Discussion Groups / English Usage / January 2004

Interesting question.

I know that any wave of any arbitrary complexity can be separated into a
(large) number of sine waves of differing frequencies and amplitudes. The
process is called Fourier Analysis, I seem to recall.

I suspect that there is sufficient difference between different people's
voices saying the same word(s) to tell them apart, even if one is mimicking
the other. I think that organisations such as the FBI can identify voice
prints quite accurately. The more recent versions of Word can accept voice
input, after you have 'trained' the software by repeating certain phrases
to it until it 'learns' your method of pronunciation.

You could certainly pick up the different vowel sound, if only because the
southern variant physically takes more time.
 I do some work on this and it's certainly possible to pick up regional
variation in the wave form of a particular word.  A couple of years ago I
did an analysis of a Southern BrE speaker and a New Yorker (both women).  
We were looking at intrusive j (yuh) sounds - anecdotally you'd expect the
New Yorker to say 'Noo' for 'New' and a southern brit to say 'Njoo' - in
fact it wasn't remotely that simple (in the 'new' example exactly opposite)
, they both used both forms and had them equally - though differently -
distributed.  Anyway, the extra j phoneme in the waveform of the brit's
realisation of 'new' is clearly visible.  I've also just done some
segmental work in which speakers read a list of body parts (don't ask).  
One of my speakers (a Californian) spoke possibly a bit too slowly and
precisely for the purpose I needed the clip for - pronouncing 'chest' with
a long gap before the final t sound.  I tried shortening the gap - taking a
little out made for a more authentic sound. (IMHO).  Certainly the extra
length gap would show up clearly on the waveform.

They'd look similar, but noticeably different.  It would also be possible
given patience and a burning need to avoid doing something more important
(like finishing the paper I'm currently desperately trying to avoid working
on) to cut and paste the vowel sounds and change one to the other.  This is
more easily done with consonants - 'chest' to 'chess' is obviously a piece
of cake (you could even do it on magnetic tape given no life whatsoever and
a sharp pair of scissors).

There's a piece in David Crystal's Cambridge Encyclopaedia of Language
about 'voiceprints', their history as being legal evidence in the US
(checkered) and the work of one Lawrence Kersta (has to be worth a Google).

Is it the case that there are noticeable

Yes.  Each waveform will have a similar shape (what synthesiser musicians
call an envelope) but variations in pitch, intonation and volume will
inevitably make each unique and probably noticably different to the naked
eye - as you up the resolution of the waveform (zoom in on the graphic)
you're bound eventually to spot differences.

Hope this is of interest
DC Cat

John,

The majority in USA think they say Mary, marry, merry the same.  What
I wonder about is whether in/ing/ink have the same vowel and how to
test this.

tom z

When I took Phonology in about 1983, the prof discussed their use in
court.  Essentially, there are sufficient differences between speakers
that it is often trivial to take two spectrograms of the same sentence
(say one off a recording and one given by the defendent) and say
confidently "These were not made by the same person, even though they
are indistinguishable to the ear".  On the other hand, if someone says
on the basis of sound spectrograms "This recording was made by the
defendent and could be no other", they're either incompetent or
committing perjury.

There are definite traits you can find having to do with formant
height, voicing onset time, and such, but you have to be careful, as
what people do is adjust their speech until it "sounds like" that of
the community they grow up in.  This is a problem that admits of more
than one solution, so even within a given dialect, you can expect to
see differences that, taken together, wind up sounding "the same".

I'm a good case in point.  For some reason, all of my /l/s, when
speaking English, are velar.  (Most English speakers, including those
I grew up with, use an alveolar [l] before and between vowels and a
velar [l~] after vowels.)  It's not that I have any trouble
pronouncing an alveolar [l], and, indeed, I picked it up without
noticing when I learned Spanish.  It just sounded right to me and was
obviously "close enough" for others.  It wasn't until I took Phonology
in college and learned that I was *supposed to be* using an alveolar
that I realized that I was different.  This, of course, would show up
quite clearly on a sound spectrogram, but it has nothing to do with my
dialect.