Responses to my Linguist-List Posting:
Echo-Word Reduplication

Hal Schiffman
South Asia Regional Studies
University of Pennsylvania

I received many responses to my enquiry of Sept. 17th, and have tried to compile them into useful rubrics, below. It's divided into subjects such as

  1. What do we call this phenomenon?

  2. Data

  3. People (contributors);

  4. Bibliography on the subject;

  5. the phenomenon in Languages other than English ;

  6. a paper contributed by Terry Regier

  7. Some unsorted and unedited but useful commentaries.

My original purpose in making this query was to be able to describe this phenomenon better to novice linguistics students and others not conversant with the data, and especially people who see reduplication only as a phenomenon in baby-talk or 'lesser' forms of language, not in modern English(!). With the help of Vasu Renganathan, I have started a kind of data-base of new items, espl. those not already found in extant publications, etc.

This can be viewed at: Here you can view items I have entered, and if you wish to contribute your own, scroll down and use the data-entry page.

  1. What do we call this phenomenon?

  2. Bibliography

    (not yet in alphabetical order)

  3. References to the phenomenon in other Languages

    1. Richard Noss's excellent discussion of this for Thai in PDF format; if you don't have his Thai Reference Grammar I can arrange to yank and post the relevant pages for you. (Doug Coope)

    2. Spanish: (Carsten Sinner) I've been collecting examples of reduplication in Romance languages (Spanish, Portuguese, Catalan, Galician) and in German for quite a while, always keeping the problem in mind and waiting for more time... I'm in the final phase of my PhD in Romance Linguistics. I collect examples for very different word formation procedures and very different meanings, but in the Romance languages the reduplication is especially used for underlining of meanings, description of longlasting effects or procedures, high intensity, etc. My interest is more the formal linguistic aspect of reduplication and lexicalization of reduplicated elements, but your approach is also fascinating.
    3. Other Spanish: Spanish has a lot of expressions such as No decir chus ni mus[something like "To not say Jeez nor bees"] (chus can be a nickname for Jesus -- but it's not sure that's what this expression refers to) Del cabo al rabo [from head to tail] De la Ceca a la Meca [from la Ceca (mosque in Cordoba) to Mecca]

    4. Lloyd Anderson: I can include also six pages on Turkish intensive adjectives (which resemble the Yiddish pattern you mentioned) and a couple of pages on Spanish. (Also a paper on vowel harmony which includes some odd phonological patterns resembling those in the English rhyming doublets data, suggesting that /b/ and "esh" are paired, and /p/ and "j" are paired, crossing the voicing dimension.

    5. Scottish Dialects:!) There's a good few forms in Scottish English I use occasionally which people down here in England don't seem to have heard before - I think many derive from Scots. Here are some examples: peelie-wally (to rhyme with 'valley'), 'weak, pale, insipid' mixter-maxter/mixie-maxy/mixty-maxty 'a random assortment of things' (also 'clamjamfrie', meaning the same?) eeksy-peeksy 'even stevens, six and half a dozen, exactly alike' feek-fike 'needless bustle' whim-wham, 'a dessert resembling trifle' niffle-naffle/nif-niff/niffy-naff 'to waste time' niff-naff 'anything very small, a small person, fussiness'

    6. The problem you describe fits in nicely with a phenomenon that is much more widespread in Africa, where it goes under the heading of ideophones. These are words that frequently are subject to special phonotactic constraints and that frequently involve some form of reduplication. In many Niger-Congo languages they can be members of any lexical category but they will differ morphological from non-ideophonic members of that category. The phenomenon was first identified in Bantu and has been studied in some depth there. However, a sometime visiting faculty member at Penn, Yiwola Awoyale, is conducting an ongoing study of ideophones in Yoruba and has, according to my last email from him, collected a body of about 25,000 ideophones in Yoruba. There is some productivity in their formation, but he would have a lot more information on that. Herb Stahlke Ball State University

