other words, there is a system behind the assignment of nouns to their gender categories across languages. This system, proposed by Steinmetz and initially tested only on German in the 1980’s turned out to rely on a default hierarchy of gender categories which underlies the operations of language-specific phonological, morphological, and semantic rules (after Onysko 2007, p. 155). As far as the interaction of GA rules is concerned, the theory presupposes a non-ranking according to whether the rules are semantic, morphological, or phonological (Onysko, 2007, p. 155). If a rule applies, it always provokes its respective gender, regardless of the existence of other rules. If rules triggering different gender apply at the same time, they all “count” equally and they have the same associative strength (see further Section 7 for information on GA conflict resolution). This is how Rice recalls the initial stage of developing this theory of GA:
‘[…] We remember the period in which his [Steinmetz’] theory was being developed, including a day when he proudly announced that he had been through an entire German dictionary, and had identified only eight nouns that defied the predictions of his approach - and of course he had explanations for these, as well. We also note with interest that the theory presented there anticipated all of the key properties of Optimality Theory as developed in the 1990s: Violable constraints, hierarchical constraint ranking, competition among candidates […].’
(Rice et al. 2010)
Interesting results were also obtained for Russian (Rice, 2004, 2005, 2006). On the basis of gender information of the 53,892 nouns found in Obratnyj Slovar’ Russkogo Jazyka [Reverse Dictionary of the Russian Language] (Lazova et al. 1974), Rice incorporates morphological rules related to the declinability and indeclinability of nouns, among others, to account for GA (for masculine-feminine-neuter division).2 Under the heading ‘Some Russian GA rules’, Rice (2004, 2005) proposes the following
2 As a preliminary remark, it should be pointed out that such an approach to GA (empirically testifiable by means of carefully designed experiments) stands in sharp contrast to the premises in this study (see further Section 3). For Rice (2004, 2005, 2006), the masculine-feminine-neuter division is evidenced in the grammar of Russian in
‘[…] Languages frequently present nouns that show surface violation of their GA principles. This gives the superficial appearance of GA tendencies - rather than absolutes - and may cast doubt on the enterprise of gender category prediction. […] we demonstrate that mere tendencies on the surface do not indicate the absence of reliable GA principles. Instead, tendencies are simply the expected consequence of resolution among conflicts between violable constraints […].’
4
rule set fragment: morphological (inflectional class) rules: -a=f, -o=n, -e=n, indec.=n; semantic rules: +male=m, +female=f, +animate=mf and claims that this simple system of rules along with a default hierarchy m > f > n, suffices to account for the vast majority of GA in Russian.3 Rice (2004, 2005) claims that as much as 97% of nouns in Russian may be accounted for in this way.4
The variety of the language under analysis in this study is Contemporary Standard Polish (henceforth, CSP), known as the Cultural Variant, or the Polish Literary Language. CSP is taught in schools and used in the media, and is uniform throughout Poland. In the north, there is a small community speaking Kashubian, nowadays recognized as a separate (regional) language. In a few regions, there are groups of people who still speak dialects of Polish at home, for example the Tatra Mountains’ highlanders use the Podhalanian Dialect, and the Silesian Dialect is spoken in south-western Poland. However, most of these people can also speak the standard variety of Polish. Generally speaking, regional variations of spoken Polish involve minor phonetic and lexical differences.5
CSP is also an interesting language from the point of view of gender studies as the interplay of GA rules is intricate. Not only do rules interact, but they also compete. The conflict is usually resolved,
the Nom.Sg. case (i.e. gender is expressed on adjectives, possessives and demonstrative pronouns, as well as verbs in the past tense, and it is grammatically marked on the noun: its semantics and/or its (in)ability to decline) (compare with Corbett, 1991, assignment based on a single form in phonological criteria). Rice (2004, 2005, 2006) fails to consider the Russian noun itself as a place where pre-grammar classification takes effect, despite the fact that Russian has been frequently cited as a language which has developed (largely) overt gender marking on nouns (i.e. there is a strong correlation between the morphophonological properties of nouns in the nominative singular and their gender, e.g. Rodina and Westergaard, 2017, p. 199) (see however a special case учитель “teacher”, vs. учительница “female teacher”, etc., and further complexities involving кале́ка “cripple”-type nouns). Note also that dictionary grammatical GA made by lexicographers may be normative in nature and hence not reflect actual usage.
3 Rice’s formulation for Russian is that “nouns ending in a segmentable morpheme +a are feminine” and this formulation is “a notational variant of the claim that nouns of the 2nd declension are feminine” (Rice, 2006, p. 1401). Accordingly, the (segmentable) inflectional ending +o, and the (segmentable) inflectional ending +e in the nominative singular typically assign neuter gender. Indeclinable (inanimate) nouns are predominantly neuter. Nouns denoting an animate thing tend to be assigned non-neuter gender. For sex-differentiable nouns, those denoting biological males are masculine, and those denoting biological females are feminine (Rice, 2004, p. 3-5; Rice, 2005, p. 4-6). The posited default hierarchy m > f > n means that masculine is the grammatical default in Russian, followed by feminine, followed by neuter, and this hierarchy is meant to resolve instances of gender competition in Russian (Rice, 2004, p. 2; Rice, 2005, p. 3).
4 See also Corbett and Fraser (2000). Galbreath (2010) enumerates several deficiencies in Rice’s (2004, 2005, 2006) analysis of GA in Russian. For example, Galbreath (2010, p. 66) points out that Rice (2004, 2005, 2006) does not address the issue of derivational suffixes, or successfully animate neuters, but it is outside the scope of the present study to thoroughly discuss these deficiencies and their possible effects on the accountability score (for details see Galbreath (2010) and his complete account of grammatical GA in Contemporary Standard Russian in which he very much relies on phonology (not (in)declinability) as one of the the prime indicators of grammatical gender).
5 This information comes, in large part, from a booklet published by the Council for the Polish Language (Rada
Języka Polskiego) in 2007 and designed to popularize knowledge about CSP. Link:
http://www.rjp.pan.pl/images/stories/pliki/broszury/jp_angielski.pdf. Other interesting facts about CSP have been found in Awramiuk, Krasowicz-Kupis and Smoczyńska (2017, p. 139).
