Analogy in Word-formation

A Study of English Neologisms and Occasionalisms

Elisa Mattiello

• Publisher: De Gruyter Mouton
• Year: 2017
• eBook ISBN: 9783110549737
• Edition: 1
• Topic: Languages & Linguistics
• Subtopic: Linguistics
• Index: Languages & Linguistics

This book fills a gap in lexical morphology, especially with reference to analogy in English word-formation. Many studies have focused their interest on the role played by analogy within English inflectional morphology. However, the analogical mechanism also deserves investigation on account of its relevance to neology in English. This volume provides in-depth qualitative analyses and stimulating quantitative findings in this realm.

1Introduction

1.1An overview of the term ‘analogy’ in linguistics

Analogy is a very old concept in linguistics, and one that has attracted a plethora of interpretations and definitions. Originally, the Greek term αναλογία (analogía) denoted a real mathematical proportion, which was used by Greek grammarians, from Aristophanes of Byzantium and Aristarchus of Samothrace onwards, in order to categorise morphological forms (Schironi 2007). Then analogy was introduced into Latin grammar and became a basic criterion for working out grammatical rules. Hermann Paul (1846‒1921), the nineteenth-century Neogrammarians’ leading theorist, as well as the structuralist Ferdinand de Saussure (1857‒1913) inherited the word in this sense. Interestingly, the term ‘analogy’ was either kept as analogia, or translated into Latin as proportio, comparatio, secundum rationem, or even as regula. In addition, Latin grammarians adopted the Greek proportional descriptive technique, especially used for morphological inflection, according to which A : B = Aı : X (X = Bı). In other words, if one knows that the plural of shoe is shoes, one may reasonably deduce that the plural of tie is ties in line with the following analogical reasoning: shoe : shoes = tie : X (X = ties).

The concept of proportion and its correlated notion of substitution were pervasive in the varied literature which took analogy into consideration. Bloomfield (1933), for instance, claimed that “[t]he utterance of a form on the analogy of other forms is like the solving of a proportional equation” (p. 276), as in his example Charlestoner [1927] ‘one who performs the dance called Charleston’ (OED2),1 coined after dancer [c1440], waltzer [1811], two-stepper [n.d.] (see two-step [1900] OED2, s.v. two), and so on. Bloomfield (1933) viewed regular analogies of this type as “habits of substitution” (p. 276), i.e. as paradigmatic substitutions of the variable part in the analogical proportion (e.g. Charleston in Charlestoner substitutes waltz in waltzer).

In the 1960s and 1970s, however, analogy became a rather illegitimate topic in linguistics, expressly banned – especially in the United States – by generative grammarians such as Noam Chomsky (1957) and, later, Paul Kiparsky (1974) and Mark Aronoff (1976), and replaced by other more adequate notions. According to Kiparsky (1974: 259), for instance, the general concept of analogy should be refuted because, although it allows the following proportion: ear : hear = eye : *heye, there is no verb *to heye (meaning ‘to see’) in English. Unlike Bloomfield (1933), generativists analysed words such as Charlestoner as formed via rules – i.e. the syntagmatic concatenation of the morphemes Charleston and -er – rather than on the model of words such as waltzer.

Nonetheless, the neogrammarian notion of analogical formation (“Analogiebildung” in Paul 1880) had not disappeared and, against the Chomskyan generative tradition and American structuralism, it came back as a legitimate area of enquiry. Charles Hockett, in particular, was the first to defend Bloomfield’s (1933) concept of analogy:

An individual’s language, at a given moment, is a set of habits ‒ that is, of analogies. Where different analogies are in conflict, one may appear as a constraint on the working of another.

(Hockett 1968: 93)

Thus, we do not form *swimmed as the past tense of swim, on the analogy represented by sigh : sighed (and similar weak verbs), because the regular pattern here meets a constraint in the form of the subregularity represented by sing : sang.2 Anttila (1977) was another unequivocal defender of analogy, which he also called “pattern” or “structure” (p. 25).

The end of the twentieth century was characterised by an increased interest in the concept of analogy, especially connected to the psycholinguistic studies of the 1980s (see Bybee 1988), or to computational models where analogy is the only morphological mechanism available. In such models, analogical algorithms are developed which can be used to predict, generate, or select new forms on the basis of the similarity of a given base with existing forms in the lexicon. Three well-known such algorithms are Skousen’s Analogical Modeling (AM, Skousen 1989, 1992; Skousen, Lonsdale, and Parkinson 2002; Skousen and Stanford 2007), the Tilburg Memory Based Learner (TiMBL, Daelemans and van den Bosch 2005; Daelemans et al. 2007), and the Generalized Context Model (Nosofsky 1986, 1990).