  4. A Paper on the subject of reduplication in general:

    Reduplication and the Arbitrariness of the Sign

    Terry Regier ( Department of Psychology; University of Chicago
    5848 South University Avenue; Chicago, IL 60637 USA
    Abstract: The meanings expressed by reduplication, or linguistic
    doubling, are similar across a wide array of languages. Interestingly,
    some of these shared meanings do not concern doubling,repetition, or
    plurality. This non-arbitrariness of the sign may be attributable to
    the interplay of two forces: iconicity, and conceptually-based
    semantic extension. Cross-linguistic evidence supporting this account
    is presented. More generally,this paper argues that the interaction of
    iconicity and semantic extension constitutes a potentially powerful
    source of non-arbitrariness in the mapping between sound and meaning.
    While the relationship between sound and meaning in language is
    largely arbitrary (de Saussure, 1966), there are exceptions to this
    general rule. One of these can be found in reduplication, or
    linguistic doubling. While this form expresses an array of meanings
    cross-linguistically, there is a core set of meanings which recur so
    frequently and in so many languages as to demand explanation.
    Consider for example the English words dum-dum, claptrap, nitwit, and
    riffraff. These words have two things in common: a reduplicative form,
    and an element of contempt in the meaning. One can also cite examples
    such as arbitraryshmarbitrary, based on a productive borrowing from
    Yiddish. Moving further linguistically afield we find Uzbek
    ikir-chikir (petty, trivial, idle) and Farsi chart-o-part (irrelevant
    talk, nonsense), among many others. This phenomenon is intriguing
    because the connection between contempt and doubling is not an obvious
    Other senses that recur cross-linguistically include small, scatter,
    lack of control, plurality, intensity, affection, baby, continuity,
    and completion (Moravcsik, 1978; Niepokuj, 1991; Regier, 1994). There
    is no simple abstraction over the set of meanings expressed by
    reduplication. But the set of meanings is not boundless either, and in
    fact covers only a relatively small region of semantic space
    (Moravcsik, 1978). Interestingly, the same fairly specific meaning is
    often expressed by reduplication in unrelated languages. It is
    exceedingly unlikely that this would occur either by chance, or
    through widespread borrowing. What is happening, then? And what are
    the ramifications for the doctrine of the arbitrariness of the sign?
    Iconicity and Semantic Extension
    This paper suggests that the observed regularities may result from the
    interplay of two forces. One force is iconicity or sound symbolism - a
    "direct linkage between sound and meaning" (Hinton et al., 1994). For
    example, many languages use reduplication in baby talk, that register
    of the language used when addressing small children (Ferguson, 1964;
    Haynes and Cooper, 1986). Since babies themselves reduplicate
    extensively in learning to speak (Fee and Ingram, 1982; Ferguson,
    1983; Schwartz et al., 1980), it is understandable why doubling and
    babies would come to be associated, and this association reflected in
    a variety of unrelated languages. Similarly, the widespread use of
    reduplication to express repetition and plurality is
    self-explanatory. But what about other widely attested meanings, such
    as contempt? In such cases, there is no clear mirror of the meaning in
    the doubled form.
    This is where the second force comes into play, building on the
    first. Meaning may spread, either synchronically or over the history
    of a language. Once reduplication marks one meaning in a language, it
    may come to also mark others that are closely conceptually
    related. And this process of semantic extension may then repeat
    itself, chaining out to yet other meanings (Bybee et al., 1994; Heine
    et al., 1991; Lakoff, 1987; Sweetser, 1990). For example, the
    iconically grounded notion of baby is clearly related to the notion
    small. Thus, we might expect to find reduplication expressing small in
    some languages - and we do. In its turn, small is conceptually close
    to contempt, as small things tend to be dismissible (small fry,
    peanuts). Thus, reduplication may mark contempt in some languages
    because of a trail of semantic extension from the iconically grounded
    sense of baby, through small, to contempt. Categories formed of senses
    chained together in this fashion have been termed 'radial categories'
    (Lakoff, 1987).
    Figure 1 illustrates the central theoretical claim being advanced
    here: that the interaction of these two forces can account for what
    might otherwise be a puzzling set of crosslinguistic regularities in
    the semantics of reduplication. We begin by noting that dissimilar
    senses can be iconically grounded in the same form. In this case,
    baby, repetition, and plurality are all taken to be sound-symbolically