5
or the noun vacillates in gender. To the best of my knowledge, there are several accounts of grammatical GA in CSP. Stefańczyk’s (2007) monograph is clearly the best attempt so far. The vantage point is a three-way gender division reflected largely in a three-way formal distinction instantiated by the nominative singular (of both adjectives, verbs in the past tense, and nouns). Related work in this field is Kilarski and Orzechowska’s (2007) contribution who claim that often it has been taken for granted that phonological/phonetic regularities are based on morphological regularities, although it seems that they altogether have never received as much attention as, for example, the question of the number of genders in CSP. Kilarski and Orzechowka’s (2007) work show that the major generalizations for CSP can be stated purely in terms of phonology/phonetics, but both Stefańczyk (2007) and Kilarski and Orzechowska (2007) do not take a stand on how rules interact in a systematic way, and no theoretical model is pursued. Proceeding from this observation and taking the successes of Steinmetz’ theory established for German and Russian, this study presents a preliminary analysis of GA in CSP using Steinmetz’ (1986) fundamental work and Rice’s (2003, 2004, 2005, 2006) continuation within the framework of Optimality Theory (henceforth, OT). The repertoire of nouns is taken from Słownik języka polskiego PWN (henceforth, SJP PWN) (Drabik et al. 2014) (see further Section 8). In contradistinction to Rice’s (2004, 2005, 2006) papers, this study takes an experimental approach to its subject matter because the gender values: masculine, feminine, or neuter are based on my knowledge about the gender of particular nouns. I advance (relying on the principle of economy within generative linguistics) a set (though not complete) of morphophonological-phonetic and semantic GA constraints to account for GA in CSP, and this is irrespective of the behaviour of associated words (see further Section 3). I assume that this morphophonological-phonetic and semantic information is part of adult native Polish speakers’ linguistic competence, it is stored in their mental lexicon, and it helps them determine the gender of native and loaned nouns (Corbett, 1991).
2. A reiteration of the purpose of the thesis
The purpose of this thesis is to explore Optimal Gender Assignment Theory (henceforth, OGAT) and, in particular, to pursue the claim (not directly, of course) that universally, gender is assigned based on a set of crucially non-ranked gender features constraints and markedness constraints.6 Under this condition, I address the following question: why are, for example, nouns like tato, żabka, jabłuszko (all Nom.Sg. case nouns, all nouns of CSP) masculine, feminine, and neuter, respectively? The extent to which the adult Polish native speakers know the gender of a noun remains to be substantiated by other researchers (see further Section 3), but I will attempt to present a general and preliminary theoretical
6 It is not my intention to argue for or against this claim but a word is due when it comes to the problem of falsifiability as it might be relevant in the case of the OGAT. For example Enger (2009, p. 1287) rightly argues that OGAT opens the door to the following possible scenario: “whenever linguist X thinks he/she has found a counter-example, linguist Y may come up with an additional rule that “solves” the problem and “saves” the model”.
6
outline of GA framed in OT based on a corpus of 18,488 nouns of CSP found in SJP PWN. The main focus is on gender rule interaction and gender conflict resolution. Such a preliminary study can give an impression of how the mental model of GA might be organized in native CSP.
3. Nouns of CSP and gender value assignment
Hockett proposes the following definition of linguistic gender: “Genders are classes of nouns reflected in the behaviour of associated words” (Hockett, 1958, p. 231, cited in Corbett, 1991, p. 1), and “the relevant reflection in the associated words is agreement” (Corbett, 2013, p. 89). In other words, agreement can be safely used in order to define what gender actually is, it is the criterion for the presence/absence of grammatical gender in a given language, and it serves to establish the number of grammatical genders the given language has and which nouns belong to which gender.There may be some overt grammatical signs to indicate gender differentiation, however “no amount of marking on a noun can prove that the language has a gender system; the evidence that nouns have gender values in a given language lies in the agreement targets that show gender” (Corbett, 2013, p. 89).7 With a special reference to CSP, Swan (2015, p. 13) claims that given this definition, the actual number of genders distinguished “has ranged from the traditional three to four, five six, seven, eight, nine, or to an almost unlimited number”.8
Gender in CSP is an interesting feature for several reasons. On the one hand, it shows up in the morphology of its agreement targets, and in that respect it is a complex morphosyntactic phenomenon. On the other hand, gender in CSP largely shows up in the nominative singular form of the noun itself, irregardless of associated words. By way of example: in duż-yM stół-+ØM “large table”,
7 In terms of the analysis, agreement patterns give information about the gender value of a noun. In terms of the generation of a surface structure, however, causality goes the other way: the noun is marked grammatically for gender and the gender value of a noun is said to consistently 'control' the form of particular 'agreement targets' in a noun phrase. Agreement targets vary surprisingly across gender languages (see Corbett, 1991). Note that there exist languages (e.g. Brazilian Portuguese), for which the agreement method does not appear to be fail-safe (Carvalho, 2016, p. 14).
8 I quote below the observation made by Bogusławski (2009, p. 13) with respect to the number of grammatical genders present in the system of CSP:
To date the debate on the amount of grammatical genders still continues. These concerns do not bear any significance for the present analysis and the intricacies of the gender system of CSP defined on syntactic grounds lie far beyond the scope of the present study. For an overview, see for example Nowosad Bakalarczyk (2009, p. 12-18); Koziarski and Krysiak (2012, p. 23-25 ); Wierzbicka (2014, p. 157-163); Seretny and Stefańczyk (2017, p. 72-77).
‘No end seems to be in sight to the controversy over “grammatical genders” in Polish. Ever new lists of “gender” or “genderlike” labels with the concomitant classes of words (usually nouns) or their forms as well as examples thereof are produced, and ever new objections to what other authors propose are raised.’
7
the gender marked on the adjective duż-yM “large”, which indicates that stół-+ØM “table”, falls in the set of masculine nouns, also turns up on stółM “table”, itself. Just from hearing a noun in isolation,9 adult native speakers of Polish are usually able to identify many thousands of singular nouns as unarguably masculine, feminine, or neuter in the singular.10 My sample noun stół “table”, is unarguably masculine, prośba “request”, is unarguably feminine, czytanie “reading”, is neuter, without hesitation. The adult native speakers of Polish know that (phonetically/phonologically speaking), in the majority of cases, consonantal endings (Ø) assign masculine gender, the final vocalic ending /+a/ assigns feminine gender and the final vocalic endings /+o/, and /+e/ assign neuter gender. Therefore, in my study I follow the assumption that the gender specification of nouns of CSP is lexical, as opposed to grammatical, which means that this information: masculine, feminine, or neuter comes from the lexicon and is encoded in the lexical items.11 More precisely, it will be argued that singular nouns in CSP (native and loaned) are largely assigned gender via a lexical (presyntactic) rule referring to word final sounds in the nominative singular.12 Formal criteria, as will be argued throughout the thesis, play a particularly important role in GA in CSP, but the gender value: masculine, feminine, or neuter is often determined by an interaction of GA rules (e.g. mężczyzna “man”, see also the special case profesor “professor”, sędzia “judge”, etc). Crucially, closer synchronic examination reveals a great deal of regularity in the gender system of CSP andit can be observed that adult native speakers also assign gender with ease to any loanword that happens to be inserted into the model. They do so by quite simply taking advantage of the (already existing) regularities in GA of native nouns, e.g. clubbing (<E. clubbing),13 recykling (<E. recycling), etc.14
9 In isolation (i.e. out of context) the nouns are in the Nom.Sg. case whose one of the syntactic functions is “naming”. Because of its naming function, the Nom.Sg. case is used as the form of Polish noun entries in dictionaries or on vocabulary lists.
