Parts of speech and their classifications
1. Parts of speech are the functional classes of lexical units in a natural language, identified on the basis of either grammatical or semantic-grammatical criteria. The idea of dividing words into classes distinguished on the basis of their grammatical roles dates back to Ancient Greece. A Greek grammar by Dionysius Thrax defined eight parts of speech (name, pronoun, word (verb), participle, adverb, preposition, conjunction, article). In Latin grammar (where no articles occur), the list was extended by adding interjections. This classification survived in the grammars of many modern languages until the 18th century, when the name
was subdivided into noun, adjective and numeral, the participle was reclassified into a form of verb and the category of a particle was added (Jodłowski 1971: 12; cf. also Bogusławski, Drzazgowska 2016: 598–622, 632–634; Arnauld, Lancelot 1991; Stankiewicz 1994; Bartmiński, Nowosad-Bakalarczyk 2003).
In the history of grammar, Polish grammar included, we come across other classifications: some are based on logical categorisations, other consist in the enumeration of classes (typologies), while some combine both of these approaches. It is impossible to reconcile the logical correctness of a classification (given frequent deviations from the rules), the homogeneity of the classification criteria, and their substantive reasonability (both sufficient generality and specificity, as well as precision). For instance, nouns are used to denote concepts other than things (cf. e.g. biel, siwienie, wierność), not all nouns are inflected for case and number (cf. e.g. atelier, żelazo, sanki), and not all adjectives are inflected (cf. e.g. bordo, orange, khaki). Prepositions, conjunctions, and relative pronouns are quite commonly described as connective lexemes, but – given the distinct grammatical nature of the objects connected by each of these classes – the term ‘connective lexemes’ is not explanatory, but rather serves as a general label replacing a description. ‘Independent/ dependent lexemes’ are similarly empty
‘shell’ labels. Classifications by enumeration, which arouse major – and justified – methodological doubts, may be more exact in terms of reflecting the classes of units and the differences between them than categorisations into mutually exclusive subsets.
No grammar can be constructed without classifying words into parts of speech. To determine the rules for combining the units of a language into structures, one needs to have unit classes and descriptions of their properties. On the other hand, every attempt at identifying word classes must presume that neither the nature of specific classes nor their sizes would be identical. Some classes of parts of speech are open-ended, meaning that new words may be added to them; typically such words refer to certain elements of the world – for instance, nouns or verbs may be coined depending on the needs of language users, while the categorisation of new units into such classes is forecast-based (Laskowski 1981). By contrast, other classes of parts of speech, e.g. particles or conjunctions, are closed. These classes, distinguished on the basis of linguistic research, are relatively stable and have no referents in the non-linguistic world (the words revealing the speaker have a distinct status in this respect), and refer to other linguistic objects, either co-existing with or required by them. This distinction dates back to the Aristotelean opposition of categorematic and syncategorematic expressions, popularised in Polish academia by Tadeusz Kotarbiński (1961: 645), and in Polish linguistics by Stanisław Karolak (1990, 1999: 639–641, 646–647, 650, 2002). Nevertheless, we should emphasize that there is no exact correspondence as regards the opposition between expressions with extratextual and intratextual function (one of the many pairs of terms used by Karolak in this context) and the opposition between the parts of speech classes (cf. also Grochowski 2014b).
A grammatical categorisation deals with lexical units (lexemes), regardless of the number of graphic/phonological words they are composed of (Bogusławski 1976, 1978). Before attempting a classification of units, it is necessary to solve the problem of the grammatical homonymy of lexemes, at least for the purposes of the specific classification (cf. e.g. Danielewiczowa 2012ab; Grochowski 2014a; Grochowski, Kisiel, Żabowska 2014). It is an undeniable fact that certain units have an identical form but belong to different classes.
However, no single commonly accepted grammatical classification of such units has been developed. The history of Polish linguistics confirms the view that the number of grammatical unit classes depends on the categorisation criteria, the extent of their granularity and precision, the purpose of the research and the researcher’s intent. The evolution of research on parts of speech in Polish language studies is reflected in the literature on general and Polish language linguistics (cf. Bogusławski, Drzazgowska 2016; Heinz 1978; Jodłowski
1971; Laskowski 1999ab; Porzeziński 1923; Skarżyński 1994, 2001; Ulitzka 2008; Urbańczyk 1993). This study outlines the evolution of the criteria applied to classify the language units and a review of such classifications proposed in the 20th century, with a special focus on its second half, and at the beginning of the 21st century.
Yet another metagrammatical problem is to ensure that the distinguished lexeme classes are appropriately defined. Even if the opposing classes are mutually exclusive and the units that have not yet been examined may be assigned to them, this is insufficient to conclude that it is feasible to develop independent definitions for specific classes. Usually, a definition of a class contains the names of other classes to either reveal the contrast between the two or to demonstrate that a specific type of relationship with objects from another class is a constitutive characteristic feature of the class being defined. Vicious circles in definitions of specific parts of speech could be avoided (direct vicious circles, like “an adverb is a part of speech that connotes a verb, an adjective or another adverb” being the most easily avoidable) on condition of constructing an appropriate metalanguage of grammar and its application to describe parts of speech. It is quite likely that the authors of classifications were completely unaware of this problem.
2. In the 19th and in the first half of the 20th century, the prevailing criteria for the classification of words (lexemes) into parts of speech relied on semantics. One of the most wellknown categorisations of words into ten parts of speech, based on their enumeration, was introduced into Polish grammar in 1921 by Stanisław Szober (1957: 92–105). It continued to be in widespread use in linguistic literature until the end of the 1960s. Szober distinguished between the signs of feelings (interjections) and the signs of the objects of thought, which he further divided into signs of objects (nouns and pronouns), properties of objects (adjectives, numerals and pro-adjectives), states and activities of objects (verbs), properties of properties (adverbs, pro-adverbs and numeral adverbs) as well as signs of relations (conjunctions, prepositions and particles). This categorisation overlaps with the other two – one based on the functional value of words and on their relation to other words in a sentence. From the perspective of functional value, Szober distinguished between signifying words (interjections, nouns, adjectives, verbs, adverbs except for numeral adverbs, conjunctions, and prepositions), deictic words (pro-forms and particles) and signifying-deictic ones (numerals and numeral adverbs). The words whose “function consists in expressing the relations between other words
in a sentence” (Szober 1957: 92) are referred to as non-independent and include prepositions, conjunctions, and particles, all other words being considered as independent.