    related to the form itself - this is shown by the dashed lines
    crossing the sound-meaning divide. Motivation has already been given
    for baby, and the other two (repetition and plurality) do not require
    much. Each of these three senses then serves as the root node for a
    tree of related concepts. Links between senses are initially posited
    on the basis of apparent conceptual relatedness, and motivation is
    then sought for each of the connections. The senses listed in the
    graph (repetition, continuation, small, baby, attenuation, contempt,
    affection, plural, intensity, completion, lack of control,
    non-uniformity, spread out or scatter) are the primary ones attested
    in the literature (Bybee et al., 1994; Moravcsik, 1978; Niepokuj,
    1991). The hypothesis is that they appear because of this interaction
    between iconic and conceptual structure.
        reduplication       SOUND
                                /    |      \
                   /-----------/     |       \---------\
                  /                  |                  \
                /                    |                    \
               |                     |                    |        MEANING
               |                     |                    |
      baby               repetition              plural
             /   \                   |                  /      \
            /     \                  |                 /        \
        small   affection        continuation     intensity ---  spread,
          /\                                         |         - scatter
         /  \----\                                   |        /    |
        /         \                                  |       /     |
    attenuation  contempt                       completion  |  lack of control
        Figure 1: The interaction of iconicity and semantic extension
    How can this hypothesis be tested? It makes two related predictions,
    both stemming from the assumption that much of the observed regularity
    is conceptually rather than iconically mediated. The first prediction
    is that we should expect to find each of the displayed inter-concept
    links implicated in the semantic extension of other linguistic
    forms. For example, in the expressions small fry and peanuts we have
    already seen that terms for small things sometimes assume a
    contemptuous or dismissive meaning. This observation supports the
    small to contempt conceptual link in Figure 1, since it is a
    linguistic manifestation of that link that has nothing whatever to do
    with reduplication per se. Such evidence strongly suggests that the
    connection is both real, and genuinely conceptual. It will be the
    burden of this paper to demonstrate that some such non-reduplicative
    motivation can be adduced for many of the individual links shown here,
    such that this first prediction is largely met. This prediction will
    be referred to as the local prediction, since it concerns individual
    links on a one-by-one basis.
    The second, or global, prediction is that we may expect to find a
    non-reduplicative linguistic form that expresses an entire subtree of
    the senses in this graph. Why should this be expected? Because if
    conceptually-mediated semantic extension produces a chain from baby
    all the way out to the apparently unrelated notion of contempt in the
    case of reduplication, this same conceptual chain should be available
    to other linguistic forms. Thus, one might expect to find, for
    example, a form associated with all senses in the circled subtree on
    the left, or in the circled subtree on the right. This global
    prediction is met as well, as we shall see. This paper will argue that
    the entire subtree of senses rooted at baby is shared with the
    diminutive (Jurafsky, 1996): cross-linguistically, diminutives express
    a range of senses other than just small, including each of the senses
    in the baby subtree. This systematic semantic overlap strongly
    suggests that these senses cohere for conceptual reasons, rather than
    reasons related to linguistic form. A similar argument will be
    advanced concerning the other circled subtree, rooted at spread,
    scatter. Each of the senses in this subtree is expressed by the
    Russian verbal prefix raz- (Regier, 1994). Again, this semantic
    sharing of an entire cluster of senses with a linguistic form that is
    not itself doubled suggests a conceptual basis for the particular set
    of senses observed. It is significant for this argument that while the
    diminutive and raz- each express a range of senses, that range is
    fairly limited in both cases (Jurafsky, 1996; Regier, 1994). Thus, the
    sharing of specific sets of senses cannot be attributed simply to
    extreme broadness of coverage of either the diminutive, or raz-.
    The central ramification of this account is simple. If there is indeed
    such a conceptual rippling-out to senses that are not themselves
