Poll Response

Margo's Magical Letter Page

Following are replies to an earlier query in a different format. I used the responses given here to reformulate the query.

Responses to New Query

Response 1: linguist with the Summer Institute of Linguistics in both SEA and Africa. I am presently teaching linguistics at the Graduate Institute of Applied Linguistics in Dallas, TX.
Response 2: NJ sans doctorate
Response 3: I am strictly an amateur philosopher with no way to type myself as to employment, academic or otherwise.
Response 4: Germany, Professor em. of psychology
Response 5: LIVING IN ISRAEL since July 1970. Lived in Florida/Georgia USA from birth in 1937 until 1970. BA (math/psych), JD (law) both from Univ of Florida. Worked in the States as an insurance mathematician. Worked in Israel as a programmer and software engineer. Am currently working as a technical writer (in English).
Response 6: I am currently in Chungli, Taiwan. However, I am a native U.S. citizen, born and raised in Michigan. I'm currently an Associate Professor of English. I hold a B.A. and and M.A. in Anthropology, and a Ph.D. in Linguistics.
Response 7: Ontario, psychologist
Response 8: Australia, full professor = "lecturer", in my dialect
Response 9: Austria no, full prof. linguistics Berkeley, at the moment lecturing on pronominals, with a major component on the phonological content of pronominal systems (NB systems, not particular persons). I argue that it's the "audibility" of oppositions that is crucial and that this predicts both: 1) the unmarkedness of most segments involved pronominal morphology, and 2) the fact that a majority of languages have pronominal systems with phonological content displaying major class distinctions among pronominal morphemes--most commonly obstruence vs. nasality vs. glide or laryngeal.
Response 10: PhD in linguistics, American
Response 11: Yours truly, 4 years in MIT linguistics grad school. 10 years running natural language software business. 4 years freelance phonosemantics. autumn, defend PhD at U of Trondheim.


R3: I assume without statistics that the general and majority trend is against this, given the sway of statistical and other naturalistic thinking at which I, to be blunt, scoff heartily.
R8: I'd say virtually zero. All would agree with Saussaure that there is some degree of iconicism in the lexicon/grammar. It's just that some think that it is not a truly linguistic element, and should therefore be safely ignored. Others disagree. If I wanted to get a _representative_ sample of what linguistis think, I wouldn't post the questionnaire to the S-Symbol list, as you're obviously going to get a highly skewed sample.
R11: My impression 4 years ago when I started this site was that a handful of linguists in the world admitted some level of iconicity, but that 98% believed the sign to be completely arbitrary. Now it seems to me either my impressions have changed or opinion is changing. I'd say more like 90% of linguistics believe the sign to be completely arbitrary.