Almost in parallel to Szober, Jan Łoś (1923: 287–408) constructed his own system of parts of speech. This was laid down in the first chapter of his “Składnia” (the fifth part of “Gramatyka języka polskiego” by T. Benni, J. Łoś, K. Nitsch, J. Rozwadowski, H. Ułaszyn of 1923) – an improved version of its first edition of 1915 (when it made part of the two-volume collective study entitled “Język polski i jego historia z uwzględnieniem innych języków na ziemiach polskich”). Importantly, Łoś enumerates the very same ten parts of speech as Szober, in separate paragraphs, without organising them into groups in any way. He starts every description with an important property of the specific part of speech. Such properties are not subordinated to any single specific category. A noun is the name of an object, an adjective denotes a property and, in a sentence, it either functions as a modifier or as a predicative expression. Ordinal numerals have the same syntactic functions as adjectives; by contrast, cardinal numerals jeden, dwa, trzy, cztery belong to the category of adjectives, whereas the remaining ones are nouns. Pro-forms express the concept of substance or property, and are subdivided into pronouns and pro-adjectives. The properties of a verb include voice, manner (aspect), mood, tense, person, and number. Adverbs modify verbs. A preposition is a formal word, with the very same syntactic function as a case-marking ending. Conjunctions bind words or sentences and highlight the coordinate or subordinate relationship between them. Interjections are not parts of a sentence, but rather a linguistic symbol of a speaker’s feelings (Łoś 1923: 288, 325, 326, 327, 332, 355, 360, 364, 370).
Henryk Gaert ner (1938: 144–206) developed a system of parts of speech or, in other words, the categories of words in grammar, which differs markedly from Szober’s grammar in terms of its much more extensive yet original terminology, and the granularity of categorisation, rather than by the substantive concept in itself. Among the words that may be independent expressions, Gaert ner (1938: 5) distinguishes intentional utterances, also referred to by him as znaczniki (literally ‘markers’), which are “the result of a willful intent of the speaker,” and reflective utterances (interjections) “which are physiological reflexes arising as a result of a sentimental state.” The former include nouns, adjectives, numerals, proforms, verbs, and adverbs. Independent connectors (conjunctions and prepositions) cannot form utterances alone, as they create utterances only when combined with one more ‘marker’, e.g. przy domu, aby kupić, niech zrobi (Gaert ner 1938: 6). Certain word classes have been categorised in detail from the perspective of their meaning and function. Gaert ner classifies nouns, for instance, into concrete and abstract, animate and inanimate, common and proper, always singular and always plural (singularia and pluralia tantum). Meanwhile, the gradabil-
ity of adjectives and adverbs is treated as the function of these word classes. Pro-forms are divided on the basis of their relation to names into pronouns, pro-adjectives and pro-numerals, as well as separately on the basis of their semantic function – here the scholar lists, among other things, personal pronouns, deictic pronouns, possessive pronouns, distributive
pronouns, negative pronouns, and relative pronouns. He further introduces a multi-level classification of verbs based on such categories as aspect, repeatability and the type of activity. On top of this, he discusses numerous functions of impersonal verb forms. Adverbs are subdivided into dozen or so classes relying on the type of the examined circumstances, e.g. adverbs of degree and quantity, frequency, order, effect.
The system of parts of speech outlined in the grammar by Zenon Klemensiewicz (1939, 1960), belonging – according to its author – to word studies, as opposed to sentence (syntax) studies – does not differ substantially from the point of view of methodology from the concepts put forward by Szober, Łoś, and Gaert ner. Klemensiewicz (1960: 49–66) distinguishes ten parts of speech by enumerating and describing them primarily from the semantic point of view. A noun denotes an object in a broad sense, both perceptible by the senses and inaccessible to them, but it is also a name for a detached quality (Klemensiewicz 1960: 49–50). Adjectives denote the properties of objects, and since the quantity of properties is immense, the scholar extends his description by giving examples: properties include colours, taste, smell, sound, propensity, the property of intent and feelings. An adverb is a part of speech that denotes a property of an activity, a state or yet another property, or specifies their place and time. A numeral is used to denote the number of objects. Verbs denote activities or states. A preposition is “an invariable word that constitutes a syntactically inseparable semantic unit with a noun or a pronoun.” A conjunction is “an invariable word that connects two words or two sentences.” An interjection “externalises the emotional state or the intent of the speaker.” A particle “prepares the listener to correctly understand an entire utterance or emphasizes, highlights the importance of its parts” (Klemensiewicz 1960: 64–65). Additionally, the scholar distinguishes pronouns, pro-adjectives, pro-adverbs and pro-numerals, but fails to provide a general description of a pro-form as a part of speech. The descriptions of these subclasses, however, clarify that a pro-form replaces a specific part of speech, and performs an analogous function in a similar way.
Witold Doroszewski (1952: 132) classifies words into parts of speech on the basis of the criterion of inflection. He divides them into inflected (variable) and uninflected (invariable) ones (prepositions, conjunctions, particles). The first group is subdivided into words inflected for case (these include names: nouns, adjectives, pro-forms and numerals) as well as for person and tense (verbs). Unlike nouns, adjectives are inflected for gender, while unlike numerals both nouns and adjectives are inflected for number. Nouns form a class of modified words, while adjectives, numerals, pro-forms and verbs are modifying words. Adjectives perform a modifying function with respect to a noun and agree with the noun in terms of gender, case and number. Verbs in personal forms sometimes perform a modifying function with respect to the subject of a sentence. The syntactic equivalents of the inflection-based classification of parts of speech into names and verbs are the subject and the predicate (Doroszewski 1952: 201, 235, 282).