    iconically grounded, this can generate a greater degree of
    non-arbitrariness in the mapping between sound and meaning than we
    would otherwise expect. Reduplication serves only as an instance of
    this more general point. The same principle could also operate
    elsewhere in language.
    This paper takes the following form. It considers each of the senses
    in Figure 1 in turn, and provides linguistic evidence for the local
    and global predictions with respect to the current sense. This
    supports the overall view that iconicity and semantic extension may
    both contribute to the observed regularities in the mapping from
    reduplicative form to meaning. Finally, the data are discussed as they
    bear on the question of the arbitrariness of the sign.
         The Senses of Reduplication
    Each of the senses listed above is treated in turn. Each entry is
    subdivided into three parts. First, evidence is given exemplifying the
    cross-linguistic use of reduplication to express the sense in
    question. Sometimes, this data will indicate that reduplication
    expresses a particular sense in the grammar of some language. However,
    coverage is also deliberately broadened to include non-productive uses
    in the lexicon, such as nitwit, knick-knack, and the like. Many of
    these express the same senses as those expressed by more properly
    grammatical reduplication. The second part of the entry lists
    non-reduplicative evidence for any links that may tie this sense to
    its predecessors in the semantic network - in support of the local
    prediction. Finally, the third part lists evidence for any overlap
    this sense may exhibit with the diminutive, or with raz-, in support
    of the global prediction.
    Baby: Reduplication: English baby, French b'eb'e, and Tamil papa
    (baby). In addition to these lexical examples, many languages use
    reduplication productively to mark baby register, that register used
    when addressing babies (Ferguson, 1964; Haynes and Cooper, 1986). This
    is exemplified by the English forms Georgie-Porgie, doggy-woggy, and
    there there. Links: The iconic grounding of this sense has already
    been motivated. Overlap: Baby is also a central sense of the
    diminutive cross-linguistically, along with the very closely related
    notions child and young. We can see this in examples such as Ojibwa
    kwezens (girl; kwe - woman), Tibetan dom-bu (bearcub; dom - bear), and
    Nez Perce 'icey' (young coyote; 'icey' - coyote).
    Affection: Reduplication: Arabic Ramr_um (the affectionate form of the
    name R_ima), English honey-bunny and bye-bye (a more intimate version
    of bye). Moravcsik (1978) also notes that this sense appears
    cross-linguistically. Links: The link from baby to affection is a
    pragmatically natural one. Evidence for this link comes from the use
    of the term baby itself as an endearment in English. Overlap: The
    diminutive is also very commonly used to mark affection. Examples
    include Russian belochka (dear little squirrel), Afrikaans
    oorgrootjies (dear great-grandparents), and English Terry (the
    affectionate form of the name Terrance).
    Small: Reduplication: Agta walawer (small creek; wer - creek), Comox
    djidjidis (little tooth; djidis - tooth), and English tidbit and
    itty-bitty. Links: The link from baby to small is motivated by the
    perceptually very salient fact that babies are small, and may well be
    the most conceptually salient class of small things. The use of
    linguistic forms meaning baby to also mean small is illustrated in
    such English sentences as Look, a baby airplane, meaning a small
    airplane. Overlap: The notion small is also a central sense of the
    diminutive, and appears widely across languages, such as Ewe kp'e-v'i
    (small stone; kp'e - stone), and Hungarian felh"ocske (small cloud;
    felh"o - cloud).
    Attenuation: Reduplication: Swahili maji-maji (somewhat wet; maji
    wet), Thai k`aw-k`aw (oldish; k`aw - old), Tagalog mahiyahiya (be a
    little ashamed; mahiya - ashamed) (Moravcsik, 1978). Links: The link
    from small to attenuation is exemplified in non-reduplicative
    expressions such as It's a little cold today, in which the word
    little, which means small, is used to attenuate the force of the
    utterance. Overlap: The diminutive is widely used to express
    attenuation, for example Hungarian nagyocska (fairly large; nagy -
    large) and Greek ksinutsikos (sourish; ksinos - sour).
    Contempt: Reduplication: Bengali bhethor-shethor (in, but that is
    insignificant; bhethor - in), Dutch mik-mak (worthless collection),
    English claptrap, hillbilly, German Pille-palle (insignificant
    things), and Yiddish layfen-shmayfen (running is beside the point;
    layfen - to run). Links: We have already seen motivation for the link
    from small to contempt in the English expressions small fry and
    peanuts. Overlap: As one might expect, given the existence of such a
    link, the diminutive is also often used to express contempt. Consider
