Documentation

Linglib.Features.Prosody

Features.Prosody #

[Pie80] [BP86]

Theory-neutral prosodic types: pitch accents, phrase accents, boundary tones, and the prosodic hierarchy, following the autosegmental-metrical (AM) framework established by [Pie80] for English and extended cross-linguistically by [BP86].

[Ste00a] uses these types to connect prosodic structure to CCG derivations and information structure.

Overview #

These types are used across multiple theories:

Pitch accent types ([Pie80], ToBI conventions).

The full inventory of English pitch accents. A starred tone (*) is phonologically linked to the stressed syllable; an unstarred tone in a bitonal accent precedes or follows it at some given space in time.

[Pie80] identified seven possible pitch accent shapes; the H*+H accent was eliminated as a possible pattern, leaving six ([BP86] §2.1).

In Japanese, by contrast, the single possible pitch accent shape is a lexically linked H, analyzed as H*+L. The accent location is lexically distinctive, but the shape is fixed. English uses the full inventory to contrast different intonational meanings (e.g., declarative vs. surprise-redundancy vs. impatient reassertion on the same word orange in an orange ballgown; [BP86] Fig. 1).

Instances For
    Equations
    • One or more equations did not get rendered due to their size.
    Instances For
      @[implicit_reducible]
      Equations

      Phrase accent: terminal tone of the intermediate phrase.

      [BP86] §4.2–4.3 decompose what [Pie80] called the "phrase accent" into a tone that spreads from the last pitch accent to the edge of the intermediate phrase. The phrase accent is H or L, independent of the boundary tone.

      In Japanese, the accentual phrase boundary L is always L; the only variation is whether an optional H boundary tone follows at the intonation phrase edge.

      Instances For
        Equations
        • One or more equations did not get rendered due to their size.
        Instances For
          @[implicit_reducible]
          Equations

          Boundary tone: terminal tone of the intonation phrase.

          [BP86] distinguish the boundary tone (edge of the intonation phrase) from the phrase accent (edge of the intermediate phrase). Together, the phrase accent and boundary tone produce four terminal configurations: LL%, LH%, HL%, HH%.

          In Japanese, the boundary tone at an intonation phrase edge is always L except in yes/no questions, where H is optional ([BP86] §4.2).

          Instances For
            Equations
            • One or more equations did not get rendered due to their size.
            Instances For
              @[implicit_reducible]
              Equations

              Full terminal contour of an intonation phrase: phrase accent + boundary tone. [Pie80]: four possible combinations.

              Instances For
                Equations
                • One or more equations did not get rendered due to their size.
                Instances For
                  def Features.Prosody.instDecidableEqTerminalContour.decEq (x✝ x✝¹ : TerminalContour) :
                  Decidable (x✝ = x✝¹)
                  Equations
                  • One or more equations did not get rendered due to their size.
                  Instances For

                    How pitch accents relate to the lexicon ([BP86] §2.5).

                    • lexical: accent location specified in lexicon, shape fixed. Japanese: H*+L at lexically specified mora. Swedish: accent 1 vs 2. The accent shape cannot signal different intonational meanings.
                    • postlexical: accent shape chosen by intonation, location by prominence. English: 6 pitch accent shapes, location determined by focus/stress. The shape contrasts different intonational meanings (declarative, surprise, impatience, etc.).
                    Instances For
                      Equations
                      • One or more equations did not get rendered due to their size.
                      Instances For
                        @[implicit_reducible]
                        Equations

                        How a morpheme interacts with the prosodic specification of its base. [KH77] [Rol18]

                        Orthogonal to AccentSpecification, which classifies word-level accent determination (how is the accent location decided?). ProsodicDominance classifies morpheme-level prosodic interaction (does this morpheme override the base's accent/tone, or respect it?).

                        The dominant/recessive distinction originates in the accentual morpheme classes of [KH77] (deaccenting vs non-deaccenting suffixes in IE) and was generalized to tonal morphology by [Rol18] as the GT dominance typology.