A1. Word meanings are completely arbitrary.
R1. True of some words, not of all.
A2. All word meanings are nearly entirely arbitrary except for a few words whose meaning to some extent reflects their sound.
R1. Disagree
R10: certain languages or vocabularies within a language or periods in the development of language exhibit (1 or 2) Other vocabularies, languages or periods exhibit the other,
A3. A substantial percentage of word meanings are affected by their sound in some languages. In other languages sound rarely affects semantics.
R1. Disagree with the second part
A4. A substantial percentage -- though less than 50% of word meanings are affected by their sound in all languages.
R1. Probably true of some lgs.
R4: Agrees, but ""affected by their sound" is perhaps ambiguous. Do you mean "word meaning" as realized by individual language users (subjective or "impressive" sound symbolism, SSS) or do you mean that the phonetics of words of particular languages, during language evolution, has been affected by meaning (objective or "expressive" sound symbolism (ESS)? The percentage of word meanings affected by SSS is much larger than the percentage of words affected by ESS.
R5: Agrees.
R6: I'll agree with this statement.
A5. A substantial percentage -- over 50% of word meanings are affected by their sound in all languages.
R1. Could be true of some lgs., but still needs a lot more study.
R7: Agrees
R8: Agrees. But we probably don't agree on what we mean by "affected by their sound"
R9: None of these is quite what I believe. Actually, I don't particularly like the implied directionality of the questions in A, since I'm more inclined to think in terms of (semi-)arbitrary association as well as true iconicity. So I want to say something like A5, but in terms of sound meaning correlations. I'd further want to hedge about this to say "most languages", because there are many languages, like English, which have massive borrowings that muddy the waters considerably. So in the terms of the survey I'd want: In most languages more than 50% of words show sound meaning correlations, but in the remainder, a substantial percentage--though less than 50% do.
A6. All word meanings are affected to some extent by their sound in all languages.
R1. I disagree
R11: No words are wholly determined by their sound, but all word meanings are affected by their sound. In general, each form has a unique visceral effect on the psyche, and that holds of speech sounds as well. The fact that speech sounds are also used in language does not prevent them from having this effect. The universal character of this effect applies equally to all words. However, the more concrete or specific the referent, the less salient is this effect. This can be verified empirically beyond a reasonable doubt.
A7. All word meanings are wholly determined by their sound in all languages.
R1. Definitely disagree
R3. I gravitate toward full sound symbolic meaning for all words mostly as an offshoot of my reverence for the thinking of Plato, Aristotle and, especially, Pythagoras. In modern terms this means sympathy for the new complexity theorists, who, unlke the evolutionists, believe that only coherent forms can come about and thrive, i.e. that randomness and true chaos are not possible. Naturally the same applies to language; it is not chaotic and random/conventional only, but meaningful. Thus I take for granted that words and the larger language system of which they are a part are meaningful sound patterns, and that much of the truth of this remains to be found by those who will take the trouble to look for it.
R10: None of the above. My view is that a large proportion of words whose meanings are in domains for which users perceive metaphorical concrete meanings (including as concrete also time and space, not merely objects) have forms affected to varying degrees by their meanings.
R11: No word meaning is wholly determined by its sound. Obviously not... if it were, then you'd be able to tell what some words meant by listening to it, and that holds of no word.


B1. Sound affects meaning in onomatopoeic words, in words which refer to sounds themselves or which refer to things which make sounds. The phonetics of these words resembles the sound of the referent.
R1. Agree
R2: ok to an extent, but we really are dealing here with a code, not absolute acoustic resemblance.
R4: Agree
R8: yes
R9: Again I'd want some wiggle room, even though I pretty much agree. How languages reduce "raw" sound to "tame" sounds is partly arbitrary. (This is actually a strong argument against the extreme iconist stand.)
R11: A few
B2. Early languages were more iconic. As language evolves, it becomes increasingly abstracted from its form. Therefore those languages which have changed less over time are more iconic.
R1. Speculation, I certainly would not agree that languages with more sound symbolism are therefore more primitive or less evolved.
R2: with a caveat- because of parallelism between language history and ontogenesis, you have to be careful with "early", and cyclicity historically is something to consider- simpler structure with more phonosemantics can actually be "later" than more complex structure with less iconicity.
R4: I found that empirically (see my Psychophonetik book)
R7: Agrees
R8: no
R9: This so fries my shorts that I have to say it. Absolutely not. It's just possible that the very, very first human language had more iconicity, but what is the sound of 'in'? Or for the strong iconists, which of the five (or is it six) possible iconic representations of 'in' wins?
R11: I don't really know what this means exactly, but it's a position I've often heard expressed... certainly in order for a language to be a language at all, it has to have reference. If a language did not have reference, it would be like an art form, like dance. It could express a mood, but it could not be used to say anything like, "Please pass the potatoes." In order to say anything, words have to be used to refer to something other than themselves, and that choice of referent cannnot be predicted from form or sound. So every language must have both elements of meaning (iconic meaning and reference) active and functioning in every word. But whether languages in earlier societies tend to be more poetic or have less concrete referents, I have no idea.
B3. There is a natural process in human psychology which conspires to associate with any form a coherent meaning. Therefore there is a tendency (for example in a language learner) to try to ascribe a single coherent meaning to a word and a morpheme with the result that words and morphemes are said to have 'meanings'. This process (I'll call it Semantic Association) also applies on the lower level of the sub-morpheme (partial syllables -- to onsets and rhymes, say, as described by e.g. Bolinger, Rhodes, Lawler, McCune).
R1. Some level of validity, esp. in some areas of a lg.
R4: I don't know whether the ambiguous term "word" in " try to ascribe a single coherent meaning to a word and a morpheme" is useful here. SO I do not exactly understand your question and what you mean by "Semantic Association", is it perhaps the same as the denotation of my "impressive sound symbolism"?
R5: Agrees
R6: I'll agree, with hesitation, to this.
R8: yes
R9: I, of course, still agree with myself, my co-author, and my student, and my late, esteemed colleague and intellecual forerunner.
R11: Yes, Semantic Association is a universal tendency like the visceral, immediate effect of sound, but it is a separate phonosemantic process. This doesn't mean that all words starting with a given submorpheme all have a commen element of reference (i.e. all refer to some aspect of light or cohesion), but it does mean that the mind tends to ascribe to /gl/ a coherent semantic range. In the same manner, often the various senses of a word do not share a common referent, but the mind tends to interconnect the sense nevertheless by analogy, metaphor, etc..
B4. Semantic Association applies also on the level as low as the phoneme.
R4: Agree
R8: yes - in a far more limited set of cases. e.g. smallness and "i"
R9: Yep
R11: Yes... it holds of all phonemes in the same manner it holds of all submorphemes
B5. Semantic Association applies as low as the phonetic feature (i.e. all labials or stops have some common element of meaning).

