The Processing of Lexicon and Morphosyntax 1443857114, 9781443857116

Citation preview

The Processing of Lexicon and Morphosyntax

The Processing of Lexicon and Morphosyntax

Edited by

Vincent Torrens and Linda Escobar

The Processing of Lexicon and Morphosyntax, Edited by Vincent Torrens and Linda Escobar This book first published 2014 Cambridge Scholars Publishing 12 Back Chapman Street, Newcastle upon Tyne, NE6 2XX, UK British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Copyright © 2014 by Vincent Torrens, Linda Escobar and contributors All rights for this book reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owner. ISBN (10): 1-4438-5711-4, ISBN (13): 978-1-4438-5711-6

TABLE OF CONTENTS An Overview of Language Processing ........................................................ 1 Vincent Torrens and Linda Escobar Part I. Lexical Processing A Theoretical Account of the Effects of Acoustic Variability on Word Learning and Speech Processing ................................................................. 7 Joe Barcroft and Mitch Sommers Comprehending Metaphors of Different Types: Evidence from Russian..... 25 Natalia Cherepovskaia and Natalia Slioussar Resolving Tip-of-the-Tongue States with Syllable Cues........................... 43 Nina Jeanette Hofferberth and Lise Abrams The Interpretation of Brazilian Portuguese Quantifier todo in Distributive and Collective Contexts: An Experimental Study ..................................... 69 Mercedes Marcilese and Erica dos Santos Rodrigues Part II. Morphosyntax Processing An ERP Study on Aspectual Mismatches in Converbial Contexts in Polish ..................................................................................................... 89 Joanna Báaszczak, Patrycja JabáoĔska and Dorota Klimek-Jankowska To Store or not to Store, and if so, Where? An Experimental Study on Incidental Vocabulary Acquisition by Adult Native Speakers of German ................................................................................................ 127 Denisa Bordag, Amit Kirschenbaum, Andreas Opitz and Erwin Tschirner Developmental Dyslexia and Unaccusativity .......................................... 149 Silvia Martínez-Ferreiro The Minimal Structure Principle and the Processing of Preposition Stranding ................................................................................................. 165 Ellen Thompson and Naomi Enzinna

AN OVERVIEW OF LANGUAGE PROCESSING VINCENT TORRENS AND LINDA ESCOBAR

It is well known that research on cognitive development is very wide; in order to narrow this subject matter, this book intends to contribute to a general interest by highlighting recent research on speech perception and production concentrating on two key areas: lexical and sentence processing. Among others, we cover some crucial topics of word learning, the tip-of-the-tongue phenomenon, lexical access and metaphors in the arena of first language acquisition, foreign language learning and speech processing. 1 In contrast to most books dealing with research on cognitive studies, this book is an attempt to provide an overview from the combination of the perspective of both psycholinguists and linguists. The reader will find the scientific method by which certain hypotheses are tested by online experimental methods, self-paced reading, semantic priming and event related potentials. In addition, the results will also be interpreted taking into account some linguistic approaches to sentence processing. The first area covered in the book deals with the crucial topics of word learning, the tip-of-the-tongue phenomenon, lexical access and metaphors. The second area concerns the intricate question of whether syntactic properties of different languages affect language processing looking at evidence of different linguistic phenomena such as verbal aspect, logical form, quantifiers, argument structure and movement. The first paper by Barcroft & Sommers reviews research on how acoustically varied input affects speech processing in L1. In addition it 1

This volume includes a selection of the papers presented at the Experimental Psycholinguistics Conference that took place in Madrid with the aim to include a large number of studies conducting research in the field of psycholinguistics. The editors of this volume would like to thank the help of the members of the scientific committee: Vasiliki Chondrogianni, Nina Kazanina, Victoria Marrero, Robert Reichle, Eva Soroli and Monica Wagner. They are also grateful to Albert Costa, Conxita Lleó and Frank Wijnen for their participation in the conference organised thanks to the financial support of the Universidad Nacional de Educación a Distancia.

2

An Overview of Language Processing

analyses new word learning in both L1 and L2 in order to draw conclusions about the effects of acoustic variability in the evolving quality of lexical representations. They consider how word identification in L1 English is negatively affected by phonetically relevant sources of variability such as talker, speaking style, and speaking rate but not by amplitude and fundamental frequency. They also explain how these five sources of variability affect new word learning, by showing how phonetically relevant sources that pose costs to L1 word identification also positively affect new word learning whereas amplitude and fundamental frequency yield null effects in both domains. Bordag, Kirschenbaum, Opitz & Tschirner, next, present a study on the incidental vocabulary acquisition during reading of syntactically complex and simple texts by adult native speakers of German. A comparison of the L1 data with the L2 data reveals that syntactic complexity positively affects L2 acquisition of new meanings, but does not influence acquisition while reading in L1. The results further indicate that while new lexical units are added to the open, incomplete L2 system with relative ease, adding new items to the L1 lexicon is more restricted. While the L2 data show evidence of an interaction of newly acquired meanings with earlier acquired representations, no evidence for such integration into the semantic network can be found for the L1 data. The results suggest that new L1 representations are stored in episodic memory rather than in semantic memory. Cherepovskaia & Slioussar, then, study which mechanisms are involved in understanding metaphors in online processing. These authors attempt to shed new light on the question whether metaphors are more difficult to process than literal expressions. In order to do so, they choose syntactically diverse materials in Russian to replicate previous studies analyzing how two types of metaphors are processed in comparison with the same expressions of literal meanings. They argue that the first type of metaphor is engrained in the speakers’ minds on the conceptual level, but can be expressed in different ways. In contrast, the second type of metaphor turns out to be fixed both on the conceptual and on the linguistic level. Their findings also show that metaphors in these two groups are processed differently. Yet, they do not find any evidence that literal meaning is assessed first. However they conclude that unlike the first type of metaphor, the second type is stored in the mental lexicon as a whole. The paper that ends the first part of the book by Hofferberth & Abrams presents a study on Tip-of-the-tongue phenomenon (TOT). They support the idea that TOT represents a speaker's temporary and typically frustrating inability to retrieve a known word, which results from

Vincent Torrens and Linda Escobar

3

weakened connections between a word's lexical representation (lemma) and its phonology (lexeme). They also argue that encountering phonologically-related cues during a TOT specifically related to words containing the first syllable helps to resolve such an inability. The chapter then discusses different models of speech production where the locus of TOT is found using various methodologies investigating TOT states in laboratory studies, including a new experiment using a syllable in isolation as the cue. The second part of the book starts with Báaszczak, JabáoĔska & Klimek-Jankowska’s chapter concerning one of the most contentious issues in new psycholinguistics ERP experiments about the P600 index related to the interpretation difficulty reflecting both semantic and syntactic integration problems. The goal of this paper is “to report new psycholinguistic (Event Related Potentials, ERP) results related to the processing of grammatical aspect in converbial contexts in Polish”, given the case that “Polish converbs (gerunds) impose two kinds of restrictions: (i) specific morphological selectional requirements as well as (ii) specific semantic/pragmatic constraints on temporal ordering.” The authors test different morphological mismatches in contexts of selectional restrictions imposed by anteriority and simultaneity converbial morphemes on their verbal stems. On the standard assumption that a P600 value reflects an integration difficulty at the morpho-syntactic level, the authors discuss the experimental question, among others, “why the P600 is stronger in the case of simultaneous ungrammatical converbs”. The study presented by Marcilese & dos Santos Rodrigues explores in detail how Brazilian Portuguese adults interpret distributive and collective readings when the universal quantifier “todo” is in play. These authors use online methodology to provide evidence in favor of the Good enough approach according to which the language processor can sometimes generate only partial, incomplete representations of the linguistic input (Christianson et al. 2001; Ferreira and Stacey 2000; Ferreira 2002; Ferreira and Patsos 2007). These authors make use of online methodology Psyscope, and according to their results, they further claim that differently from language production, comprehension seems to be strongly influenced by a “fast problem-solving” principle. These authors suggest that theoretical semantic and syntactic descriptions of linguistic phenomena do not necessarily explain the effects observed during language processing. The last two chapters by Martínez-Ferreiro and by Thompson & Enzinna explore in detail how processing relates to syntactic derivations in the case of argument structure and movement. Their conclusions are, however, very different. Martínez-Ferreiro put forward “the crucial role of

4

An Overview of Language Processing

argument structure, and, more specifically, the relative complexity of unaccusative verbs in comparison to unergative forms in developmental dyslexia along with non-brain-damaged subjects”, whereas Thompson & Enzinna conclude that processing seems not to be affected by the same syntactic constraints that are present in sentence derivations. In particular, the Minimal Structure Principle (MSP) by which the representation w ith mo r e projections is discriminated in favor of the one with fewer ones is claimed not to be active in language processing. As evidence, they show that “P-Stranding and Pied-Piped Sentences in general have similar processing costs” despite the fact that the former derivation has less projections than the latter ones. In sum, both parts in this book attempt to cover most of the core concerns of psycholinguistics considering lexical and sentence processing and its implications for syntactic assumptions, while keeping the reader interested in new methodological approaches and new paradigms from experimental psycholinguistics, neuroscience and cognitive psychology. Moreover, most chapters do not only focus on empirically-driven theory advancement, but also on the impact of these advancements in applied contexts such as first language acquisition, foreign language learning and speech processing,

PART I. LEXICAL PROCESSING

A THEORETICAL ACCOUNT OF THE EFFECTS OF ACOUSTIC VARIABILITY ON WORD LEARNING AND SPEECH PROCESSING JOE BARCROFT AND MITCHELL S. SOMMERS1

Abstract In this paper we review research on how acoustically varied input affects speech processing in a first language (L1) and new word learning (in both L1 and second language, L2) in order to draw conclusions about the effects of acoustic variability on the evolving quality of lexical representations. First, we consider how word identification in L1 English is negatively affected by phonetically relevant sources of variability such as talker, speaking style, and speaking rate but not by amplitude and fundamental frequency (F0) (Mullennix, Pisoni & Martin, 1989; Sommers, Nygaard & Pisoni, 1994; Sommers & Barcroft, 2006). Second, we explain how these five sources of variability affect new word learning (in L2, Barcroft & Sommers, 2005; Sommers & Barcroft, 2007; Sommers; and for talker variability, in L1, Sommers, Barcroft & Mulqueeny, 2008). Those phonetically relevant sources that pose costs to L1 word identification also positively affect new word learning whereas amplitude and F0 yield null effects in both domains. Finally, we present a model of how phonetically relevant acoustic variability affects lexical acquisition and speech 1

Joe Barcroft Department of Romance Languages and Literatures Washington University in St. Louis St. Louis, MO 63130 E-mail: [email protected] Mitchell S. Sommers Department of Psychology Washington University in St. Louis St. Louis, MO 63130-4899 E-mail: [email protected]

8

Effects of Acoustic Variability on Word Learning and Speech Processing

processing across the lifespan. It depicts how phonetically relevant variability (a) positively affects new word learning by producing more distributed lexical representations and (b) poses costs during lexical processing because variant forms must be mapped onto the canonical word forms.

