Frequency linguistics — 2

by pieterseuren

My last blog, on John Taylor’s frequency linguistics, provoked a few reactions. One was from my good friend Camiel Hamans, who, in an email of April 8th, raises the question of the relation between frequency and productivity of morphological rules, especially in the context of morphologically articulated new products, which become accepted in a speech community as new items in their own right, usually with specialized meanings. Take compounds like tennis elbow, where the semantic relation between the two component parts is entirely different from what it is in, e.g., tennis racket. All they have in common is the fact that they restrict the class of objects denoted by the second element (the ‘head’) by means of the first element: a tennis elbow is some kind of elbow, and a tennis racket is some kind of racket. The relation between the two is only very partially defined by the semantics of the language and depends mostly on world knowledge: all the first element of the compound does is serve as a suggestive prompt or memory aid for the hearer to get the right conventionalized meaning. Such compounds thus count as lexical items in their own right. Other examples are lexical neologisms such as cinemascope, horoscope or microscope, where the -scope element suggests something to do with seeing and the first element is vaguely connected with what it meant in ancient Greek.

Hamans’s own specialism is the area in morphology characterized by creative combinations and/or abbreviations of existing forms. Abbreviations (‘clippings’) are forms like  blog for weblog, or French pneu for pneumatique (‘tyre’), or info for information, or plane for air(o)plane. ‘Blending’ consists in the putting together of significant parts of existing words, often combined with re-analysis, as in docutainment, formed from documentary and entertainment, or cheeseburger, from cheese plus the -burger of hamburger, re-analyzed as ham + burger, or infopreneur, formed on the pattern of entrepreneur (yet we don’t find e.g. mortipreneur, probably because the word undertaker already exists). Or the French franglais, from français plus anglais, used for heavily anglicized French, or Nazi German Gestapo from geheime Staatspolizei (‘secret police’). Such processes are of all times, but they have become much more frequent over the past one hundred years.

In all such cases, the acceptance or rejection by a speech community depends to a large extent on frequency, says my friend Camiel Hamans. True, but not only on frequency. Salience also seems to be a necessary condition, together with public favour, even to the point where a single but highly salient occurrence may give rise to a new word, idiom or expression. Thus, when Winston Churchill, in a speech held on 5 March 1946 at Westminster College, Fulton, Missouri, pronounced the sentence From Stettin in the Baltic to Trieste in the Adriatic an iron curtain has descended across the continent, the expression iron curtain, which had been known before but never gained currency, became part of the English language overnight, giving rise to variants such as bamboo curtain (surrounding the People’s Republic of China) or ice curtain (between the Soviet Union and the US, i.e. the Bering Sea), not because it was frequent (it had not been) but because it was salient. Its frequency came later, as a result of its salience and favourable reception. In general, frequency is always the result of something else causing it, whereby it may itself again become the cause of further effects.

Crucially, the acceptance of such neologisms (in a wide sense) depends on public favour. In the Netherlands, a vigorous advertizing campaign has been afoot for some time, stimulated and paid for by the Dutch hearing aid business world, to force the word audicien down people’s throats, in analogy to the existing word opticien for an ophthalmic optician (retailer of spectacles). Yet the public refuses to use the word and the campaign has so far been a notable failure, probably because it is feared that hearing aid retailers are assuming too much status, which will lead to higher prices.

It follows that for such cases the notion of morphological productivity must be severely restricted, just as any morphologically articulated neologism in the lexicon is always a form of restricted regularity. On the one hand, there are rule-governed morphological and syntactic restrictions, such as the use, in many languages, of the suffix -o for new clippings, as in the (Australian) English garbo from garbage man (‘refuse collector’). At the syntactic level, the rule-boundedness of neologisms is nicely exemplified by the fact that the Dutch lexicalized form leren kennen (literally ‘learn-to-know’ but actually meaning ‘get to know, make the acquaintance of’), which corresponds semantically and partly also syntactically to the German lexicalized kennen lernen, differs from the German form in that the order of the component verbs is reversed: Dutch has ‘learn-know’, while German has ‘know-learn’. This difference is not arbitrary but a consequence of the general syntactic fact that Dutch verb-clusters are mainly right-branching but German verb-clusters mainly left-branching. So, on the one hand, there are rules. On the other hand, however, the rules at issue can be successfully applied in the formation of conventionalized composite expressions only if the public agrees. This ‘will of the people’ is what I call ‘settling’ in the first chapter of my forthcoming book From Whorf to Montague (which will be out in September, OUP assures me).

I don’t think Hamans’s observation invalidates the point of view I expressed in my previous posting, since what Hamans refers to are all cases of language change, as they add new elements to the language, and what I said, in my previous posting, is that, in my view, language use, change and acquisition may be subject to frequency factors, but not language competence, to the extent that it is stable and internally uniform (one specific, sociolinguistically well-defined register). Frequency is not a parameter in grammar, is what I said, and say.

