This article is about morphology, that part of grammar that deals with how new words are formed from other words and how they are inflected. Thus, the English words boy-hood, boy-ish, boy-friend are formed from the noun boy and boy is inflected in the plural in the form boy-s. All of these four English morphological patterns are productive. Now, morphological productivity is often considered to be present or absent in a simple two-way distinction. English plural formation in girl-s from the singular girl or in boy-s from the singular boy is productive, whereas plural formation in English ox-en, child-r-en, brethr-en, and feet from the singulars ox, child, brother, foot is not. Productive morphological patterns are thus considered to be those that freely apply to new words, as in our example of English plural formation, where any new English noun has an -s plural, as in laser, Pl. laser-s, etc. In contrast, the unproductive plural pattern of ox-en, etc, never extends to a new English noun. Let us call such productive morphological patterns productive rules, as is done in many linguistic and psycholinguistic models.

Such productive morphological rules apply freely, i.e. extend in a general and systematic way and are often coined with a very low degree of language awareness, i.e. without intentional, conscious awareness of the morphological pattern.

Most of the time, however, or at least very often, competing morphological patterns cannot be identified so easily as either productive or unproductive. Take, for example, the following, not exhaustive list of nine common ways of forming the plural of a German noun:

The last three patterns are evidently unproductive, but the others clearly have various degrees of productivity.

Now, some of our research questions are: what makes a morphological pattern productive? What role does productivity play in language processing, in language acquisition, as well as in language impairment, and in writers’ often audacious coining of new words by poetic license? (These last two questions will not be answered here for lack of space.)Reference Dressler, Stark, Pons, Kiefer, Kiss and Mészáros^1–Reference Dressler and Panagl³

Before tackling these questions of morphological productivity, however, let us briefly look in a transdisciplinary spirit at the notion of productivity in fields other than linguistics. The creator of a scientific concept of productivity is the famous eighteenth-century Scottish economist Adam Smith. In his economic theory, Adam SmithReference Hwaletz⁴ insisted on the following three factors of productivity in economics:

1. (1) skill of the worker within division of labour,

2. (2) economy through saving of time,

3. (3) mechanisation.

What was added later on in economics is the developmental aspect of the economic growth rate.

We will see later that these criteria also apply to morphological and linguistic productivity in general, at least in a metaphoric way.

In biology and agriculture, productivity means yield in biomass, i.e. something similar to fertility. In the history of arts, of literature and music, productivity refers to human creativity, i.e. to the production of new artistic works in a given time. Even here quantity counts more than quality. Note that in all these other fields, economics, biology and the arts, productivity is a gradual concept. Thus, the dichotomous view of linguistic and morphological productivity as ‘all or nothing’ becomes suspect. Still, one must acknowledge that there exists much less argumentation for the gradualness of productivity with phonological and syntactic rules than with morphological rules.

Recent years have seen a flurry of linguistic publications on morphological productivity,Reference Bauer^5–Reference Rainer, Štekauer and Lieber¹¹ which is a sign of the important role that concepts of productivity play in the area of linguistic morphology and among cognitive concerns in the field. Indeed, notions of productivity are relevant for basic questions of the mental lexicon, i.e. for the representation and processing of words and morphological patterns. If a psycholinguistic model assumes a symbolic format of rules,Reference Pinker^12–Reference Clahsen, Sonnenstuhl, Blevins, Baayen and Schreuder¹⁴ then the productivity of a morphological pattern is a major motif for assuming a rule, i.e. it is assumed that, for example, the English plural boy-s is computed by a plural formation rule that applies to the singular boy, which is represented as such in the mental lexicon, or, at least, it is assumed that such plurals can be computed in production and decomposed in receptive processing. Inversely, unproductive morphological patterns are often assumed to be not amenable to modelling by rules, but are assigned to mere storage. In some so-called dual route modelsReference Pinker^12, Reference Clahsen¹³ the distribution between computation or (de)composition and lexical access to stored items is strictly complementary; that is, it is assumed that, when hearing the plural boy-s, this form is decomposed into the singular form boy and the pluraliser -s, because only the singular boy is stored, and not so the plural. On the other hand, if one hears the unproductive plural ox-en, then this plural is stored along-side the singular ox and thus subject to direct lexical access. And no rule is available which could decompose ox-en into the singular ox and a pluraliser –en.