Exemplar-based models suggest that exemplars that are stored in our memory can help predictions on language behaviour and the analogical algorithm capitalises on the multiple relationships that words in the lexicon may have. In these models, analogies are based on sets of words of varying sizes, where a set may consist of a single word or thousands of words. In this approach, analogical algorithms can model both rule-like behaviour (with very large analogical sets) and local analogies (with a single item as model) (see Chapman and Skousen 2005 and Arndt-Lappe 2014 for applications of Analogical Modeling within the realm of derivational morphology).

In the same years, Becker (1990) reopened the debate on the notion of proportional analogy as developed by Paul (1880), thus attracting the attention of prominent morphologists, such as Plag (1999) and Bauer (2001). In the present century, Itkonen (2005) has additionally explored analogy from a cognitive-philosophical perspective. Hill (2007) has discussed the subject of proportional analogy from a historical perspective. Recently, van den Bosch and Daelemans (2013) have elaborated a computational approach called Memory-based Language Processing (MBLP), which can be viewed as an implementation of the analogical, exemplar-based strand of linguistic theories developed throughout the twentieth century.3

In the pertinent modern literature, analogy is considered one of the main mechanisms or guiding principles in language learning and change (Hock [1938] 1991; Anttila 2003; Hopper and Traugott [1993] 2003; Fischer 2007; Bybee 2010; Aronoff and Fudeman [2005] 2011; Fertig 2013; cf. Amiot 2008 for French), and a key notion in inflectional morphology and its diachronic evolution. For instance, many English strong verbs have been regularised by analogy with weak forms: regular help‒helped‒helped, from the previous irregular paradigm help‒holp‒ holpen, is a case in point (Fertig 2013: 8). Instead, an example of a verb which has resisted analogical change is bring‒brought (Fertig 2013: 9). Finally, a counter-trend to morphological regularisation is the change from dive‒dived to American English dive‒dove, which is formed by analogy with drive‒drove (Online Etymology Dictionary).

Another relevant realm for analogy concerns psycholinguistic studies and experiments on first language acquisition and speech errors. For example, the impact of types of analogy on L1 acquisition has recently been studied by Dressler and Laaha (2012) for German. Indeed, unconscious analogies are often the source of children’s overgeneralisations according to fully productive, partially productive, and even unproductive rules. For English, various studies by Eve Clark (e.g. Clark 1981, [2003] 2009) have demonstrated the role of analogy in the formation of novel compounds by young children, such as *garden-man or *plant-man (vs. less transparent gardener), coined by analogy with well-known forms ending in -man (e.g. dustman, fireman, etc.).

Moreover, Victoria Fromkin’s works on speech errors (Fromkin 1973, 1980) have shown that many errors are caused by phonological (and semantic) similarity between two items: for instance, when ambitious is pronounced in the place of ambiguous (Fromkin 1980: 3), or when two similar words are blended (e.g. *shaddy ← shabby and shoddy) (Fromkin 1973: 36). Interestingly, the latter example is analogical with attested lexical blends, such as slang fantabulous [1959] ← fantastic and fabulous, ‘of almost incredible excellence’ (OED2), or ginormous [1948] ← gigantic and enormous, ‘very large, simply enormous’ (OED2) (see analogy in speech errors in 3.4.4). Studies on second language acquisition, by contrast, seem to disregard analogy as a relevant mechanism in language learning, with very few exceptions (e.g. Lyster and Sato 2013), which integrate an analytic rule-based system with a memory-driven exemplar-based system (Skehan 1998), but still on the learning of English inflection.

1.2Aims of the book and its contribution to word-formation theory

Despite its versatile character, the common (mis-)conception that most scholars still have of analogy as a rare or unimportant process largely departs from its original meaning and current relevance. This book aims to disprove the myths that: a) analogy is inconsequential in English word-formation (Bauer 1983; Plag 2003), and b) it is devoid of relations to morphological rules (Dressler and Merlini Barbaresi 1994; cf. Dressler and Laaha 2012; Mattiello 2016).

In word-formation, different schools of thought consider analogy, and its correlated notion of creativity, either as an exceptional process that stands in contrast to the productive formation of novel words using rules (Plag 1999), or as the most important means of derivation including both productive and unproductive patterns (Zemskaja 1992). The notion of schema with various degrees of abstraction and productivity, which was introduced by Bybee (1988) and refined by Köpcke (1993, 1998), makes this already complex picture even more complicated. Although Köpcke (1993) developed his model of schemas (vs. rules) for inflectional morphology, this concept is also applicable to word-formation (see also Arndt-Lappe 2015 for an overview of analogy in contemporary word-formation theories, and its relation to other mechanisms in synchronic morphology, such as rules or schemas).

