Morphological productivity is a widely discussed topic in English word formation. What it means for a word formation process to be morphologically productive is controversial and various views exist concerning the definition of morphological productivity. The present essay aims to shed some light on the matter. A number of definitions of morphological productivity will be presented and discussed and especially the difference between productivity and creativity, and if there is such a difference at all, will be highlighted. Moreover, the question of whether productivity can be measured will be addressed, and different types of productivity measure will be compared and discussed. Is there such a thing as a good productivity measure, and how can the presented methods be improved? Rather than giving a superficial overview of many aspects of morphological productivity, this essay will focus on only a few, but nevertheless very important viewpoints.
2 What is morphological productivity?
The question of what morphological productivity is cannot be answered explicitly as there exist many different viewpoints in the literature. While Bauer (1983) states that a word-formation process is productive “if it can be used synchronically in the production of new forms” (18), Plag (1999) argues that productivity is “the property of an affix to be used to coin new complex words” (44). Now, according to Plag productivity only accounts for inflectional and derivational processes, but Bauer does not share this opinion. In fact, he points out that “there are some word-formation processes which are non-affixal, but which may nevertheless be productive” (Bauer 2001:12), and he lists a number of ablaut-motivated compounds like chitchat, dilly dally and fiddle faddle to support his argument. From his point of view, productivity does not refer solely to affixes but rather to morphological processes (cf. Bauer 2001:13). There are several other definitions of morphological productivity, like that of Spencer (1991) who regards a rule as productive if it is “regularly and actively used in the creation of totally new words” (49). Spencer’s definition resembles that of Bauer in that he focuses on the synchronic creation of new words, but also differs in the respect that Bauer highlights the potential formation of new words while Spencer points out that productivity is an active process. Therefore, according to Spencer, morphological productivity is not concerned with the possibility of a word-formation process to form new words but there has to be actual evidence for this. A further definition was proposed by Schultink (1961) on which the more modern definitions are based. He presented a more sophisticated account of morphological productivity:
Productivity as a morphological phenomenon is the possibility which language users have to form an in principle uncountable number of new words unintentionally, by means of a morphological process which is the basis of the form-meaning correspondence of some words they know. (qtd. in Plag 1999:13)
Schultink’s point of view is opposed to that of morphological creativity. On the basis of learned rules, speakers form new words. The application of these rules, for example that the suffix -ness can be attached to adjectives to form nouns, is crucial for the process of morphological productivity. In the following sub-section I will give an overview of the difference between creative and productive word-formation processes.
2.1 Productivity vs. creativity
Schultink’s ‘unintentional’ or also unconscious nature of use of productive rules is typical: when a rule is very productive, neologisms on its basis will hardly be noticed as they look so familiar and not innovative. For example, the suffix -ness can be used to form nouns from a large number of adjectives, and speakers have internalised this rule to such an extent that neologisms with -ness are not particularly striking and are more often formed than neologism with the suffix -ese (cf. Haspelmath 2002: 101). On the other hand, “creative neologisms are always intentional formations that follow an unproductive pattern” (Haspelmath 2002: 100).
This view was originally derived from Lyons (1977) who put forward the notion of rule-governed and non-rule governed word-formation. According to Lyons, productivity is a defining property of language, allowing a native speaker to produce an infinitely large number of sentences, to be accounted for by the rules of grammar (cf. Lyons 1977: 549), whereas creativity is “the native speaker’s ability to extend the language system in a motivated, but unpredictable (non-rule governed) way” (Bauer 2001: 63). Bauer gives an example of this difference on the basis of the word headhunter. If it is seen as a formation invented to designate a member of a tribe which keeps the heads of its victims, the word-formation process is productive because it is governed by syntactic rules. However, if it is seen as a metaphorical expression, referring to one who recruits executives for a large corporation, the word-formation process is regarded as creative because the meaning of the word is semantically opaque and if one does not know the meaning of the second sense of headhunter, it is not possible to derive it from the word alone (cf. Bauer 2001: 63). Hence, analysability and semantic transparency seem to be pre-requisites for morphological productivity, but they are by no means sufficient conditions, e.g. the suffix -ess is analysable and transparent but not productive. In addition to that, creative processes can be analysable and transparent as well. This becomes clear when we look at analogical formations such as trialogue, which is derived from dialogue, or the German word Hausmann, which is derived from Hausfrau (cf. Haspelmath 2002: 102).