In his classification of words into parts of speech, Stanisław Jodłowski (1976: 13–24) relies on mixed criteria, combining the semantic values of words and their formal properties. The former include the “stylistic attitude of the speaker,” the “epistemological categories of words” and the “denotation technique,” while the latter contain the syntactic functions and morphological features. According to Jodłowski (1976: 14), each word represents one of three stylistic attitudes: expressive (e.g. ach, och, niestety, ba), impressive (e.g. no, hej, hola, hop, halo) or denotative (all the other words that have none of the two previously mentioned functions). Epistemological categories concern the manner of expressing content, and are further divided by the author into objective (e.g. zieleń, niepokój, smutek), qualitative (e.g. zielony, bliski, wesoło), quantitative (e.g. cztery, sto, mało, dużo), activity-based (e.g. śpiewa, paruje, rozpacza) and relational (e.g. tu, tam, obok, nad, pod, albo, ale, ponieważ). The denotation technique
is the way in which a word is linked to its signified, and involves prodding the recipient towards his or her memory resources (mnemonic or memory technique) or an external situation (non-mnemonic, or pro-form technique). Jodłowski (1976: 17–18) reduces the syntactic criteria to the primary functions of specific parts of speech in a sentence, and whenever he refers to morphological properties, he means the properties typical of a given class.
In view of these criteria, Jodłowski (1976: 18–23), assigns the following sets of properties to the eleven parts of speech that he differentiated by enumeration: 1. A noun is a denoting, mnemonic and naming word, capable of performing the role of a subject. 2. An adjective is a denoting, mnemonic, non-naming word, expressing a property of an object, capable of performing the role of a noun modifier. 3. A numeral is a denoting, mnemonic, non-naming word, expressing a quantity or order, capable of performing the role of a noun modifier. 4. A verb is a denoting, mnemonic, non-naming word, expressing an activity, process or state; personal forms of verbs may perform the role of a predicate. 5. An adverb is a denoting, mnemonic, non-naming word, expressing a property of an activity, capable of performing the function of an adverbial. 6. A pro-form is the name of a class that is non-homogeneous according to each of the criteria; it consists of pronouns, pro-adjectives, pro-adverbs and pro-numerals; the author generally characterises pro-forms as denoting non-naming words. 7. A preposition is a denoting, mnemonic, non-naming word, expressing subordinate relationships between words. 8. A conjunction is a denoting, mnemonic, non-naming
word, “marking coordinate relations between words and relations between sentences other than denoted by relative and interrogative pronouns introducing a dependent clause”; it performs the function of a marker of conjunction. 9. A modal expression (in Polish: modulant, literally ‘modulator’) (“an element of the broader framework of an utterance,” a substitute for the term “particle”) is a mnemonic, non-naming word, “expressing a situation, logical value, modal attitude, emotional state, or linguistic commentary of an utterance.” Within this class, the author distinguishes situating modulators (e.g. tylko, także, właśnie), valorising modulators (e.g. tak, nie, owszem, naprawdę, istotnie), modal modulators (e.g. na pewno, chyba, niech, oby), and affective modulators (e.g. no, raczej, aż, na szczęście) (cf. also Jodłowski 1971:
97–114). 10. An interjection is a mnemonic, non-denotative, expressive word (e.g. ach, oj, brr). 11. An imperative particle (in Polish: nakaźnik) (e.g. no, nuże, hej, hola, halo) is a mnemonic, non-denotative, impressive word.
3. The theory of parts of speech developed by Jerzy Kuryłowicz (1936/1979) has had a major impact, whether direct or indirect, on all further concepts and classifications of parts of speech developed within European structuralism. Kuryłowicz assumed that 1. parts of speech have two common properties – they have a symbolic function and are words (within the meaning outlined by Antoine Meillet, 1921, they combine a lexical and a grammatical value) and that 2. certain specific correlations occur between parts of speech and the constituents of a sentence, as every part of speech performs more than one function in a sentence. The basic, unmarked function is referred to by Kuryłowicz (1979: 148) as the primary function, the remaining ones being secondary. The scholar constructed a system of basic parts of speech composed of four classes: nouns, adjectives, verbs and adverbs. Their primary functions are those of a subject, a noun modifier, a predicate and an adverbial respectively. Neither interjections nor pro-forms meet the first condition to be considered as a category of parts of speech, as they do not have a symbolic function, but rather perform an expressive (the former) and a deictic (the latter) role. The second condition laid down in the definition of parts of speech is not met by prepositions, conjunctions, cardinal numerals and articles, as, rather than words, they are morphemes (Kuryłowicz 1979: 154–155).
Tadeusz Milewski (1965) put forward a classification of parts of speech based on the syntactic-semantic criteria. For teaching purposes, the classification is presented in the most approachable way in his linguistics textbook, where he used the example of the Polish language to discuss the semantic and syntactic systems of a natural language. The components of the semantic system comprise naming, deictic, and ordinal words. A word, by contrast, is defined as a syntactic component constructed in a specific language according to the same scheme as the other syntactic components; in the Polish language, it is a “set of morphemes which, as a whole, performs the function of a sign for a specific phenomenon, and, at the same time, defines its position within a sentence” (Milewski 1965: 86). According to this scholar, to be classified as being the same part of speech, words must “perform the same function in the connotation system and have the semantic features corresponding to it” (Milewski 1965: 85, 97, 115). On the basis of the (categorial) connotation criterion, Milewski divided naming words into primary (nouns), secondary (adjectives and verbs) and tertiary (adverbs). Nouns
(e.g. dom, wilk, białość, pisanie) are semantically complete and do not connote anything, but are connoted by secondary words. Adjectives connote and modify nouns, creating nominal groups, e.g. biały dom, zgniły owoc. Non-transitive verbs connote a subject in the nominative case, e.g. pies siedzi, while transitive verbs further connote a direct object, e.g. matka kocha córkę or additionally also an indirect object, e.g. ojciec daje książkę synowi. Secondary words, adjectives and verbs, are not connoted by anything. Adverbs, e.g. dobrze, wysoko, bardzo, górą, connote adjectives and verbs, producing two syntactic strings: adverb – adjective – noun, e.g. wysoko kwalifikowany robotnik, and noun – verb – adverb, e.g. ptak leci wysoko. Deictic and ordinal words perform the same function as naming words. Milewski (1965: 107–108) divides both categories applying the same rules. Primary deictic words are pronouns, e.g. ja,
ty, ten, on, the secondary ones are possessive pronouns, e.g. mój, twój, nasz, wasz, while the tertiary deictic words are pro-adverbs, e.g. tu, tam. Primary ordinal words are cardinal numerals, e.g. dwa, trzy, cztery, pięć, the secondary ones are ordinal numerals, while the tertiary ordinal words are adverbial numerals, e.g. dwojako, trojako (all examples after Milewski) (cf. also Milewski 1952/1971).
