Margo's Magical Letter Page
by Margaret Magnus
copyright 1998
all rights reserved
I have shown that /b/ words fit in certain /b/-sensitive classes called phonethemes, but unless it can be demonstrated that other words do not fit in these classes, I have shown nothing at all. The theory cannot then be falsified. Fortunately, such an experiment can be carried out straightforwardly. I have run larger scale tests of this type later and in a more efficient way. But let me now run such a test on a small scale. This will both serve as an example of the type of test which can be used to falsify the findings of phonosemanticists, and it will put perspective on the other data I have presented.
I have dealt with two kinds of classifications for the /b/ words. First I outlined a number of major classes which were tailor-made for /b/, in which one would expect /b/ to perform well if indeed its pronunciation directly affects the semantics of the words which contain it. I showed that some sense of a nearly all monomorphemic words beginning with /b/ do fit into these classifications. Nearly all of the remaining words had very concrete reference.
Several of the concrete nouns could have been classified either in a phoneme-neutral or a /b/-favorable way. For example, I can classify words referring to 'Place' in such a way that as many /b/ words fit into some class as possible. This classification shows itself to resemble the pronunciation of /b/ - a result which supports the hypothesis:
/b/ Places (b-biassed classification scheme)
Exceptions: 1%
A phoneme-neutral classification oriented toward the structure of the world would also be expected to include a large percentage of words in a relatively small number of classes. Such a classification might look like this:
/b/ Places (phoneme-neutral classification scheme)
Exceptions: 3%
If the Socratic hypothesis is correct, you would expect that if you sort the words of place beginning in /g/ or /n/ according to the phoneme-neutral scheme, they too will fit in nicely. But if you sort them according to the classification which is biassed toward /b/, you would expect that they fit in less nicely:
/g/ Places (neutral)
Exceptions: 4%
/g/ Places (b-biassed classification scheme)
Exceptions: 54%
The phoneme /g/ along with the other velar and labial stops has a fair number of containers. /g/ is the phoneme of the gullet. Notice that one class that /g/ really wants to form into is valleys:
Valleys/Depressions - gill, glen, gorge, grave, groove, grotto, gulch, gulf, gully, gutter
Some of these can be forced into the 'backlogged' and 'bottom' classes of /b/, but the 'valley' class is better, because it includes a greater percentage of /g/ place words. Another class into which a large percentage of /g/ place words fall is parks and other open green areas, which are completely unrepresented in /b/. Notice that these words tend to contain an /r/ which in general frequently correlates with growth:
Parks - garden, glade, grange, grant, green, grounds
Let me run the experiment again with the places in /n/:
/n/ Places (neutral)
Exceptions: 5%
The phoneme /n/ has even more bumps than /b/, but the /n/ bumps are small and knob-like. /n/ has nearness (narrows, near, neck, next, niche, nigh, nook) and smallness (knob, knoll, narrows, neck, niche, nipple, node, notch, nest, nook) about it. It is also very negative (nay, never, nix, no, none, nope, not, etc.). The 'base/bottom' words in /n/ come from its negativity. The connecting words in /n/ have a different geography from those in /b/. In /b/ there is a bridge between two points or a fastening. In /n/ connection shows up as intersecting lines:
/n/ Places (b-biassed classification scheme)
Exceptions: 58%
Let me conduct one further and tougher test to verify the hypothesis. If I am correct that it is the pronunciation of /b/ which accounts for the above statistics, then we would expect that were we to take the words of 'place' from a different language and conduct exactly the same experiment, we would get very similar results to what we get in English. We will use Russian as the test language, which is far enough from English that cognates and borrowings will not have such a powerful effect on the data:
/b/ Places - Russian (b-biassed classification scheme)
Exceptions: 8%
/b/ Places - Russian (neutral)
Exceptions: 5%
And now let me try the place words beginning with /t/. I do not choose /g/, because there are an overwhelming number of foreign words in Russian /g/, and I do not choose /n/, because 4 of the major prefixes begin with /n/, and there are very few root words left - only 6 /n/ place words. However, despite the paucity of data, both /g/ and /n/ do yield results similar to those presented here for /t/:
/t/ Places - Russian (b-biassed classification scheme)
Exceptions: 47%
/t/ Places - Russian (neutral classification scheme)
Exceptions: 11%
Notice that Russian words fit the English phoneme neutral classification less well than English words. And the Russian /b/ words fit the English /b/-favorable classification less well than English /b/ words. Still, the tendency is clearly for Russian /b/ words to favor the English /b/ favorable classification more than do Russian /t/ words. If it's true that the metaphors which predominate in a language help determine which classes exist in a given phoneme, then you would expect this result. For example, if it's true that the class of 'angry' words in English /b/ exists because of the existence of the English metaphor 'Explosion IS Anger', then you would expect an 'angry' /b/ class only in those languages that have this metaphor.
This test is repeatable. What is more convincing than reading the data presented here, of course, is for the reader to conduct the test independently in any language using /b/ and another phoneme. (hint, hint) The test need not be conducted for words of 'place', of course. It can be any class.
Margo's Magical Letter Page