Another problem is that it is difficult to distinguish clearly between (levels of) consciousness and intentionality in individual cases. It might be the case that there is a mix of intentional and unintentional and unconscious word-formation processes. Haspelmath notes that because of the fact that we do not know what a speaker intends and thinks when he forms new words, it is impossible to say that productive processes are always unconscious (Haspelmath 2002: 101). Haspelmath illustrates this with the example of the word mentalese, which was coined by a philosopher in the mid 20th century. The formation of mentalese was creative, but the question arises why he did not use the word thoughtese or mindese which would have been equally acceptable from a semantic point of view. The answer to this question is that the suffix -ese prefers to follow bases with a strong-weak stress pattern (like in mòtherése or Jàpanése) and since thought and mind are monosyllabic, they do not conform to this pattern (cf. ibid.). It is unlikely that the philosopher considered this as he coined mentalese, but he might have internalised this rule and made his choice unconsciously.
Thus, it cannot be confirmed that productivity and creativity are two independent processes which are mutually exclusive, but they rather seem to influence each other and both processes can contribute to the formation of new words at the same time. This is emphasized by the fact that creative processes can turn into productive processes; for instance, the suffix -scape was first used as an analogy and later became productive (cf. Claridge 2008). This shows that there is no clear boundary between productivity and creativity and the question remains where creativity ends and productivity starts, or if a distinction between these processes can be made at all.
3. Measuring productivity (?)
In connection with the above mentioned problem that it is problematic to distinguish between productivity and creativity, it is plausible that productivity is not a process which is either there or not but it rather seems to be the case that productivity can be gradually measured on a scale. Thus, we cannot explicitly say that a word-formation process is productive or unproductive, but it might be highly productive or less productive. But is it actually possible to measure how productive a given word-formation process is? This question will be discussed on the basis of various measures which have been proposed in the past.
Bolinger (1948) suggested that productivity is “the statistical readiness with which an element enters into new combinations” (qtd. in Plag 2003: 52). This implies a quantitative notion of productivity and means that information about the type frequency of a certain process and about the number of new words which are formed by this process are necessary. Type frequency is actually the most frequent type of measure, but it is at the same time highly disputed (cf. Plag 2003: 52). It measures the number of different words formed according to a certain pattern, for example, how many different words are created with the suffix -ness. This can be done using a good dictionary, e.g. the Oxford English Dictionary (OED). The problem with this is, however, that this will not tell us anything about the synchronic use of the suffix -ness, and we only learn how productive the suffix was in the past (cf. Bauer 2001: 144). It might be the case that the process made its way into the mental lexicon in such a way that speakers do not regard a word-formation which was once new as unusual anymore, and therefore they do not use this process to form new words with it. Plag mentions the suffix -ment, which was employed for the coinage of many new words in the past and is still very frequent, but no longer used in the formation of new words (cf. Plag 2003: 52). Hence, if a certain type is frequent in the dictionary, it does not mean that it is productive since productivity is a phenomenon of the synchronic use of language.
Another method is to count the number of neologisms that arose over a certain period of time. Again, the OED is a suitable source for this as it shows, for instance, how many new words entered the lexicon in the 20th century with the suffix -scape. This way it can be seen whether a given affix was only productive in the past or still is. However, sometimes a pattern is so productive that neologisms are likely to be overlooked (cf. Plag 2003: 53). Moreover, we have to rely on the lexicographers who composed the OED and there may be a large number of neologisms they have not seen. It is often the case that neologisms occur only once and the OED does not attest every word which has ever been written and there are certainly a large number of productive nonce-formations which never made their way into the lexicon. Moreover, neologisms which occur in spoken language should not be underestimated. The OED might be a good starting point but other sources like corpora should be taken into consideration as well.