According to Milewski’s concept (1965: 87), prepositions, conjunctions, particles or interjections are not syntactic components, and as such they are not words. They are not parts of speech, but rather loose morphemes, performing the function of markers of grammatical categories. The relationship between the constituent sentences of a complex sentence is expressed through relative pro-forms and conjunctions. A relative pro-form is a component of the subordinate clause that it connotes, while pointing to one of the components of the main clause. Conjunctions “do not point out, but rather connote, and they do so with respect to two slots specifically – the one before and the one after themselves,” (Milewski 1965: 113).
Henryk Misz (1967) was the first to put forward a distributive classification of syntactemes with regard to the Polish language. The distributive criterion was applied as a result of an analysis of the connotation of syntactemes using substitutive-transformational methods. Syntactemes form a class of words and word forms with an identical syntactic value. This syntactic value of a word and a word form is defined as a set of properties that are syntactically relevant, and thus used to create correct syntactic structures (Misz 1967: 26–28, 30–31). The process of creating structures involves syntactemes performing the determinative (head) and determined (dependent) function. The lack of such a function means that certain words have only (a) form-formative (e.g. by, niech in: można by, niech zginie), (b) word-formative (e.g. bądź, indziej in: gdzie bądź, gdzie indziej) or (c) dictionary function (e.g. może, nawet in: może nie ma pani na to czasu, zakazywano nawet oświetlania ulic), but have no syntactic
value (Misz 1967: 40).
In his classification (by enumeration), Misz distinguished seventeen classes of syntactemes and the three classes without a syntactic value referred to above. Leaving aside the classes and subclasses (of parts of speech) identified in traditional grammars, one needs to mention the three classes that had not appeared in the earlier grammar descriptions. These include intensifiers (dependents of adjectives and adverbs), e.g. bardzo, zbyt, trochę, całkiem, dosyć, niemal, mocno, lekko (Misz 1967: 55), identifying functors to, to jest, to znaczy, równa się, znaczy (cf. Czas to pieniądz., Zrozumieć to wybaczyć.) and comparative functors jak, jakby, niby, niczym, niż, niźli, aniżeli (e.g. tańczy niby baletnica, zgrabniejsza niż baletnica) (Misz 1981; Bałabaniak 2013).
4. The first classification of lexemes compliant with the rules of logical division based on grammatical criteria (with reference to inflected – morphological – and non-inflected – syntactic – parts of speech) was proposed in 1974 by Zygmunt Saloni. It was applied, with very few modifications, in several consecutive editions of the formal structuralist syntax textbook by Zygmunt Saloni and Marek Świdziński (1981/1998). At the top level of the classification of inflected lexemes, Saloni contrasted lexemes inflected for case with lexemes which were not. The first group, the one inflected for case, was further subdivided by him into lexemes not inflected for gender (nouns) and those inflected for gender, which in turn were subdivided into lexemes inflected for number (adjectives) and not inflected for number (numerals). Lexemes not inflected for case were, on the other hand, classified into lexemes inflected for person (main or personal verbs, in Polish czasowniki właściwe) and lexemes not inflected for person (defective or impersonal verbs, in Polish czasowniki niewłaściwe literally ‘improper’). At
the highest level of classification of non-inflected lexemes, Saloni (1974: 98) contrasted lexemes which are not used independently against lexemes used independently (interjections). The first group was divided into lexemes with and without a connective function (particle-adverbial lexemes). Connective lexemes were subdivided into those that do (prepositions) and do not trigger a specific case (conjunctions).
A commentary is needed to modify the scope of the classes distinguished applying the abovementioned oppositions. Adverbs derived from adjectives, such as jasno, jaśniej, najjaś-
niej are interpreted by Saloni (1974: 11) as a subclass of adjectives with neutral forms with respect to case, number, and gender, used with a verb or an adjective. The class of adjectives further includes adjectival participles, pronouns inflected for case, gender and number (such as ten, taki, mój), ordinal, manifold (e.g. piąty, trojaki), multiple (e.g. pięciokrotny), multi-
plicative (e.g. potrójny) numerals, as well as adverbs derived from them, such as trojako, pięciokrotnie, potrójnie. The class of numerals is limited to “traditional” cardinal numerals and pro-numerals (such as ilu, ile). Collective and fractional numerals are classified by Saloni (1974: 99) as nouns, while lexemes such as dwakroć are seen as particle-adverbs. Collective numerals were included into the cardinal numerals paradigm slightly later (Saloni, Świdziński 1998: 197). Finally, adverbial participles and the infinitive are interpreted as neutralised
verb forms (Saloni 1974: 94).
Roman Laskowski (1984, 1998) put forward two functional classifications of lexemes based on syntactic criteria (motivated semantically, rather than formally) in the two editions of the GWJPM. The second was developed as a result of a broader debate on the nature of inflection (Bogusławski 1987; Grzegorczykowa 1987; Bańko 1987; Laskowski 1987) and on the first classification (cf. Grochowski 1986; Gruszczyński 1987; Bobrowski 1995; Wróbel 1996), of which the scholar himself also became aware many years later, when discussing the history of research on the theoretical foundations of Polish grammar (Laskowski 2012; see also Grochowski 2019). The author introduced a general division of lexemes into syntagmatic units (in Polish syntagmatyki; units capable of entering into syntactic relationships
with other units) and asyntagmatic units (in Polish asyntagmatyki; units which do not enter into syntactic relationships with other units); cf. Laskowski (1984: 30, 1998: 58). The latter are primarily syntactically independent expressions, being independent utterances (interjections) or expressions with the primary function of an utterance, but contextually dependent, implying the existence of a text (appositions, e.g. owszem, istotnie, właśnie, czyżby?). It is thanks to the GWJPM that the term “apposition” (in Polish – dopowiedzenie) as a part of speech gained popularity in literature (cf. e.g. Dobaczewski 1998; Grochowski 1997; Żabowska 2011); however, the classification of appositions as a class of units is being disputed (Wajszczuk 1997: 21–60, 2005: 73–104). Interjections were subdivided into onomatopoeia and appeals, based on their secondary syntactic functions; the former may occur secondarily in the predicative function (cf. Szklanka bęc na podłogę., Kula tylko bzz koło ucha.), while the latter may enter into syntactic relations with the vocative, and some also with the imperatives (cf. O rany, chłopie!, Halo, Marysiu, poczekaj!). The predicative function of onomatopoeia is disputable; onomatopoeia in the position of the head of a phrase it cited (Grochowski 1996), but not applied, which is confirmed e.g. by the fact that it cannot be assigned any grammatical categories relevant to a verb.