R1. Lgs. adapt sound symbolism to their phoneme level, but the basis is probably at the phonetic level.
R2: B3 to B5- because of cyclicity (again), you can go as low as a feature in some languages, but perhaps only as far as multifeature/phoneme clusters in others. I'm not sure is Association is what really happens here, in fact I lean towards loss of inherent structure over time (so the brain more and more is grasping at old straws that aren't there anymore, at least until cyclicity of type reinvigorates the system).
R4: Agree
R8: it's possible, I'm not aware of any examples offhand.
R9: I was ready to say yes, but then you went and used the universal quantifier. I think there are semantic associations at subphonemic levels, but at no level do I believe the association effect is all inclusive, not even at morpheme level. (How many morphemes in English are there with the form /z/? Are genitive, plural, and 3rd sg. present reflexes of the same meaning?)
R11: Yes... it holds of all phonological features in the same manner it holds of all submorphemes and phonemes.
B6. Phonosemantic effects are truly iconic. I am thinking now of something like what Plato describes in the Cratylus, where things in the world to which words refer have an 'essence' and the sound meanings somehow synesthetically reflect that essence. It need not be that this synesthetic association be 'right' or 'wrong' like Plato thought. It can be like Nodier thought in his later years that there's an interpretive element to this synesthetic correspondence... in other words, different people or cultures and 'feel' this essence differently and hence express it differently.
R1. The notion of an 'essence' is much too abstract and philosophical. It is more a matter of sensory areas of meaning where iconicity enters in.
R2: There are typological differences in the organization of phonosemantics, not cultural- the iconicity shifts in predictable and coherent fashion from type to type.
R4: his is too much and cannot therefore be endorsed fully. In the fiirst place I do not exactly understand the meaning of "Phonosemantic effects are truly iconic". I also hold "there's an interpretive element to this synesthetic correspondence.." But this does in no way imply that "different people or cultures and 'feel' this essence differently and hence express it differently".
R6: I won't exactly agree to this statement, but feel it is worth exploring further in more elaborated forms.
R8: no; probably impossible to prove or falsify.
R9: Nope, but I should point out that under on generous reading, the Nodier version is like a tighter form of B3.
R10: Again, none of the above. Really two different questions. One is in B1 and B2, where my view is that certain parts of the vocabulary (not languages as wholes, "primitive" or other) are more subject to sound symbolic effects than other parts. So as vocabulary grows with increasing numbers of technical jargon words which do NOT have concrete metaphors (as in A. above), the degree of sound-symbolic effects on the vocabulary as a whole may decrease. But since people remain rooted in gravity, day vs. night, etc. the core I expect to retain a tendency to reflect sound symbolism. One could also phrase it that there is a strong correlation between word length and abstractness, so that shorter words are more likely to show the effects of sound-symbolism than longer ones, on average, not in all cases. Not because they are short, but because those words which are closer to having gut-reactions tend to be short. So it was a very good pragmatic choice for Margaret and others (almost all of us?) to focus in some of our own work on monosyllables, even if polysyllables can also contribute. A second question is in B3, B4, and B5. What units have sound-symbolic semantic associations? Word, morpheme, phoneme, or phonetic feature? I think some of each, to be determined empirically (as the preceding questions also). But not all in any given level. Required is a word with sufficient concrete metaphorical associations, and then any level MIGHT in a given word in a given language reflect the sound-symbolic semantic associations in some portion of its form. No universals here, *because* historical change can break old sound-symbolic semantic associations (that's most of the reason for "arbitrariness of the sign", which is just as pervasive as sound-symbolic associations).
R11: They are this, and they are also subject to Semantic Association, a separate process or law. Semantic Association or clustering differs from iconic meaning in that it involves reference and hence has an arbitrary dimension. It is the tendency for a certain range of referents to cluster toward certain phonological forms, and the tendency for the mind to want to interconnect these referents in some cognitive, rational way.


