Play anna1.wav: - download anna1.TextGrid


Example of H*L-H%

In the example <annal.wav>, the high pitch accent (H*) on An- in Anna is followed by a low phrase accent (L-) that rises again into a high boundary tone (H%). Even though the f0 is smooth across the boundary, the changes in f0 and the lengthening in -na give the perceptual impression of a boundary between Anna and married. Notice that the second intonational phrase in this utterance is the now familiar L*H*L-L% (Examples XX and YY). As we have seen, the intonational phrases are separated by the 4 Break Index. The juncture between married and Lenny is the default phrase medial inter-word boundary and is labelled with a Break Index 1.

The next example of H*L-H% occurs on a single word intonational phrase <Dan-L1>. Note the rise in f0 on the vowel of Dan (H*) followed by a fall (L-) that ends in a small rise (H%).

Play Dan-L1.wav: - download Dan-L1.TextGrid


Example of H*L-H%

Often, mixed phrase accent/boundary tone combinations (e.g. L-H% as shown in these two examples above, or H-L% described below) do not end in the more extreme high or low levels that are typical of H-H% or L-L% respectively. In this example, the L- makes it unlikely that the H% will rise to the top of the speaker's range. Without the extreme f0 excursions, other boundary cues, like the lengthening at the end of the last syllable in the intonational phrase, becomes more important.

The last phrase accent/boundary tone sequence to fill out the inventory is H-L%, shown here preceded by an L*. (H* H-L% sequences exist as well). In this next example <hiAlejna-L.wav>, the L* on -le- (Alejna is pronounced uh-LAY-na) is followed by a non-pitch-accent-lending f0 movement up to an H- that is then followed by a slight fall (L%). Other examples of H-L%, including <banana-B824a2.wav> below, do not end in a falling f0. Although the phrase accent/boundary tone sequence is also an H-L%, the final f0 is nearly flat in a middle range. This is more typical of an H-L% than a slight fall.

Play hiAlejna-L.wav: - download hiAlejna-L.TextGrid


Example of L* H-L%

Play banana-B824a2.wav: - download banana-B824a2.TextGrid


Another example of L* H-L%

Because the two tones (in these H-L% and L-H% examples) are being realized over a potentially short duration, sometimes only over the end of the final syllable, it is often difficult to see the short realizations of the final boundary tone. However, sometimes a labeler will get a perceptual impression even if it isn't visible in the F0 track; or can hear that it is not another category even if she can't hear or see a clear rise and fall.

The next example is cautionary: the pitch tracking software indicates a rise in f0 at the end of the utterance. However, the pitch periods have become irregular, which often happens as a speaker ends an utterance, causing a spurious small second peak in the waveform. The software that calculates the f0 mistakes this small peak for a pitch period and determines the pitch period to be smaller (as a result the f0 frequency is calculated as higher than it should be). The best test in cases like this is to compare f0 from earlier in the utterance with the end of the utterance by listening carefully for which sounds lower. The most reliable measuring tool in these cases is the listener's auditory system: listen carefully to whether it sounds as if the speakers pitch sounds as if it were intended to go up at the end (specifying an H%) or to stay low (specifying an L%).

Play anna1: - download anna1.TextGrid


Example of H* L-L% with a spurious rise at the end (this is an excerpt of <anna1.wav>.)

Like the boundary tones introduced earlier, there is no constraint on what pitch accent can precede which boundary tone, so more challenging sequences like L*L-H% or H* H-L% are also possible. The examples above (H* L-H% and L* H-L%) were introduced first in this tutorial because they are maximally contrastive, so it is easier to see the two different tones in the final sequence. Next, we'll show examples where the pitch accent choice does not provide maximal contrast.

Play banana-T824a1.wav: - download banana-T824a1.TextGrid


An example of L* L-H%

Compare the L-H% contour in <banana-T824a1> to both the L-L% and the H-H% in the following three renditions of Marianna's money. In the first (left most) phrase the f0 rises to produce the H* on -anna, then falls for the L- phrase accent on -na's money until it culminates in a rise for the H% boundary tone on -ney. The H% tone is clearly higher than, e.g. the f0 in the L-L% of the middle rendition. On the other hand, the H% in this L-H% combination does not rise as high as it does in the last (right most) rendition of Marianna's money, where the L* is followed by an H-H%. Here, the f0 rises very high in the speaker's range, as is frequently seen in productions of H-H%. Understanding these interactions between phrase accents and boundary tones, as well as changes in voice quality at the end of phrase and their effects on the visible f0 track, is an important part of learning to use the ToBI system.

Play money1.wav: - download money1.TextGrid


Examples of H* L-H% and H*L-L%, to contrast with H* H-L% below

Play money2.wav: - download money1.TextGrid


Example of H* H-L% to contrast with H* L-H% and H*L-L% above