    for example Latin Graeculus (miserable Greek), and English limey (a
    derogatory term for an Englishman, considered a diminutive by analogy
    with doggy, Jimmy, and the like).
    This concludes the set of senses that overlap with senses of the
    diminutive. Diminutive data listed here was obtained from a recent
    cross-linguistic treatment of the semantics of the diminutive
    (Jurafsky, 1996). In this work, we find 16 separate senses of the
    diminutive that are commonly found in the languages of the world. The
    fact that only 16 senses were identified suggests that the diminutive
    is relatively restricted in the range of meanings it may assume. This
    grants the fivesense overlap with reduplication that we see here its
    significance, for it means that the overlap cannot be attributed to
    broad semantic coverage on the part of the diminutive. This in turn
    suggests that the overlap may result from a network of shared
    conceptual links.
    Repetition: Reduplication: English boogie-woogie, Mongolian bayn bayn
    (often, constantly), Sundanese guguyon (to jest repeatedly; guyon - to
    jest), and Tzeltal -pikpik (to touch it lightly repeatedly; -pik - to
    touch it lightly). Links: This sense is taken to be sound-symbolically
    grounded directly in the doubled form - the repetition of the stem in
    the doubled form maps easily onto the meaning of repetition. Overlap:
    This sense is not a part of a systematic semantic overlap with any
    other linguistic form currently under study.
    Continuity: Reduplication: The use of reduplication to express
    continuity has been noticed by many researchers. Reduplication
    expresses continuative aspect in Tagalog (French, 1988) and Javanese
    (Niepokuj, 1991). In addition, the continuative sense of reduplication
    is evident in Hindi kit kit (monotonous droning on) and arguably
    English dilly-dally. Links: The use of reduplication in the sense of
    continuity is easily motivated from repetition. Lakoff (1987) has
    noted that multiplicities are often spoken of as masses, as we see in
    the sentence The guards were posted all over the hill. Here, the
    multiple guards are implicitly viewed as a mass which covers the
    hill. Such examples may occur because when a multiplicity is viewed at
    a coarser level of resolution, it will appear to be a mass. While this
    sentence concerns physical mass and multiplicity, it motivates the
    multiplicity-to-mass transformation generally. The link from
    repetition to continuity is the temporal analog of this physical
    example: we blur the individual repetitions of an action together such
    that conceptually it becomes a single continued action. Overlap: This
    sense is not a part of a systematic semantic overlap with any other
    linguistic form currently under study.
    Plurality: Reduplication: The use of reduplication to express
    plurality is widespread, appearing in Dakota, Agta (Niepokuj, 1991),
    Comox (Sapir, 1915), Papago, Samoan, and numerous other languages
    (Moravcsik, 1978). Links: This sense is taken to be sound-symbolically
    grounded directly in the doubled form - the plurality of elements in
    the form maps cleanly onto the meaning of plurality. Overlap: This
    sense is not a part of a systematic semantic overlap with any other
    linguistic form currently under study.
    Spread Out, Scatter: Reduplication: A number of languages use
    reduplication to express the notion of spread out or scatter. Examples
    are Japanese tokoro-dokoro (scattered) and Mongolian aravgarsaravgar
    (spread out). Links: The link from plurality can be motivated by
    noting that the result of scattering is a plurality of objects in a
    plurality of locations. However, this paper does not present
    non-reduplicative linguistic evidence for this link. In this case, we
    must rely rather on the intuitive plausibility of such a
    connection. Overlap: Many Russian verbs beginning with the prefix raz-
    also have a semantics related to scattering or spreading out. Examples
    are razgonjat' (to disperse), razmetat' (to scatter (s.t.); to spread
    (s.t.) out), raznosit'sja (to spread), raskidyvat' (to scatter;
    spread), rasprostranjat' (to spread, distribute), rasseivat' (to
    disperse, scatter), rassredotochivat' (to disperse), and rasstilat'
    (to spread).
    Intensity: Reduplication: Reduplication is very widely used to express
    intensity. Consider for example English a whole whole lot, very very
    good, Hindi lal-lal (very red; lal - red), and Mongolian aray ^caray
    (just barely; aray - barely) and "o^cn"o"on t"o^cn"o"on (a great
    deal). In addition to these examples, this sense appears in Dakota,
    Turkish, Dagur, Perak Malay, Tangale (Niepokuj, 1991), Sundanese, and
    Thai (Moravcsik, 1978). Links: There are at least two possible sources
    of motivation for this sense. One of these is a link from plurality,
    highlighted by the English expressions many thanks (Moravcsik, 1978)
    and a thousand pardons, in which quantity is used to express
