Nov 29
======

Öhman:1966
----------

In Fig. 1 we see a significant drop in the voice bar when the /d/ and
/g/ sound is produced. Why do both voiced consonants occurring between
two vowels lose their voicing?

The description of the data processing methods suggests that the
author eyeballed the average line and drew it by hand, claiming that
statistical computations would take too much time and wouldn't be much
better. Do we know how accurate this approach is?

What would change with a more natural instead of a monotone speaking
style?

Why did the author focus on voiced plosives instead of other types of
consonants?

Why did fricatives, unlike stop consonants, not show a significant
locus shift?

The study uses one speaker per language? Is this a practical matter
ie. Availability of participants or they assume that there is no
within language variation? That would be a strong claim .How can it be
justified?

The authors suggest that the vowels in final positions are
neutralized. Does this relate to the fact that they are also
shortened?

Why are third formants difficult to measure?

My question has to do with the acquisition of coarticulation patterns,
especially as we know that they are, to an extent,
language-dependent. My eight month old makes a lot of VCV sounds - to
what extent is she picking up on my (or her dad's!) coarticulation
patterns? Do bilingual babies' coarticulation patterns different from
monolingual babies'? Anyway, it got me thinking. I guess we'll have to
look at them on a spectrogram.

While I understand that this paper was published in 1966, is it really
true that superimposing tracings to visually determine the average
formants for each word is not much inferior to statistical
calculations?

I'm generally intrigued by the use of cps as a unit of measurement. I
see from a quick online search that it stands for "cycles per second”
and is the same as Hz, so I am wondering if it was simply more common
at the time/if there is a particular reason the field of phonetics has
switched to using Hz?

Fig. 1 -> Just as a first impression, I'm surprised that the third
formant of /ø/ lowers for /øgy/ but not for /øga/. To me there is no
apparent reason why this should be the case based on the final vowels'
formants. (I recall talk of a "velar pinch"... perhaps something
related to that?)

Fig. 5 -> It strikes me as counterintuitive that the order of
frequencies for second-formant transitions for different vowels does
not remain constant across consonants. For example, it's easy to
understand that in general F2 for /ø/ drops less before /d/ than
before /b/, but it's not so easy to understand why for /d/ the order
of drop magnitude is /dy/ < /dø/ < /du/ < /do/ < /da/ whereas for /b/
it is /bø/ < /by/ < /bo/ < /ba/ < /bu/. Is there some sort of rhyme or
reason behind this or does it just speak to the idiosyncratic nature
of coarticulation?

"It is possible [...] to synthesize two VCV utterances that have the
same VC and CV first-, second-, and third-formant terminal frequencies
but that nevertheless differ with respect to the auditory impression
of the consonant that they give to a Swedish listener. At least one of
the vowels must be chosen differently in the two cases" -> Why does
one of the vowels need to be different? I think I may be missing part
of the logic/train of thought.