This chapter turns to deletion rules, as well as other kinds of allomorphy, some of which can be recovered only historically. The first set of changes to be considered produces allomorphic variation due to the dropping of one or more phonemes from the edges of the original forms of roots or affixes. Like replacement rules, deletion rules must be defined in terms of both the phonological environment and the type, position, and number of morphemes involved in the change.

Consonant deletion

No double consonants. The general principle which governs the simplification of double consonants to single ones, phonetically, is that English, unlike Italian, Finnish, and many other languages, does not allow “long” or “double” consonants within a word. Our discussion of double consonant spelling in section 4.2, Chapter 6, pointed out the important distinction between double consonants in the spelling, which can arise as a consequence of total assimilation, and “long” consonants in the pronunciation. Word-internal double consonant letters are not pronounced “long”: irrational and eradicate have the same [r], symmetry and cemetery have the same [m].

English allows real long, or geminate, consonants, only at clear morpheme/word boundaries: file log, roommate, bus-schedule, bookcase, letter writer, etc. In such instances longer duration is a phonetic phenomenon only; it is unrelated to spelling. Nevertheless, there is a parallel between real long consonants and the double letters resulting from the total assimilations of Chapter 6: both are signals of morpheme boundaries.

The two general themes of this book are the historical origins and the structure of English words. Our word stock is huge. It is useful to divide it up between words that belong to the common language that everybody knows from an early age and words that are learned in the course of our education. The former, the core vocabulary, is nearly the same for everyone. The latter, the learned vocabulary, is peripheral and certainly not shared by everyone. The core vocabulary is not an area where we need special instruction – the core vocabulary is acquired at a pre-educational stage. Our learned vocabulary is a different matter. It varies greatly in size and composition from one individual to another, depending on education and fields of specialization. No single individual ever controls more than a fraction of the learned vocabulary. Often the extent of one's vocabulary becomes a measure of intellect. Knowledge about the history and structure of our words – both the core and the learned vocabulary – is a valuable asset.

The vocabulary of English is not an unchanging list of words. New words enter the language every day, and words cease to be used. The two sources of new words are borrowing and word-creation. In fields of higher learning, like the life sciences, physical sciences, medicine, law, and the social sciences, English has usually borrowed words from other languages to get new words to cover new concepts or new material or abstract phenomena.

Fossilized allomorphy

Unlike allomorphy resulting from phonetically motivated replacement, deletion, or epenthesis, the allomorphy discussed in this section cannot be attributed to the operation of an active and transparent phonetic rule. “Fossilized” allomorphs can be deceptively unlike each other in form, yet they are historically related – they are cognates. The morpheme variants in this group arose as a consequence of systematic changes in pre-Old English times, going all the way back to Proto-Germanic and even to Indo-European. Within English, i.e., after the fifth century, the conditions for these early changes became obscured, and their results can no longer be seen as regular or highly predictable. If we set up a scale of predictability for allomorphic variation, allomorphs due to assimilation or vowel weakening will rank very high; mostly, the conditions for these changes are transparent and recoverable. At the low end of the scale would be the non-productive allomorphic variation described in historical terms, for which the conditions have ceased to exist. At the bottom of the scale are completely unpredictable alternate forms of cognates, the subject of Section 5. Historical, or fossilized, allomorphy tends to follow several general patterns; familiarity with those can help us discern etymological cognates in spite of their overt differences.

Two such fossilized processes are gradation and rhotacism. Originally they affected root morphemes, and occurred regularly in conjunction with specific grammatical changes within a paradigm, e.g., the present vs. the past-tense form of one and the same verb, the nominative vs. the genitive case of the same noun.

In 1995, when introducing a talk based on material later published as Bauer (1996) at the Free University of Amsterdam, I commented that I had been working on morphological productivity for over twenty years, but still did not feel that I had a coherent picture of the topic. In the intervening five years, I believe that I have achieved some degree of coherence in my view, and this book is an attempt to communicate that understanding. I still regard the picture presented here as provisional, and I expect it to change in the next five or ten years. Twenty-five years ago, morphological productivity was not at the forefront of linguistic theorising, but today, thanks to the dispute between connectionists and purveyors of alternative views of morphology, it has taken on a far greater importance. This not only makes it timely to provide a statement of a position, but guarantees that further research in the near future will outdate some of the comments that are made here. It is my hope that the work presented here will provide a stepping-stone in the development of a new deeper understanding – even if it is one which contradicts my own cherished positions.

Introduction

The aim of this chapter is to introduce, discuss and define a number of notions that appear to be necessary in the study of morphological productivity. To a certain extent, this chapter continues the work of the last chapter, in that it considers distinctions and the terms for them that have been introduced in the literature on morphological productivity. But it goes beyond the last chapter in considering matters in much more detail and suggesting terminological distinctions where none have previously been made or where those that are made appear unsuitable.

The issue

This book deals with productivity as it affects morphological systems. Unfortunately, in the present state of morphological studies, this opening statement may mean different things to different people, even a definition of ‘productivity’ being a matter of some dispute. In order to develop a position from which conclusions can be drawn, it is thus necessary to begin from the very beginning, and gradually to construct a secure foundation of notions to support the enterprise. We can begin this construction by considering one fundamental definition of ‘productivity’.

Hockett (1958: 575) gives the label ‘productivity’ to that property of language which allows us to say things which have never been said before, the design feature that Chomsky (1965: 6) calls ‘creativity’. We do not have to go any further to see that this is an area in which there is, at least, some terminological difference of opinion. To add to the confusion, Chomsky (1965: 5) also talks about syntactic processes being ‘productive’ without making it clear whether ‘productive’ and ‘creative’ are the same or different things, while others, such as Lyons (1977: 76–78), distinguish carefully between the two terms, though not all scholars draw the same distinction.