10 What counts is the phonetics, not the orthography, which is a concept based on visual cues the Polish native speakers might rely on. Exploring the pottential effects of orthographic as opposed to morphophonological-phonetic information is left as topic for future research. The idea that orthographic information may play a role in GA in CSP can be downplayed if we consider a language acquiring child. Also, given the modest amount of data of Polish children acquiring Polish, no final claims can be made with respect to the validity of various GA hypotheses.
11 This assertion is in line with Swan’s (2015, p. 13) claim that Hockett’s (1958) definition ignores “considering lexical sub-categorization by itself, which probably any language has, without the accompanying behaviour of associated words, to be a kind of gender”.
12 See also Wierzbicka’s article (2014, p. 156-157) and her discussion on the partial workings of semantic GA rules within the lexicon of Polish. She concludes that in a group of all nouns only some show semantic motivation. For example, kalosz (originally a female French noun) could have been assigned to masculine gender as a result of association with the Polish hypernym but (m) and whisky <E. whisky has become feminine because of Polish
wódka (f). Summing up, Wierzbicka points out that it would not be very useful to analyze GA in Polish from the
semantic point of view.
13 The noun is masculine because it falls in the (Ø) slot as it is pronounced more or less as [klabjink] in CSP (see Section 5.1.).
14 This ability to classify nouns according to gender can be further examined in experiments, conducted orally, and involving real nouns. Such experiments could investigate whether adult native speakers of Polish, upon hearing the noun, use the information carried by the noun’s “ending” and use semantic information in the noun
8
The phenomenon of the expansion of a particular gender (at any given time) by taking in new noun formations has been well described in the literature on GA, and quite interestingly, it has been taking place in native CSP. Take for example some feminine nouns with stems in (Ø), which have ‘atypical’ gender markers and which seem to be in the process of gradual morphophonological reanalysis. Nowadays, in the lexicon of many adult native speakers of Standard Polish many or most of these nouns “go with the flowˮ and have been replaced by formations in /+a/ (e.g. cerkiew and cerkwia “Orthodox Church”, konew and konewka “jug”, marchew and marchewka “carrot”, rzodkiew and rzodkiewka “radish”) (Stefańczyk, 2007, p. 31).
In this study I refer to the marking on the noun itself as an expression of gender. Nevertheless, the gender values: masculine, feminine, and neuter do not stand on their own, their effects transcend the lexicon. For example, when a Polish adult native speaker wants to use a singular noun in the nominative case in CSP, as previously stated, he/she can ‘work out’ its gender and he/she knows what to do with the noun. The gender values interact with agreement and the morphophonological gender of a noun as expressed by the nominative singular case in CSP serves as the basis for the classification of nouns into declensional patterns, according to which the inflectional endings throughout paradigms are formed (Janssen 2014, p. 109; Janssen, 2016, p. 30). Not knowing the gender of a noun in the nominative singular leads to problems for example with adjectival agreement, and subject-verb agreement in the past tense (Janssen, 2016, p. 38; Stefańczyk, 2007, p. 17; Seretny and Stefańczyk, 2017, p. 72).
Two groups of nouns challenge this ideal picture of lexemes of CSP having a single gender value in the lexicon. From the point of view of the adult Polish native hearer/speaker, many what I call ‘profession nouns’ (i.e. an umbrella term for nouns denoting professions, names of honorary titles and academic titles, academic positions, names of offices in their generic, predicative, and titular meaning), nowadays, can be assigned either gender value: masculine and feminine at the lexical level depending on whether the reference is to a male individual, or a female individual. More to the point, morphophonologically, these nouns do not exhibit variation, but can be semantically ambiguous. Nowadays, many professions are open to both men and women. In the case of some ‘profession nouns’, the referent may culturally be assumed to be a male (e.g. proboszcz “parson”, górnik “miner”, hutnik “steel factory worker”, ślusarz “locksmith”), whereas the referent of niania “nanny” is highly
to decide on the gender in the absence of any phrasal information that would help them to determine gender based on agreement. Also nonsense nounscan be created combined with different “endings” (e.g. hoszek, wepa,
lęko) so that the role of the morphophonological-phonetic information encoded on the noun endings could be
investigated independently of any semantic effects. Adult native speakers of Polish could be asked to use these nonsense nouns by providing an agreeing adjective, thus indicating the corresponding gender value.
9
likely to be thought of as woman.15 A different type of gender variation can be attested with kaleka
“cripple”-type nouns, whose gender is underspecified in their strict nominal form (i.e. at the level of the noun) but determined contextually via identification with a human referent in discourse. These nouns can trigger either masculine or feminine agreement in controlled elements, usually in accordance with the referent’s gender (see further Section 9 and Section 10.3.).
4. Organization of the remainder of the study
My discussion has so far concentrated on the gender system of CSP and I have pointed to its complexity. I have presented different theoretical approaches to the study of gender, the number and quality of genders present in the system of CSP. The main aim was to show where this study is situated as far as different GA mechanisms are concerned. The realities of pre-syntactic GA have been touched upon and its relevance in native CSP. In the part that follows (Section 5) I introduce the relevant characteristics of the phonemic inventory of CSP. In Section 5.1. I make assumptions about the usual native pronunciation patterns of nouns of CSP covering the most important issues such as the realisation of nasal <ą> and <ę> in CSP, among others. In Section 6 I discuss the phonological and morphological structure of nouns in CPS in the nominative singular. Section 7 is a presentation of Steinmetz’ theory of GA, most clearly stated in his (1986) paper ‘Two principles and some rules for gender in German: inanimate nouns’ and it contains a thorough discussion of two main principles GENDER TALLY andGENDER ECLIPSIS deemed important in GA. Section 8 is a presentation of the dictionary
and corpora serving as sources of data for this study. In Section 9 I refer to groups of nouns which are problematic for the model of GA dealt with in the study, that is: (1) nouns vacillating in gender, (2) kaleka “cripple”-type nouns. In Section 10 I present the tenets of OGAT and in a follow-up discussion the markedness hierarchy of gender categories in CSP. I advance a rule set that generates gender for the nouns under investigation and I make occasional observations about other Indo-European languages. Each rule is formalized as a constraint, in anticipation of my OT GA a la Rice (2003, 2004,
15 Again, it would be interesting to examine the level of the difficulty adult native Polish speakers show with pre-syntactic assignment of a gender value to ‘profession nouns’ and/or the preference of adult Polish native speakers for one gender value over another. More than likely, adult native Polish speakers would spontaneously assign correct gender to all overtly gender marked nouns (e.g. kelnerka kelner{-k-}/+a/ “waitress”, nauczyciel{-k-}/+a/ “female teacher”, etc.), and they might be less proficient with non-overtly marked nouns (e.g. hutnik hut-nik+Ø “steel factory worker”, profesor profes{-or}+Ø “professor”, sędzia sędzi/+a/ “judge”). Of course, there exists an alternative account of GA in the case of ‘profession nouns’. Rather than claiming that ‘profession nouns’ exist in two (competing) versions, it can be argued that the gender value of a ‘profession noun’ can be either masculine or feminine based on the context and cannot be determined out of context. In other words, such nouns are underspecified with respect to gender and this underspecification persists until the noun has reached syntax. In consequence, it can be argued that the gender system of CSP cannot be straightforwardly explained within one assignment theory (i.e. while OGAT can deal with GA of such nouns as tato “dad”, contextually assigned gender of the nouns profesor, sędzia, for example, remains problematic for the mechanisms of the theory; see also Dahl (2000, p. 106) for the category of ‘referential gender’)).