This dichotomy of the computing path of rules on the one hand, and the lexical access to the storage path of unproductive patterns on the other, becomes problematic if one assumes a gradable concept of degrees of productivity, as will be done here.

First, there exists an old tradition of distinguishing in a dichotomous way regular and irregular verbs in English, German, Dutch, etc. For these three languages the notion ‘regular verbs’ coincides with the totally productive class of verbs such as E. seem – seem-ed, kiss – kiss-ed. All the other different verbs belong to unproductive patterns. But are all of them really irregular, as tradition has it? The isolated verbs go – went – gone, bring – brought clearly are irregular, but the dozen verbs of the class sing – sang – sung, drink – drank – drunk show some regularity in the sense of homogeneous relations between present, past tense and past participle. Thus, they are better called sub-regular verbs rather than irregular ones.

Second, the productive and regular verbs of English, German and Dutch present also the default case, in the sense of the normal case, from which only a limited number of verbs deviate in well-defined ways.

This approach of collapsing productivity, default and regularity is already inadequate for plural formation in Dutch, which contains two competing productive rules, as in boek ‘book’, Pl. boek-en versus auto ‘car’, Pl. auto-s. Thus, none of the two plural formations is the default. Furthermore, should both be identified as regular? This approach fails even more in the strongly inflecting Slavic and Baltic languages or in Latin,Reference Dressler, Bolkestein, Kroon, Pinkster, Remmelink and Risselada^15–Reference Dressler, Kilani-Schoch, Gagarina, Pestal and Pöchtrager¹⁸ where many productive inflection classes compete with each other. Russian examples for fully productive verb classes areReference Dressler, Gagarina and Reishman¹⁷ illustrated with loans from English:

less productive, but very sizable is the class: kompjuter-izir-ova-t’ ‘computerise’.

Actual grammatical productivity of morphological patterns is gradual. The hierarchical criteria for degree of productivity, developed by DresslerReference Dressler, Bolkestein, Kroon, Pinkster, Remmelink and Risselada^15, Reference Dressler^16, Reference Dressler and Ladányi¹⁹ can be briefly summarised in the following way.

1. (a) The most important criterion is represented by the integration of loan words with unfitting properties, which have to be fitted (accommodated) to the system of the borrowing language. This criterion is the most important one, because a pattern must have maximum productivity in order to overcome two obstacles:

   The first obstacle is foreignness: a new word loaned from a foreign language still showing various symptoms of foreignness must be treated like an indigenous word for an indigenous rule to apply.

   The second obstacle is represented by unfitting properties, which have to be accommodated in order to fit the conditions of rule application.

   For example, most Romance languages (other than French) have thematic vowels in all major classes of verbs, e.g. in Italian, a thematic vowel –a- in am-a-re ‘to love’, past participle am-a-to, or a thematic vowel –i- in fin-i-re ‘to finish’, past participle fin-i-to. English (and German and Dutch) verbs have no thematic vowels, e.g. E. love, past participle lov-ed = G. lieb-en. Therefore, when an English loan-verb is adapted into a Romance language, a thematic vowel must be added: to dribble > It. dribbl-a-re, Sp. dribl-a-r; E. to flirt > It. flirt-a-re, Sp. flirte-a-r.

   The thematic vowel inserted in these loan-verb adaptations is always –a-. This shows that in Italian and Spanish only the conjugation with the thematic vowel –a- is maximally productive.

2. (b) A hierarchically lower criterion of productivity appears in the integration of loan-words with already fitting properties. Here, integration must overcome only the obstacle of foreignness.

3. (c) Hierarchically still lower productivity criteria are represented by various types of inflection of new indigenous words. Thus, rules that apply only to new indigenous words but not to any loan words are less productive.

But this is only a first step. A second step consists of distinguishing productive patterns or rules without and with competition in their respective domains of application. A productive rule without a productive competitor is more productive than a rule that competes with another in the same domain of applications. For example, within German feminine nouns, plural formation with –s and –(e)n are both productive. On the one hand they compete with each other both after word-final consonants and full vowels, such as -a. Note the rivalry between the variants:

On the other hand, to feminine nouns ending in the neutral unstressed vowel schwa [], as in die Liebe ‘love’, only the –(e)n plural can apply. Thus, the plurals of the French loan words die Sorte, Reserve cannot be anything else than die Sorte-n, Reserve-n. Therefore, the –(e)n plural is more productive after word-final schwa, as in Sorte-n, than after word-final full vowels and consonants, as in Pizz-en, Farm-en.