1. Background Acoustic variability refers to variations in dimensions of the speech signal other than the specific linguistic content. For example, the same linguistic information can be spoken by different talkers or at different speaking rates, resulting in what Pisoni (1985) has termed variations in indexical, as opposed to linguistic, features of speech. In this paper, we consider the role of acoustic variability across the lifespan by reviewing research findings on the effects of different sources of acoustic variability-talker, speaking style, speaking rate, amplitude, and fundamental frequency-in two areas: (a) first language (L1) speech processing and (b) second language (L2) and L1 vocabulary learning. After identifying the unique pattern of research findings that has emerged within these two areas of research, we propose a model that depicts cognitive responses related to vocabulary learning and speech processing when one is exposed to input that contains phonetically relevant acoustic variability over time. In the research review in this section we first consider effects of five sources of variability-talker, speaking style, speaking rate, amplitude, and fundamental frequency-on L1 speech processing. We do so while paying particular attention to studies on word identification in noise wherein target words are presented in acoustically consistent versus acoustically varied formats. We then move on to studies on the effects of the same five sources of variability on L2 vocabulary learning and, in one case, the effects of talker variability on L1 vocabulary learning. This organizational structure is largely historical in nature given that much (but not all) of the research on acoustic variability and L1 speech processing preceded the research on acoustic variability and L2 vocabulary learning. However, once the research review is complete and the intriguing pattern of findings is summarized, we transition to a more developmental perspective when presenting our model of the relationship between acoustic vocabulary, vocabulary learning, and speech processing across the lifespan given that learning a word precedes the ability to identify that word in noise from a developmental perspective.2 2

Whereas this review focuses on effects of acoustic variability on speech

Joe Barcroft and Mitchell S. Sommers

9

2. Previous studies on speech processing and vocabulary learning 2.1. Research on Acoustic Variability and Speech Processing Many studies on acoustic variability and L1 speech processing point to the cognitive costs of processing acoustically varied as opposed to acoustically consistent stimuli. For example, presenting tokens produced by multiple as opposed to single talkers decreases L1 vowel perception (Assmann, Nearey & Hogan, 1982), word recognition (Mullennix, Pisoni & Martin, 1989; Ryalls & Pisoni, 1997; see also Creelman, 1957; Peters, 1955, as cited in Pisoni, 1997), and word naming (Mullennix, Pisoni & Martin, 1989).3 In addition, Sommers and Barcroft (2006, Experiment 1) found that word identification in L1 was negatively affected by speakingstyle variability. In this case, the no-variability condition involved listening to target words in one of the six speaking styles (produced by a single talker): normal, excited, denasalized, elongated (produced at a slow speaking rate), child-like (talker was asked to produce items “as if speaking like a young child”), and whispered voices. The variability condition, in contrast, involved listening to target words in all six of these speaking styles. The results of that study can be seen in Figure 1. Sommers, Nygaard, and Pisoni (1994) and Sommers and Barcroft (2006, Experiment 3) also demonstrated negative effects for speaking-rate variability. All of these findings are consistent with the position that listeners process indexical features of input containing talker, speakingstyle, and speaking-rate variability, which poses a cost that increases the overall needed for L1 word identification.

processing and vocabulary learning, key findings in other areas, such as research on the effects of talker variability on memory for L1 words and L2 phonemic training, also will be briefly noted. 3 There is also evidence indicating that at least some types of acoustic variability can improve memory for L1 words. Mullennix, Pisoni, and Martin (1989) found that memory for words spoken by multiple talkers was significantly better than memory for words spoken by only a single talker. Similarly, Goldinger, Pisoni and Logan (1991) found that, when listeners are given sufficient time to encode voice characteristics, serial recall is better for items spoken by multiple than single talkers. Thus, acoustic variability is associated with both costs and benefits in L1. The costs are reflected in reduced speech intelligibility for acoustically varied input; the benefits are reflected in improved memory for items produced with variable acoustic features.

10

Effects of Acoustic Variability on Word Learning and Speech Processing

Figure 1. Effects of speaking-style (voice-type) variability on spoken word identification in L1

Not all sources of variability have been found to produce negative effects on L1 word identification, however. Sommers, Nygaard, and Pisoni (1994) assessed the effects of overall amplitude and found that it had little effect on L1 word identification. They explained this finding by proposing that only sources of variability relevant to phonetic perception will impair spoken word identification. According to this phonetic relevance hypothesis, variability in overall amplitude, in contrast to talker, does not alter acoustic properties that are used for phonetic perception (in this case for speakers of English) and therefore does not pose the cost to L1 word identification that talker does. Sommers and Barcroft (2006, Experiment 2) corroborated this position by demonstrating that variability in overall fundamental frequency (F0), argued not to alter significantly the acoustic properties used for phonetic perception for speakers of English (in contrast to what would be the case for speakers of tone languages), also produced null effects on L1 word identification performance. To summarize, three sources of variability -talker, speaking style, and speaking rate- have been found to produce negative effects on spoken word identification whereas two sources of variability -amplitude and F0have not, at least for L1 speakers of English. According to the phonetic relevance hypothesis proposed by Sommers, Nygaard, and Pisoni, this pattern of findings makes sense because talker, speaking-style, and

Joe Barcroft and Mitchell S. Sommers

11

speaking-rate variability alter phonetically relevant properties within the speech stream whereas amplitude and F0 variability do not, at least for speakers of English. Clearly, it would be worthwhile to examine other sources of acoustic variability and to examine the effects of overall F0 variability on word identification performance with participants who speak a tone language and, therefore, for whom F0 is phonetically relevant at the lexical level. In such a case, one might predict negative effects, in contrast to the null effects observed for F0 variability among speakers of English.

2.2. Research on Acoustic Variability and Vocabulary Learning The same five sources of variability (talker, speaking style, speaking rate, amplitude, and F0) that have been tested with regard to their effects on spoken word identification in L1 have been assessed with regard to their effects on L2 vocabulary learning (Barcroft & Sommers, 2005; Sommers & Barcroft, 2007; Barcroft & Sommers, accepted pending revisions), and in the case of talker variability, with regard to their effects on L1 vocabulary learning (Sommers, Barcroft & Mulqueeny, 2008). Does processing indexical properties of input with phonetically relevant acoustic variability lead to more distributed lexical representations of word form and in this way positively affect vocabulary learning? If so, is this benefit restricted to sources of variability that are phonetically relevant to the participants attempting to learn the new vocabulary? In this section, we review the studies that were designed to address these questions.4 To begin, Barcroft and Sommers (2005) examined the effects of variability in speaking style (voice type) and talker on naïve listeners' ability to learn Spanish (L2) vocabulary. Speaking style and talker were selected as sources of variability because they represent one instance of intraspeaker variability (speaking style) and once instance of interspeaker variability (talker) that occur naturally. The general methodology used in 4

Another body of research suggests that acoustically varied presentation formats can be useful when teaching L2 phonemic contrasts, such as when training native Japanese speakers on the English contrast between liquid consonants /r/ and /l/. A number of studies have found benefits of acoustically varied input when training on this contrast (Logan, Lively, & Pisoni, 1991; Lively, Logan, Pisoni, 1993; Lively, Pisoni, Yamada, Tokura, et al., 1994; Bradlow, Pisoni, Akahane-Yamada & Tokura, 1997; see also Hardison, 2003). These findings suggest that access to indexical features of speech can facilitate establishing phonemic categories. According to Logan, Lively, and Pisoni, the use of acoustic variability during training may help listeners develop "stable and robust phonetic categories that show perceptual constancy across different environments" (p. 876).

12

Effects of Acoustic Variability on Word Learning and Speech Processing

these experiments was to compare vocabulary learning in conditions with no variability (one talker or one speaking style), moderate variability (3 talkers or 3 speaking styles) and high variability (6 talkers or 6 speaking styles). For both sources of variability, speed and accuracy of picture-toL2 recall and L2-to-L1 translation were the dependent measures. The findings for both sources of variability indicated positive and additive effects for acoustic variability. The participants were faster and more accurate in both types of recall when they learned the words in acoustically varied as compared to acoustically consistent formats. Moreover, learning with high variability was better (faster and more accurate) than for moderate variability, which in turn was better than with no variability. The results for talker variability can be seen in Figure 2. Figure 2. Effects of talker variability on L2 vocabulary learning

The researcher attributed this pattern of findings to the encoding of indexical features with the novel word forms during each presentation of a target word, which leads to a more distributed (robust) developing representation of the word forms in question. A model of this process appears in Figure 3.

Joe Barcroft and Mitchell S. Sommers

13

Figure 3. Model of the positive effects of variability on developing lexical representations

In another study, Sommers and Barcroft (2007) assessed boundary conditions for the benefits of variability by looking at the effects of three different sources of acoustic variability -overall amplitude, fundamental frequency, and speaking rate- on L2 vocabulary learning. Sommers and Barcroft hypothesized that overall amplitude and fundamental frequency might not affect L2 vocabulary learning in consideration of the phonetic relevance hypothesis and its predictions regarding L1 spoken word performance (Sommers, Nygaard & Pisoni, 1994) and the lack of effect of amplitude and F0 variability previously observed by Sommers, Nygaard, and Pisoni and Sommers and Barcroft (2006), respectively. Recall that, according to the phonetic relevance hypothesis, listeners encode and retain indexical properties of the speech signal such as talker characteristics and speaking rate that affect acoustic features, such as formant frequencies and transitions, that are used for phonetic identification whereas sources that do not alter phonetically relevant features, such as overall amplitude (affecting perceived loudness without altering formant frequencies or other phonetically relevant parameters), are ignored or processed more automatically (as in the case of an automatic gain control).