10
2005, 2006). In Section 10.3. I present a preliminary OT analysis of GA in CSP, building on the data (the rule set). I discuss GA in morphologically simple and complex nouns (derived and compound, non-expressive and non-expressive derivatives). The special focus is on GA conflict resolution in CSP. In Section 11 I provide an accountability score for GA and I address apparent exceptions. The study is rounded off with a short summary and concluding remarks.
5. Phonemic inventory of CSP
To begin with, I make some assumptions as far as the repertoire of phonemes (consonantal and vocalic) in present-day Standard Polish is concerned. Chart 1 shows the vowel inventory and Chart 2 shows the consonantal inventory categorized by place and manner. The phonemic status of the vowels and consonants in brackets is often debated (see further Section 5.1.).
Chart 1. Vowel inventory
i (ɨ) u
ɛ ɔ
(ɛ̃) (ɔ̃)
a
(summarized and adapted from Sussex and Cubberly, 2006, p. 153-162) Chart 2. Consonant inventory
labial palatalized labial
dental alveolar palatal velar palatalized velar Stops -v +v p b (pʲ) (bʲ) t d k g (kʲ) (gʲ) Affricates -v +v ʦ ʣ ʧ ʤ ʨ ʥ Fricatives -v +v f v (fʲ) (vʲ) s z ʃ ʒ ɕ ʑ x Nasals m (mʲ) n ɲ Approximants w l r j *+v=voiced, -v=voiceless
(summarized and adapted from Sussex and Cubberly, 2006, p. 163-166, and Dukiewicz and Sawicka, 1995, p. 111)
5.1. Phonetic transcription of nouns in CSP and pronunciation issues
As a general comment, it should be pointed out that the orthographic form of nouns in CSP corresponds to a great degree with the actual pronunciation (with a few exceptions). However, some comments are in order here. My sample nouns below come from SJP PWN. The question whether or not for example [p] and [pj] or [b] and [bj] are separate consonantal phonemes does nor bear any
11
significance to the problem and analysis presented in this study. Nevertheless, I follow Kilarski and Orzechowska (2007) in that pilśń “felt”, for example, is pronounced as [pʲilɕɲ], biel “white” as [bjɛl] in present-day Standard Polish. The vowels <ą> and <ę> do not appear in any other contemporary Slavic language and they are also rare in European languages. In Contemporary spoken Polish the nasality disappears in most positions (see my discussion below). Word-finally, the nominative singular ending <ę> is pronounced [ɛ] in colloquial speech (Feldstein, 2001, p. 17; Swan, 2003, p. 7-8) and [ɛ̃] has been considered present in emphatic, or rather, artificial speech (e.g. Stankiewicz, 1986, p. 64). Throughout the study, the phonetic specification of noun entries in Nom.Sg. case reflect my pronunciation of these nouns that I adopt in daily use, which to a great extent mirrors the pronunciation patterns specified in SJP PWN. For example, the orthographic <é>, appearing in loanwords from French (e.g. attaché [ataʃέ] <F. attaché), is realized as [ɛ], and a bit of foreign accent is added to indicate a change of pronunciation.16 Furthermore, the following are assumed to reflect the usual native pronunciation (Feldstein, 2001: p. 15-16; Rada Języka Polskiego, 2007, p. 2; Wtorkowska, 2011, p. 245-246):
<ą> as [om] before <p>, <b>, <t>, <d>, <k>, <g> (e.g. kąpiel is phonetically [kɔmpjɛl] “bathing”)
<ą> as [ɔn] before <t>, <d>, <c>, <dz>, <cz> (e.g. pieniądz [pjɛɲɔnʦ] “money”, gorączka [gɔrɔnʧka] “fever”)
<ą> as [ɔɲ] before <ć>, <dź> (e.g. żołądź [ʒɔwɔɲʨ] “acorn”)
<ą> as [ɔŋ] with velar [ŋ] before <k>, <g> (e.g. łąka [wɔŋka] “meadow”)
<ą> conserves its nasal character word-finally (e.g. bon ton [bɔ̃tɔ̃] “fashionable manner or style”) <ę> as [ɛm] before <p>, <b>, <t>, <d>, <k>, <g> (e.g. sęp [sɛmp] “vulture”)
<ę> as [ɛɲ] before <ć>, <dź> (e.g. chęć [xɛɲʨ] “desire”)
<ę> as [ɛŋ] before <k>, <g> with velar ŋ (e.g. ręka [reŋka] “hand”)
<ą> and <ę> have a nasal-vocal character before <w>, <f>, <z>, <s>, <sz>, <ż>, <sz>, <ś>, <ź>, <ch> (e.g. węch [vɛ̃x] “olfaction”).
Instead of going into too much detail, the comment below concerns voicing assimilations within consonant clusters (e.g. Zagórska Brooks, 1975, p. 13-14):
C+vC-v becomes C-vC-v (e.g. ławka [wafka] “bench”, kłódka [kwutka] “padlock”)
C-vC+v (except for [v], [vj], and [ʒ]) becomes C+vC+v (e.g. prośba [prɔʑba] “request”, liczba [liʤba] “number”)
C-v + [v], [vj], [ʒ] becomes C-v +[f], [fj], [ʃ] (e.g. twarz [tfaʃ] “face”, potwór [pɔtfur] “monster”)
16 The pronunciation of foreign sounds can, nevertheless, vary from one individual to the next. There might be a very slight difference in pronunciation, at the very edge of recognition available to humans.