    intensity. The other is a link from spread out, or the result of
    spreading out, namely enlargement. For example, we can see from the
    Russian expression bol'shoe spasibo (literally, big thank-you) that
    size can also be used to express intensity. Overlap: The Russian
    prefix raz- can be used to express intensity of a condition or
    feeling: razbalivat'sja (to be or become properly ill), razobidet'sja
    (to take great offense; obidet'sja - to be offended), raskaljat' (to
    make scorching hot), and rasshalit'sja (to get very playful).
    Completion: Reduplication: Many languages use reduplication to express
    completion or perfectivity, particularly in the Indo-European family
    (Moravcsik, 1978; Niepokuj, 1991), including Sanskrit, Greek, Latin,
    and Gothic. Links: A possible motivation for the use of reduplication
    in this sense stems from intensity: performing an action intensely can
    lead to completion of the overall action. Thus, working on a project
    intensely will lead directly to completion of the project, while a
    more lackadaisical approach to the same project will result in its
    remaining in a state of incompletion for a considerable time. However,
    this paper does not present non-reduplicative linguistic evidence for
    this link. In this case, we must rely rather on the intuitive
    plausibility of such a connection. Overlap: The Russian prefix raz-
    can be used as a perfective marker: rasserdit'sja is the perfective
    form of serdit'sja (to become angry). In addition, a number of Russian
    lexical entries beginning with raz- express completion: raskupat' (to
    buy up), raspivat' (to empty a bottle drinking), razgadyvat' (to
    solve; to get to the bottom of), razljubit' (to stop loving),
    razrjazhat'sja (to run down, be used up), and razygryvat' (to bring to
    a conclusion).
    Lack of Control: Reduplication: Lack of control and disorder are
    expressed by reduplication in a number of languages. Consider for
    example Danish misk-mask (mess, disorganized jumble), Dutch
    schelle-belle (overly independent young woman), English
    helter-skelter, higgledy-piggledy, pell-mell, willy-nilly, Farsi
    g. ati-pati (pell-mell), Hebrew tohu-va-bohu (chaos), and Russian
    tjap-ljap (anyhow, in a slipshod manner). Reduplication is also used
    grammatically to express lack of control in some Salish languages
    (Carlson and Thompson, 1982). Links: This sense may be motivated by
    noting that lack of control can easily cause spatial scattering or
    spreading out. This link is highlighted in the English word
    scatterbrained: the notion of scattering here is used to express
    mental lack of control or absent-mindedness. We see this link again in
    the colloquial American English sentence She's a very together person,
    which uses a word whose central sense is the opposite of dispersion or
    distension to express poise and self-control, i.e. the opposite of
    lack of control. Overlap: A number of Russian verbs beginning with the
    raz- verbal prefix have meanings that concern lack of
    control. Consider for example: razbushevat'sja (to rage, get violent,
    start lashing out), razvolnovat' (to upset (s.o.)), razlazhivat'sja
    (to go wrong), razmechtat'sja (to be lost in dreams), razozlit'sja (to
    get furious), razrydat'sja (to burst into sobs), raskapriznichat'sja
    (to become very naughty, act up), and rasserdit' (to annoy, to make
    Non-Uniformity: Reduplication: Non-uniformity is sometimes expressed
    by reduplication. Consider colloquial Levantine Arabic nus.-nus. (half
    and half, a mixture of the two; nus. - half), English hodge-podge,
    bric-a-brac, mish-mash, Neoaramaic rangerange (in several different
    colors; range - color), and Tamil ithe-athe (this and that). Links:
    This sense may be linked to scatter, in that non-uniformity of state
    may be metaphorically viewed as non-uniformity of location - which is
    the natural result of scattering. We can see this in such colloquial
    expressions as This guy's grades are all over the place, meaning that
    they are not at all uniform. There is a reliance here on a very
    general and widespread metaphor that views abstract states as
    locations (Lakoff, 1987). This sense may also be iconically grounded
    to some extent. It is interesting to note that the sense of
    non-uniformity is often expressed by nonuniformity of form. Many
    English reduplicatives with this sense seem to exhibit variance in
    form between the first and second instantiations of the stem, such as
    mish-mash, hodgepodge, knick-knacks, bric-a-brac, zig-zag. This is not
    the case in all languages however; Arabic nus. -nus. and Neoaramaic
    range-range are counterexamples. Overlap: The Russian prefix raz- can
    be seen to also express non-uniformity in examples such as: razdumat'
    (to change one's mind), razlichat'sja (to differ, be distinguished),
    raznit'sja (to differ), raznoobrazit' (to diversify), and
    raskhodit'sja (to disagree, differ).