Syntagmatic lexemes are parts of phrases; they primarily perform the function of a component of a phrase or the indicator of the syntactic relation between components. The former are autosyntagmatic, whereas the latter are synsyntagmatic lexemes (Laskowski 1984: 31); in
the second edition of the GWJPM, the terms “syntactically independent” and “syntactically non-independent lexemes,” much less popular in contemporary grammar, are used (Laskowski 1998: 59). The concept of synsyntagmacity and its relationship to the concept of a (non) independent word (component) and (the lack of) syntactic/utterance-related independence have been analysed by Ewa Walusiak (2011).
Autosyntagmatic lexemes primarily perform the function of the head of a phrase or the function of a dependent component in a phrase. Laskowski (1984: 33–34) divides the first group – predicatives – into personal and impersonal ones (e.g. trzeba, widać, szkoda, rad, go-
tów). Lexemes that perform the function of a dependent component in a clause are the components of a noun phrase that play the role of the head (nouns and pronouns), a dependent component of a noun phrase (adjectives and numerals) or are not part of a nominal group at all (adverbs and modalizers). Unlike nouns, pronouns do not accept attributive expressions, are not used predicatively, and have the selective category of number. The fundamental difference between adjectives and numerals (this class being limited to cardinal and collective numerals) lies in the fact that while the former are unilaterally grammatically determined by the head of the nominal group, the latter occur in a bilateral grammatical determination relationship with the head of the phrase: the numeral determines the case and the number of the noun, while the gender of the numeral is determined by the noun (Laskowski 1984: 36).
Adverbs are autosyntagmatic lexemes that primarily perform the function of a dependent in a phrase. The class of so-called modalizers (modal expressions, modalizatory in Polish) – autosyntagmatic lexemes that “are characterised by freedom in terms of filling various syntagmatic positions within a phrase” (cf. also chyba, może, widocznie, akurat, dopiero, nawet,
tylko, zwłaszcza, prawie, niemal) – was distinguished in opposition to adverbs as late as in the classifications put forward in the GWJPM (Laskowski 1984: 31, 1998: 59). Although a broader class of modal expressions was brought up in an earlier classification (Jodłowski 1976: 20–22), it was defined on the basis of semantic criteria. In the first edition of his “Morfologia,” the author identified one common property of adverbs and modal expressions which do not occur in a noun phrase, while assigning “a limited freedom to enter into syntactic relationships” to adverbs only, modalizers being free from such restrictions. Influenced by the discussions at the turn of the 1980s and 1990s (cf. Grochowski 1986; Wróbel 1996), Laskowski verified his views on the above-mentioned opposition. He characterised adverbs as lexemes that do not enter into any syntactic relations with nouns, while describing modalizers as lexemes for
which such combinations are admissible (Laskowski 1998: 59).
In the first issue of the GWJPM, Laskowski (1984: 31) divided synsyntagmatic lexemes into connectors (markers of syntactic relationships) and non-connectors (particles). As regards the former, he further distinguished the markers of a syntactic dependency of a clause (relative connectors and subordinate conjunctions) and of a noun phrase (prepositions), as well as the markers of the lack of syntactic dependency (coordinating conjunctions). Relative connectors, traditionally referred to as relative pro-forms and later as complementizers (in Polish – relatory, Laskowski 1998: 59) are the components of one of the clauses which together make up a complex sentence, in opposition to conjunctions (subordinate and coordinate) which do not make part of any clause. Motivated by the discussion on the first edition of the GMJPM, Laskowski (1998: 58) reviewed the division of synsyntagmatic lexemes that perform the connective function (connectors) and subdivided them into the indicators of the syntactic dependence of the nominal phrase (prepositions) and the indicators of dependence between clauses, contrasting conjunctions (which are not obligatory components of any of the clauses being combined) with complementizers, which are the components of one of the clauses that make up the complex sentence. Particles, described in the first edition of the GWJPM as lexemes without properties inherent to other units (as not syntactically independent and not performing any connective function), are characterised in the second edition as
“lexemes” with a “non-free order.” Laskowski (1998: 65) divides them into adverbial particles, which bind with a verb (e.g. oby, niechże, bodajby, no, nuże) and adnominal ones, syntactically related to the noun phrase (e.g. byle, lada, niespełna, tuż, zaraz).
Laskowski’s (1984) functional classification was followed by Maciej Grochowski’s (1985, 1986) syntactic classification of uninflected lexemes published slightly afterwards, which elaborated on the classification of these lexemes as made by Saloni (1974) and, to a certain extent, Laskowski. Following Laskowski, Grochowski (1985: 81–84) contrasted appositions and interjections, dividing the latter into four subclasses: those capable of constituting declarative (onomatopoeia, e.g. bęc, fik, lu, łubu-du, pstryk) and non-declarative utterances (predicative interjections, e.g. huzia, precz, wara, wynocha), which may either co-occur with vocative nouns (appeals, e.g. cześć, dobranoc, hej, pa) or do not have this property (interjections such as e.g. cholera, do diabła, kurwa, rany boskie).
Grochowski (1985: 84–87) divided the remaining uninflected lexemes, which are not used independently as utterances, into those that either have or do not have a connective function. The first group, requiring a specific case (prepositions) is contrasted with lexemes without this requirement and with conjunctions. These two classes are distinguished on the basis of the criterion of whether they fill (relative pro-forms) or do not fill (conjunctions) a syntactic position of a component of the clauses being combined, which was already applied in Laskowski’s classification. In Grochowski’s classification, the complementizer class is narrower, as it only consists of uninflected lexemes.
