Play nominated-Hstar.wav: - download nominated-HStar.TextGrid


<nominated-Hstar.wav> with accompanying Praat TextGrid.

This example has 3 words, and is produced as a single intonational phrase with one pitch accent. Let's look at the transcription for the Tone tier first. As mentioned above, the Tone tier marks two types of events: those associated with accented syllables and those associated with phrasing. The most prominent syllable in this utterance is the second syllable (no-min-a-ted) of nominated. The f0 contour shows a rise that peaks at the end of the syllable -min- and then falls in the following syllable. This is an example of a canonical H*. In the ToBI transcription, the H stands for a High tonal target and the * indicates that the H tone is associated with an accented syllable. Perceptually, this H* syllable is more salient than other syllables around it. There are other types of pitch accents, including bi-tonal types, that make up the full ToBI inventory. We'll look at these other types in turn.

The other intonational event that is labeled in the ToBI transcription of this utterance is the falling f0 at the end of the utterance. This low pitched region is marked with L-L% where the L- indicates that there is a Low phrase accent that is followed by a Low boundary tone (indicated by the L%) on the final syllable.

Another short example is illustrated in another short utterance <won-a>. Again, this contour is a single intonational phrase with one pitch accent. The pitch accent is an H* and the phrase accent/boundary tone is an L-L%.

Play won-a.wav: - download won-a.TextGrid


<won-a.wav> with accompanying Praat TextGrid.

This L-L% sequence is often found at the end of spoken declarative sentences, especially final sentences in a turn or discourse. Later we will see phrase boundaries of a different type that have only a phrase accent. Moreover, like pitch accents, there are other combinations of high and low tones that make up the full inventory of phrase-related tonal markers. Before introducing other kinds of pitch accents, phrase accents and boundary tones, however, let's turn our attention to labels in the Break Index tier. A Break Index marks the level of disjuncture between two words. Since intonation is one of the acoustic cues to disjuncture, there is some redundancy between the Tone and Break Index tiers. For example, an L-L% boundary tone signals the maximal level of disjuncture (i.e. at the end of an intonational phrase) and would be marked with a 4. Other boundaries are marked with smaller break indices. For example, typical word boundaries in a fluent sequence of words within a phrase are marked with a 1. In this example, all break indices are 1's except for the final 4 break index. There are levels of intermediate phrasing within intonational phrases that will be marked with break indices 2 or 3 STOPED, as we will show later. The 0 break index is reserved for the case when two words are produced so that the boundary between them is indeterminate. For example, the word sequence did you is frequently produced as didja, where the final /d/ of did and the initial y of you together form a j-like sound. (In example , "He said you would", listen to the final /d/ of said and the y of you.)

Play said_would1.wav: - download said_would1.TextGrid


Example of Break Index 0 between said and you.

The following examples show more instances of H* L-L%, with the same contour on utterances of varying lengths.

Play minimum1b.wav: - download minimum1b.TextGrid


example <minimum1b> a 2-word utterance with H* L-L%.

Here, the initial low tone might suggest an alternate interpretation of the pitch accent ( L+H*) that will be discussed later. A newly adopted tier, the alternative tier, allows labellers to capture these reasonable alternate parses of the intonational contour.

Section 2.1 Exercises: Praat & H* L-L%

A. listening exercises:
1. Just listen. The following file has various speakers producing various texts, all with H* L-L% (ex1a1.wav)
2. Listen and choose. 5 of the 7 versions of "a minimum" are produced with H* L-L%, and 2 are produced with different contours. Can you pick which? (ex1a2.wav has the choices with letters to keep track, ex1a2-unlettered.wav has the same choices in the same order, but without the intervening letters.)

B. labelling exercises:
For each file 1 through 6 (ex1b1.wav, ex1b2.wav, etc.)
- create a blank TextGrid in Praat with the following tiers: tones, words, breaks, misc. (Make sure that words tier is an interval tier, and that the other three are point tiers.)
- Label the words, break indices, the phrase accent-boundary tone combination and pitch accent on the appropriate syllable for each Intonational Phrase.
- Save your TextGrid files—include your initials in the filename (eg. ex1b3amb.TextGrid)

Issues to consider while labelling:
--alignment of labels
--break indices aligned to end of word
--phrase accent and boundary tone aligned to end of last word of phrase --pitch accent in correct syllable, at correct location. The best place (in most cases) is the highest intensity and the "fattest" (in the waveform) part of vowel. If there isn't an obvious single "fat" point, the approximate center of the vowel will suffice.

C. further exercises: Produce and record your own H* L-L% phrases

1. listen to the file ex1c1.wav, which contains the words "a minimum" produced by different speakers with H* L-L%
- record (in Praat or another recording application) and imitate the intonation contour on the same words ("a minimum") in your own voice. Try several different pitch ranges, speeds or volumes: in your "normal" range, in a higher- or lower-pitched voice, softly, loudly, slowly, quickly.
- look, listen and compare. Do your versions look like the H* L-L% contour? If not, think (descriptively) about how your versions differ. (Note: It is often tricky to produce a given contour on demand: we're used to using prosody communicatively, and without conscious planning. Don't worry if you've produced something that sounds different: you can always save your files to label later on, once you've learned more the of the ToBI inventory.) - You may want to create a TextGrid to label those versions you believe are H* L-L%. Don't worry about labelling others.
- save your recording as a .wav file with the name "ex1c1" followed by your tobi class username (ie the name in your email address, eg. ex1c1amb.wav). If you have made a TextGrid, save it also using the same naming scheme.

2. Try producing several H* L-L% versions on the following words and short phrases (feel free to add to the list):
me, you, banana, another banana, a lime, arugula, watermelon, an umbrella, marmalade
- save your soundfiles as .wav files, and any associated TextGrids, named with "ex1c2" and your initials, and a number or word if you produce more than one sound file (eg. ex1c2amb2.wav or ex1c2amb_me.wav)

3. Try producing H* L-L% versions on a longer phrase or two. Listen to files from section 2.1, and try your own renditions on the same text:
"He said you would." (said_would1.wav)
"Marianna won it." (won-a.wav)
"He was nominated." (nominated-Hstar.wav) - save your soundfiles as .wav files, and any associated TextGrids, named with "ex1c2" and your initials, and a number or keyword if you produce more than one sound file (eg. ex1c3amb2.wav or ex1c3amb_won.wav)