Don't Ask

For Assignment 1 you looked at the phonetic realizations of many (read) versions of the phrase "don't ask". The phoneme sequence /nt#æ/ has a wide range of phonetic correspondences -- and you saw that it's often unclear how many of which "phones" (= phonetic segments) are present in a given example, and also what aspects of the lexical input (or other contributions to the communication process) the observed phonetic features correspond to.

For example, when there's a nasal murmur (or a nasal tap, if that's different) it seems clear that it comes from the /n/ of "don't". But in the pronunciation in FCLT0_SI2068 of the orthographic sequence "don't know", the two vowels are separated by a 54-msec nasal murmur. Has the /t/ of don't been deleted, or assimilated to the surrounding /n/ segments? Or is the speaker using a lexical entry for don't in which the final /t/ is already missing? Is the phonetic result of the "don't ask" sequence two nasal phones, [nn]? Or has the syllable-final nasal segment been deleted, after nasalizing the preceding vowel? Or have the two segments merged into one? How could we tell (say from a larger production or perception study)? Are these even coherent questions?


And when there's a silence and/or a region of irregular voicing (= "glottalization") between the vowels, it's natural to see it as a reflex of the /t/ of "don't", since syllable-final /t/ is often glottalized. But later in the same SA2 sentence we have the orthographic sequency "an oily", and the onset of oily is often glottalized (or started with an full-fledged glottal stop), as in FJCS0_SA2.wav:


So as we also noted, a glottal stop, or glottalized nasal murmur or vowel, might also be simply part of the phonetic interpretation of the vowel-initial word "ask". Again, there are questions of identity and process:

  1. Are glottal stop, glottalized nasal murmur, and glottalized vowel different "phones" in such examples? Or different versions of the same phone? Or non-segmental aspects of the phonetic interpretation of the underlying phonological representations?
  2. When glottalization occurs in a particular instance of "don't ask", should we attribute it to the /t/, to the word-initial vowel, or to both? Is there one phonetic segment or two (or more) in such cases? How can we tell?

Looking at a related question from a slighly different angle, here's TIMIT sentence SX441 "Don't look for group valuables in a bank vault", as produced by speaker MPRF0:


In this case, the /n/ in the syllable coda of "don't" and the initial /l/ of "look" have apparently merged into a sort of nasalized lateral, with glottalization of its medial portion.

And in the same sentence produced by speaker MJDC0, the same thing has happened, except that there's no apparently glottalization:


These are very simple examples, drawn from recordings made in a formal setting where speakers are reading lists of decontextualized sentences. The phonetic situation is pretty clear from the spectrograms and waveforms. But even here, there are many open questions about how to represent the pronunciations, and even more questions about how to relate the pronunciations to the lexical phonology. For discussion of another set of relevant examples, see "On beyond the (International Phonetic) Alphabet".

You should think about these questions as we continue through the semester.