Lexemes that do not perform the connecting function were further subdivided into those occurring in declarative clauses and those for which this criterion is non-defining, as they may occur both in clauses of a declarative and non-declarative nature. The first group is further classified into lexemes that imply certain grammatical forms of the verb (mood operators) and lexemes that are not marked (as a class) in this respect (declarativeness modifiers). Mood operators include, for example, units such as byle, byleby, niechby, oby in utterances like Byleś wrócił przed zamknięciem bramy., Byleby przyłożyć głowę do poduszki i zasnąć., Niechby pan z nim porozmawiał chociaż przez chwilę. Obyś żył długo i szczęśliwie. while declarativeness modifiers include such lexemes as aby, a nuż, czyż, no in utterances like Piotr się
aby nie przeziębił w drodze?, A nuż Ewa o tym zapomni?, Czyż on przestanie w końcu gadać? Usiądź no trochę dalej!
Grochowski divided lexemes occurring in declarative and non-declarative utterances (1985: 89–97) into those that enter into a syntactic relation with a noun and those that do not have this property (adverbs, e.g. bardzo, natychmiast, wkrótce, znienacka). When it comes to the first group, the scholar distinguished between lexemes that bind with a verb and those which do not (adnominal operators). Lexemes capable of binding with verbs were divided into those which enter into a syntactic relation with proper names in the nominative form (particles) and those which do not have this property (adnominal-adverbal operators). The proposed criterion was based on the assumption that, if a specific lexeme is capable of entering into a syntactic relation with a proper name in the nominative case typical of the grammatical subject of a sentence, it may perform the function of a dependent of any noun, regardless of its grammatical form and syntactic position (Grochowski 1986: 59). The class of adnominal operators contains, for instance, such lexemes as: bez mała, lada, niespełna, ponad in utterances e.g. Święta już za bez mała miesiąc., Kładka uginała się pod lada krokiem., Dzieliła ich odległość niespełna metra., Piotr napisał do rodziców w ponad rok po wyjeździe., the class of particles includes e.g. lexemes nawet, również; cf. utterances Nawet Ewa dostała od dziadka zabawkę., Również Nowak przywitał się z kominiarzem., while the class of adnominal-adverbal operators includes e.g. ledwo, niemal, prawie in utterances like Woda sięgała mu ledwo do pasa. Ubranie niemal wyschło., Ewa jest już prawie doktorem.
When preparing a reviewed version of the syntactic classification of uninflected lexemes, developed on the basis of a semantic-syntactic analysis of tens of units, Grochowski (1997: 11) concluded that the hypothesis of unlimited connectivity of particles with nouns is too strong. As a result, he proposed several new criteria regarding the classification of lexemes without a connective function, distinguishing between lexemes with either a variable (particles and adverbs) or a stabilised linear position. Unlike particles, adverbs do not bind with nouns.
Meanwhile, lexemes with a stabilised order enter into syntactic relations with verbs, implying
(mood operators) or not implying (declarativeness modifiers) some of the verb’s grammatical forms. The scope of these two classes, labelled with same name, and identified earlier on the basis of slightly different criteria (Grochowski 1985: 87–89, 1997: 26–27), is the same.
Lexemes that cannot bind with verbs were divided into those that enter into relations with numerals (adnumerative operators) and nouns (adnominal operators); cf. Grochowski (1997: 27–32). The class of adnumerative operators includes, for instance, bez mała, niespełna, plus minus, z, z górą in utterances like Tym samochodem przejechał już bez mała pięć tysięcy kilometrów., Do domu idzie się stąd niespełna dziesięć minut., Musisz na niego poczekać plus minus pół godziny., Kup z pięć kilogramów ziemniaków., Wódki zamówili z górą trzy litry., while the class of adnominal operators includes e.g. byle, gdzieś, lada, tuż, wprost in utterances:
Byle drobiazg był dla nich powodem do kłótni., Wsadzili go do więzienia gdzieś przed świętami., Chwaliła się swoimi sukcesami przy lada okazji., Komar przeleciał mi tuż koło nosa., Pił wino wprost z butelki. For more on adnumerative operators, see e.g. Bogusławski 2010; Doboszyńska- Markiewicz 2013; Duszkin 2010.
Vyara Maldjieva (1995) adopted the classification by Saloni (1974) and Grochowski (1985, 1986) and proposed to extend the categorisation of uninflected lexemes by applying new, detailed classification criteria applicable to the lexemes from the Russian and Bulgarian languages as well. Relying on the distributive and functional properties of lexemes, the author distinguished as many as eighteen classes of uninflected units. The most extensive classes belong to adverbs sensu largo (cf. Małdżiewa, Bałtowa 1995; see also Grzegorczykowa 1975; Wojdak 2004; Bogusławski 2005; Bałabaniak 2013; Bałabaniak, Mitrenga 2015).
In 1996, inspired by the critical analysis of the first functional classification by Laskowski (1984) and the classification of uninflected lexemes by Grochowski (1985), Henryk Wróbel put forward a new classification of lexemes, referred to by him as a formal-syntactic one, distinguishing fourteen parts of speech. The author assumed the superiority of the opposition between lexemes used and not used independently as utterances, known from Laskowski’s first classification (1984), and divided the class constituting the first of the confronted groups into lexemes that are context-dependent utterances (appositions) and context-independent utterances (interjections).
Lexemes which are not used independently were further divided into those which function as sentence components and those which do not have this property. This opposition corresponds in terms of scope to the opposition between autosyntagmatic and synsyntagmatic lexemes found in Laskowski’s first classification (1984). However, both classes of units were divided dichotomously according to the very same syntactic criterion into lexemes with and without a connective function. Wróbel (1996: 55) assumed that lexemes classified into the second group within both of the distinguished categories perform other relevant syntactic functions. Lexemes which are sentence components while also performing the connective functions include a group of relative pro-forms referred to as względniki and complementizers. The units of both of these classes correspond to traditionally construed relative pro-forms. The difference between the two lies in the fact that the former are inflected and accommodated (cf. e.g. który, jaki, kto, co, ile), while the latter are not (cf. e.g. gdzie, dokąd, kiedy, jak). Insofar as the lexemes that are sentence components without any connective function are concerned, Wróbel (1996: 56–57) distinguished those which do (verbs) and do not perform the role of the head of a clause. The latter group is further subdivided on the basis of the criterion of accommodation – in this case, the author differentiated between accommodated (nouns, numerals, adjectives) and non-accommodated lexemes (adverbs, particles).
