[b-hebrew] Phonemes and minimal pairs
jc.bhebrew at googlemail.com
Sun Jan 24 05:53:09 EST 2010
Despite my earnest attempts to make the blindingly simple eyeopeningly clear
I still get the occasional offlist email showing that key points of my
discussion are still being misunderstood. Just recently I got one or two
onlist as well and so feel it's worth composing a few words to straighten
The best way to make this clear is by discussing phonemes and minimal pairs.
Linguists have developed tests for ascertaining the human perception of
phonemes like that of minimal pairs. For example, can the native speaker
hear the difference between 'ship' and 'sheep'? If they can then we have
good reason to believe we have discovered a phoneme. However, at this stage
it may be worth observing an aside. Just because native speakers can hear
and reproduce this difference doesn't mean speakers of other languages can.
I will always remember my efforts to get some of my Italian students to hear
the difference between 'ship' and 'sheep'. Only the best and youngest were
ever able to hear the difference (a preliminary stage before having a prayer
of being able to produce it). In production tests they would typically
produce pretty much consistently something between 'ship' and 'sheep' for
both words. i.e. with an Italian 'i'. For me, someone trained by full
immersion in both systems, the difference is strikingly clear. For those
with full immersion training in only the one environment perhaps the
difference is not so clear. In the Ukraine we can see other examples. They
have a soft l which took me a long time to hear. I can hear it now no
problems and differentiate it from a normal 'l' but my production is far
from perfect. The best I can get to it is to produce an almost 'l' with the
beginnings of a 'y' after it. The shorter and softer the beginnings of that
'y' sound get the closer I get to being accepted as a native speaker.
Anyway, this is just an aside. On with the real points. The minimal pairs
test has sound basis and is a well respected formalised test in linguistics
for the establishment of phonemes. No problems here. But what needs to made
absolutely crystal clear is this. The test can only establish one level of
perception because it is limited to sequences of sounds which are naturally
produced. That is to say it cannot highlight the differences between well
formed speech where the micro signals between phonemes are correctly formed
and a speech signal with corrupt versions. However, we could modify the
minimal pairs test to highlight this level of perception. We could take a
speech signal and corrupt it at these points and present the natural and
corrupt versions to a native speaker and ask them to rank them. Once we have
seen that the recognition task is so routine to the native that they
consistently and effortlessly make correct ranking distinctions we see
compelling evidence that even these minute changes are perceived at some
level by the native speaker.
So why doesn't the native speaker of any language feel compelled to encode
these signals in there phonemic alphabets? The answer is quite simple. Let's
take a look at 'at' and 'ad'. There are differences in the signal between
'a' and 't' and between 'a' and 'd'. But the native speaker does not need to
have a written representation of them. The information of the 'a' and the
following 't' is enough for the reader to unambiguously know which sound
they need to make when migrating from the 'a' sound to the 't' sound.
Likewise, for the migration from 'a' to 'd'. In conclusion, native speakers
around the world with various phonemic alphabets have never and likely will
never feel the need to encode the bridges between phonemes because while
they are quite clearly perceived at some level they are completely
unambiguous and so there is no need for any form of graphical representation
(only a mental one).
And so now we get back to the implementors of the Hebrew alphabet. Or more
accurately, we should probably be talking about the implementors of the
Could they perceive at some level the minute signals between phonemes? Of
course they could! They were human beings with the same level of
sophisticated hearing! Could they perceive phonemes including consonants and
vowels as highlighted by the minimal pairs test? Of course they could! How
would they be able to produce the different vocalisations of MLX if they
couldn't?!? Did they feel the need to encode these differences in their
alphabet? No! Why not? Because they didn't feel the level of perception was
ambiguous enough to warrant implementing an encoding for similar reasons
that we don't feel the need to encode the bridges between phonemes. i.e.
they did not use the minimal pairs test in the formulation of their
alphabet. Let me say that again for emphasis. They did *not* use the minimal
pairs test to formulate their alphabet. They were working at a level of
perception one stage higher. That is to say, they were working at the level
of syllables just like everyone else who had had enough of logograms in the
When working at the syllabic level there are evidently two major approaches.
1) We can create a unique symbol for each syllable conforming to some
minimal pairs description of phonemes.
2) We can operate at a higher level of abstraction and economically select
symbols that represent a group of syllables unified in some respect (e.g.
the first phoneme of the syllable)
There are clear advantages and disadvantages to system 1. A well formed
system can be far less ambiguous. We can make differentiate Ba (CV) from Bab
(CVC). However, in practice developing such a system can be a messy business
because of the difference between the way we enunciate words in isolation
and the way they actually get spoken in natural language. We also have to
learn a much greater number of symbols. This is a clear disadvantage.
System 2 has its merits and disadvantages. Its clear disadvantage is its
potential for ambiguity. Its clear advantage is its economy.
Of course, list members have implied a third possible system:
3) Implement the consonants only.
And so the debate for how the proto-Sinaitic alphabet was implemented boils
down systems 2) and 3). That is to say did they actively decide to implement
consonants only or was this just the by product of implementing an
economical syllabic alphabet? Evidently the implementors had some level of
perception between the consonant phonemes and the vowel phonemes. This is
evidenced by the symbols they chose. The syllables that each symbol
represents are only unified by their initial consonant sound and so it is
more than abundantly clear that the implementors perceived these
differences. But how did it come to be that the implementors unanimously
relegated vowel sounds with explicit linguistic knowledge of the difference
between a vowel and a consonant? Evidently, this kind of knowledge was
implicit. Let's look again at our earliest attested example of
Things worthy of note are the following:
1) All symbols are in their natural upright position (most visible with the
ox head and the man)
2) It doesn't seem to matter which direction they are written in
3) There are no spaces between words
Let me say that again for emphasis. There are *no* spaces between words.
That's right! This looks like a more professional handling of a speech
signal than modern alphabets. Speech signals don't have gaps in them where
we make almost arbitrary word distinctions. The lack of spaces in early
inscriptions reflects this nature of spoken natural language. We are looking
at a representation of the speech signal with the following features:
1) Syllabic segmentation
2) Economic use of symbols encoding explicitly only the consonant phonemes
So why weren't the vowels encoded? What is the difference between vowels and
consonants? Using minimal pairs testing they would seem to be of equal
importance, right? But evidently they are not. Syllables and their
beginnings are perceived at a higher level of perception than the vowels
that merge them. It was this level of perception that the implementors of
the proto-Sinaitic alphabet were evidently working at.
More information about the b-hebrew