C1. The semantics of a word is analyzable into discrete identifiable components -- some of these are arbitrary and some are not.
R1. Partly agree
R2: OK
R4: Here again I have difficulty because of the ambiguity of the term "word". I hold that the phonetic units of words (Lautgebilde) are analyzable into components (but that is merely like saying that morphemes consist of phonemes). I also hold that these components (phonemes) contribute to complex SSS. And as far as ESS is concerned I hold that phonetic complexes of words of natural languages need not be fully determined by associated lexical meaning, some arbitrariness is generally involved. But this is a statistical question. ESS problems cannot be resolved by single word demonstration.
R7: Agrees
R8: yes (first half). what do you mean? (second half). If you mean "arbirarily associated with a phonetic/phonological form" then yes.
R11: Yes, word semantics is analyzable. It's not an amorphous blob. Reference is clearer to pinpoint in words with concrete reference, but that doesn't mean the semantics of other words can't be analyzed. In general, it's dangerous to make claims about what can't be done.
C2. The semantics of a word cannot be cleanly enough analyzed into discrete components that it makes sense to talk of some components being arbitrary and others not.
R1. Partly agree, as in morphology, we can make clear cuts for some, but find others to be fused to varying degrees.
R4: The concept "semantics of a word" should be clarified. I think that lexical or denotative meaning and connotative meaning (what I called "Allgemeinqualitt later)is a useful distinction, even though connotative meaning of words may become hardly distinguishable in case of words denoting feeling tones on the lexical meaning level. If "semantics of a word" is applied to connotative meaning then clearly I cannot endorse your statement. Osgood`s Semantic Differential" is a technique suited to analyse connotative meaning into components. That's what I did in my Psychophonetik.
R8: This again is really two questions. No, no.
R10: It is not that semantics is a sort of amorphous blob, but rather that the forces applicable may be acting at the same time with various strengths, so that however finely one does semantic analysis (even in cases where it may be relatively more valid than in other cases, and that is always a question), the semantic components one ends up with may still be affected by various visceral-metaphor forces with varying strengths. I think it is the INDIRECT relation between forces and results that I am emphasizing, because forces can also contradict or even just slightly nudge results off course of what some one force alone would do. So using a different semantic analysis doesn't change that, it merely means we are dealing with parts of a word's semantics rather than the whole of it. If one is segmenting a word or morpheme or whatever into semantic components, and one says that one of the components is affected by some visceral force, then that is no longer saying that the word or morpheme [the whole] is affected by some visceral force. I guess even more bottom line, I simply don't believe that one can classify words or morphemes into discrete, mutually exclusive categories, some labeled sound-symbolic(ally influenced) and others not. Relatively more or less, in various ways, sure. But not discrete categories. Yet we still can try to make distinctions of degree and kind, there is every reason to pursue that.
C3. The semantics of some words are analyzable, while the semantics of other words cannot be fully analyzed.
R6. Agree
R9: Oy! Semantic analyzability varies enormously from morpheme to morpheme/word to word/phrase to phrase. Words and even longer expressions tend to accrue aspects of meaning that are not just the sum of their parts. Thus I'd say that the semantics of a morpheme/word/phrase cannot always be cleanly analyzed, but that clear sound symbolic effects can nonetheless be observed.
R10: Again, some cases *more like* C1, some cases *more like* C2, without my having any wish to choose between C1 and C2 as generalizations. This is the issue in another guise whether "generative semantics" has validity. For some parts of semantic space / vocabulary, it comes closer to being valid, for other parts, it is farther from being valid.
R11: All words can be analyzed.