14

Effects of Acoustic Variability on Word Learning and Speech Processing

In applying the phonetic relevance hypothesis to studies of acoustic variability and L2 vocabulary learning, Sommers and Barcroft (2007) reasoned that if the beneficial effects of talker and speaking-style variability are a result of encoding and retaining phonetically relevant sources of variability (i.e. they serve as an additional retrieval cue), then sources of variability that are not retained (because they are not phonetically relevant for single-word stimuli among English speakers, as was the case in the study) should not provide a benefit in recall performance. Consistent with this prediction, Sommers and Barcroft found that neither variability in overall amplitude nor in fundamental frequency affected L2 vocabulary learning as measured by the same dependent variables used by Barcroft and Sommers (2005). Speaking-rate variability, on the other hand, produced positive effects, which is also consistent with the phonetic relevance hypothesis and Barcroft and Sommers (2005) explanation of how processing of indexical information in acoustically varied speech can lead to more distributed developing representations of novel word forms and in this way improve L2 vocabulary learning. Sommers and Barcroft (2007) proposed the extended phonetic relevance hypothesis to explain this pattern of effects observed for different sources of acoustic variability and L2 vocabulary learning. According to this hypothesis, English speakers benefit from speaker, speaking-style, and rate variability because these sources of variability are phonetically relevant (indexical properties of speech that affect acoustic features, such as formant frequencies and transitions, which are used for phonetic identification) to them. They do not benefit, however, from amplitude variability and FO variability, at least in the case of English speakers, because these sources of variability are not phonetically relevant to them. Sommers, Barcroft, and Mulqueeny (2008) assessed whether the benefits of talker variability would extend to (a) L2 words in another L2, Russian (Experiment 1); (b) novel words (pseudowords) that represented known objects versus novel objects (nonobjects) (Experiment 2); and (c) L1 words, in this case very low frequency concrete nouns so that they would be novel to adults (Experiment 3). In all three experiments, the participants were native speakers of English. They attempted to learn 8 new words in conditions of no variability (1 talker), moderate variability (3 talkers), and high variability (6 talkers). In all three experiments, positive and additive effects of talker variability were observed for accuracy and latency of vocabulary recall, suggesting that the benefits of acoustic variability extend to multiple types of vocabulary learning in both L1 and L2.

Joe Barcroft and Mitchell S. Sommers

15

Barcroft and Sommers (2011) helped to isolate the causal mechanism of the benefits of acoustic variability by assessing whether the benefit is due to difficult encoding demands (cognitive effort hypothesis) or the development of a more distributed (robust) developing lexical representation. Experiment 1 in the study compared the effects of learning new words in normal (easier encoding) versus denasalized (more difficult encoding) voice. The normal-voice condition led to more vocabulary learning, providing direct evidence against the cognitive effort hypothesis. The second experiment assessed the robustness of the lexical representations of words learned in acoustically consistent versus acoustically varied formats by measuring the accuracy and latency of L2to-L1 translation at four different signal-to-noise ratios. At all four ratios, words learned in the acoustically varied condition produced better (more accurate and faster) performance, and the degree of improved performance of the single- over multiple-talker condition increased as a function of the amount of decrease in the signal-to-noise ratio. These findings, which are consistent with the representation quality hypothesis and inconsistent with the cognitive effort hypothesis, provide strong evidence that the positive effects of phonetically relevant sources of acoustic variability such as talker are not due to increased cognitive effort but are the result of phonetically relevant acoustic variability promoting more robust formal representations for the target words in question. In another study, Barcroft and Sommers (accepted pending revisions) sought to test the extended phonetic relevance hypothesis (as described above) by examining effects of overall F0 variability among (a) bilingual speakers of Zapotec (a tone language) and Spanish, for whom overall F0 should be phonetically relevant at the lexical level due to their knowledge of Zapotec, and (b) speakers of Spanish who did not speak a tone language, for whom overall F0 should not be phonetically relevant at the lexical level. All participants attempted to learn 24 Russian concrete nouns while hearing words and viewing pictures on a computer screen. Three levels of F0 variability were compared: (a) no variability, or 1 F0 for 6 repetitions of a word; (b) moderate variability, 3 F0s x 2 repetitions; or (c) high variability, or 6 F0s x 1 repetition. The specific F0 levels used in the no variability condition were counterbalanced across participants. The results indicated that F0 variability significantly improved L2 vocabulary learning for speakers of the tone language (Zapotec) but not for the others, for whom null effects were obtained. These findings provided strong new evidence favoring the predictions of the extended phonetic relevance hypothesis. To summarize, research on acoustic variability and vocabulary learning

16

Effects of Acoustic Variability on Word Learning and Speech Processing

has demonstrated positive and additive effects for acoustic variability when the sources of variability in question are phonetically relevant to those learning new vocabulary, including the sources talker, speaking style, speaking rate, and, in the case of tone language speakers only, F0. Acoustic variability based on sources that are not phonetically relevant to those learning the new vocabulary produce null effects. Examples of these sources are amplitude and F0 variability for speakers of English and Spanish learning novel vocabulary, but not in the case of F0 variability for speakers of a tone language (Zapotec). Other findings, such as those of Sommers and Barcroft (2011), strongly support the position, as proposed by Barcroft (2005), that at least some sources of acoustic variability lead to more distributed (robust) representations of the novel words being studied.

3. Theoretical proposals 3.1. A Unique Pattern of Effects for Different Sources of Variability As can be viewed clearly in Table 1, an interesting pattern has emerged with regard to the effects of different sources of variability on L1 word identification and L2 vocabulary learning. The sources of variability that decrease L1 word identification performance are the same as those which produce positive effects on L2 vocabulary learning, and L1 vocabulary learning in the case of talker. As can be seen in the bottom two rows, establishing phonetic relevance depends on the previous linguistic experience of an individual. Whereas for tone language speakers overall F0 is phonetically relevant, for non-tone language speakers, it is not, at least at the lexical level when it comes to contrasting between minimal pairs, such as the contrast between /‫ݐ‬ùt‫ݕ‬à‫ޝ‬/ meaning ‘change’ and /‫ݐ‬ùt‫ݕ‬ӽ‫ޝ‬/ (the second syllable having a rising tone) meaning ‘reheat’ in Isthmus Zapotec. The question mark in the table indicates research still needing to be conducted on the effects of F0 variability on spoken word identification in L1 among speakers of a tone language. Both the original and extended versions of the phonetic relevance hypothesis would predict that F0 variability should increase processing time (increase reaction times) during word identification among this group of speakers. Additionally, given that the positive effects of phonetically relevant acoustic variability on vocabulary learning have only been demonstrated once for talker variability only, future research should examine the effects of other

Joe Barcroft and Mitchell S. Sommers

17

sources of variability on L1 vocabulary in order to determine whether the pattern of effects in L1 follows the pattern that has been demonstrated for L2 vocabulary learning. Table 1: Effects for Different Sources of Variability on L1 Word Identification and L2 Word Learning Source of Variability Talker Speaking Style Speaking Rate Amplitude F0 for tone language speakers F0 for non-tone language speakers

Phonetically Relevant? Yes Yes Yes No Yes

L1 Word Identification Negative1 Negative2 Negative3 Null2 ?

L2 Word Learning Positive5 Positive5 Positive6 Null6 Positive7

No

Null4

Null6,7

1 Mullennix et al. (1989); 2Sommers & Barcroft (2006); 3Sommers, Nygaard & Pisoni (1994); Sommers & Barcroft (2006); 4Barcroft & Sommers (2006), 5 Barcroft & Sommers (2005); also Sommers, Barcroft, & Mulqueeny, 2008 for talker and L1 vocabulary learning; 6Sommers & Barcroft (2007); 7Barcroft & Sommers (accepted pending revisions)

What does the pattern of effects depicted in Table 1 mean when it comes to the effects of acoustic variability across the lifespan? First, it clearly means that not all types of acoustic variability are alike and cannot be expected to produce the same effects. Whereas talker, speaking-style, and speaking-rate variability can be expected to produce negative effects (increased reaction time) on spoken word identification and positive effects on vocabulary learning (in L2 and in L2, at least for talker variability), amplitude and F0 variability can be expect to have no significant effect on either word identification or vocabulary learning. These findings are consistent with the extended phonetic relevance hypothesis. Second, the pattern of effects also indicates that not it clearly means that not all speakers should be expected to respond in the same manner to different sources of acoustic variability. As the bottom two rows in Table 1 illustrate, speakers experienced with using tonal contrasts at the lexical level experience benefit from F0 variability when attempting to novel vocabulary whereas speakers who are not experienced in using tonal contrasts in this manner do not benefit from F0 variability. Finally, the pattern of effects points to the need for a theoretical explanation of the effects of phonetically relevant acoustic variability at different stages of

18

Effects of Acoustic Variability on Word Learning and Speech Processing

linguistic development and language use. Why is it that acoustically relevant acoustic variability has positive effects on vocabulary learning, increasing accuracy and shortening response times, but poses costs during spoken word identification, increasing response times? The model proposed in the following section was designed to address this critical question.

3.2. Modeling the Effects of Phonetically Relevant Acoustic Variability across the Lifespan In this final section, we present a model (Figure 4) designed to account for why phonetically relevant acoustic variability at once positively affects vocabulary learning (increasing accuracy and decreasing response times) and negatively affects L1 speech processing (increasing response times on spoken word identification). The model assumes that the type of acoustic variability being provided in the input is phonetically relevant to the listener(s) in question, which implies the listener(s) will process the acoustically varied input (and not simply ignore or gain-control it out) in a manner that allows for the indexical information in the acoustically varied input to be encoded and stored. The three circles on the left side of the model (Word Learning) represent the degree to which the formal component of a developing lexical representation is distributed or robust (see Figure 3 for a depiction of how increases in variability can lead to more distributed representations). Therefore, what is depicted here is how the more distributed (robust) representations associated with having processed acoustically varied input leads to better cued recall of target words. In essence, the left side of the model is depicting what happens during a posttest phase when a participant is presented with a picture and is asked to attempt to retrieve the target word in question. In such a case, the input consists of a picture, which in turn activates the semantic representation for the referent in question (as represented by the arrow from the input box to the semantic representation box. Because the semantic representation is distributed (such as when the semantic space after seeing a picture of a squirrel is activated), the nature of the “search” for the appropriate novel form (such as ardilla, the Spanish word for 'squirrel') is also distributed in nature, as depicted by the downward stream of six lines, which are not uniformly spread across the semantic representation box.