    This concludes the subset of senses that overlap with senses of
    raz-. Earlier work had identified seven different senses of this
    verbal prefix: scatter, lack of control, non-uniformity, intensity,
    completion, split, and analysis (Regier, 1994). Of these, the first
    five appear to be shared with reduplication. As in the case of the
    diminutive, this semantic overlap suggests a conceptual basis for this
    subset of senses, as the senses cohere across linguistic forms.
    This paper has argued that the regularities in the semantics of
    reduplication stem from an interaction of iconicity and semantic
    extension. On this account, some senses are iconically grounded, and
    others derive from these through a conceptual spreading-out. This
    account predicts that we will be able to find non-reduplicative
    manifestations of each of the links in Figure 1 (the local
    prediction), since if the links truly exist and are conceptual in
    nature, they should be available, one by one, to other linguistic
    forms. As we have seen, such linguistic evidence has been cited for
    links to all but two of the senses under consideration. Thus, this
    local, link-based, prediction is in large measure substantiated. The
    account also predicts, more globally, that entire clusters of senses
    will be marked by the same non-reduplicative linguistic form. As we
    have seen, there is a subset of five senses that are shared with the
    diminutive, and another subset of five senses that are shared with the
    Russian verbal prefix raz-. This sharing of entire clusters of senses
    across linguistic forms strongly suggests a conceptual, rather than
    purely iconic, basis for the phenomenon.
    This argument is not without its weaknesses. There are three that
    demand immediate treatment, and future work will be directed at
    them. The first weakness is that evidence from more languages is
    clearly required, to further determine just how widely shared the
    listed senses are. The current paper has relied heavily for its data
    on earlier published studies of reduplication. On that basis it has
    simply demonstrated that each sense is fairly widely shared - this is
    perhaps a reasonable beginning, but it is only that. The second
    weakness is that there are some senses that have not yet been
    incorporated into the network of senses. An apparently fairly
    widespread example is the notion game: consider English ping-pong,
    pall-mall, tic-tac-toe, Basque (and now Spanish and English) jai-alai,
    Hindi holi-koli, and ^se^s-be^s, the name for backgammon in large
    regions of Turkish- and Persian-influenced southwestern Asia. One
    might imagine a link to the baby cluster on the basis of playfulness,
    or since the games tend to pit one player against another, a link to
    plurality. But given the current lack of either local (link-based) or
    global (cluster-based) support for this sense, such connections must
    remain speculative at this point. The third weakness is that it is
    possible, even probable, that some of the links in the proposed graph
    are incorrect. This is particularly true in light of the fact that
    non-reduplicative linguistic evidence could not be found for two of
    these links. A more satisfying approach to this issue would be to
    provide further independent motivation for the links, by basing the
    links in the graph on non-linguistic evidence, such as conceptual
    relatedness judgments collected from subjects.
    Returning to the central claim, however, even given such possible
    flaws, it still seems quite probable that these regularities in the
    sound-meaning mapping can be attributed to an interaction between
    iconicity and semantic extension. For even if our cross-linguistic
    coverage is somewhat limited, and not all senses have been worked into
    the semantic network, and some of the links have been incorrectly
    ascertained, there remain two critical facts. First, across a range of
    languages, the same general form is used to express the same fairly
    tightly circumscribed set of meanings. Some of these meanings are
    clearly iconically grounded in the form itself, and interestingly,
    some of them appear not to be. Second, subsets of these senses are
    also expressed by other linguistic forms, suggesting that the senses
    travel together for reasons of conceptual relatedness. Thus, we have a
    clear exception to the arbitrariness of the sign, and a strong
    suggestion that conceptually mediated semantic extension may play a
    role in this exception.
    Ultimately, reduplication is simply a case study; it establishes a
    principle. It is an instance in which iconicity and semantic extension
    seem to have conspired to violate de Saussure's doctrine in a less
    than transparent fashion. There may be other instances, iconically
    rooted at very different spots in semantic space. It remains to be
    determined to what extent such other instances exist. But the larger
    ramifications of this study lie precisely in the possibility of such