Now let us compare this model of morphological productivity with the notion of productivity in economics, starting with Adam Smith’s three criteria.

1. (1) His first criterion ‘skill of the worker within division of labour’ corresponds to division of labour of the most productive morphological rules: they apply in a complementary way to different words, as is the case with German plural formation with –n to feminine nouns after word-final schwa, as in die Sorte-n and with plural formation with –s after word-final full vowels, as in die Auto-s ‘car-s’, Photo-s ‘photographs’, Taxi-s ‘cab-s’. But the most productive morphological rules clearly cooperate in another type of effective division of labour: they can apply one after another. From the English noun department one can form the adjective department-al; from this adjective one can form the verb to department-al-ise, and from this verb in turn the noun department-al-is-ation, and finally one can pluralise this noun to department-al-is-ation-s. Here we have a fixed order of rules that may remind us of industrial work on a conveyor belt in the assembly line.

2. (2) Adam Smith’s second criterion of productivity is ‘economy through saving of time’. To this I’ll refer both with the next criterion and later on.

3. (3) Adam Smith’s third criterion of productivity ‘mechanisation of work’ can be compared with the mechanical character of automatically applied productive rules. This automatic and, in general, subconscious rule application guarantees economy in terms of the time-saving of Adam Smith’s previous, second criterion.

4. (4) In addition to Adam Smith’s three criteria, I have also mentioned for productivity in economics the fourth criterion of the growth rate on the developmental time axis. This corresponds to what morphological productivity achieves in the morphological domain of word formation. Productive morphological rules are used to create an incessant stream of new words, e.g. one of the most productive English word-formation rules is agent and instrument formation with the suffix –er, as in the agent nouns produc-er, cutt-er, peace-break-er and in the instrument nouns ice-break-er, grass-cutt-er, sound re-produc-er; that is, they have a high lexical growth rate, so to say.

Let us start with evidence from online processing, a field that has developed in the last few decades and has provided us with the least indirect evidence on how the brain processes linguistic information on a functional level. Clearly, there exist several psycholinguistic models for the analysis of online processing tests. The model I adhere to is the so-called race model.Reference Baayen and Schreuder^20, Reference Frauenfelder and Schreuder²¹ This model belongs to the family of models that assume an overlap of morphological (de)composibility by rule and storage;Reference Pinker and Ullman²² that is, it is not assumed that in production and reception, for example, productive ‘regular’ English verbs are processed only by rule, whereas non-productive ones are only stored in all their inflectional forms. Instead, the race model assumes a large overlap: all frequent verb forms are also stored, even if they can be accounted for by productive rules, and also unproductive sub-regular patterns may be processed by rule; that is, the direct lexical access path to stored forms often competes with an access by rule both in production and reception, and both competing types of access are engaged in a race. Among the factors that decide who wins the race, is the degree of productivity. The higher its degree of productivity, the better chance a rule has to trump direct lexical access. On the other hand, the more frequent is a stored inflectional form, the more likely it is for direct lexical access to win the race. This model of productivity may remind us of Adam Smith’s second criterion of productivity, i.e. ‘economy through saving of time’.

Among the various online processing tests that I did with various Canadian, Dutch, German and Austrian scholars, I want to mention just one. In a yet unpublished experiment, presented at the 13th international morphology meeting in Vienna,Reference Korecky-Kröll, Stempfer, Wiesinger, Reinisch, Bertl, Libben and Dressler²³ we performed an online processing test on German plural formation. This was a grammatical judgement test where subjects had to decide whether a plural form presented on a computer screen did or did not exist. The measuring variable was accuracy and reaction time. Viennese adults and children from 6 to 10 years of age had to judge three types of German noun plurals: first, existing plurals, e.g. Buss-e from the masculine singular der Bus ‘bus’ or Küss-e from the masculine singular der Kuss ‘kiss’; second, non-existing illegal, i.e. ungrammatical plurals of the type *Kuss-en ‘kiss-en’ (so to say formed like the English plural ox-en); and third, non-existing, but potentially grammatical plurals of the types Kuss-e and Büss-e, formed by competing productive rules. The results are compatible with our productivity model.