Joe Barcroft and Mitchell S. Sommers

19

Figure 4. Model of the positive and negative effects of phonetically relevant acoustic variability across the lifespan WORD LEARNING

WORD IDENTIFICATION OUTPUT

INPUT

(Word form)

(Picture of word referent activates semantic representation)

SEMANTIC REPRESENTATION

No Variability

1

2

Moderate Variability

High Variability

No Variability

Moderate Variability

High Variability

INPUT OUTPUT (Learner produces word form)

(Word form) 1 = Canonical word form (based on statistical learning over time) 2 = Ongoing reshaping of canonical form (albeit minimal) from new less canonical input in moderate and high variability conditions

At this point we can observe why the more distributed representations of words learned with more variability are more likely to be retrieved when cued by a picture. The reason is that the more distributed representations are more likely to come into contact with the “search” lines emanating from (or simply being part of) the semantic representation. As can be seen in the model, a word learned with no variability makes contact only once, whereas a word learned with moderate variability makes contact three times, and a word learned with high variability makes contact six times. Each contact constitutes a successful meaning-to-form mapping and allows the participant to produce the recently learned (or being learned) word form in question. Therefore, as indicated by the larger number of arrows in the output for high variability (six arrows) over moderate (three arrows) and low variability (one arrow) and the larger number of arrows in the output for moderate over no variability, the developing representation of word forms learned in more acoustically varied formats (as more acoustically varied input) is naturally retrieved more often, allowing them to be produced more often as output.

20

Effects of Acoustic Variability on Word Learning and Speech Processing

The right side of the model, on the other hand, depicts the effects of phonetically relevant acoustic variability on word identification and the continuing development of the formal component of lexical representations. In contrast to the left side of the model, the right side concerns cognitive procedures that happen long after the word to be identified has been learned. Input is depicted on the bottom, and output on the top as it was deemed easier to follow visually in this manner, but it would make no difference if the two were inverted, providing that the arrows always go from input toward output. The three circles represent the canonical representation of the form of any given word. The lines going upward represent instances of the word form in question being produced in spoken input. Instances in which an upward-moving line makes contact with the solid part of the circle represent instances in which a word is accurately perceived and identified. The degree to which the upward-moving lines are spread apart represents the degree to which the listener is being exposed to acoustically varied spoken input. In the case of no variability, the lines are close together, depicting the lack of variability in the input to which the listener is exposed. In this case, all six lines make contact with the solid circle and therefore are successfully retrieved and identified. In the case of moderate variability, the lines are not as close together, representing more acoustically varied input, and two of the lines do not make contact with the solid circle, leading to less accurate retrieval and worse word identification. Finally, in the case of high variability, the lines are even more spread apart, representing even more acoustically varied input, and three of the lines do not make contact with the solid circle, leading to even less accurate retrieval and even worse word identification. In this way the model depicts why phonetically relevant acoustic variability leads to worse performance on spoken word identification. In light of the fact that language users are also language learners, the model also depicts how acoustically varied input for words, even when they were first learned a long time ago, has an impact upon their canonical formal representations. The three solid circles represent the canonical form of a word. When listeners are presented with acoustically varied input that does not match this existing canonical form, the canonical form reshapes in response to the noncanonical forms to which the listener (listenerlearner) has been exposed. The dotted circles represent this reshaping. As can be seen in the model, assuming that input without acoustic variability is consistent with the existing canonical form, no reshaping will take place. Moderately variable input, on the other hand, can cause the canonical form to reshape to a certain degree, as represented by the single

Joe Barcroft and Mitchell S. Sommers

21

dotted circle around the solid circle. Finally, highly variable input can cause the canonical form to reshape even further, as represented but the two dotted circles. Of course we are assuming a distributed representation of word form in both the word-learning and word-identification sides of the model, therefore, the size of the circles on the word-identification simply represent that a certain degree of reshaping will take place in this distributed representation of the word form in question. The reshaping may be very minor in some cases, but in other cases it may be more substantial, such as in cases when one is exposed to a new variety (dialect) of their native language over an extended period of time and the process of accommodation begins to take place. One of the strengths of the model presented here is that it provides a mechanistic account of the effects of phonetically relevant acoustic variability across the lifespan, beginning with how it affects vocabulary learning when we are first exposed to new lexical items and extending to how it affects how we process speech during word identification. The more distributed lexical representations associated with acoustically varied input during vocabulary learning depict why more instances of target word form can be retrieved when cued by the activation of the conceptual/semantic space of the referent of the target word in question. The lack of one-to-one mapping of acoustically varied forms of a known (previously acquired) word and the canonical form of the word depict why acoustically varied input poses costs during word identification. Clearly, future work can help to provide more fine-grain accounts of the processes involved in both word learning and word identification, including a more quantitative account to how canonical word forms respond to different degrees of non-canonical acoustically varied input, but in our estimation the present model provides a well-founded framework that is consistent with the current body research on the effects of variability on both vocabulary learning and speech processing.

4. Summary and Conclusion In this paper, we have presented a review of research on the effects of acoustic variability on (a) L1 speech processing, and in particular L1 word identification; and (b) vocabulary learning, mostly L2 vocabulary learning, but at least in one case, L1 vocabulary learning. The research in these two areas indicates that acoustic variability based on sources that are phonetically relevant, such as talker, speaking style, and speaking rate, negatively affect L1 speech processing but positively affect vocabulary

22

Effects of Acoustic Variability on Word Learning and Speech Processing

learning (mostly L2 vocabulary learning to date, and at least in the case of talker variability, also L1 vocabulary learning). Acoustic variability based on sources that are not phonetically relevant to listeners, such as amplitude and fundamental frequency (at least for native English speakers), have produced no significant effects on L1 speech processing, however. Interestingly, the same phonetically relevant types of acoustic variability (talker, speaking style, speaking rate) that produce negative effects on speech processing also produce positive effects on (mostly L2) vocabulary learning whereas acoustic variability based on other sources, such as amplitude and fundamental frequency (F0) (at least for native speakers of English), yield null effects, mirroring the pattern observed for speech processing precisely when it comes to effects versus null effects for all of the different types of acoustic variability observed to date. The predictions of Sommers and Barcroft's (2007) extended phonetic relevance hypothesis are consistent with this overall pattern of effects, as are the findings of Barcroft and Sommers (accepted pending revisions) demonstrating that speakers of a tone language experience improved L2 vocabulary learning with F0-varied speakers whereas non-tone language speakers do not. We also have presented a connectionist-oriented (emergentist) model that depicts how phonetically relevant acoustic variability positively affects vocabulary learning, poses costs during speech processing, and continues to shape the canonical forms of lexical items in the mental lexicon across the lifespan. Clearly, new research is needed to expand upon the existing body of research on different sources of acoustic variability, vocabulary learning, and speech processing. Such research can continue to test the predictions of the extended phonetic relevance hypothesis by examining, for example, whether F0 variability negatively affects L1 word identification among speakers of a tone language such as Zapotec, as the hypothesis would predict. Other research can test predictions of the model presented in this paper, help to refine the model, or both. Research in this area should continue to improve our understanding of the fascinating patterns that have been observed to date regarding the relationship between acoustic variability, vocabulary learning, and speech processing.

Joe Barcroft and Mitchell S. Sommers

23

References Assmann, P., Nearey, T. M., & Hogan, J. (1982). Vowel identification: Orthographic, perceptual, and acoustic aspects. Journal of the Acoustical Society of America, 71, 975-989. Barcroft, J., & Sommers, M. S. (2005). Effects of acoustic variability on second language vocabulary learning. Studies in Second Language Acquisition, 27, 387-414. Bradlow, A. R., & Pisoni, D. B. (1999). Recognition of spoken words by native and non-native listeners: Talker-, listener-, and item-related factors. Journal of the Acoustical Society of America, 106, 2074-2085. Bradlow, A.R, Nygaard, L.C. & Pisoni, D. B. (1999). Effects of talker, rate, and amplitude variation on recognition memory for spoken words. Perception and Psychophysics, 61, 206-219. Church, B. A. & Schacter, D. L. (1994). Perceptual specificity of auditory priming: Implicit memory for voice intonation and fundamental frequency. Journal of Experimental Psychology: Learning, Memory and Cognition, 20, 521-533. Goldinger, S. D., Pisoni, D. B., & Logan, J. S. (1991). On the nature of talker variability effects on recall of spoken word lists. Journal of Experimental Psychology: Learning, Memory, & Cognition, 17, 152162. Mullennix, J. W., Pisoni, D. B., & Martin, C. S. (1989). Some effects of talker variability on spoken word recognition. Journal of the Acoustical Society of America, 85, 365-378. Nygaard, L. C., Sommers, M. S., & Pisoni, D. B. (1995). Effects of stimulus variability on perception and representation of spoken words in memory. Perception & Psychophysics, 57, 989-1001. Palmeri, T. J., Goldinger, S. D., & Pisoni, D. B. (1993). Episodic encoding of voice attributes and recognition memory for spoken words. Journal of Experimental Psychology: Learning, Memory, & Cognition, 19, 309-328. Sommers, M. S., & Barcroft, J. (2005). Effects of variability in voice quality and fundamental frequency on spoken word identification. Manuscript submitted for publication. Sommers, M., & Barcroft, J. (2007). An Integrated Account of the Effects of Acoustic Variability in first language and second language: Evidence from Amplitude, Fundamental Frequency, and Speaking Rate Variability. Applied Psycholinguistics, 28, 2, 231-249. Sommers, M., & Barcroft, J. (2011). Indexical Information, Encoding Difficulty, and Second Language Vocabulary Learning. Applied

24

Effects of Acoustic Variability on Word Learning and Speech Processing

Psycholinguistics, 32, 2, 417-434. Sommers, M.S., Nygaard, L. C, & Pisoni, D. B. (1994) Stimulus variability and spoken word recognition. Effects of variability in speaking rate and overall amplitude. Journal of the Acoustical Society of America, 96, 1314-1324. Takayanagi, S., Dirks, D. D. & Moshfegh, A. (2002). Lexical and talker effects on word recognition among native and non-native listeners with normal and impaired hearing. Journal of Speech, Language and Hearing Research, 45, 585-597.