    other instances. For if these two forces, iconicity and semantic
    extension, can yield non-trivial regularities of the sort we have seen
    in the case of reduplication, they should in principle also be able to
    do so elsewhere in language. In this manner, they constitute a
    potentially powerful source of non-arbitrariness in the mapping
    between sound and meaning.
    Acknowledgements: Thanks to Kathryn Campbell-Kibler, who has assisted
    in the research project reported here. Thanks also to the many many
    other people who have contributed to this work, through enjoyable and
    helpful conversations.
    Reduplicative data: Arabic (Ferguson, 1956; Wehr, 1961); Bengali
    (Subutai Ahmad); Comox (Sapir, 1915); Dutch (Dirk Geeraerts); Farsi
    (Nikki Mirghafori); Japanese (Matsuda, 1974); Mongolian (Bosson,
    1964); Neoaramaic (Krotkoff, 1982); Russian (Taube et al., 1987);
    Sundanese (Robins, 1959); Tamil (Srini Narayanan); Uzbek (Waterson,
    1980). All other reduplicative data (Bybee et al., 1994; Moravcsik,
    1978; Niepokuj, 1991). Diminutive data (Jurafsky, 1996).
    Bosson, J. E. (1964). Modern Mongolian - A Primer and Reader. Indiana
    University Publications, Bloomington, Indiana. Uralic and Altaic
    Series, volume 38.
    Bybee, J., Perkins, R., and Pagliuca, W. (1994). The Evo lution of
    Grammar: Tense, Aspect, and Modality in the Languages of the
    World. University of Chicago Press, Chicago.
    Carlson, B. F. and Thompson, L. C. (1982). Out of control in two
    (maybe more) Salish languages. Anthropological Linguistics,
    de Saussure, F. (1966). Course in General Linguistics.  McGraw-Hill,
    New York. Translation and notes by Wade Baskin.
    Fee, J. and Ingram, D. (1982). Reduplication as a strategy of
    phonological development. Journal of Child Language, 9:41-54.
    Ferguson, C. (1956). Arabic baby talk. In Halle, M., editor, For Roman Jakobson: Essays on the Occasion of his 60th Birthday. Mouton, The Hague. 
    Ferguson, C. (1964). Baby talk in six languages. American
    Anthropologist, 66(6):103-114.
    Ferguson, C. (1983). Reduplication in child phonology. Journal of
    Child Language, 10:239-243.
    French, K. M. (1988). Insights into Tagalog: Reduplication,
    Infixation, and Stress from Nonlinear Phonology. Summer Institute for
    Linguistics and University of Texas at Arlington, Arlington, Texas.
    Haynes, L. M. and Cooper, R. L. (1986). A note on Ferguson's proposed
    baby-talk universals. In Fishman, J., Tabouret-Keller, A., Clyne, M.,
    Krishnamurti, B., and Abdulaziz, M., editors, The Fergusonian Impact:
    In Honor of Charles A. Ferguson, Volume 1: From Phonology to Society,
    pages 127-134. Mouton de Gruyter, Berlin.
    Heine, B., Claudi, U., and H"unnemeyer, F. (1991). Grammaticalization
    : A Conceptual Framework. University of Chicago Press, Chicago.
    Hinton, L., Nichols, J., and Ohala, J. J. (1994). Sound
    Symbolism. Cambridge University Press, Cambridge. 
    Jurafsky, D. (1996). Universal tendencies in the semantics of the
    diminutive. Language, 72:533-578. 
    Krotkoff, G. (1982). A Neo-Aramaic Dialect of Kurdistan: Texts,
    Grammar, and Vocabulary. American Oriental Society, New Haven,
    Lakoff, G. (1987). Women, Fire, and Dangerous Things: What Categories
    Reveal about the Mind. University of Chicago Press, Chicago. 
    Matsuda, K. (1974). Kenkyusha's Japanese-English
    Dictionary. Kenkyusha, Tokyo. 
    Moravcsik, E. (1978). Reduplicative constructions. In Greenberg,
    J. H., editor, Universals of Human Language, Volume 3: Word Structure,
    pages 297-334, Stanford, California. Stanford University Press.
    Niepokuj, M. K. (1991). The historical development of reduplication,
    with special reference to Indo-European. Ph.D. dissertation,
    University of California, Berkeley.
    Regier, T. (1994). A preliminary study of the semantics of
    reduplication. Technical Report TR-94-019, International Computer
    Science Institute, Berkeley, California.
    Robins, R. H. (1959). Nominal and verbal derivation in Sundanese.
    Lingua, 8:337-369.
    Sapir, E. (1915). Noun Reduplication in Comox, a Salish Language of
    Vancouver Island. Government Printing Bureau, Ottowa, Canada. Canada
    Department of Mines, Geological Survey, Anthropological Series, number
    6, memoir 63.
    Schwartz, R., Leonard, L., Wilcox, M. J., and Folger, M. K.  (1980).
    Again and again: reduplication in child phonology. Journal of Child
    Language, 7:75-87.
    Sweetser, E. (1990). From Etymology to Pragmatics: Metaphorical and
    Cultural Aspects of Semantic Structure. Cambridge University Press,
    Cambridge, England.
    Taube, A. M., Litvinova, I. W., Miller, A. D., and Daglish, R.  C.
    (1987). Russian-English Dictionary. Russky Yazyk Publishers, Moscow.
    Waterson, N. (1980). Uzbek-English Dictionary. Oxford University
    Press, Oxford. 
    Wehr, H. (1961). A Dictionary of Modern Written Arabic.  Otto
    Harrassowitz; Wiesbaden.
  5. Useful Comments and Discussion, last modified 10/10/99.