But apart from that, what the precise role of frequency is in language use, change and acquisition and how it interacts with other factors in these areas, is an interesting question, which, to my knowledge, has not been investigated to any depth, despite the very frequent appeals to ‘frequency’ in linguistic studies of all kinds. So, as I was resting after my cataract operation (it all went very well, thank you), I developed, in a still tentative way, two causal models, one for language change and one for language acquisition, where frequency is assigned a specific role in conjunction with other factors. I do not claim that this is the answer to the question at issue. I may be entirely, and probably am partially, wrong about this, but I hope it can be the starting point of a fruitful discussion, leading to what I called a reduction of the frequency craze to rational proportions.

To facilitate understanding, I draw a parallel with fire, which is caused by a heat source affecting a combustible mass. The accumulated effect of the heat source leads to a steady increase of the temperature. At a given moment, this process reaches a critical threshold resulting in a flame and thus fire, staying at a constant temperature ceiling. This fire has multiple effects, among which the destruction of the combustible mass, but also the production of more heat when the heat is partly fed back into the process. The fire is thus increased till all the combustible mass has been destroyed—unless measures are taken to make the mass less combustible or no longer combustible at all, for example by dousing it with water. I represent this recursive process by means of a spiral model, as in the figure below.


In like manner we can envisage a frequency spiral in a community of speakers for the use of lexical items, idioms, construction types, and all the other elements that are part of a given language L. This spiral is fed by a variety of frequency sources. Thus, an expression or expression type may be used frequently because it is semantically related with frequently occurring situations or objects (‘facts of life’), or because it substitutes for alternative expressions or expression types on grounds of one salient use that has caught on, as in the case of iron curtain. If it is lucky, it will reach a threshold value at which it acquires the relatively stable status of being part of L—that is, of being ‘settled’ in the technical sense introduced by me.

‘Settling’ is thus comparable to the status of fire in fig. 1. Trivially, a well-settled language system L has multiple effects on and in the actual use of L, but in addition there will be a feedback effect on the frequency source, resulting not in destruction, as in the case of fire, but in the stable settling of the newly added element of L in the community of speakers. Thus, the use of the definite article in English as the default means of selecting a specific reference object came about historically as a settled weakening of the non-default demonstrative determiner this/that/these/those. The resulting increase in frequency has meanwhile led to a very solid and stable entrenchment of the definite article in English noun phrases that make a reference to a specific entity or set of entities. Additional consequences of the frequency spiral, reflected in the language as the result of settling, are loss of salience, semantic bleaching and conventionalization.

As with water poured on fire, the settling can be undone if a frequency source falls away, as when certain situations no longer occur or certain objects are no longer used. Thus, the once current word astrolabe has virtually disappeared from English—it does not occur in the corpus-based Collins Cobuild English Language Dictionary of 1987, but it does occur in Webster’s—simply because astrolabes are no longer used (if you don’t know what an astrolabe is, go to google). But loss of frequency can also be caused by widespread disfavour, which can be based on anything whatsoever, given the fickleness of popular favour or disfavour. Disfavour acts on language as water does on fire. This analysis gives us fig. 2, which analyzes language change in a way that runs parallel to fig. 1 above.


The model is likewise applicable to language acquisition. Here, the medium is the learner’s mind + motor control system and the input to what is now the competence spiral consists of acoustic sensations interpreted as semantically significant speech sounds. In this case, competence is clearly enhanced by  frequency—though, as I said in my previous blog, frequency can only have an effect if competence is already sufficiently developed for the child to recognize frequency. But frequency is not the only source of increasing competence. Salience is another important factor. And sometimes the child acts against frequency and prefers to apply a rule in cases of irregular formations.

Thus, last December I was on the phone from Brazil with my nine-year old granddaughter (I am a totally besotted grandfather), who told me, in Dutch, that the streets were all frozen over. Now, instead of using the correct word bevroren (‘frozen’), which is the settled but irregular form, she hesitated a short moment and said bevriesd (‘freezed’), which is the regular but incorrect past participle of the verb bevriezen (‘freeze’). She must have heard and understood the correct word bevroren many times, especially during those days, yet she decided that, when talking to her grandfather, she should behave properly and apply the rule. This was very nice of her, especially because she thereby demonstrated that children, in acquiring their language, are often faced with the problem of having to choose between a rule and a, possibly often heard, violation of the rule. If you want to say, as our friend John Taylor, as one of many, does (see previous blog), that in language acquisition all depends on frequency and rules play no role, you are up against this kind of data, which is fairly common, as colleagues assure me.

In fig. 3 you see how competence is presented as arising from a competence spiral fed by  a variety of factors, including, besides other factors, frequency. The enhanced competence, in its turn, enhances competence, recursively, till a relatively stable level of active competence is achieved—though the height and kind of such ‘stable’ levels vary a great deal from person to person according to life history and social class, and, as is well known, the level is never completely stable as one goes on acquiring new elements of one’s language till the end of one’s life, albeit in a less dramatic way.


If I am at all right in the matter of these spiral models, it is clear that the role of frequency, though real, is not only restricted but also different in language change from what it is in language acquisition. Those who agree that this is so, face the challenge of specifying in what precise ways it is different. The spiral models shown above are one way of getting closer to an answer.