When measuring productivity with the help of corpora, the relation of the number of words formed by the relevant process which occur only once in the corpus (hapax legomena / hapaxes) and the total number of lexemes formed with that same process in the corpus can be examined. This approach was put forward by Baayen and Lieber (1991) who argue that “given a suitable text corpus the productivity of a morphological process is the quotient of the number of hapax legomena n1 with a given affix and the total number of tokens N of all words with that affix” (Plag 2004: 9), or in mathematical terms:
In other words, the more hapaxes there are in the corpus, the higher is the productivity rate, and the more words with higher frequencies there are, the lower is the productivity (cf. Plag 2003: 57). Thus, productivity is “characterized by large numbers of low frequency words and small numbers of high frequency words” (Plag 2004: 9). The following table with productivity rates for six suffixes, calculated from the written part of the British National Corpus (BNC), illustrates this:
(Plag 2003: 57)
Compared to the relatively small number of tokens, the affix -wise has a high number of hapaxes and therefore a high productivity, while the affix -able is least productive because with regard to the number of tokens there are comparatively few hapaxes in the BNC. This sounds reasonable and it is indeed “among the hapax legomena that the greatest number of neologisms appear” (Plag 2003: 55), which was empirically tested by Plag. However, when following Baayen and Lieber’s approach which is based on the assumption that hapaxes correlate with neologisms, it has to be considered that in fact not all hapaxes are productive formations. What matters is the size of the corpus. In a relatively small corpus, a large number of words will be hapaxes and the majority of these hapaxes will not be the result of productive word-formation processes but well known words of the lexicon. In a corpus that is sufficiently large, like the BNC, the probability that hapaxes are unknown words and, moreover, neologisms is much higher (cf. Plag 2003: 57), but for a small corpus the calculated productivity rates would not be representative of the actual language in use.
A further problem is that Baayen and Lieber ignore type frequency. Does it mean that as soon as an item occurs more frequently in a corpus it is not the result of a productive word-formation process anymore? This is contrary to the type frequency approach which, although highly disputable, should at least be taken into consideration. When we look at the above table, we can see that the suffix -ness is attached to 2466 different bases, but the suffix -wise is only attached to 183 different types of bases. Nevertheless, -wise is supposed to be more productive because it has more hapaxes in relation to the number of tokens. Despite this, it might be the case that a large number of the remaining types occurs only few times. Types that occur only two or three times may still be productive word-formations. In fact, a query on the written part of the BNC revealed that the suffix -ness is attached to 426 types of bases which have only two tokens! Words which fall into this category are for example japaneseness, silveriness and windiness. They cannot be found in the dictionary and are clearly the result of a productive word-formation process. In contrast, the suffix -wise is only attached to 20 different types of bases which occur two times in the BNC. Baayen and Lieber do not take these cases into consideration. Does it mean that words which occur more than one time do not fall into the scope of productivity anymore? This assumption is fairly myopic! In order to achieve a reliable productivity measure, the proportion of both types and tokens should be taken into account, which might be achieved with the use of weighted interpolation. For example, a higher weight is assigned to hapaxes, a lower weight to word-formations which occur two times and again a lower weight to word-formations which occur three times, and so on. The sum of these weights which are multiplied by the number of tokens are then divided by the sum of tokens, which might give us a more reliable measure of productivity:
where Î» is a different weight set for each number of tokens, n is the number of tokens and N the number of total tokens of the given word-formation process.
The question which remains to be answered is where to set the threshold of productivity, i.e. how many tokens can a word have without becoming unproductive and thus, word-formation processes up to which token frequency should be included in the formula? Moreover, how should these weights be set? Answering these questions goes beyond the scope of this essay, but might be interesting for further research in the field of morphological productivity.
As has been shown, the notion of morphological productivity is difficult to pinpoint. Many different viewpoints about the definition of productive word-formation processes exist throughout the literature. It already starts with the types of morphemes which are involved in word-formation and there are differing opinions about whether only formations with affixes are productive or if compounds can be included as well. Furthermore, there is controversy concerning the difference between productivity and creativity and whether productive processes are always unintentional and rule-governed while creative processes are intentional and non-rule governed. However, in this essay it was concluded that these processes are not independent but may influence each other to a certain degree.
Another important question, which still needs to be answered, is whether productivity can be measured and if yes, which is the appropriate method to do so. Various methods have been presented which all have their advantages and shortcomings. While many researches focus on type frequency, this method is highly controversial and although alternative measures have been proposed they do not seem to capture the notion of productivity appropriately. The problem is that there are so many different viewpoints about what productivity actually means that it is difficult to arrive at a suitable measure. We can only measure something on the basis of a proper definition and if there are various definitions there are various corresponding measures, but it is impossible to determine which one is the ultimate. An extension to Baayen and Lieber’s formula has been suggested as a starting point for further research, but there still remain open questions as to how it can be applied properly.