Stump (2006): Heteroclisis and paradigm linkage #

This file formalizes Stump's account of heteroclisis, the property of a lexeme whose paradigm contains forms built on stems of two or more inflection classes. A rule of paradigm linkage gives a content cell of a lexeme a form correspondent built on one of the lexeme's stems: the universal default assigns the root, and narrower language-specific rules override it by Pāṇini's principle, assigning a coradical, a stem that differs from the root in form or in inflection class. The paper's rules (5), (14), (15), (17) and (18), with the endings of the Czech declension fragment, after Short's tables and the rules of Mluvnice češtiny, realize every form of the Czech nouns in its Tables 1, 6 and 8, read from the forms data, once the masculine a-stems' nominative plural in -ové is read from their own class and FILOSOF's, which neither grammar gives, is supplied. The heteroclisis of Czech PRAMEN is morphosyntactically conditioned, its inflection class fixed by the property sets its cells express, whereas that of Sanskrit AHAN is morphologically conditioned, following a pattern of alternation among Strong, Middle and Weakest stems. Stump's Sanskrit rules of paradigm linkage (20a,b) "are independently motivated by the need to account for the pattern of stem alternation in nonheteroclite paradigms such as the masculine and neuter paradigms of PRATYAÑC" (p. 295): one list of rules gives the grades of PRATYAÑC's stems, which share a declension, and the stems of AHAN, which do not. The paper measures how closely a heteroclite paradigm's class boundary follows an inflectional category, and claims that the categories serving as absolute correlates are privileged: every rule of paradigm linkage sensitive to some category is sensitive to a privileged one.

Main definitions #

attested: the forms of a Czech cell, every matching row of the forms data.
Decl.toClass, CzNoun.noun: the paper's declensions as classes of Short's tables, and the nouns as entries of the Czech noun fragment.
Stem, Entry, inferSoftMasc: stems individuated by lexeme and declension, lexical entries read off the nouns, and the rule of stem inference (18).
CellRule, rulesFor, stemFor: the rules of paradigm linkage (14), (15) and (17) as domain rules, resolved by the Elsewhere engine with the root of rule (5) as the fallback.
classNomPl, nomPlOverride, suppliedNomPl: the masculine a-stems' nominative plural read from their class, FILOSOF's, which neither grammar gives, and the two together.
czRealize, czLinkage: the realization of form cells and the Czech paradigm linkage.
degreeNum, degree: the degree of correlation of a category with a class split.
IsAbsoluteCorrelate, IsMaximalCorrelate, IsCloven, IsFractured, IsIntersectiveCorrelate: the paper's classification of heteroclite paradigms, over any cells pairing a case with a number.
SensitiveTo, Privileged, SatisfiesPCR: the privileged category restriction.
SktForm, f₁: Sanskrit form property sets and the property mapping (10).
strength, rule20a, rule20b, sktLinkage: the strong and weak property sets of (19), the rules (20a,b), and the linkage of PRATYAÑC and AHAN.
hrdLinkage, matLinkage: Sanskrit HṚD(AYA) and Russian MAT'.

Main results #

realize_matches_tables: the rules realize every attested form of the nine Czech nouns.
czRealizeTable_stemFor_iff: the fragment's endings alone miss exactly the six nominative and vocative plurals of PŘEDSEDA, SLUHA and FILOSOF.
predseda_eq_hrdina, sluha_eq_hrdina_iff: rules (14) and (15) rebuild Short's paradigm of the masculine a-stems, from which SLUHA's departs only in the locative plural.
sluha_locPl, filolog_locPl, filologu_voc: the grammars' velar forms, derived, the locative plurals alike through the soft-masculine coradical and the hard-masculine stem.
pramen_eq_koren_pattern: rule (14) gives PRAMEN the classes the grammars give kořen.
heteroclite_iff, segments_invariant, pramen_suppletive: PRAMEN, PŘEDSEDA, SLUHA and FILOLOG are heteroclite on form-identical stems, and heteroclisis is suppletion.
degreeNum_eq_card_iff: a category is an absolute correlate exactly when the class split factors through it.
pramen_degrees, predseda_degrees, sluha_degrees, filolog_degrees: the paper's figures.
isAbsoluteCorrelate_of_rules: rules seeing only a category make it an absolute correlate.
pcr, not_satisfiesPCR_hypothetical: the Czech rules satisfy (40), and a case-only rule would violate it.
fractured_minimal_intersective_privileged: claim (39) for the Czech nouns.
rules_match_table9, rules_match_table10: (19) and (20a,b) give PRATYAÑC's grades and AHAN's stems at every cell of Tables 9 and 10.
pratyanc_not_heteroclite, ahan_heteroclite, ahan_grade_eq_pratyancN: one list of rules makes AHAN heteroclite and PRATYAÑC not, with the same grades.
ahan_degrees, ahan_fractured: AHAN's paradigm is fractured, its case correlation .71.
hrd_cloven, privileged_case, pcr_skt: case, the absolute correlate of HṚD(AYA)'s cloven paradigm, is privileged, and (20a,b) satisfy (40).
hrd_form_suppletive, mat_suppletive_not_heteroclite: suppletion with and without heteroclisis.

Implementation notes #

A Czech noun's segments are the stem of its entry in the Czech noun fragment, the citation form less the nominative singular ending of its class, so PŘEDSEDA's root and coradical share the segments předsed. A stem's endings are the fragment's endings of its noun in the class that the stem's declension is for the noun's animacy, with the velar vocative singular -u and locative plural -ích and the noun's lexical nominative plural, MUŽ's -i or -ové and FILOLOG's -ové, and the stem is inflected as the fragment inflects it, palatalized before the masculine locative plural -ích and the animate nominative plural -i. Where the paper and the grammars differ the paper's forms stand and the difference is stated: the nominative and vocative plurals in -ové of PŘEDSEDA and SLUHA, realized on the hard-masculine coradical, whose class gives -i, are read from their own class, and FILOSOF's, which neither grammar gives, is an override; Stump's hard-feminine roots for the masculine a-stems rebuild Short's separate hrdina declension; and his soft-masculine coradical of velar stems gives the locative plurals in -ích that Mluvnice češtiny's rule gives the hard-masculine stem. Sanskrit endings are read off exemplars, chosen by declension among PRATYAÑC, NĀMAN and MANAS. Rules compete by domain inclusion, the rendering of Pāṇini's principle, and selection falls back on the root, by (5) and (12). Gender is taken as lexical: PRATYAÑC's masculine and neuter are two lexemes, clause (ii) of f₁ is vacuous, and the feminine, whose ī-stem pratīcī (n. 18) would make PRATYAÑC heteroclite, is left out, as are n. 21's stem indexing and n. 22's alternants. Sanskrit sandhi (aho-bhis, pratyag-) is unstated, so no forms are built. Rows are filtered by language.

Stump (2006): Heteroclisis and paradigm linkage #

Main definitions #

Main results #

Implementation notes #

References #

Cells and the attested forms #

Stems and lexical entries #

Rules of paradigm linkage #

Realization #

Heteroclisis #

The degree of correlation #

Correlates of heteroclisis #

Correlates of Czech heteroclisis #

Rules sensitive to a category #

The privileged category restriction #

Sanskrit cells, stems and form property sets #

Sanskrit HṚD(AYA) (Table 4) #

Sanskrit stem grades (§3.2) #

The rules against Tables 9 and 10 #

Heteroclisis from stem alternation #

The property mapping f₁ #

The privileged category restriction in Sanskrit #

Russian MAT' (Table 5) #

The property mapping `f₁` #