COMPREHENDING METAPHORS OF DIFFERENT TYPES: EVIDENCE FROM RUSSIAN NATALIA CHEREPOVSKAIA * AND NATALIA SLIOUSSAR

Abstract Despite extensive research, what mechanisms are involved in understanding metaphors is still a matter of debate. This paper focuses on one aspect of this problem: on the sequence of steps in online processing. Some authors think that to understand a metaphor, we must assess and reject the literal meaning of the expression first, so metaphors should be more difficult to process than literal expressions. Some claim that certain metaphors are stored in the mental lexicon as a whole, so they should be easier to process. And yet the others believe that metaphors are processed compositionally, but there is no primacy of the literal meaning. This paper sheds new light on this question, presenting data from a self-paced reading experiment on Russian. We chose syntactically diverse materials to complement previous studies looking primarily at “an x is a y” metaphors and analyzed how two types of metaphors are processed compared to the same expressions used in the literal meaning. The first type is engrained in the speakers’ minds on the conceptual level, but can be expressed in different ways. The second type is fixed both on the conceptual and on the linguistic 1

N. Cherepovskaia Saint-Petersburg State University, Russia e-mail: [email protected] N. Slioussar Utrecht institute of Linguistics OTS, the Netherlands, and Saint-Petersburg State University, Russia e-mail: [email protected]

26

Comprehending Metaphors of Different Types: Evidence from Russian

level. Our experiment showed that metaphors in these two groups are processed differently. In both cases, we did not find any evidence that the literal meaning is assessed first, but concluded that the latter, unlike the former, are stored in the mental lexicon as a whole.

1. Introduction The study of metaphors and other non-literal expressions provoked several important debates. The first concerns the status of metaphor in the language. Aristotle who was the first to address this question discusses it primarily in The Poetics and The Rhetoric. He contrasts metaphor with the ‘ordinary language’ and assumes that its main function is ornamental. Aristotle’s theory was widely criticized, but many subsequent works still share its basic insight: that metaphors are ‘special’ (e.g. Black 1962, 1993; Searle 1979, 1993). An alternative view was suggested by George Lakoff and his colleagues (Lakoff and Johnson 1980, 1999; Lakoff 1993). According to it, metaphors permeate our language and are foundational for our thinking. Namely, since we cannot sense the denotata of abstract concepts, we think and speak of these concepts in terms of concrete concepts. For example, time can be understood in terms of space (time as a road) or conceptualized as a precious resource (time as money). The second problem that became crucial to the study of metaphors concerns the relation between literal and non-literal meanings. Searle (1979) suggested that to understand a metaphoric expression, the speaker must assess and reject the literal meaning of this expression first. The predictions of this theory can be tested experimentally. In particular, if it is true, metaphors should take longer to be processed than the same expressions in the literal meaning. Various behavioral studies dedicated to this question did not find any evidence for the primacy of the literal meaning (e.g. Gildea and Glucksberg 1983; Glucksberg 2003; McElree and Nordlie 1999; Ortony et al. 1978). But there are some notable exceptions: for example, Brisard et al. (2001) found that metaphors were processed slower than corresponding literal expressions. It can also be noted that most experiments looked only at one type of metaphors: “an x is a y” (e.g. Brisard et al. 2001; Glucksberg and Keysar 1990; Glucksberg et al 1997; Jones and Estes 2006; McGlone and Manfredi 2001; Lai et al. 2009). So the debate is still open. In addition to that, the results of neurophysiological experiments addressing the problem of hemispheric specialization and trying to localize brain regions selectively involved in literal or non-literal language pro-

Natalia Cherepovskaia and Natalia Slioussar

27

cessing are mixed (e.g. Bottini et al. 1994; Eviatar and Just 2006; Giora 2007; Mashal et al. 2007; Pynte et al. 1996; Yang et al. 2009). Clinical studies point to different processing mechanisms for metaphors and literal expressions (e.g. Champagne-Lavau and Stip 2010; Iakimova et al. 2005). Moreover, although several alternatives to Searle’s theory were suggested (e.g. Bowdle and Gentner 2005; Gentner and Wolff 1997; Giora 2003), none of them became widely accepted. The third problem that is discussed in the field revolves around the mental lexicon. Many authors assume that at least some non-literal expressions are stored in the mental lexicon as a whole (Giora 1999). On one hand, this provides an easy explanation to the question how non-literal meanings are retrieved. On the other hand, it does not seem plausible that all metaphors are stored as a whole, so the border must be drawn somewhere. We supposed that conflicting results obtained in different studies and hence the answers to the above-mentioned questions might depend on the type of the chosen metaphors. In this paper, we present the results of a self-paced reading study based on Russian language where two types of metaphors were analyzed. The first type is engrained in the speakers’ minds on the conceptual level, but can be expressed in different ways. The second type is fixed both on the conceptual and on the linguistic level. To complement previous studies, we chose metaphors with the syntactic structure other than “an x is a y”.

2. Method Participants Participants were 26 native speakers of Russian (14 male and 12 female) with no history of language disorders. Design and materials In their seminal book Lakoff and Johnson (1980) identified a number of metaphors that form the basis of our abstract thinking. To give an example, we conceptualize time as money or in general as a precious resource. This is why in English, like in Russian and many other languages, we can invest time into something, spend it, save it, waste it, loose it, there can be a lot of time or too little time, enough or not enough time to do something, we call time precious and valuable, there are expressions like ‘time is money’ or ‘could you spare me some time?’ etc. This is not the only meta-

28

Comprehending Metaphors of Different Types: Evidence from Russian

phor we use for time. We can also conceptualize time as space, in particular, as a road or as some moving creature (‘the time has come’, ‘time is going too fast or too slow’). In total, these metaphors help us to speak and think about different aspects of this highly abstract concept. Not all metaphors are like that. For example, the English expression soft shoe is used to refer to a speech or a remark delivered in a conciliatory manner in order to persuade, distract, or otherwise influence someone. The comparison is transparent: conciliatory speeches make you feel comfortable, exactly as soft shoes do. However, it is not true that we in general conceptualize speeches as shoes. No other English expressions are based on this comparison. It is also notable that there is no parallel expression in Russian. One way to capture the difference between these two types of metaphors is to say that the latter fixed both on the conceptual and on the linguistic level: there is only one expression (or a couple of them) exploiting a given metaphoric comparison. The former are engrained in the speakers’ minds on the conceptual level, but many diverse expressions are based on them. We supposed that this distinction might be crucial for processing and decided to compare these two types of metaphors in our experiment. Let us analyze a couple of metaphors that we used in our stimuli. Lakoff and Johnson note that we can conceptualize our mind and our mental self as a machine or a brittle object (these concepts partly overlap because a machine is a brittle object that can break down etc.). These metaphoric comparisons stand behind the following Russian expressions used to describe people’s mental and psychical states: þuvstvovat’ sebja razbitym ‘to feel slack’, literally ‘to feel broken’, razvalivat’sja na þasti, ‘to be tired or ill’, literally ‘to fall to pieces’, slomat’sja ‘to break down’, xrupkij ‘delicate’, literally ‘brittle’, razvalina ‘wreck’, zanosit’ ‘to lose control over oneself’, literally ‘to skid’, so skripom ‘with squeaks’ (about doing something slowly and reluctantly), bez tormozov ‘without breaks’ (usually about a careless or impolite person), among many others. Syntactically, these expressions are very diverse. The examples above include verbs, nouns and adjectives, as well as prepositional phrases and verb phrases of different structure. We aimed to reflect this diversity in our experimental materials. This seems to be important, especially given the fact that most previous behavioral experiments relied only on one type of metaphors: “an x is a y”. To give an example, we chose the expression razvalivat’sja na þasti, ‘to be tired or ill’, literally ‘to fall to pieces’ from the list above for the first group of metaphors in our stimuli. The second group of metaphors included examples like raskusit’ krepkij orešek ‘to overcome a tough person’, literally ‘to bite through a hard

Natalia Cherepovskaia and Natalia Slioussar

29

nut’, and bol’šaja šiška ‘important person’, literally ‘big (tree) cone’. The latter has an English analogue, a hard nut to crack, and the former does not. Crucially, in general we do not conceptualize people as nuts or cones, so the relevant metaphoric comparisons are exploited only in a couple of expressions. In Russian, the expression krepkij orešek ‘a hard nut’ can be used by itself and combined with the verbs raskusit’ ‘to bite through’ and raskolot’ ‘to crack’, while bol’šaja šiška is the only expression based on the relevant metaphoric comparison. We chose ten metaphors of both types for our study. The aim of the experiment was to analyze how they are processed compared to the same expressions used in the literal meaning: whether they would be read faster or slower. Therefore, all target expressions were placed in two sentences setting up different contexts for them. One presupposed a literal meaning and the other a metaphorical meaning. Thus, 20 pairs of target sentences were designed. To measure reading times, they were divided into fragments (the details of the experimental procedure are explained below). There were at least three identical fragments in each pair: the one containing the metaphor or the same expression in the literal meaning (the crucial fragment, or CF), the one before that (the preceding fragment, or PF) and the one after that (the following fragment, or FF). Examples with two types of metaphors are given in (1)(2). Slashes show how they were divided into fragments. The full list of the target sentences can be found in the Appendix. 1. a. Sergej Ivanoviþ / k užasu vseɯ rodstvennikov (PF) / razvalivalsja na þasti, (CF) / prostojav tri þasa (FF) / pod prolivnym doždem. Sergei Ivanovich / to the horror of all relatives / was falling to pieces / after standing for three hours / in the pouring rain b. Ljubimyj babuškin stul / k užasu vseɯ rodstvennikov (PF) / razvalivalsja na þasti, (CF) / prostojav tri þasa (FF) / pod prolivnym doždem. the grandma's favorite chair / to the horror of all relatives / was falling to pieces / after standing for three hours / in the pouring rain 2. a. Sotrudniki policii / nikak ne mogli (PF) / raskusit’ krepkijj orešek (CF) / kak ni staralis’ (FF) / vybit’ u nego priznanie. policemen / absolutely could not / bite through a hard nut / however hard they tried / to beat a confession b. Ryžie beloþki / nikak ne mogli (PF) / raskusit’ krepkijj orešek (CF) / kak ni staralis’ (FF) / dobrat’sja do zɺrnyška.