12
Furthermore, CSP displays a systematic word-final devoicing of voiced consonants:
Phonetically, final [b] sounds like [p], [d] like [t], [g] like [k], [w] like [f], [z] like [s], [ʑ] like [ɕ], [ʣ] like [ʦ], [ʥ] like [ʨ], [ʒ] like [ʃ], [ʤ] like [ʧ] (e.g. obraz is phonetically [ɔbras] “picture”).17
6. Morphological & phonological-phonetic make-up of nouns in CSP
When considering the phonological-phonetic regularities in GA in a language one must start by characterising the (possible) phonological-phonetic forms of words. The important distinction in this respect, which is necessary for CSP and for the present study, is that between derived and non-derived singular forms of nouns. Nouns in CSP consist of a stem combined with endings. The stem itself can be either morphologically simple (e.g. walka wal{-k-}/+a/ [val-k+a] “fight”), or complex (e.g. rozbieralnia rozbier{-alni-}/+a/ [rɔzbjɛr-alɲ+a] “cloakroom”), and the number of post-stem endings (i.e. final elements in word formation) typically does not go beyond three in the Nom.Sg. form in CSP. Non-derived nouns are those which have not had a word-formational suffix appended while Non-derived nouns do. By way of example, non-derived nouns include those like stół stół+Ø [stuw+Ø] “bread” which has a zero desinence (Ø) in the nominative singular (the stems of Ø-nouns end in a consonant), or woda wod/+a/ [vɔd+a] “water”, wino win/+o/ [vjin+ɔ] “wine” with a vowel ending (also called the desinence) /+a/ [+a] and /+o/ [+ɔ], respectively. Along with walka wal{-k-}/+a/ [val-k+a] “fight”and rozbieralnia rozbier{-alni-}/+a/ [rɔzbjɛr-alɲ+a] “cloakroom”, derived nouns include, for example, those like czytanie “reading”, from czytać “to read”, which (now discussed in more detail) breaks down into the derived stem czyt-ani,made up of the verbal root czyt- with the derivational suffix {-ani-} [-aɲ-], combined with the desinence /+e/ [+ɛ] that marks the nominative case for a neuter singular noun. The morphophonological-phonetic make-up of compounds of native (e.g. Wielkanoc Wielkanoc+Ø (lit. great+night) “Easter”), and foreign origin is discussed later (Section 10.3.) when I analyze GA of morphologically complex nouns.
7. A presentation of Steinmetz’ (1986) theory of GA using German language data
As previously stated, Steinmetz’ model of GA crucially relies on the notion that there is no ranking of semantic, morphological and phonological rules. From a psycholinguistic perspective, the main idea behind the model is that GA rules (which, of course, are based on observed regularities) work on the background of a default hierarchy which functions as a last resort in case of rule based GA conflict. In consequence, the gender of all nouns of a given language can be accounted for. Interestingly, all language-specific rules count the same and they have the same associative strength. In what follows I
17 This section covers only the most important issues of orthography deviating somewhat from IPA-interpretation. There are a few more but they do not have any significance for the present analysis.
13
bring to the fore the main task for a linguist working within this theoretical framework. Importantly, Steinmetz (1986) relies heavily on a generative approach to GA. From a generative perspective, the best analysis is the one accounting for the data with the smallest number of rules (Trosterud, 2006, p. 1442, on Steinmetz’ (1986) system). The GA system under analysis in this section is that of German, which has three grammatical genders in the singular.
The core strategy of Steinmetz’ theory is to assign nouns to the least marked category by default. Steinmetz (1986) does not provide any theoretical discussion of the criteria for establishing a markedness hierarchy of gender categories in any given language, that has grammatical gender. However, the major indicator that masculine is the least marked gender in German, is according to Steinmetz (1986), the fact that the number of rules required to cover the set of inanimate masculine nouns is far greater than the number of rules needed for feminine and neuter. In consequence, the nouns belonging to the masculine category are less homogenous than other, more marked categories. The set of rules that would be required to cover all inanimate masculine nouns can be replaced with a single rule, namely assignment to masculine by default (Hunt, 2018, p. 104).
Another indicator appears to be type freqency. At least, Steinmetz (2006, p. 1419) presents statistical frequency of the genders for the purpose of identifying the marked and unmarked. The idea is simple enough: the gender with the highest type frequency, that is the gender with most members, is the least marked. In German, approximately two-thirds of the monosyllabic nouns have been found to be masculine in Köpcke (1982).
This dominant gender in German eclipses the other genders and this principle is called Gender Eclipsis. Defined as m > f > n, the principle implies that masculine gender eclipses both feminine and neuter gender, while feminine gender eclipses neuter. Stated differently, the principle of Gender Eclipsis implies that a German noun will have masculine gender unless there is a rule that assigns feminine or neuter gender instead. If a rule applies, it always triggers its respective gender. Within the framework of masculine as the default gender for any given language, rules calling for masculine GA are largely redundant, unless they are needed to explain otherwise unaccountable exceptions.
According to Steinmetz (1986, p. 191), there are three basic types of rules for assigning gender to a noun in any given language:
(1) M-rules, which assign gender on the basis of morphological or phonetic shape,
(2) S-rules, i.e. semantic rules, which, as the name implies, assign gender on the basis of meaning, and (3) SC-rules, i.e. subcategorization rules, which apply only within the domain of certain S-rules and assign gender either on the morphological/phonetic shape or on the basis of meaning, depending on the domain involved (see Section 10.2.1. for the proposal of the SC-rule operating in CSP).
Furthermore, Steinmetz (1986) introduces an important restriction as far as the interaction of these rules in the German gender system is concerned. Namely, the theory employs the convention
14
that for each noun all applicable language-specific rules are listed to form a gender table, but the gender table of any given noun may contain no more than one S-rule. If two S-rules are possible, only the more general of them may be applied (Steinmetz, 1986, p. 192). It follows then that some semantic rules active in the German gender system are redundancy rules.
The problem of gender rule competition, Steinmetz (1986, 2006) resolves by introducing the principle of Gender Tally. This principle can be expressed as an instruction to count the number of times each gender is assigned, and if one gender has a higher value, that gender is assigned to the noun. If two or more genders have the same value, the competition is resolved according to the principle of Gender Eclipsis. In German the postulated markedness hierachy is masculine before feminine before neuter (m > f > n). Whenever in German masculine and another gender have the same value, masculine gender is assigned; in cases feminine and neuter have the same value, feminine gender is assigned. Consider gender tables in (1)-(5) (adapted from Steinmetz, 2006, p. 1425) for selected German nouns and the operation of Gender Tally and Gender Eclipsis in the German GA system:
Noun Gender rules Gender score
(1) boom “boom (in business)”, klotz “block” [no rules apply] 0m 0f 0n=m (2) gespräch “conversation”, gefühl “feeling” [ge-=n] 0m 0f 1n=n (3) gemeinde “community”, gebärde
“gesture”
[ge-=n, -e=f] 0m 1f 1n=f
(4) gebäude “building”, gemüse “vegetable” [ge-=n, -e=f, superordinate=n] 0m 1f 2n=n
(5) gefängnis “prison” [ge-=n, -nis=fn] 0m 1f 2n=n
When considering the gender tables, bear in mind that the GA rules are non-ranked. For example, the placement of the rule ge-=n to the left of -e=n is not to be interpreted as a hierarchical dominance. In the case of boom “boom (in business)” and klotz “block”, as no morphological nor phonological nor semantic rule applies to these nouns, as masculine is the default gender, and as masculine gender eclipses both the feminine gender and the neuter gender, the noun is masculine. In the case of gespräch “conversation” and gefühl “feeling”, a morphological-phonetic rule (i.e. nouns beginning with the morpheme ge- are neuter) applies to these nouns, and as no rule calling for masculine or for the feminine gender applies here, the Gender Tally results in neuter gender. In the case of gemeinde “community” and gebärde “gesture”, as one morphological-phonetic rule (nouns ending in the morpheme schwa are feminine) and one morphological-phonetic rule (nouns beginning with the morpheme ge- are neuter), apply, computationally we have a tie (0m 1f 1n) which is resolved by working with a left-to-right principle of Gender Eclipsis. Since feminine gender eclipses neuter gender in German, the noun is feminine. With gebäude “building” and gemüse “vegetable”, the tie is broken by a semantic rule superordinate noun=n (a superordinate noun refers to an object at a high
15
level of categorization, after Steinmetz, 1986, p. 190). In the case of gefängnis “prison”, two rules apply to this noun: nouns beginning with the morpheme ge- are neuter, and nouns ending with the morpheme -nis are not masculine. The Gender Tallyresults in neuter gender.