                        • dominant: overrides the prosodic specification of the base. Accent: Japanese -teki removes stem accent ([Kaw15]). Tone: Mwaghavul verbalisers replace base melody ([AF26]).
                        • recessive: applies only when the base is prosodically unmarked. Accent: Japanese -si 'Mr.' preserves stem accent. Tone: Giphende floating tones dock only to unvalued TBUs.
                        • neutral: concatenates without prosodic interaction; the general phonological grammar determines the output. Tone: Hausa referential -ⁿn ([Rol18]).
                        Instances For
                          Equations
                          • One or more equations did not get rendered due to their size.
                          Instances For
                            @[implicit_reducible]
                            Equations

                            Dominant morphemes override the prosodic specification of their base.

                            Equations
                            Instances For
                              def Features.Prosody.ProsodicDominance.combineAccent (baseAccent : Option ) :
                              ProsodicDominanceOption

                              Combine a base accent with a suffix's prosodic dominance.

                              The accent position (Option Nat) represents a mora- or syllable-indexed accent; none = unaccented.

                              Equations
                              Instances For

                                Transparadigmatic uniformity ([Rol18]): dominant morphemes produce the same output regardless of whether the base is accented or unaccented. This is the defining property of dominance — it neutralizes the base contrast.

                                Recessive morphemes preserve the base contrast: an accented base stays accented, an unaccented base stays unaccented.

                                Fine-grained affix accent classification ([Kaw15] §6).

                                The 3-way ProsodicDominance (dominant/recessive/neutral) captures only one axis of morpheme–accent interaction. [Kaw15], building on [poser-1984] and [vance-1987], identifies eight distinct affix accent behaviors in Japanese, differing in whether the affix carries its own accent, whether it deletes or preserves root accent, and whether it inserts a new accent at a particular position.

                                toProsodicDominance projects back to the coarser 3-way classification: types that preserve root accent when present map to recessive; types that override root accent map to dominant.

                                • recessive : AffixAccentType

                                  Suffix bears accent; loses to root accent when root is accented. E.g., Japanese -tara (conditional): accented root → root accent preserved; unaccented root → suffix accent surfaces.

                                • dominant : AffixAccentType

                                  Suffix bears accent; always overrides root accent. E.g., Japanese -ppoi (-ish): root accent deleted, suffix accent surfaces regardless.

                                • recessivePreAccent : AffixAccentType

                                  No own accent; inserts accent on root-final syllable only when root is unaccented. Preserves root accent when present. E.g., Japanese -si (Mr.): ono → ono'+si; u'ra → u'ra+si.

                                • dominantPreAccent : AffixAccentType

                                  No own accent; always inserts accent on root-final syllable, deleting root accent. E.g., Japanese -ke (family of): ono → ono'+ke; mu'raki → muraki'+ke.

                                • accentShifting : AffixAccentType

                                  No own accent; shifts existing root accent to pre-suffix position. Unaccented roots remain unaccented. E.g., Japanese -mono (thing): ka'k(+u) → kaki'+mono; nor(+u) → nori+mono.

                                • postAccenting : AffixAccentType

                                  Inserts accent immediately after the affix (typically a prefix). E.g., Japanese o- (honorific): huro' → o+hu'ro.

                                • deaccenting : AffixAccentType

                                  No own accent; deletes root accent. Output is unaccented. E.g., Japanese -teki (的 -like): ke'izai → keizai+teki.

                                • initialAccenting : AffixAccentType

                                  No own accent; inserts accent on root-initial syllable. E.g., Japanese -zu (group/plural): okamoto → o'kamoto+zu.

                                Instances For
                                  @[implicit_reducible]
                                  Equations
                                  Equations
                                  • One or more equations did not get rendered due to their size.
                                  Instances For

                                    Deaccenting is a special case of dominance (overrides root accent).