30

Comprehending Metaphors of Different Types: Evidence from Russian

red squirrels / absolutely could not / bite through a hard nut / however hard they tried / to get to the seed Numerous reading studies show that we keep processing a portion of text for a while when our eyes have already moved to the next portion (e.g. Rayner and Duffy 1986; Rayner et al. 1989). This is known as spillover effects. In our experiment, identical PFs were needed to control for such effects in the crucial region. Identical FFs allowed us to analyze later processing stages of the crucial fragments. Before including target sentences in our experiment, we subjected them to a pretest. 16 native speakers of Russian who subsequently did not take part in the experiment rated them on two four-point scales. Firstly, they were asked whether a given sentence sounded natural to them. Secondly, they were asked whether it was easy to understand. 0 was the lowest mark, and 3 was the highest. Each informant saw only one sentence from a pair, so every sentence was rated by eight speakers. All sentences received average ratings between 2.1 and 2.9 on both scales, except for one sentence from the second group where the target expression was used in the literal meaning. Its average ratings were 1.7 and 2.0. We decided to remove this sentence and its metaphoric counterpart from the stimulus set and also excluded one pair of sentences from the first group for the balance. As a result, we had 18 pairs of target sentences in our experiment. During the experiment, each participant read one target sentence from every pair. Thus, we had two experimental lists. They included 18 target sentences and 20 fillers in a random order. In addition to that, three practice items were presented before the main experimental session started so that participants could familiarize themselves with the procedure. All fillers and practice items were approximately of the same length as the target sentences. Procedure The experiment was run on a PC using Presentation software. We used a self-paced moving window paradigm (Just et al. 1982). Each trial consisted of the participant’s silently reading a target or filler sentence and answering a comprehension question. When sentences first appeared on the screen, all words in them were masked by dashes while spaces and punctuation remained intact. The participant pressed a key to reveal a fragment of text such that each key press revealed further text and masked the previously revealed text. Presentation software allows measuring the time be-

Natalia Cherepovskaia and Natalia Slioussar

31

tween each key press with 1 ms accuracy, thus registering reading times for every fragment. After each sentence, a comprehension question was shown. Then, after the next key press, two possible answers appeared on the right and on the left of the screen, and the participant could choose by pressing the key labeled ‘left’ or ‘right’. The left-right distribution of correct and incorrect answers was randomized. Questions and answers were not masked. Then the participant pressed a key to move onto the next item. Participants were instructed to read at a natural pace and answer the questions as accurately as possible (comprehension questions are used in such studies to make sure that participants pay attention to what they are reading). Analysis We analyzed participants’ question-answering accuracy and reading times. All participants answered at least 86.8% of the questions correctly (94.7% on average). Approximately half of the incorrect answers were about filler sentences. Given the low number of relevant errors, a breakdown of reading times into correct and incorrect question trials was not done. We analyzed reading times for the crucial fragments of target sentences and for the fragments preceding and following them (only these fragments were directly comparable in metaphoric and literal conditions). Raw reading times (per fragment) were trimmed in the following way. If they exceeded a threshold of 2.5 standard deviations, by region and condition, they were adjusted to this threshold. In total, about 2.7% of the data was adjusted.

3. Results and discussion Average reading times for the crucial, preceding and following fragments of target sentences in different experimental conditions are given in Table 1. Average reading times for the crucial and following regions where the effects of experimental manipulations can be expected are also presented in Fig. 1 and Fig. 2. Let us remind that the sentences in the first group contained metaphors that are engrained in the speakers’ minds on the conceptual level, but can be expressed in different ways. The sentences in the second group contained metaphors that are fixed both on the conceptual and on the linguistic level. Analyses of variance (one-way ANOVAs) were performed on the data. Separate ANOVAs were conducted with participants and with items as random factors (F1 and F2, respectively). In the first group, there was no

32

Comprehending Metaphors of Different Types: Evidence from Russian

significant reading time difference between metaphors and literal expressions in any region (Fs < 0.6, ps > 0.4 for all comparisons). In the second group, the crucial fragments were read significantly faster in the metaphoric condition than in the literal condition (F1[1,51] = 3.95, ɪ = 0.05; F2[1,17] = 5.22, ɪ = 0.04). The difference between the following fragments was not significant, as well as the difference between preceding fragments (Fs < 0.2, ps > 0.6 for both comparisons). Preceding fragment literal metaphoric Group 1 792.9 757.7 Group 2 807.8 792.2

Crucial fragment literal metaphoric 732.5 737.7 848.2 734.7

Following fragment literal metaphoric 759.3 807.9 774.1 771.3

Table 1. Average reading times for the crucial, preceding and following fragments of target sentences (in ms)

Fig. 1. Average reading times for the sentences in the group 1 (in ms)

Natalia Cherepovskaia and Natalia Slioussar

33

Fig. 2. Average reading times for the sentences in the group 2 (in ms)

Thus, in both groups we could not find any evidence for the primacy of the literal meaning, which would suggest that to understand a metaphor, the literal meaning must be accessed and rejected first. At the same time, our findings indicate that metaphors of different types are processed differently. The fact that metaphors from the second group are read faster than the corresponding literal expressions, while the metaphors in the first group are not can be taken to indicate that the former are stored in the mental lexicon as a whole, so their meanings can be accessed directly, while the latter are processed compositionally. The facilitatory effect in the second group manifests itself at early processing stages, as expected.

4. Conclusions This paper addresses the question how metaphors are processed. Existing behavioral studies show controversial results. In some cases, metaphors were read slower than corresponding literal expressions, and this was taken to support Searle's (1979) theory and similar approaches according to which to understand a metaphor, we must assess and reject the literal

34

Comprehending Metaphors of Different Types: Evidence from Russian

meaning of the expression first. In several other studies, metaphors were processed faster, which led to the conclusion that they are stored in the mental lexicon as a whole. And some authors did not find any difference between metaphors and literal expressions. We suggest that at least some of these controversies might be due to the fact that different types of metaphors are processed differently. Our selfpaced reading experiment on Russian demonstrates that metaphors that are fixed both on the conceptual and on the linguistic level are read faster than the corresponding literal expressions. We conclude that such metaphors are stored in the mental lexicon as a whole, which facilitates their processing because their meanings can be accessed directly, rather than computed compositionally. Metaphors of a second type, which are engrained in the speakers’ minds on the conceptual level, but can be expressed in different ways, do not exhibit similar facilitatory effects. At the same time, they are not read slower than the corresponding literal expressions. Finally, studies like (Brisard et al. 2001) show that the third type of metaphors, novel metaphors that are not familiar to the readers either on the linguistic or on the conceptual level and look unusual to them, take more time to be processed than literal expressions.1 We conclude that these two types of metaphors are processed compositionally. As it seems to us, taken together, these findings also indicate that we do not access and reject literal meanings when understanding metaphors. This approach would make sense if it could be applied to all cases where metaphoric expressions are not stored in the mental lexicon as a whole. If we can compute metaphoric meanings without accessing literal meanings first—and our results with the second type of metaphors point into this direction—nothing precludes the conclusion that we always do so. How do we explain that metaphors of the third type are read slower than literal expressions? Searle’s (1979) theory and similar approaches are not the only ones that can predict that. In fact, many authors assume that literal and non-literal meanings are processed differently, but this difference is qualitative rather than quantitative and does not involve computing ones before the others (e.g. Rummelhart 1993). Results from neurophysiological and clinical studies are compatible with this view. If it is correct, we do not expect that metaphors will always be read as fast as literal ex1 It can be noted that in our experiment, metaphors in this group were read slightly slower than literal expressions. However, we do not think that this difference was similar to what Brisard et al. (2001) found. The effect they observed was early: it was significant in the region containing the target expression, while in our study, a small non-significant difference could be seen only in the following region.

Natalia Cherepovskaia and Natalia Slioussar

35

pressions or slower than they are—this may depend on additional factors such as metaphor novelty.

Appendix: Experimental materials Target sentences from the first group: 1M. Sergej Ivanoviþ / k užasu vseɯ rodstvennikov / razvalivalsja na þasti, / prostojav tri þasa / pod prolivnym doždem. Sergei Ivanovich / to the horror of all relatives / was falling to pieces / after standing for three hours / in the pouring rain 1L. Ljubimyj babuškin stul / k užasu vseɯ rodstvennikov / razvalivalsja na þasti, / prostojav tri þasa / pod prolivnym doždem. the grandma's favorite chair / to the horror of all relatives / was falling to pieces / after standing for three hours / in the pouring rain 2M. Posle svad'by / naši otnošenija / kak-to nezametno / zašli v tupik, / teper' nužno bylo / iskat' vyɯod / iz složivšejsja situacii. after the wedding / our relationship / somehow imperceptibly / hit a dead end / now it was necessary / to find a way out / from this situation 2L. Guljaja po ulicam, / Petja i Vasja / kak-to nezametno / zašli v tupik, / teper' nužno bylo / iskat' novuju dorogu. walking along the streets / Petya and Vasya / somehow imperceptibly / hit a dead end / now it was necessary / to find a new way 3M. Sily / v konce putešestvija / prišlos' ơkonomit', / þtoby ɯvatilo / na poslednij pereɯod. efforts / at the end of the journey / had to be saved / so that we had enough / for the last passage 3L. Den'gi / v konce putešestvija / prišlos' ơkonomit', / þtoby ɯvatilo / na obratnyj bilet. money / at the end of the journey / had to be saved / so that we had enough / for a return ticket 4M. Obryvok frazy, / neostorožno skazannoj / staršim synom, / gluboko ranil / Mariju Petrovnu, / i ej prišlos' / obratit'sja k psiɯologu. a fragment of the phrase / carelessly said / by the eldest son / deeply wounded / Maria Petrovna / and she had / to consult a psychologist

36

Comprehending Metaphors of Different Types: Evidence from Russian

4L.