Even more important, grammaticality judgments on non-existing but potential plurals were less accurate and triggered longer reaction times than judgments on existing grammatical and non-existing ungrammatical plurals. These results support the concept of potential grammatical forms, a concept that stems from a rule approach to morphology. Thus, these results support the concept of productive morphological rule, as presented in my linguistic model of morphological rules.

This experiment leads us to a second psycholinguistic field: first language acquisition. In comparing children’s results with adult results, we found that only 8-year-olds approach adult proficiency. This may mean that, at least at this age, children have extracted productive rules for accounting for German plural patterns.

Still, offline production tests that we administered to younger children, ranging from 2 years 6 months to 6 years of age showed similar results much earlier. Offline production tests provide more indirect evidence on how we use language, because they are performed with less time pressure and allow reflection about the task a participant has to perform; that is, they are done with a higher level of language awareness.Reference Laaha, Ravid, Korecky-Kröll, Laaha and Dressler²⁴

Again, the results supported our productivity model in showing the best results for plural rules with maximum productivity, the second best results for competing productive plural rules and the worst results for unproductive plurals. Among erroneous plurals, children produced many more potential plural forms than illegal ones. This tallies very well with the also still unpublished results of spontaneous production data in two longitudinal studies of two girls from age 1 year 3 months to 2 years 7 months and then to 6 years.Reference Korecky-Kröll, Dressler, Stephany and Voeikova²⁵ In these corpora the children stopped at 2 years 6 months of age to produce, within erroneous plural forms, illegal plurals that combine umlaut with the –(e)n plural suffix, e.g. ungrammatical *Elefänt-en instead of Elefant-en, plural of Elefant ‘elephant’. That is, afterwards they produced spontaneously among erroneous forms only potential plurals.

I have not yet answered the two following questions. The first question is: why is there such gradualness in the degrees of productivity in morphology, whereas in phonology and syntax, rules are either productive or not? Interestingly, there exist some degrees of productivity in those areas where phonology and syntax interact with morphology, namely in morphonology and morphosyntax. Take the famous rule of velar softening in electric with word-final [k] and its derivative electric-ity with an [s] instead of the [k]. This morphonological rule is semiproductive. Thus, in the late 1970s, linguists invented the term formulaic-ity, a derivative of formulaic, but immediately quarrelled whether it should be pronounced formulai[s]ity or formulai[k]ity. My tentative answer to the question of why gradualness of degrees of productivity is a hall-mark of morphology only, is that only morphological rules are deeply involved with many, both general and specific, features of words in the lexicon, purely morphological but also syntactic, segmental phonological, prosodic, semantic, pragmatic features. This multiplicity of factors produces gradualness.

The second question is: what makes a morphological rule productive? One answerReference Wurzel²⁶ has been that a morphological rule must be transparent for being productive. This solution is wrong, because there exist unproductive transparent rules and productive non-transparent rules, although there is an evident and motivated tendency for a correlation between productivity and transparency.

My answerReference Dressler¹⁶ has been that morphological rules are prototypically productive, like phonological and syntactic rules. Productivity is a primary property of rules. Why should children during their language acquisition process construct automatically applying rules, if they are not productive? As a consequence, I consider unproductive rules as not real rules at all, i.e. they are usually dead rules that have lost their basic property of productivity in historical development.

But do we need morphological rules at all? Many linguists, and an even larger percentage of psycholinguists, have negated the assumption of morphological rules, and with a lot of plausible arguments and often with successful simulations of morphology acquisition. Let me just mention schema models, family size models, and connectionist and other statistic approaches to language learning. I would like to pose just two questions to the adherents of such counter-proposals:

First, do they negate also phonological and syntactic rules? An example is the English phonological rule of aspiration of stops, unless they follow sibilants. Thus, the /p/ in the pin is aspirated, whereas the /p/ in to spin is not. In my view, to negate morphological rules but not syntactic and phonological rules as well, is incoherent.

Second, how do these counterproposals model the notion of potential words or potential morphological word forms? Consider for example the potential German plurals that we subjected to successful online tests with both adults and older children. We have also done offline grammaticality judgment tests on German, Polish, French and Italian morphology: the results vouched for a clear separation of non-existent potential morphological forms from non-existent illegal forms.