In the discussion above, I have touched upon the issue that no more than one S-rule may apply to a specific noun. In what follows, I use an example from the semantic domain “fruit”, as exemplified in German by Steinmetz (1986, p. 193) on the basis of a noun like Obst “fruit”. German has a semantic rule which assigns neuter gender to a superordinate noun (superordinate=n) (see above) and a semantic rule assigning a zero gender to a noun denoting a fruit (fruit=0), specifying that a noun denoting a fruit does not take any particular gender. Since the rule superordinate=n is the more general of the two, only this rule will appear in the gender table for this noun.
As stated in the introductory part of this thesis, Steinmetz’ (1986, 2006) model of GA gets most of the German language data right. That is, the two principles Gender Tallyand Gender Eclipsis operate in conjunction and resolve all instances of gender rule competition in favour of the correct GA (Steinmetz, 1986, p. 191).18
8. Data
The following dictionary and corpora served as sources of data for this study: a hard copy version of the monolingual dictionary SJP PWN and the Korpus Języka Polskiego PWN (PWN Corpus of Polish) (henceforth, KJP PWN).19
SJP PWN provides a comprehensive grammatical description of Polish words. It covers over 40,000 most frequently used lexemes of Standard Polish along with their most frequent meanings. The
18 The question arises whether the human mind actually calculates the putative gender relevant rules and selects the gender with the highest value within the Gender Tally system, before it activates a default mechanism to decide on the gender of a noun (Onysko, 2007, p. 156 ). In Rice (2004, p. 11-12), a parallelism is drawn between Gender Tally and connectionist processing (e.g. McClelland and Eman, 1986):
‘When a target (in our case a noun) activates certain units in a network (assignment rules in our case), one of the factors facilitating the selection of a certain unit is the amount of conceptual overlap. The higher the degree of overlap, the greater are the chances that a certain unit is selected. This is analogous to Gender Tally. When a majority of rules competes with a minority, the majority represents the higher degree of conceptual overlap. In other words, Gender Tally assigns gender on the basis of conceptual overlap. While this does not indicate that one has to believe in connectionism in order to adopt Gender Tally, the parallelism is nevertheless interesting.’
The psychological reality of GA in CSP, or in any other language, and in particular further elaboration of Steinmetz’ model of GA with psycholinguistic experimentsis left as topic for future research. As far as German is concerned, Schwichtenberg and Schiller (2004) find out on the basis of a series of experiments on gender congruency effects, that semantic rules such as predatory animals=m and musical instruments=f play a role in grammatical GA of native German speakers.
19 Here is the link to the on-line version of this dictionary: https://sjp.pwn.pl/. The KJP PWN corpus is searchable at https://sjp.pwn.pl/.
16
core of the SJP PWN vocabulary consists of words found in the above mentioned KJP PWN corpus which contains 100 million words. The dictionary includes nouns which have entered the lexicon of Standard Polish relatively recently, e.g. clubbing (<E. clubbing), lobby (<E. lobby), recykling (<E. recycling), zorbing (<E. zorbing), as well as colloquial words such as for example ściema “windup”, wrzutka “snapback”, zakupoholizm “shopaholism”.
Retrieved and stored in a spreadsheet in the Excel file were: all singular nouns, nouns which SJP PWN treat as singularia tantum (e.g. obuwie “footwear”, pierze “feathers”, miłość “love”, zdrowie “health”), and those which the dictionary treat as occurring usually in the plural form (e.g. krakers “cracker”, tips “artificial nail”, klapek “flip-flop”, kierpec “mountaineer mocassin”, ssak “mammal”). Included are obsolete (e.g. świekra “mother-in-law, (obs.)”, lowelas “playboy”), dialectal (e.g. ceper “low-lander”, dzięcielina “clover”) and otherwise stylistically marked nouns. Not extracted were pluralia tantum like spodnie “trousers”, drzwi “door”, nożyczki “scissors”, etc. With a special reference to loanwords, it needs to be pointed out that although the graphemic systems of the recipient language (CSP) and the target language (English or any other language) are different, the nouns once loaned from other languages usually adapted easily into the CSP graphemic and morphophonemic systems, e.g. fontanna (<I. fontana), koncert (<I. concerto), nominacja (<E. nomination), dżojstik (<E. joystick), kamamber (<F. camembert), gmina (<G. gemeinde), konstytucja (<L. constitutio).20 The dictionary provides additional information concerning nouns, that is, (alternative) spelling and (alternative) pronunciation with stress patterns, when applicable. I have explained in Section 3 that it makes sense to investigate GA in CSP incorporating gender values based on the adult native speakers’ knowledge about the gender of particular nouns. I classified the nouns as either masculine, feminine, or neuter, or both masculine and feminine (see further Section 9). Cases of vacillation are rare and include 11 nouns (see Section 9).21
9. Scope of the Analysis
In any language, nouns belong to the major part-of speech classes. Obviously, SJP PWN does not contain all Polish nouns. The selection of nouns is, however, quite representative as it contains nouns with high frequency scores. Some lexemes have various meanings. SJP PWN indicates each separate meaning by means of a digit. A decision had to be made which meaning of the nouns to include in the analysis and it is the first one of the possible meanings listed which was analyzed. Analyzed were then oko in the meaning “eye”, and not “an oily circle in soup”; babka in the meaning “grandma”, and not
20 There are some notable exceptions, e.g. clubbing (<E. clubbing).
21 The author of this study is a 36 year old female native speaker of Standard Polish, living in Poznań (Wielkopolska Region, West Poland), temporarily staying in the Netherlands, with no history of speech or hearing disorders. The extent to which the gender values mirror the spontaneous GA of the rest of the population is ultimately an interesting question to me, and it may be of interest for other researchers to explore in future studies.