Oskolok butylki, / neostorožno razbitoj / staršim synom, / gluboko ranil / Mariju Petrovnu, / i ej prišlos' / obratit'sja k vraþu. a fragment of the bottle / carelessly broken / by the eldest son / deeply wounded / Maria Petrovna / and she had / to consult a doctor

5M. Glavnyj argument / Andreja Alekseeviþa / byl razbit / iz-za ego neumenija / jasno izlagat' / svoi mysli. the main argument / of Andrey Alekseevich / was crushed / due to his inability / to express clearly / his ideas 5L. Bamper mašiny / Andreja Alekseeviþa / byl razbit / iz-za ego neumenija / pravil'no parkovat'sja. the bumper of the car / of Andrey Alekseevich / was crushed / due to his inability / to park correctly 6M. Uɯaživaja za Polinoj, / Pavel byl vynužden / sbavit' oboroty, / þtoby ne ispugat' / robkuju devušku / svoej nastojþhivost'ju. courting Polina / Pavel was forced / to slow down / so as not to frighten / the timid girl / with his persistence 6L. Na krutom spuske / Pavel byl vynužden / sbavit' oboroty, / þtoby ne ispugat' / eɯavšuju s nim / požiluju tetju. on the steep downhill / Pavel was forced / to slow down / so as not to frighten / his elderly aunt / riding with him 7M. Novaja teorija, / stavšaja populjarnoj, / byla postroena / na osnove nedavno obnaružennyɯ, / no mnogokratno proverennyɯ / dannyɯ. a new theory / which became popular / was built / on the basis of newly discovered, / but repeatedly checked / data 7L. Novaja krepost', / stavšaja populjarnoj, / byla postroena / na osnove nedavno obnaružennyɯ / drevniɯ fundamentov. a new fortress / which became popular / was built / on the basis of newly discovered / ancient foundations 8M. Množestvo / ɯorošiɯ idej / v ơpoɯu zastoja / umerlo tol'ko potomu, / þto im ne okazali / dolžnogo vnimanija. a lot / of good ideas / in the period of stagnation / died only because / they did not get / enough attention 8L. Množestvo / ɯorošiɯ ljudej / v ơpoɯu zastoja / umerlo tol'ko potomu, / þto im ne okazali / svoevremennoj pomošþhi. a lot / of good people / in the period of stagnation / died only because / they did not get / help in time

Natalia Cherepovskaia and Natalia Slioussar

37

9M. Sotrudnik, / prinjatyj v štat /v prošlom mesjace, / bystro slomalsja / iz-za nepreryvnoj raboty / i ušel / na bol'niþnyj. an employee / hired / last month / quickly broke down / due to working without pauses / and went away / on medical leave 9L. Televizor, / kuplennyj babuškoj/ v prošlom mesjace, / bystro slomalsja / iz-za nepreryvnoj raboty / i byl otdan / na remont. a TV set / bought by the grandma / last month / quickly broke down / due to working without pauses / and was given away / to be repaired Target sentences from the second group: 1M. Sotrudniki policii / nikak ne mogli / raskusit’ krepkijj orešek / kak ni staralis’ / vybit’ u nego priznanie. policemen / absolutely could not / bite through a hard nut / however hard they tried / to beat a confession 1L. Ryžie beloþki / nikak ne mogli / raskusit’ krepkijj orešek / kak ni staralis’ / dobrat’sja do zɺrnyška. red squirrels / absolutely could not / bite through a hard nut / however hard they tried / to get to the seed 2M. Na novoj rabote / vmesto družnogo kollektiva / Ol'ga obnaružila / zmeinoe gnezdo / i neožidanno dlja sebja / prišla v užas. at the new job / instead of a friendly team / Olga found / a nest of vipers / and unexpectedly for herself / was horrified 2L. V gustoj trave / vmesto gribov / Ol'ga obnaružila / zmeinoe gnezdo / i neožidanno dlja sebja / prišla v užas. in the thick grass / instead of mushrooms / Olga found / a nest of vipers / and unexpectedly for herself / was horrified 3M. Kogda Kolja opazdyval, / uþitel'nice matematiki / priɯodilos' slušat' / babuškiny skazki / vmesto ɨžidaemyɯ / iskrenniɯ izvinenij. when Kolya was being late / the math teacher / had to listen / to granny's tales / instead of the expected / sincere apologies 3L. Kogda otkljuþali svet, / pjatiletnej Olen'ke / priɯodilos' slušat' / babuškiny skazki / vmesto ožidaemyɯ / veþerniɯ mul'tfil'mov. when lights went off / five year old Olechka / had to listen / to granny's tales / instead of the expected / evening cartoons

38

Comprehending Metaphors of Different Types: Evidence from Russian

4M. Opytnyj psiɯolog / znaet, kak ne popast' / v bol'noe mesto, / kogda nado postavit' / pacientu diagnoz. an experienced psychologist / knows how not to hit / a sore spot / when (she) has to diagnose a patient (literally: when (she) has to put / to a patient a diagnosis) 4L. Opytnaja medsestra / znaet, kak ne popast' / v bol'noe mesto, / kogda nado postavit' / pacientu kapel'nicu. an experienced nurse / knows how not to hit / a sore spot / when (she) has to put / a patient on a drip 5M. V štabe bastujušþiɯ raboþiɯ / k vþheru pojavilas' / bol'šaja šiška, / i srazu stalo ponjatno, / þto iɯ trebovanija / budut uþteny. at the headquarters of the workers on strike / in the evening there appeared / a big strobile/bump / and it became clear at once / that their demands / will be considered 5L. Na lbu u Vani / k veþeru pojavilas' / bol'šaja šiška, / i srazu stalo ponjatno, / þto on ne skazal mame / vsej pravdy. on Vanya’s forehead / in the evening there appeared / a big strobile/bump / and it became clear at once / that he did not tell / the whole truth to his mother 6M. Uþastniki konferencii / bezžalostno metali / jadovitye strely / v ocepenevšego ot užasa / neopytnogo dokladþika. conference participants / mercilessly threw / poisonous arrows / in a petrified with horror / inexperienced speaker 6L. Na znamenitoj kartine / drevnie oɯotniki / bezžalostno metali / jadovitye strely / v ocepenevšego ot užasa / ogromnogo mamonta. in the famous picture / ancient hunters / mercilessly threw / poisonous arrows / in a petrified with horror / huge mammoth 7M. Kogda novyj proekt Niny / poterpel neudaþu, / Polina pervoj / brosila v nee kamen' / i, ne razdumyvaja, / potrebovala eɺ uvolit'. when Nina’s new project / failed / Polina was the first / to throw a stone at her / and without hesitation / demanded her to be dismissed 7L. Kogda sobaka / kinulas' na rebjat, / Polina pervoj / brosila v nee kamen' / i, ne razdumyvaja, / vstala meždu nej / i malyšami. when the dog / attacked the children / Polina was the first / to throw a stone at it (literally: at her) / and without hesitation / stepped between it / and the kids

Natalia Cherepovskaia and Natalia Slioussar

39

8M. Korrupcija i bjurokratija / v našej strane / bystro sozdali / blagodatnuju poþvu, / ideal'no podɯodjašþuju / dlja rosta prestupnosti. corruption and bureaucracy / in our country / quickly created / a fertile ground / ideally suited / for the growth of crime 8L. Ispol'zuja udobrenija, / južnye fermery / bystro sozdali / blagodatnuju poþvu, / ideal'no podɯodjašþuju / dlja vyrašþivanija kukuruzy. using fertilizers / southern farmers / quickly created / a fertile ground / ideally suited / for growing corn 9M. Byvšim suprugam / niþego ne ostalos', / kak sžeþ' korabli, / þtoby v budušhem izbežat' / utomljajušþiɯ iɯ / ssor i konfliktov. former spouses / had nothing left / but to burn the ships / to avoid in the future / tiresome / quarrels and conflicts 9L. Posle poraženija / vice-admiralu / niþego ne ostalos', / kak sžeþ' korabli, / þtoby v budušþem izbežat' / zaɯvata flota vragom. after the defeat / vice admiral / had nothing left / but to burn the ships / to avoid in the future / the capture of the navy by the enemy

Acknowledgments The study was supported in part by the grant 14-04-00586 from the Russian Foundation for Humanities.

References Black, Max. 1962. Metaphor. In Models and metaphors, ed. Max Black, 25-47. Ithaca, New York: Cornell University Press. —. 1993. More about metaphor. In Metaphor and thought, ed. Andrew Ortony, 19-41. Cambridge: Cambridge University Press. Bottini, Gabriella, Rhiannon Corcoran, Roberto Sterzi, Eraldo Paulesu, Pietro Schenone, Pina Scarpa, Richard S.J. Frackowiak, and Christopher D. Frith. 1994. The role of the right hemisphere in the interpretation of figurative aspects of language: A positron emission tomography activation study. Brain 117: 1241-1253. Bowdle, Brain F., and Dedre Gentner. 2005. The career of metaphor. Psychological Review 112: 193-216. Brisard, Frank, Steven Frisson, and Dominiek Sandra. 2001. Processing unfamiliar metaphors in a self-paced reading task. Metaphor and Symbol 16: 87-108.

40

Comprehending Metaphors of Different Types: Evidence from Russian

Champagne-Lavau, Maud, and Emmanuel Stip. 2010. Pragmatic and executive dysfunction in schizophrenia. Journal of Neurolinguistics 23: 285-296. Eviatar, Zohar, and Marcel Adam Just. 2006. Brain correlates of discourse processing: An fMRI investigation of irony and conventional metaphor comprehension. Neuropsychologia 44: 2348-2359. Gentner, Dedre, and Phillip Wolff. 1997. Alignment in the processing of metaphor. Journal of Memory and Language 37: 331-355. Gildea, Patricia, and Sam Glucksberg. 1983. On understanding metaphor: The role of context. Journal of Verbal Learning and Verbal Behavior 22: 577-590. Giora, Rachel. 2007. Is metaphor special? Brain and Language 100: 111114. —. 2003. On our mind: Salience, context, and figurative language. New York: Oxford University Press. —. 1999. On the priority of salient meanings: studies of literal and figurative language. Journal of Pragmatics 31: 919-929. Glucksberg, Sam. 2003. The psycholinguistics of metaphor. Trends in Cognitive Sciences 7: 92-96. Glucksberg, Sam, and Boaz Keysar. 1990. Understanding metaphorical comparisons: Beyond similarity. Psychological Review 97: 3-18. Glucksberg, Sam, Matthew S. McGlone, and Deanna Manfredi 1997. Property attribution in metaphor comprehension. Journal of Memory and Language 36: 50-67. Iakimova, Galina, Christine Passerieux, and Marie-Christine Hardy-Baylé. 2006. La compréhension des métaphores dans la schizophrénie et la dépression. Une approche expérimentale. L'Encéphale 32 : 995-1002. Jones, Lara L., and Zachary Estes. 2006. Roosters, robins, and alarm clocks: Aptness and conventionality in metaphor comprehension. Journal of Memory and Language 55: 18-32. Just, Marcel Adam, Patricia A. Carpenter, and Jacqueline D. Woolley. 1982. Paradigms and processes in reading comprehension. Journal of Experimental Psychology: General 3: 228-238. Lai, Vicky Tzuyin, Tim Curran, and Lise Menn. 2009. Comprehending conventional and novel metaphors: An ERP study. Brain Research 1284: 145-155. Lakoff, George, and Mark Johnson. 1980. Metaphors we live by. Chicago: University of Chicago Press. Lakoff, George, and Mark Johnson. 1999. Philosophy in the flesh: The embodied mind and its challenge to western thought. New York: Basic Books.