In general, traditional generative approaches to word-formation, such as Aronoff’s (1976) and Spencer’s (1991), tend to think of rules as the basis of broad generalisations, reserving analogy for local, lexically restricted patterns. Generativists have recently recognised the importance of analogy in inflectional morphology, as a means of generalisation/extension of a morphological pattern across paradigms (Aronoff and Fudeman [2005] 2011: 92), but they do not confer any regularity on analogical word-formation. In such a rule-based approach to morphology, analogy is conceived as a surface means to produce neologisms or occasionalisms (see 2.1 for a distinction) via particular defaults of individual (complex) words, rather than productive rules. A clear expression of this view, which is frequently found in the generative literature, is, for example, in Bauer (1983):

If instances of word-formation arise by analogy then there is in principle no regularity involved, and each new word is produced without reference to generalizations provided by sets of other words with similar bases or the same affixes: a single existing word can provide a pattern, but there is no generalization.

Bauer (1983: 294)

This definition, however, is too restrictive and inadequate to cover all the possible facets of the concept of analogy. It excludes, for instance, analogical compounds such as the well-known ear-witness [1539] ‘a person who testifies to something on the evidence of his own hearing’ (OED2), after eyewitness [1539],4 and software [1960] ‘the programs and procedures required to enable a computer to perform a specific task’ (OED2), after hardware [1947], which at the same time conform to rules.

A theoretical approach which deals with the analogical mechanism is Booij (2010), who develops a model known under the name of Construction Morphology. Constructionist theories address the interesting question of how analogy is related to schemas, which are regarded as the central mechanism in word-formation. In Booij (2010), it is claimed that schemas and subschemas may operate on symbolic features, and that the crucial difference between analogical formations and schema-based formations lies in their making reference to different degrees of abstraction. Analogy in this model is defined as a strictly local mechanism, which is complementary to schemas and may constitute an initial stage of the development of a schema (Booij 2010: 88–93; cf. 3.2.2 for a pertinent discussion). Similarly, in my approach (Mattiello 2016), analogy as a local mechanism is identified with the concept of ‘surface analogy’ (defined below and in 1.3), whereas ‘analogy via schema’ can be viewed as an extension of the former, from a single model to a group of prototype words that share the same model (see a clearer distinction in 1.3).

Therefore, the flexibility of schemas and subschemas gives Construction Morphology a conceptual advantage over many other morphological models. However, as observed by Bauer, Lieber, and Plag (2013):

[I]n addition to schemas and subschemas, analogy is needed as an additional and quite separate mechanism in order to account for the manifold isolated analogical formations that seem to be quite common in English (and other languages).

Bauer, Lieber, and Plag (2013: 633)

This brings us to the family of performance-orientated connectionist approaches, in which analogy is the only morphological mechanism available. In computational analogical theories, we observe that analogy is conceptualised as a predictive mechanism, where predictability emerges from the fact that models are selected by algorithms.

A crucial property of computational analogical models that is particularly relevant for morphological theory is that they have been claimed to be able to account for both local analogy and rule-governed behaviour. For instance, in their approach, Derwing and Skousen (1989) claim that a description based on analogy may become a real alternative to a description based on (productive) rules. Therefore, they propose an analogy-based account that excludes the notion of rule, or subordinates it to the superordinate analogical principle. In this regard, Bauer (2001: 75) argues that “[t]here are a number of reasons for believing that morphology is basically a matter of rule-governed behaviour, and a number of reasons for believing that it works basically by analogy” (see his arguments and counter-arguments for taking a rule-governed vs. an analogical approach to morphology) (Bauer 2001: 76–84). The two concepts, therefore, are not inconsistent with one another (Mattiello 2013).

In a similar attempt to attribute a prominent role to analogy, Becker (1990) – w...

Table of contents

• Cover
• Title Page
• Copyright
• Preface
• Contents
• 1 Introduction
• 2 New words, neologisms, and nonce words
• 3 Analogy in English word-formation
• 4 Analogy in specialised language
• 5 Analogy in juvenile language
• 6 Analogy in journalistic language
• 7 Analogy in literary works
• 8 Acceptability of new analogical words
• 9 Conclusions
• References
• Lexical index
• Subject index
• Fußnoten