Ling 590P Class 2.1 (09/15/26) Intro to vertical phenomena

Agenda

• Look at your comments on Barnes et al. (2020)
• Talk about some key concepts in thinking about "vertical" aspects of phonetic implementation of tone/intonation

Notes on Barnes et al.

• Lack of phonetic work on African tone languages
• Rich source of potential phonetic universals/tendencies about phonetic implementation of tone/intonation

Downtrends

• The Barnes et al. (2020) paper has a good overview of this in Section 9.4.3
• Key point from that section: Relative nature of fundamental frequency: often referred to as syntagmatic ("horizontal" relations)

Unlike musical notes, linguistic pitch specifications cannot be invariantly associated with context-independent f0 values, but rather are massively, multi-dimensionally relative. Barnes et al. (2020, p. 133)

• Control of pitch (f0) range: topline, bottom line, and differential effect of f0 range modifications on different tones

“One basic observation about this variation is that it tends to affect higher f0 targets more saliently than lower. Rather than shifting globally upward or downward, pitch range seems instead to be compressed or expanded, primarily through raising or lowering of its topline or ceiling.” (Barnes et al. 2020, p. 133)

Definitions

Downtrend

“Perhaps the most exhaustively studied pattern of contextual pitch range variability is downtrend, a term spanning a range of common, if not universal, phenomena involving a tendency for pitch levels to fall over the course of an utterance, which has long been recognized (e.g. Pike 1945: 77)”. (Barnes et al. 2020, p. 135)

• You can borrow Pike (1945) from the Internet Archive:

• Note also other early descriptions of f0 range modifications in Pike (1945) Section 4.4!

Kinds of downtrend/contributing factors to downtrend

“Our usage here is as follows:

‘declination’ refers to ‘a gradual tapering off of pitch as the utterance progresses’ (Cohen and ’t Hart 1965). Declination is putatively global, gradient, and time dependent.

‘Downstep’ refers to a variety of categorical (or at least abrupt) lowering patterns, usually, though not exclusively, affecting High tones.”

“Lastly, the term ‘final lowering’ (Liberman and Pierrehumbert 1984) refers to various (p. 136) probably distinct phenomena involving low ered pitch in domain-final contexts.” (Barnes et al. 2020, p. 135)

• More on mathematical differences between the phonetic implementation of declination vs. downstep (and also implementation of final lowering) on Thursday!
  • Key difference: declination as a "global" (and "phonetic"?) effect, downstep as a "local" (and "phonological"?) effect.
    

More on declination

Yuan and Liberman (2014) (earlier version: 4-page interspeech paper, Yuan and Liberman (2010))

How do you adjudicate between the hypotheses about whether declination is under linguistic control or purely biomechanical (i.e., "breath group"?

1. Utterance pre-planning: Correlation between length of utterance and initial F0 peak height or declination slope
2. F0 declination is affected by speaking style (e.g., read vs. spontaneous speech) and by sentence type (e.g., declaratives vs. interrogatives)

3. Language specificity in declination patterns (Yuan and Liberman (2014)'s English vs. Chinese broadcast speech comparison)

• More recent work, see e.g. Kim and Tilsen (2024)

More on downstep!

• Connell (2011) handbook article on downstep (I'm working on getting a full electronic version of this), with a perspective from Sub-Saharan African languages

Downstep is most commonly described as the lowering influence of a low tone (L) on a following high tone (H), such that a new, lower, "ceiling" is set for all subsequent Hs within a specifiable domain or prosodic unit (Connell 2011, p. 824)

• Examples of terracing and downstep: Miya demo (LING 506)

  • What motivates the introduction of downstep as a concept, rather than analyzing by introducing more phonological tonal categories?

• Excerpt from Schuh (1998) Miya grammar:

• 'Automatic' vs. 'non-automatic' downstep (Stewart 1965):

  • 'non-automatic': H is lowered by preceding underlying or floating L
  • 'automatic': H is lowered by preceding surface L
  • Note: downdrift has sometimes been used to refer to 'auotmatic downstep', see Connell (2001, 2011), but downdrift also sometimes seems to be used to refer to declination

Downstep and ToBI transcription

• In Autosegmental-Metrical Theory, also has been called catathesis (e.g., Beckman and Pierrehumbert 1986)
• Caution: in "ToBI-style" intonational transcription, transcribing downstep is typically taken to be a phonetically transparent annotation---not necessarily anything to do with whether or not it is phonologically triggered
• See for instance OCW ToBI training materials from 2006, Section 2.6:

However, the !H* (‘downstepped H star’) pitch accent label is dependent on the immediately preceding labels. It indicates that there is a specific tonal relationship between the prominence labelled !H* and the preceding pitch accent. Specifically, a downstepped H pitch accent indicates that the tone of the prominent syllable is realized by a perceptually lower f0 than that of an immediately preceding High tone: the tone has ‘stepped down’ from the preceding High.

• Also, different researchers may vary in whether they choose to mark downstep, e.g., !H, or not in the f0 scaling of an annotated recorded passage/sentence

• More generally "TobI-style" transcription has been criticized for inadequacy in describing tonal scaling, see for instance Dilley and Breen (2022) commentary in Prosodic Theory and Practice book and their proposal for RaP, another transcription system, and other work by Dilley and colleagues