Nouns in the nominative case govern the form of number (and sometimes also gender) of the verb being the head of a clause. The author stretches the class of nouns to include pronouns (such as ja, ktoś, nikt) which occur in the same syntactic positions. Numerals (cardinal and collective) and adjectives do not have this property. The former govern the case of the noun, while the latter are governed by the noun with regard to case, gender and number (Wróbel 1996: 57). Adverbs differ from particles in that the former have limited, while the latter unlimited distribution. In Wróbel’s approach (1996: 58), adverbs are dependents of verbs and/or adjectives, while particles (cf. e.g. jeszcze, już, nie, również, tylko, aż, po prostu, niemal), may be combined, at least in theory, with any part of speech. In the classification in question, this term is construed in a slightly broader way than in Grochowski’s proposal (1985) as it includes both lexemes referred to in the latter as particles, as well as adnominal and adnominal-adverbal operators.
Wróbel divided lexemes that do not function as sentence components (1996: 58–59) into those that perform a connective function (prepositions, conjunctions) and those that do not play this role (mood operators, modalizers). Both classes within the distinguished pairs are contrasted on the basis of the very same government criterion, but it is applied to dependencies between language units of different types. Prepositions govern the case of a noun, while conjunctions do not have this property. Mood operators (the meaning of this term is the same as in Grochowski’s classification of 1985) govern the form of verbs with which they bind to create a sentence component, e.g. oby requires a -ł- form, byleby requires an infinitive, while
niech – present tense forms. Modal expressions (modalizers) do not govern the form of verbs. The idea for this class, referred to in Polish as modalizatory, comes from the classification by Grochowski (1985), who called the lexemes classified into it as declarativeness modifiers. The term modalizatory had previously been used by Laskowski (1984), but in a different sense. A few years later, in his original descriptive grammar textbook of the Polish language, Henryk Wróbel modified the classification of lexemes with respect to the units that do not function as sentence components and do not have a connective function. When it comes to particles, the author found that their distinguishing property is the capacity to bind with nouns and, based on this criterion, he contrasted particles with the remaining classes of lexemes, which, in turn, were divided into the ones that bind with verbs (mood operators and modalizers) and those which do not. Mood operators and modalizers are already known from previous classifications by Wróbel (1996) and Grochowski (1985, 1986, 1997) – their scope has not changed, and the differences concern only the steps undertaken to distinguish these groups. Wróbel (2001) divided lexemes that do not bind with verbs into ad-numerals and ad-particles. The former enter into syntactic relations with numerals, e.g. niespełna, około, ponad, z, and correspond to the adnumerative operators identified by Grochowski (1997), whereas the latter enter into relations with prepositions, e.g. tuż, zaraz, gdzieś, wprost, and form a class of a slightly different scope than the adnominal operators identified by Grochowski (1997). In his grammar textbook, Wróbel (2001) extended his classification to include sixteen parts of speech.
In 2003, Zofia Zaron put forward a functional classification of lexical units based on the three types of their properties: categorial connotation (cf. Bühler 1934/2004; Gołąb 1967), categorial accommodation, and inflectional properties. When describing the categorial connotation of units of a specific class, the author defined their sentence-forming capacity and identified the class they connote. Speaking of categorial accommodation, Zaron identified the grammatical category whose values are determined by the units of the class being characterised. The inflectional properties are construed literally, e.g. a noun is typically inflected for case and number. Zaron distinguishes twelve lexeme classes (including three subclasses of
verbs). Not all of the enumerated criteria are equally relevant to devise all the opposing pairs.
The categorial connotation properties of verbs are identical with sentence-forming capacities, while interjections have no connotative properties at all. This is a property that verbs share with conjunctions and which in turn, unlike verbs that determine the category of case, do not have any classifying accommodative values. All conjunctions are uninflected. From the perspective of inflection, Zaron (2003: 184) divides verbs into personal (inflected for tense, mood, person, number, and gender), defective 1 – impersonal (such as trzeba, należy, inflected for tense, mood) and defective 2 – uninflected/invariable (such as biada, huzia, trach). Units belonging to the remaining grammatical classes do not have sentence-forming properties. Nouns and pronouns connote the category of verb or another noun (in exceptional cases – also pro-forms). Units from both groups determine the person of the verb, while nouns additionally determine the gender, and sometimes the number and the case. Numerals connote the category of noun and determine its number, while adjectives connote nouns or verbs (if they perform the function of a predicate), without determining the value of any grammatical categories.
Zaron (2003: 185) restricts the scope of adverbs to derivatives from gradable adjectives and to bardzo, characterising the units from this class as connoting the category of verb, adjective or another adverb. Adverbs are inflected for grade. Prepositions connote nouns and determine the value of the case category. Particles (the so-called ad-clausal ones, such as czy,
niech) connote a clause, while modalizers (such as chyba, nawet, prawie) may connote any syntactic units. Localizers (such as rankiem, gdzieś, tu, teraz, kiedyś) only connote verbs, (cf. also Zaron 2005).
Maciej Grochowski (2003) proposed a classification of synsyntagmatic lexemes based on the dominant criterion of the linear properties of the unit classes (e.g. the opposition between fixed and variable order, or the initial or final position) and the connection between these properties and the formal properties of these units (e.g. units implying a numeral and occurring in preposition with respect to it, units occurring in interposition with respect to the segments they combine).
When laying down the theoretical foundations for his “Słownik gramatyczny języka polskiego,” Zygmunt Saloni (2007: 123–140) adopted a modified version of Maciej Grochowski’s (1997) syntactic classification of uninflected lexemes with relatively minor modifications. Saloni (2007: 124–126, 134–136) extended the subdivision of interjections relying on Grochowski’s other work (1993) on this class of lexemes, and separated adverbs from pro-adverbs (such as tam, kiedyś, nigdy) and numeral adverbs (such as dwakroć, trzykroć).
