readings/w2/Look at Table 1 for summary, and included "prosodic" universals:
A number of universal tendencies can be observed in relation to the prosodic structure of languages. A very widespread pattern is that some phonetic components are longer in a position before a boundary, such as at the end of a sentence or an intonational phrase, than they are in other positions. Beckman and Edwards (1990) are able to separate out phrase-final and word-final lengthening processes in English. Lengthening in these types of positions functions as a means of demarcating the ends of constituents, and presumably has a perceptual rather than a physiological basis. A longer syllable at the end of a constituent not only directly helps to mark the end of the constituent but also provides a better host for other markers of constituent structure such as tonal or voice quality changes, for example, the high-tone question-marker of Hausa (Lindau 1986). Pre-boundary lengthening may apply to a segment or syllable not in absolute final position, as in many Bantu languages which lengthen the penultimate syllable at the ends of sentences and, within sentences, at the ends of some major clauses. In this case, the lengthening may be positioned earlier in order to avoid the strongest impact of amplitude and pitch declination in the absolute final position.
Declination refers to an underlying downward trend of some trait over the length of an utterance or part of an utterance. Declination of F0 and overall acoustic amplitude appear to be universal, and there are some indications that declination of articulatory amplitude may also be typical (Vayra and Fowler 1992, Krakow 1993). Local perturbations - for example, for stressed syllables - may be superimposed on this overall decline, but successive local perturbations themselves often conform to a pattern of declination. In almost every language in which pitch patterns have been studied over utterances longer than word-length, the F0 pattern follows an overall downward trend, at least within certain spans and at least for some unmarked utterance type. It seems likely that this declination is, broadly speaking, associated with the baseline decline in the subglottal pressure that occurs as lung volume decreases over the course of an utterance. Important experiments reported in Gelfer, Harris, Collier and Baer ( 1985) and Gelfer, Harris and Baer (1987) showed that the Dutch-speaking subjects they studied controlled the rate of subglottal pressure decrement, as it remained constant with different rates of lung volume decrement arising from use of low air flow segments (/ml) versus high air flow segments (/f/). F0 declination was correlated with the declination of subglottal pressure, not with lung volume declination, nor with activity of the cricothyroid muscle. They therefore conclude that subglottal declination "is a controlled variable in sentence production, and that F0 declination is a consequence." The same authors conclude that the overall decline of acoustic amplitude is also a consequence of the declination of subglottal pressure. This effect, combined with anticipation of the open glottal position of normal respiration, produces a common tendency for elements in utterance-final position to become devoiced.
Phonetic universals have also been sought for smaller prosodic units such as the syllable, but few of the marks of syllable affiliation have been found to have cross-linguistic generality. One possible exception is shortening in closed syllables. In many languages a vowel in a closed syllable is shorter than a matched vowel in an open syllable (Maddieson 1985). Thus, the /ei/ in 'grey towel' will tend to be longer than that in 'great owl'. This effect may be observed most readily in languages where medial single and geminate consonants of the same type contrast (assuming that the geminate closes the preceding syllable). In this environment the influence of syllabification is isolated from other durational modifications. Some results for Italian, where this effect is very strong, are given in Table 4. Since there are languages in which geminate consonants do not have a shorter preceding vowel than single consonants, such as Japanese (Homma 1981, Smith 1995) and perhaps Sinhala (Letterman 1994), it is clear that this is not an automatic effect of the syllabification. As suggested above, what may be at work here is a strategy for keeping words with equal syllable count similar in overall duration (A possible advantage of doing this would be to keep the information density in the signal more uniform.)
Final lengthening (go big):
...vowels were categorized into one of three position classes. Vowels were classified as Final when they occurred in the final syllable before a silent pause. Vowels were considered Penultimate when they occurred in the penultimate syllable before a silent pause. Voels were counted as Non-final when they were located in any other syllable and when they where not preceded by a silent pause (this was done to exclude effects of initial strengthening).
Fn 4:
Silent pauses were identified during manual correction of word boundaries before running the final round of forced alignments. The DoReCo team made a great effort to ensure any interval labelled as a silent pause corresponds to some sort of juncture between two prosodic units rather than being just a gap due to consonantal closure. However, as no lower threshold was used for pause length, there will inevitably be some instances where the pause status of an interval could be debated.
Word-initial lengthening Blum et al. 2024 (go small)
The main units of our analysis are phones (idscrete segments of speech); words, as defined by experts on each language; and utterances, which we define as interpausal units--that is, chunks of speech that are not interrupted by a silent pause.
Word-medial lenition Lee, Sonderegger, and Clayards 2026 (go small)
Section 4
Our results should also be interpreted with the caveat that ‘word’ is an approximation of prosodic structure: lexical word-medial stops are reliably prosodic word-medial, but word-initial stops may or may not be prosodic domain-initial. Some could be at the beginning of a larger prosodic constituent (strengthening our chance of finding a word-position effect), while others may be prosodic word-medial (weakening our chance), as in suffix-like orthographic words in Korean, which are always pronounced together with the preceding word. This variance in what ‘word-initial’ means in our data introduces noise into our word-position results rather than systematically biasing them in one direction. Despite this, our results show a clear cross-linguistic tendency for word-medial stops to be weaker than word-initial ones, suggesting the effect is robust.
Drawing from Barnes et al. 2020 (in readings/w2/)
Tonal coarticulation
Tonal alignment/association
Tonal crowding resolution, truncation/compression
readings/w2/), perspective from West African Languages (Niger-Congo, Chadic)From Hyman and Schuh (1974)...some "diachronic" natural tone rule examples
Questions raised:
What is low raising?
A low tone is often raised in pitch when followed by a nonlow tone (e.g. high or mid)
Also happens in "nontone" languages:
In a language (e.g. Finnish) where one syllable or mora is stressed per word, it should be expected that the stress with the highest pitch will (barring contrastive stress) be ordinarily the first in the sentence and each subsequent stress will be on a slightly lower pitch (i.e. downdrift). Also, it should be expected that a stressed syllable will raise a preceding unstressed syllable slightly (low-raising). We have not investigated such languages, but only speculate that something of a more universal nature is involved here.
Tone spread does not always occur
Direction: rightward, "carryover"
the spreading of tones takes place always to the right and apparently never (or at least rarely) to the left
Spreading is an assimilatory process of the progressive or perseverative type, rather than of the regressive or anticipatory type. That is, the earlier otne appears to last too long, rather than the later otne starting too early.
Does it happen between all kinds of tones in the tonal inventory?
spreading is most likely to occur where the interval between the two tones is greatest
Implicational universal
in a language, if tone spreading takes palce from a mid tone into a following low, then tone spreading also takes place from a high tone into a following low in that language
...contour tones tend to be levelled--or in other words, converted into level tones