17
“a sweet cake”, or “coll. a young, attractive woman”. SJP PWN also includes metaphorical meanings based on similarity and analogy with the basic meaning. For example, the Polish noun głąb denotes ‘cabbage stalk’, but it also has a secondary meaning: “blockhead”; Polish Kozak means “Cossack”, and “a person showing off with bravado and courage”; hiena means “hyena”, but it also denotes “an unscrupulous person who takes advantage of others, preying on someone’s misfortune”; miś denotes a “bear”, but it also refers to “a person, who has some features of a bear, especially considerable weight”; papuga means “parrot”, but it also denotes “a person who repeats another person’s words without reflection”. Again, it is the first one of the two meanings which was analyzed.
Included in the analysis will be adjective-shaped nouns (e.g. wygrana “win”, czesne “tuition”). The generalizations stated in terms of morphophonology and semantics also apply to them so they nicely enter the picture of (near) complete non-arbitrariness of GA in CSP, though it is unclear whether some adjective-like words are full-fledged nouns (McShane, 2003, p. 83).22 SJP PWN includes 60 cases of such adjectival nouns and Appendix I contains more examples. Some ‘profession nouns’ are in this set and some of them are overtly gender marked (e.g. królowa “queen”, krawcowa “dressmaker”).
Some Polish nouns may appear (in some theoretical accounts) problematic for the proposed model of GA, in particular, where it is claimed that both semantic and morphophonological-phonetic factors (i.e. derivation) are relevant in the process of attributing singular nouns of CSP to a gender category.
The first specific factor that underscores the difficulty here is gender instability (vacillation). It has already been pointed out that GA in CSP is usually made without hesitation, though the vacillating gender of such nouns as bonsai “bonsai tree” (m/n), fatamorgana “mirage” (f/n), gnu “gnu” (f/n), karibu “caribou” (m/n), pekari “peccary” (m/n), richelieu “richelieu embroidery” (m/n), taxi “taxi” (f/n), toffi “toffee” (m/n), shows that it is not always the case with loanwords.23 The native nouns chłopaczysko/chłopczysko “big fellow” (m/n), chłopisko “big fellow” (m/n), and ladaco “rascal” (m/n) also constitute cases of indecision as far as their gender values are concerned. In the context of Steinmetz’ model of GA, the vacillating gender of the above mentioned nouns is understood as resulting from competition between assignment rules (i.e. between morphophonological-phonetic and semantic regularities), or it might be instead relegated to arbitrariness and idiosyncracy. By way of example, the loanword taxi “taxi” vacillates between feminine and neuter gender as a result of the competition between an available Polish lexical-conceptual equivalent taksówka “taxi”, which is of
22 Note that once GA of singular nouns in the nominative singular has been approached strictly from the morphological level (i.e. GA is sensitive to the (in)declinability of nouns), adjectival nouns would in consequence be problematic for the model and would have to be removed from the analysis as the value of gender would necessarily have to be associated with inflectional endings from nominal declension.
18
feminine gender, and the strategy that inanimate nouns ending in a vocalic phoneme other than /+a/ [+a] are statistically likely to be neuter.
Every effort has been made to address as many different sets of the lexicon as need be addressed. Included in the analysis will be ‘profession nouns’. Appendix II contains some examples. Some ‘profession nouns’ (especially those that hold low prestige) are are overtly marked for feminine gender. These happen to be professions traditionally practiced by females only and no men seem to enter these professions nowadays (e.g. gorseciarka “corset maker”, hafciarka “embroiderer”, kosmetyczka “beautician”, manikiurzystka/manicurzystka “manicurist”, pokojówka “maid”, przedszkolanka “female kindergarden teacher”, sprzątaczka “cleaning lady”), and few feminine derivatives (e.g. alpinistka “female mountain climber”, konduktorka “female conductor/female ticket collector”, lekarka “female doctor”, narciarka “female skier”, nauczycielka “female teacher”, fryzjerka “female hairdresser”).
Not included will be animate nouns (denoting human beings) ending in /+a/ [+a] in Nom.Sg. whose gender is masculine or feminine, depending on the sex of the denotatum in a situational context, and of which kaleka “cripple” kalek/+a/ [kalɛk+a] is a representative. Unlike kaleka, however, most of these nouns are derivatives, or compounds, and they are expressives, connoting deteriorative meaning (Stefańczyk, 2007, p. 35). Reference grammars of Polish point out that they are a marginal group within the nominal lexicon of CSP. SJP PWN lists 54 such nouns. A full list is given in Appendix III. In Section 10.3. I explain, in detail, how kaleka “cripple”-type nouns are different from ‘profession nouns’ and why they pose a problem for the mechanisms of Steinmetz’ (1986, 2006) GA theory.
Finally, removed from the analysis will be two nouns: państwo in the meaning “Mr. and Mrs.”, and wujostwo in the meaning “uncle and aunt”. Semantically and syntactially, these two nouns are close to other plural nouns, such as for example małżonkowie “married people” (Nom.Pl.) or narzeczeni “an engaged couple” (Nom.Pl.). That is to say, the nouns państwo “Mr. and Mrs.”, and wujostwo “uncle and aunt” cannot be put into the following context for the nominative singular: dobr-eN ____ . In short, I distinguish in the study between państwo “Mr. and Mrs.” and państwo “state”, where the second of these homonymous nouns takes neuter singular agreement.
10. Applying the mechanisms of OGAT to CSP - A Provisional Analysis
The regular behaviour of GA in CSP provides evidence for statistical data alive in the mind of a Polish adult native speaker. In this study it is modelled via OT GA a la Rice (2003, 2004, 2005, 2006).
Building on Steinmetz (1986), Rice (2003) proposes formalizing GA within a theory designed specifically to mediate conflicts between violable constraints and he presents OGAT. Namely, he joins together Steinmetz’ Gender Tally and Gender Eclipsis with Prince and Smolensky’s (1993) OT. Rice (2004, 2005, 2006) states that OGAT is an example of crucial equal ranking, a formal option allowed by
19
Prince and Smolensky, for which they did not find evidence back in the 1990s. The tenets of OGAT can be summarized as follows:
(1) OGAT: across languages, gender features constraints and language-specific gender features constraints are crucially non-ranked with respect to each other; only the genders are ranked in a markedness hierarchy in each language; the language-specific gender features constraints are all non-ranked with respect to each other but form a block that, as a whole, is non-ranked above the language’s gender markedness hierarchy. The cumulative violations of the gender features constraints determine the gender category of a noun in the case of an imbalanced conflict. When an equal number of gender features constraints militate for assignment to different gender categories (Rice dubs it a balanced conflict throughout his papers), the noun is assigned to the least marked of the conflicting gender categories (Rice, 2004, 2005, 2006).
