On the Representation and Processing of Phonological Stem Variants of Complex Words

Abstract

In the present thesis we explored regular and irregular phonological stem variants with respect to their representation and processing. We analyzed (i) derivation involving regular stem alternations in English (trisyllabic shortening – TSS) and German (umlaut) and (ii) past tense formation of German strong verbs involving irregular stem alternation (ablaut). The first chapter provides a short discussion of morphological complexity, the existing models of perception and processing of complex words and of the ERP technique employed in the reported experiments. The hypotheses and objectives of the present thesis are also laid out in the first chapter. The second chapter introduces the methods used in the following experiments. The third chapter investigates the representation and processing of regular stem allomorphy. We hypothesized that regular stem vowel alternations, such as umlaut and TSS, should be captured within a single underlying stem morpheme with a set of morphophonological and morphosyntactic rules defining its surface phonetic form. To test this hypothesis, we examined violation related brain responses triggered by allomorph misapplication, viz. *ser[i:]nity/*Starkung (Related Derived), and purely phonological violations, viz. *ser[aɪ]nity/*Sturkung (Unrelated Derived), and *seromity/*Stögung (Nonce Complete), in a series of experiments with gradually modified experimental settings (lexical decision and memory task in word list or sentence context experiments). We argued that if regular stem allomorphs have a mental representation, the reparable pseudowords should elicit error-detection responses different from those evoked by irreparable pseudowords depending on the experimental task and design. The first two experiments explored the processing of English trisyllabic shortened deadjectival nouns (serenity) presented in isolation with a superimposed lexical decision task (Experiment 1) or a memory task (Experiment 2). Experiments 3-4 investigated the processing of German deverbal nouns derived from deadjectival umlauted verbs (Stärkung) in similar experimental settings. Throughout all word list experiments, reparable (RD) and irreparable (UD and NC) pseudowords elicited differential error-detection brain responses. RD but not UD or NC pseudowords were treated as items violating morphophonological rules. We considered this response pattern as supporting our hypothesis that regular stem allomorphs share a single underlying stem morpheme. The sentence context experiments (5 & 6) that assessed the processing of deviant stimuli (RD &UD) in a biasing context showed that repair process could be triggered by the contextual effects independent of the violation type. We also report a pilot study (Experiment 7) conducted with low-proficiency L2 learners of German that investigated the acquisition of a morphosyntactic rule in the L2. We conclude the third chapter with a detailed discussion of the results of the first part of the thesis and put forth a highly speculative model of a unified mental lexicon entry for regular stem allomorphs. The focus of the fourth chapter is the representation and processing of irregular stem allomorphs of German strong verbs. We hypothesized that irregular stem allomorphs of strong verbs should be represented separately within a unified lexical entry in an underspecified manner. Thus, past tense allomorphs and not the basic stems should be marked for tense. To investigate this topic, we compared normal processing of strong verbs with deviant processing of incorrectly inflected strong verbs: geh-t/ging (C) vs. ging-t/ging-te (EI) and geh (BaS irregular), in the present (Experiment 9) and past (Experiment 8) tense contexts. Additionally, we compared the strong verb conditions with the processing of correctly and incorrectly inflected weak verbs: schenk-t/schenk-te (C regular) vs. schenk (BaS regular). The BaS irregular (geh) condition, being a bare basic stem, systematically elicited a semantic conflict, as reflected in the N400 amplitude. We attributed this effect to the underspecification of the present tense/basic allomorph for the tense feature. Therefore, the results of the second part of the thesis provided evidence for the separate representation of irregular stem allomorphs of strong verbs within a hierarchical lexical entry based on the underspecification of morphological features. We conclude the chapter with the discussion of results of the second part of the thesis and put forth a tentative model of the mental representation of irregular stem allomorphs of strong verbs. The fifth chapter provides the general discussion of the results in the framework of the existing models of perception and processing of complex words in line with the key questions presented in the first chapter. We also address the issues that could be relevant for future research.