In this conclusion it is my aim to tease out some of the threads that have been discussed in the previous chapters and present a unified view of morphological productivity. I shall also consider briefly the extension of the notion of productivity into other areas of linguistics.

Productivity summarised and defined

Availability and coinage

We are now in the position to divide morphological productivity into two distinct themes: availability and profitability. A morphological process is available if it can be used to produce new words as they become necessary. Availability is a yes/no question: either a morphological process is available or it is not. We might, however, ask more precise questions about availability and ask whether or not a particular morphological process is available with a certain type of base. For example, we might ask whether -ment is available with bases ending in -ise (answer in current English: no), or whether -ation is available with the same set of bases (answer in current English: yes). In principle a yes/no answer will be available for classes of base defined by any of the constraints discussed in section 5.2. In practice it might be more difficult, or there might be cases of uncertainty. This does not alter the fundamental principle that availability is an either-or, and that things are either productive or unproductive in this sense, with no middle ground.

Introduction

Psycholinguists are interested in morphological aspects of language behaviour because of the information that can be gleaned from it about the way in which language is processed, stored and produced. Much of the research that has been done in this area has been carried out from the point of view of research into reading, attempting to discover how readers link the text to their own linguistic competences, for instance. However, not all of this research is relevant for the purposes of this book. Here I review only two of these three aspects: storage and production.

The reasoning behind this is as follows. The comprehension of language in its written form may be complicated by different strategies of dealing with the complexities of orthographic structure that have nothing to do with the way in which meaningful units are stored in the brain. The same is true of phonological structure in dealing with the comprehension of spoken language.

Productivity and its synonyms

Although the terms ‘productive’ and ‘productivity’ appear to be relatively new, the concepts in the form in which they have been outlined in chapter 1 are old. As Schultink (1992a: 188) points out, the whole notion of grammar implicit in the work of the Sanskrit grammarians assumes the idea of productivity. This idea has been available to English grammarians at least since Palsgrave's Lesclaircissement de la langue francoyse was published in 1530. As Stein (1999: 62) says more generally of this work, ‘the concepts were there, but not yet the terms for them’, so that Palsgrave discusses affixes such as English -ness as being ‘certayne’ in their application (cited in Stein 1999). The term ‘productivity’ is much more recent. As far as Schultink has been able to discover, the term is first used by Dietz (1838: 221), who says that

Introduction

Summarising the earlier discussion, we can isolate a number of views to be found in the literature on scalar productivity.

1. There is no such thing. A morphological process is either productive or not productive.

2. Productivity is scalar, and this is an important fact about productivity. Somehow this variability is part of the linguistic competence of speakers.

3. Productivity is scalar, but this is a direct result of variable constraints that are imposed on morphological processes: the more constraints there are, the less productive the process.

4. Productivity is scalar, but this is nothing to do with linguistic competence, it is purely a matter of performance and should not be seen as an important feature of morphological productivity.

5. Productivity is scalar, and this is partly a result of morphological constraints on the processes, and partly a matter of linguistic performance. (This is a mixture of positions c and d.)

6. Productivity is not scalar. All morphological processes are productive, some are not widely used because their products are not needed often, but this is nothing to do with morphology.

Introduction

In this chapter I introduce a number of examples whose function is to illustrate various aspects of the discussion that have taken place in earlier chapters. It is my hope that the discussion of these examples will provide a focus for some of the theoretical points which have been discussed, and point to conclusions about what productivity is and how it should be dealt with.

Proto-Germanic *-dōm

Mention was made in section 1.2 of the fact that productivity of a given process can vary diachronically. The example of Proto-Germanic *-dōm is chosen to illustrate this point. It will be shown that while some formations using this process are common across different Germanic languages (and it is thus possible, though not definite, that they are shared formations from the Common Germanic period), the use of the suffix in the daughter languages shows a great deal of variation – even though it has never become wildly productive in any of them. In some cases it seems that one meaning has ceased to be productive at all; in others some type has thrived at the expense of others.

Theories of inflectional morphology

In any language exhibiting inflection, each inflected word in a sentence carries a set of morphosyntactic properties; in English, for instance, the verb form am in the sentence I am sure carries the properties ‘first-person singular (Isg) subject agreement’, ‘present tense’, and ‘indicative mood’. In very many cases, an inflected word's morphosyntactic properties are associated with specific aspects of its morphology; for instance, the properties of subject agreement, tense, and mood carried by the verb form likes in the sentence She likes reading are associated with the presence of the suffix -s. In recent years, grammatical theorists have devoted considerable attention to the nature of these associations between an inflected word's morphosyntactic properties and its morphology. Nevertheless, these efforts haven't yet led to anything like a consensus in current theories of inflection.

According to LEXICAL theories of inflection, these associations are listed in the lexicon; the affix -s, for example, has a lexical entry which specifies its association with the morphosyntactic properties ‘3sg subject agreement’, ‘present tense’, and ‘indicative mood’. Theories of this sort portray the association between an inflectional marking and the set of morphosyntactic properties which it represents as being very much like the association between a lexeme's root and its grammatical and semantic properties. This conception is rejected by INFERENTIAL theories, in which the systematic formal relations between a lexeme's root and the fully inflected word forms constituting its paradigm are expressed by rules or formulas.