Coda: What have we done?

This chapter reflects on conclusions from the study. Relational Morphology has some features that are necessary in any theory of morphology; other features are shared with other constraint-based theories. Some features, especially the centrality of relational links and the dual functions of schemas, are innovations. RM stresses the continuity between words and schemas, which confers advantages in many applications. The theory also stresses the continuity between productive patterns, nonproductive patterns with numerous instances, and truly marginal phenomena. The chapter proposes that Relational Morphology is attractive not just for its ability to account for morphological phenomena, but equally for its ability to integrate linguistic theory, psycholinguistics, and the understanding of human cognition more generally

Language processing and language acquisition through the lens of Relational Morphology

This chapter asks what is happening to linguistic representations during language use, and how representations are formed in the course of language acquisition. It is shown how Relational Morphology’s theory of representations can be directly embedded into models of processing and acquisition. Central is that the lexicon, complete with schemas and relational links, constitutes the long-term memory network that supports language production and comprehension. The chapter first discusses processing: the nature of working memory; promiscuous (opportunistic) processing; spreading activation; priming; probabilistic parsing; the balance between storage and computation in recognizing morphologically complex words; and the role of relational links and schemas in word retrieval. It then turns to acquisition, which is to be thought of as adding nodes and relational links to the lexical network. The general approach is based on the Propose but Verify procedure of Trueswell et al. (2013), plus conservative generalization, as in usage-based approaches.

Situating morphology

The book has three interwoven themes: a morphological theory, the structure of the lexicon, and an integrated account of the language capacity and its place in the mind. These themes together constitute the theory of Relational Morphology (RM), itself an extension of the Parallel Architecture of Jackendoff’s Foundations of Language, and closely related to Construction Grammar and Construction Morphology. A fundamental feature is that phonology, syntax, and semantics are independent components of language, linked by interfaces. Another feature is the continuity between lexicon and grammar. RM extends these features to the internal structure of words. In particular, morphology is constituted of a morphosyntactic component and its interfaces to phonology, phrasal syntax, and semantics. Furthermore, RM expresses regularities among words not in terms of rules that derive morphologically complex words, but in terms of declarative schemas that capture patterns of shared structure. The chapter concludes with a survey of similarities and differences between phrasal syntax and morphosyntax.

Morphologically conditioned phonological alternations

This chapter asks how affixes can affect the phonology of their stems, as in harmony/ harmonic/harmonious. Within the Parallel Architecture, phonology is an algebraic form of representation, while phonetic representation is analog in character. Their relation is negotiated by an interface that relates phonological segments and sequences to positions and trajectories in phonetic space. In these terms, the chapter explores aspiration, final devoicing, vowel shift and vowel reduction, affixes like -ity and -ious that manipulate the phonology of their bases, and affixes that can blend with their bases, for instance flattery (= flatter+ery). Again the formal machinery of sister schemas plays an important role in the account, taking over the work done in other theories by derivation (as in SPE and Lexical Phonology) and constraint ranking (as in Optimality Theory)

The functions of schemas

This chapter shows why declarative schemas are preferable to procedural rules as the means of encoding linguistic regularities. Traditionally, the function of rules is to create novel structures that are not stored in the lexicon. Declarative schemas fulfill this generative function by using Unification to create new structures. However, schemas can also perform an additional relational function, motivating patterns within the lexicon that are not fully productive. It is not evident how rules can perform such a function. The chapter’s conclusion, the Relational Hypothesis, is that all schemas encode patterns in the lexicon, and that a subset of them can be used productively to create novel structures. This changes the dynamic of linguistic inquiry from a narrow focus on productive processes to the full range of patterns, from productive to the most marginal.

Motivation in the lexicon

This chapter explores the notion of motivation. It is typically explicated in terms of inheritance: lexical items inherit properties from more general patterns. Inheritance is attractive for morphology partly because it has also been invoked in the organization of concepts. It explores three models of motivation. The impoverished entry theory claims that lexical entries contain only idiosyncratic material, and higher-order items or schemas fill in the predictable parts. The full entry theory claims that words are encoded in full, and they are motivated by virtue of being redundant with schemas. The chapter argues that the proper realization of motivation is in terms of relational links, which mark pieces of structure shared between lexical items, both between words and schemas, and both horizontally and vertically. Hence motivation is extended beyond the typical top-down relations.

Applying the tools to other domains

This chapter shows how the constructs of the Parallel Architecture and Relational Morphology can be applied outside morphology. Some extensions fall within the larger linguistic system: nonproductive schemas and sister schemas in syntax; the encoding of speech register, bilingualism, and dialect; orthography; and the mapping of phonology to meter in poetry. More speculatively, it shows that memory in other cognitive domains shares some of the character of Relational Morphology’s lexicon. The domains briefly explored are music; the understanding of the form and function of physical objects; knowledge of geography and spatial layout; and social knowledge, including customs and morality. The tentative conclusion is that these domains indeed share interesting features with language, including a vast and highly structured lexicon with multiple levels of representation, hierarchical constituency, schemas, and relational links.

Formalizing inflection

This chapter addresses the differences between derivational and inflectional morphology and how they are to be reflected in the Relational Morphology formalism. The formalization of inflectional features is illustrated in turn by English regular verbs, English irregular verbs, German weak verbs, and German strong verbs and past participles. Each case makes essential use of relational links among sister (or second-order) schemas. The analysis also offers a flexible description of inflectional classes. The chapter then discusses what verbal forms in a paradigm have to be stored in memory and how these are used to construct non-stored forms. The formalism for inflectional classes is applied to the “Same Verb Problem”: homophones with the same inflectional paradigm, for instance go/went away, go/went crazy, go/went for broke. This treatment is then extended to the polysemy of morphosyntactic tense.

Formalizing morphological phenomena

This chapter applies the tools of Relational Morphology, developed in the previous chapters, to a wide range of canonical and noncanonical morphological patterns, primarily in English, German, and Dutch. The phenomena discussed include affixes that occur in only a single word; words with legitimate affixes but nonlexical bases (a.k.a cranberry morphs); idioms; conversions and other zero morphology; linking elements; cumulative exponence; multiple exponence; sister words; sister schemas; blends and truncations; umlaut and other stem allomorphy; infixation and reduplication. In most of these cases, the complications turn out to lie in the interface between morphosyntax and phonology. In addition, many of the phenomena involve the use of sister schemas (also known as second-order schemas) and the same-except relation, which give the grammar great flexibility