D1. There is some correlation between the phonological form of words and their semantics, but that correlation is 'conventional'. That is, Sapir, for example, believed there to be a correlation between sound and meaning, but he believed that that was only due to a tendency to give words with similar sounds similar meanings. So if the basic word for 'house' in some language starts with /h/, then other words for dwellings are likely to start with /h/ as well, but whether or not 'house' starts with /h/ is arbitrary.
R1. Partly agree, some are conventional to a lg., some to an linguistic area, and some are more universal.
R4: I think here you address two different issues, namely the question whether sound meaning correlations are conventional or not, and the question, how sound meaning correlations come about. I think, to some extent ESS is conventional i.e. the phonetic comples of a word of natural language L1 might reflect (express) one connotative feature of the denoted meaning while a word of language L2 with the same denotation might reflect another connotative feature. E.g., a phonetic complex)Lautgebilde) might reflect brighness of lightning, another phonetic complex might reflect speed of lightning, both brightness and speed being features of lighning.
R8: I don't know. For a subset of the lexicon, D1 may be the case. Maybe not.
R11: Yes, it's this phenomenon that accounts for the phonesthemes, and phonesthemes, or disproportions between phonological forms and semantic classes are hold throughout all vocabularies, since Semantic Association is a natural psychological law. I've heard it often argued that no such claim can be made, because obviously I haven't classified every word in every language into phonesthemes. Indeed, even if I had, that would not be sufficient to make the claim that in all future languages, this would necessarily hold. I claim it nevertheless, because it has become obvious to me spending time with the data that a natural law is involved. So although I have never been to Pluto, and although no human probe has been to Pluto, we can still self-assuredly claim that an object when dropped near the surface of Pluto will fall toward the planet. This we know, because we recognize that anything composed of matter must be subject to a universal natural law we call gravity. I need know nothing about Pluto other than that it is composed of matter to make this prediction. And similarly, I need know nothing about a language other than that it is a language to predict that sound will affect word meaning viscerally, and that there will be a clustering effect all the way down to the level of the phoneme and the phonetic feature. However, without measuring Pluto specifically, I cannot know how fast objects will fall there, and similarly without investigating an individual language, I cannot know what the clustering classes will be.
D2. The correlation between the form of a word and its semantics is 'natural' or perhaps synesthetic. So there is some aspect of the meaning of /m/ or /l/ which will be universal (cross-linguistically).
R1. Perhaps, only true of /m/ and /l/ in some greater or lesser percentage of words, depending on the language and area. Magnitude displayed by vowel openness is found in many words in some languages and fewer in others. I would put any of these individual things at a small percentage, e.g. perhaps, the sensations of /m/ would not be involved in more than 10% of words containing /m/ in any one language.
R2: D1 and D2 come into play at different points in the type cycle. English is at the D1 point, Mongolian at the D2.
R4: This is the result of my Psychophonetik
R5: Agrees. Reservation: When comparing cross-linguistically, sounds must be grouped, for example: D/T/DH/TH/H, B/P/F/V/W, B/M/W, GH/KH/KS/X, all siblants, T/SH, G/K/Y/Cr, L/N/R, etc.
R7: Agrees
R8: For most phonemes, no. For some, e.g. "i" mentioned above, yes.
R10: Yes, I think it can be to the extent that people have assumed the results of analogy were in the "natural" class. I don't myself, though, I think the results of analogy which are NOT direct visceral effects are NOT in the natural group, they are historically arbitrary, analogy being simply one of the forces other than direct visceral effects which like standard sound changes can be unrelated to the direct visceral links and thus can produce word-families-by-analogy which are NOT sound-symbolic. How can one tell the difference, or do we assume a priori that any grouping which shows a recurring sound-meaning relation is evidence for an analogic/natural class? I thought that was equally the evidence used to show an arbitrary, conventional, morpheme (when it is sufficient in quantity and in discrete separability).
R11: Yes, this holds of all phonemes, though the universal meaning is not cognitive or denotative in essence but a 'b' which appears in a verb of contact conveys a certain fairly specific feeling tone to the word, and this feeling-tone is the same for all 'b's which sound the same.
D3. D1 for some words, D2 for others, and perhaps other additional factors involved for certain words.
R6: Agree
R8: I don't know. For a subset of the lexicon, D1 may be the case. Maybe not.
R9: I, of course, believe both, sort of. D1 is basically B3. Change D2 to "can be 'natural'" and I'm with you. I think that there are iconic meanings for sounds, probably at the level of feature clusters, e.g. apical stops. To me these provide pressure toward particular meanings for morphology of particular phonological content, but like the deep currents of the ocean, this subtle force can be easily overridden by the winds of sound change or borrowing or semantic change itself.
R10: Again, both. D1 expresses an additional force operating to shape vocabulary, which is acting together with sound-symbolic semantic associations. Probably D1 should actually be considered simply part of historical arbitrariness, though a specific sort of that.
R11: No, both apply to all words.