5. Jadwiga Wajszczuk (1992, 1997, 2005, 2010, 2011) dedicated at least a quarter of a century to developing a concept of a syntactic yet semantically motivated classification of lexemes, focusing first on a detailed analysis of conjunctions and their relationships to other unit groups, especially synsyntagmatic ones, and then moving onto the much broader problem of the markers of metatext and their relationship to the units of the substantive language level. She developed the concept of metatext outlined in the early works by Anna Wierzbicka (1969, 1971). In 2010, she proposed a classification based on the dominant role of two oppositions – the one between the connective property and the lack thereof, and the one between the syntax of dependency and the syntax of co-occurrence. First, the researcher contrasted lexemes with paralexemes, which cannot open any positions (such as interjections, appeals, phatic signals). In the set of lexemes, she distinguished between syntactemes – units which open marked semantic positions or occur in one of such positions (Wajszczuk 2010: 25) and paratactemes, which have none of these properties. Syntactemes are characterised by a weak alternative, and include units with a property determined by one or two segments of
the alternative. In her two-step process, Wajszczuk (2010: 27–28) divided syntactemes into:
1. lexemes which do not open a position, but fill positions opened by other lexemes (terms); 2. lexemes that open positions, but do not fill any positions opened by other lexemes (independent predicates) and 3. lexemes that open positions and fill positions opened by other lexemes (dependent predicates). Each of these three major groups is subdivided into several subsets of a yet narrower scope, characterised by enumeration.
Thus the terms include: (a) referential expressions: proper nouns, pronouns (e.g. ja, ty, to, tamto, ktoś, coś) and spatial-temporal pro-adverbs (e.g. tu, teraz, tam, wtedy, stamtąd), (b) simple predicates: names of natural species (e.g. brzoza, sasanka, lew, wiewiórka) and artifacts (e.g. dom, stół, filiżanka).
Independent predicates include: (a) personal and impersonal forms of verbs (e.g. zarządzono [zbiórkę], należało [to zrobić], chce mi się [płakać], łamie [mnie w kościach]), (b) predicatives (defective verbs) (e.g. miło mi, duszno tu, zimno tu; dnieje, świta, zagrzmiało), (c) adverbial participles, (d) numerals, adjectives, adjectival participles, adverbs, (e) metapredicative operators (intensifiers, limitators, approximators) (e.g. bardzo, niemal, prawie, całkiem, zbyt, dość, około [stu], tuż [za płotem]).
Dependent predicates include: (a) deverbal and relational nouns, e.g. pranie, odpoczynek, przerwa, sąsiad, środek, (b) infinitives, (c) comparatives, e.g. lepiej [niż ja], szczuplejsza [od niej], (d) prepositions, (e) includers (complementizers, in Polish: włączniki), e.g. że, żeby, aby, czy. This final term was introduced into syntax by Wajszczuk (1997: 39–46) to denote syntactemes that open two positions, determined by their meaning, for clauses only. Includers are ad-clausal operators which introduce a subordinate clause into the main sentence struc-
ture, cf. e.g. Piotr wie, że Toruń leży nad Wisłą., Anna chciała, żeby Piotr wrócił.
Paratactemes were defined by negation with respect to syntactemes, as they represent a different type of syntax – utterance syntax. Irrespective of this fact, however, paratactemes
have a number of positive properties. First of all, they open positions that are not semantically or grammatically marked; in other words, they do not complement the meaning of expressions occurring in such positions with their own meaning. The meanings of co-occurring expressions cannot be mutually exclusive, but they are independent of each other. Secondly, the structure of utterances occurring in positions opened by paratactemes is not subject to any grammatical restrictions: these expressions can take the form of sentence components (they may represents various parts of speech), syntactic groups and entire sentences. Expressions filling positions opened by paratactemes do not enter into any syntactic relationships with them, but rather simply co-occur with them. Thirdly, paratactemes are metatextual operators – they perform the function of commentaries by the speaker on the content of expressions filling the opened positions. As such, they are communicatively determined, which means that they refer to the level of the theme-rheme structure of the utterance. They either open one position, primarily the right-sided one, or two positions at once – one on each side.
The first group includes particles, while the second consists of conjunctions.
Wajszczuk (2010: 30) identified two closed subsystems of conjunctions: S1 (serial conjunctions: logical conjunction and its varieties) – units which do not operate on entire theme-rheme structures: i, lub, ni, a, albo, ani; and S2 (central conjunctions: implication and the variants thereof) – units operating on entire theme-rheme structures: czyli, to, toteż,
lecz, ale, bo, gdyż, albowiem. Particles are divided into two tiers: P1 (particles proper, in the rheme section): (a) modal, e.g. chyba, może, prawdopodobnie, na pewno, (b) proper, e.g. tylko, nawet, właśnie, głównie, właściwie, P2 (linking particles, in the theme section: (a) linking of the więc, bowiem, jednak, zatem, natomiast type, (b) linking of the skoro, jeżeli, gdyby, ponieważ, chociaż, jakkolwiek type (so-called improper conjunctions, cf. Wajszczuk 1997). The researcher did not consider the classification of paratactemes as closed. This set could be also extended to include units from other classes which to date have been less well researched, e.g. metatextual comments (such as nie ma co, i już, tak na marginesie) or thematizers (such as co do, co się tyczy, jeśli chodzi o); cf. e.g. Sulich 2008.
Shortly after the publication of the classification of lexemes discussed above, Jadwiga
Wajszczuk (2011: 277–283) introduced a number of corrections to the classification of paratactemes. First of all, she differentiated between paratactemes that bind the components of an utterance structure on both sides (conjunctions) and on one side only (metaoperators – STR comments). The author maintained her previous classification of conjunctions, elaborating only on her reasons, including the so-called conjunction reduction hypothesis in the context of Anna Wierzbicka’s criticisms (1969, 1972). Secondly, Wajszczuk (2011: 281) divided STR comments into particles – units which do not operate on entire STRs and ad-utterance operators, which do not have this property. She considered both subclasses to be heterogeneous. The first one is made up of ad-rhematic operators (e.g. tylko, także, również, nawet, właśnie, przede wszystkim) and modal particles (e.g. może, chyba, prawdopodobnie, pewnie, na pewno). The second subclass is made up of two subsystems of operators – the first one comprises units such as jednak(że), wszak(że), zatem, bowiem, przeto, tedy, natomiast, zaś, while the other such as skoro, jeżeli, jeśli, gdyby, chociaż, ponieważ. The author intentionally abstained from naming them, aware of their ephemeral nature.
All the grammatical and semantic-grammatical classifications of lexemes reflect their authors’ original concepts with respect to the division thereof in the context of the state of
research on parts of speech in the relevant period of development of the linguistic studies. Even the most precise classification is nothing other than an approximate picture of the formal-functional diversification of units and is subject to further verifications. Nevertheless, no perfect classifications have been developed and are not to be expected