10.1. Markedness relations in CSP based on SJP PWN and Index a tergo for Doroszewski’s dictionary
Since gender markedness hierarchy is such an important issue in Steinmetz’ (1986) theory and to OGAT, it is important to decide if the three genders in CSP occur in marked and unmarked oppositions and, if so, what their status is relative to each other. Steinmetz (2006, p. 1422) argues that the default hierarchy m > f > n, posited for German, also holds for all Slavic languages.
A survey of both SJP PWN and Index a tergo compiled from Doroszewski’s (1958-1969) general dictionary Słownik języka polskiego (henceforth, SJPD) (featuring approximately 125,000 entries) was made and a careful examination of the correlations between the morphophonological-phonetic regularities on the one hand, and the gender value on the other hand, provides enough support for the proposed rules in my study. In the process of extracting the regularities, the effect of final-devoicing, a phonological process described in Section 5.1. was obviously taken into account. My basis for designating masculine as a pre-syntactic default gender value in CSP, is the observation that masculine nouns, the largest set of nouns (see my discussion below), are not significantly gender marked in their Nom.Sg. form in SJP PWN.24 By contrast, feminine and neuter gender exhibit a higher degree of markedness, especially in terms of formal (morphophonological-phonetic) characteristics. In other words, the number of morphophonological-phonetic rules required to cover all masculine nouns in CSP is far greater than the number of rules needed for feminine and neuter. In consequence nouns belonging to the masculine gender in CSP are less homogenous than other, more marked categories.
24 The present analysis is inspired by Steinmetz’ (2006) structure of the argument (see for example a gender table for a noun мужчи́на “man”, Steinmetz, 2006, p. 1427), with the difference being that the grammatical default value (for Russian) is the least marked morphosyntactically. In this study, the default is the least marked morphophonologically.
20
The number of productive rules needed to account for all masculine nouns in CSP Polish can be replaced with a single default rule.
As previously claimed in Steinmetz’ default-based theory of GA the justification for and determination of markedness relations is also established based on statistical frequency of the genders (Steinmetz, 2006). Based on a repertoire of nouns listed in SJP PWN, I find the following distribution of the three genders: masculine 47% (8811/18,488), feminine 40% (7524/18,488), neuter 12% (2142/18,488), and vacillating 0,06% (11/18,488). Masculine and feminine gender are significantly distinguished by these statistics. Also, masculine gender can be considered to be the least marked gender in CSP, not least of all because of the ease of assigning nouns to the feminine gender through reference to /+a/ [+a] (see Section).25 Neuter nouns are clearly in the minority in comparison to masculine and feminine nouns.26 Interestingly, the largest group of lexemes (amounting to in SJP PWN) are morphophonologically determined de-verbal nouns such as czytanie “reading”, spanie “sleeping”, mówienie “speaking”, mycie “washing, cleaning” (cf. Stefańczyk, 2007, p. 35-36).
Using Rice’s (2004, 2005, 2005) optimality theoretic approach to markedness, I propose three markedness constraints, each of which constraints each gender:
(2) *MASCULINE: A noun is not masculine. *FEMININE: A noun is not feminine. *NEUTER: A noun is not neuter.
Since it has been postulated that masculine is the least marked category in CSP followed by feminine and then neuter, Gender Eclipsis for the whole gender system of CSP is m > f > n (i.e. masculine gender outranks both feminine and neuter gender and feminine gender outranks neuter gender). This can be formalized as in (3):
(3) *NEUTER >> *FEMININE >> *MASCULINE
In this ranking, the prohibition against assigning a noun to the neuter category in CSP, is the highest ranked.
Given the ranking in (3), the Polish noun, stół “table”, for example, is the focus for the issue of markedness as it is in the domain of no gender features constraints. It is assigned to the masculine
25 See Rice (2004, p. 2) and his formulation of this argument for the grammatical GA system of Russian.
26 Enough support for the weakness of the neuter gender can be found in Stefańczyk (2007, pp. 35-38), where in the general dictionary of CSP Uniwersalny słownik języka polskiego (USJP) (Dubisz, 2003) only around 4,000 or 10% of the 44,000 nouns are assigned neuter gender. The dictionary records around 1,960 neuter nouns with their typical and characteristic (only of them) productive nominative singular ending /+e/ [+ɛ]. The remaining 90% are masculine and feminine (Stefańczyk, 2007, p. 157).
21
category, as presented in Tableau 1. Tableau 1 shows the manner in which constraints compete with one another to ultimately select the optimal candidate. In any Tableau, input is placed into the upper left-hand corner of the Tableau and possible output candidates (gender categories) are listed below the input form in the candidate column. Any time an output candidate violates a constraint, an asterisk (*) is placed in the corresponding cell. Some violations are fatal violations, that is, they constitute violations serious enough to prevent a given output candidate from being chosen as optimal. An asterisk followed by an exclamation point (*!) indicates fatal violations. The last candidate not to receive an exclamation point is selected as optimal (marked with the manual indicator ). Throughout the data analysis section, a “-” separates the root from suffix and a “+” separates the stem from the ending (desinence). Tableau 1. stół “table” [stuw+Ø] *N *F *M a. [stuw+Ø], m. * b. [stuw+Ø], f. *! c. [stuw+Ø], n. *!
Candidates (a-c) in Tableau 1 represent the possibility of assigning the noun stół “table” to any of the three gender categories. It is the job of constraint hierarchy (which, in this study, is conditioned by formal properties, that is by final sounds that are characteristic of a noun as a lexeme operating on a pre-syntactic level) to identify which of the three candidates is optimal. Candidate (c) is promptly eliminated (indicated by !) since it violates the most highly ranked constraint of the markedness hierarchy *N, followed by candidate (b) (indicated by !) for violating the next most highly-ranked constraint of the markedness hierarchy *F. Candidate (a) also incurs a violation from the constraint *M, but nevertheless is selected as optimal, since its violation is considered the least severe among the three. In sum, the constraints in Tableau 1 moved from most marked to least marked in a left-to-right fashion (‒>). Recall that the gender value is a basic condition for gender agreement to take place in language production.
Since not all nouns in CSP are masculine, there must be ways to override the markedness hierarchy (see Section below).
10.2. Gender features constraints in CSP based on SJP PWN and Index a tergo for Doroszewski’s dictionary
In Section 10.2.1 and 10.2.2 I present a set of assignment rules in CSP (with a certain extent of rule overlap) that generates gender for the nouns found in SJP PWN and in Index a tergo for Doroszewski’s dictionary. Each rule will be formalized as constraint, in anticipation of my OT analysis. At this stage of the analysis, cases of vacillating gender, kaleka “cripple”-type nouns, have already been ignored. The rule set rests upon the assumption that CSP has masculine as its pre-syntactic default gender among