R10: In not a single case do I wish to choose one rather than another of the alternatives. I do not even want to TEND one way rather than the other, don't believe in a view that one approach is *predominantly* valid to the exclusion of the other. That last phrase, no exclusion, is crucial. Asserting one of the answers *buys us nothing*, if it is merely noting that a PORTION of vocabulary in some language or languages (not the same in each, even if partly overlapping) reflects sound-symbolic semantic associations, or does not.

Noting a pattern cannot in the slightest reduce the fact that there are always ALSO other portions of the vocabulary of the same language which do NOT reflect sound-symbolic semantic associations in the same pattern, or which DO. Noting a group behaving one way does NOT make the another group behaving another way into "exceptions". Rather, all portions of vocabularies are normal, some evidencing more sound-symbolic semantic associations, some less. Statistically, we can partly predict which parts, but not in detail, because of the arbitrariness of history. I believe choosing either extreme view on this question in general would be the same mistake made by many rule-focused generative grammarians, in trying to find rules and declare other items to be exceptions, when there may be many such so-called "exceptions" in several directions.

I believe that advancing this field will come by the gradually accumulating quantity of studies of particular instances, by investigators who are equally willing to conclude that given vocabulary DOES or DOES NOT reflect the actions of forces of sound-symbolic semantic-association, even TO VARYING DEGREES, not simple "yes" or "no" answers. In other words, by those without an axe to grind in EITHER direction, and a great ability to distinguish cases from each other, rather than lumping them all together.

For me, the most interesting question in the success or failure oflinguistics in these domains is why people are so polarized? Why is the middle ground so hard to maintain? That is an extra dimension, orthogonal to many of the questions Margaret has been asking.

Margo's Magical Letter Page