Natalia Cherepovskaia and Natalia Slioussar

41

Lakoff, George. The contemporary theory of metaphor. In Metaphor and thought, ed. Andrew Ortony, 202-251. Cambridge: Cambridge University Press. Mashal, Nira, Miriam Faust, Talma Hendler, and Mark Jung-Beeman. 2007. An fMRI investigation of the neural correlates underlying the processing of novel metaphoric expression. Brain and Language 100: 115-126. McElree, Brain, and Johanna Nordline. 1999. Literal and figurative interpretations are competed in equal time. Psychonomic Bulletin and Review 6: 486-494. McGlone, Matthew S., and Deanna A. Manfredi. 2001. Topic-vehicle interaction in metaphor comprehension. Memory and Cognition 29: 1209-1219. Ortony, Andrew, Diane L. Schallert, Ralph E. Reynolds, and Stephen J. Antos. 1978. Interpreting metaphors and idioms: Some effects of context on comprehension. Journal of Verbal Learning and Verbal Behavior 17: 465-477. Pynte, Joel, Mireille Besson, Farice-Henri Robichon, and Jézabel Poli. 1996. The time-course of metaphor comprehension: An event-related potential study. Brain and Language 55: 293-316. Rayner, Keith, and Susan A. Duffy. 1986. Lexical complexity and fixation times in reading: Effects of word frequency, verb complexity, and lexical ambiguity. Memory and Cognition 14: 191-201. Rayner, Keith, Sara C. Sereno, Robin K. Morris, A. Réne Schmauder, and Charles Clifton. 1989. Eye movements and on-line language comprehension processes. Language and Cognition Processes 4: 21- 49. Rumelhart, David Everett. 1993. Some problems with the notion of literal meanings. In Metaphor and thought, ed. Andrew Ortony, 71-82. Cambridge: Cambridge University Press. Searle, John. 1979. Expression and meaning. Cambridge: Cambridge University Press. —. 1993. Metaphor. In Metaphor and thought, ed. Andrew Ortony, 83111. Cambridge: Cambridge University Press. Yang, Fanpei Gloria, Jennifer Edens, Claire Simpson, and Daniel C. Krawczyk. 2009. Differences in task demands influence the hemispheric lateralization and neural correlates of metaphor. Brain and Language 111: 114-124.

RESOLVING TIP-OF-THE-TONGUE STATES WITH SYLLABLE CUES NINA JEANETTE HOFFERBERTH-SAUER AND LISE ABRAMS1

Abstract Tip-of-the-tongue (TOT) states represent a speaker's temporary and typically frustrating inability to retrieve a known word. Although TOTs are a type of failed word retrieval, they are useful for understanding the processes that underlie successful speech production. Specifically, TOT states are thought to result from weakened connections between a word's lexical representation (lemma) and its phonology (lexeme), a hypothesis supported by research showing that encountering phonologically-related cues during a TOT, specifically words containing the first syllable, helps to resolve the TOT. The chapter begins with a discussion of models of speech production, the locus of TOT states within these models, and various methodologies for investigating TOT states in laboratory studies. We then review previous research on syllable cueing of TOT resolution and present findings from a new experiment using a syllable in isolation as the cue. The chapter concludes by discussing the implications of these findings, including the theoretical significance of the syllable in resolving TOTs.

1

Nina Jeanette Hofferberth-Sauer Goethe-University Frankfurt, Germany E-Mail: [email protected] Lise Abrams University of Florida, USA E-Mail: [email protected]

44

Resolving Tip-of-the-Tongue States with Syllable Cues

1. Theoretical background 1.1 The tip-of-the-tongue phenomenon as a breakdown in phonological encoding The tip-of-the-tongue (TOT) phenomenon is a word finding problem, a common and universal experience when a speaker cannot immediately recall a well-known or familiar word but has the certitude that the information exists in one’s knowledge base. TOTs are reported to occur about once a week in everyday life and on 10-20% of items in laboratory studies (Brown, 2012, p. 195). TOTs are an important source of information concerning the nature of the processes and architecture of the speech production system (James & Burke, 2000, p. 1378). TOTs show us what happens when lexical retrieval fails. Models of speech production largely agree that the retrieval of a word first requires the activation of a concept to be expressed, followed by access to its semantic and syntactic properties (lemma), and lastly the activation of its phonological components (lexeme) (Dell, 1986; Dell, Chang, & Griffin, 1999; Dell & O’Seaghdha, 1991, 1992; Garrett, 1975, 1990; Levelt, 1989; Levelt, Roelofs, & Meyer, 1999; Peterson & Savoy, 1998; but see Caramazza, 1997, for a model without the existence of a lemma level). In the absence of successful retrieval, various aspects of the inaccessible target word are still frequently available in a TOT state: Speakers have a strong feeling of knowing the word, have access to semantic and sometimes syntactic information, and often have partial access to its phonological properties. For example, the speaker might have access to the word’s syntax, in particular the grammatical gender of a noun (for gender-marking languages). This prediction was demonstrated with an Italian anomic patient called Dante, who could hardly name any pictures but knew in 95% - 98% of the cases the correct gender of the target name (Caramazza & Miozzo, 1997). Similar results have been reported with healthy, native Italian speakers (Miozzo & Caramazza, 1997; Vigliocco, Antonini, & Garrett, 1997). Although speakers cannot retrieve the complete phonological form of the target word, they are often able to retrieve the first phoneme or letter, the first syllable and number of syllables, other letters or phonemes, and also the stress pattern of the target word (Brown & McNeill, 1966; for a replication of Brown & McNeill in German, see Hofferberth, 2011). TOTs provide evidence for the existence of distinct stages in retrieving a word's lemma and lexeme during speech production: they demonstrate a breakdown between these two processes (Levelt, 1989).

Nina Jeanette Hofferberth-Sauer and Lise Abrams

45

While models of speech production largely agree on the general distinction between semantic-syntactic (lemma) and word form (lexeme) retrieval processes, they differ with regard to the precise architecture of the system.2 According to the modular or discrete-serial model (Levelt, 1989; Levelt et al., 1999), the speech production process is staged: Phonological encoding starts after lemma selection is completed, i.e., after the final lexical candidate has been selected. The activation is unidirectional, i.e., the lexeme cannot intervene in the lemma activation process. Levelt’s model was revised and updated ten years later (Levelt et al., 1999) and now allows multiple lemma selection in the case of near-synonyms (as a reaction to Peterson & Savoy, 1998, see below). Furthermore, a mental syllabary, which is a repository of frequently used phonetic syllables, was added to the model. Statistics show that native speakers of English or Dutch perform 80% of their oral production with no more than about 500 different syllables, although these languages have more than 10,000 different syllables. The syllabary reposits such overused, high-frequency syllabic gestures and allows fast retrieval of ready-made motor programmes for these syllables. In contrast, low-frequency syllables are formed online according to the segmental and metrical rules of the language. The advantage of a syllabary is that it reduces the programming load relative to segment-by-segment composition of phonetic forms, in particular because the syllables of a language differ greatly in frequency (Levelt et al., 1999, p. 32). The cascade model of lexical access by Peterson and Savoy (1998) similarly proposes unidirectional activation, but allows for phonological encoding to begin before completing lemma selection under some circumstances. Peterson and Savoy (1998) as well as Jescheniak and Schriefers (1998) presented evidence showing that in the case of nearsynonymy (e.g., a picture of a sofa/couch), both the eventually produced name of the picture (e.g., couch) as well as the near-synonym (e.g., sofa) are phonologically encoded, allowing the process of lexeme retrieval to be initiated before the lemma selection process has finished. On the other hand, according to connectionist or so-called interactiveactivation models (Dell, 1986; Dell et al., 1999; Dell & O’Seaghdha, 1991, 1992), there is bidirectional activation with a permanent interaction between the lemma and the lexeme level. As soon as the conceptual level 2

The terms lemma and lexeme were introduced by Kempen and Huijbers (1983) or rather Kempen and Hoenkamp (1987). In the model of Levelt et al. (1999), the lemma only enables access to the syntactic information while the semantic information is stored in so-called lexical concepts (Hantsch, 2002, p. 12). The term lexeme is used in this paper synonymously with word form.

46

Resolving Tip-of-the-Tongue States with Syllable Cues

has been activated, lemmas whose conceptual specifications match the conceptual structure become activated and spread activation to their corresponding lexemes. As a result, even lemmas that partly match the conceptual structures in the pre-verbal message will activate its phonological representation to some extent. Unlike stage models, activation can feed back from the lexeme to the lemma level. This potential for feedback activation has been critical for challenging discreteserial models' inability to explain various phenomena, such as the statistical overrepresentation of mixed errors. When a speaker produces a mixed error, he/she replaces the intended word with a semantically and phonologically related word (e.g., rat instead of cat). Semantically and phonologically related words receive a combination of feedforward activation from the conceptual level and feedback activation from the phonological level. Therefore, rat (instead of cat, semantically and phonologically related) will be more strongly activated than dog (semantically related) or mat (phonologically related). Therefore, the chance to produce a mixed error is higher than that of producing a semantic or phonological error (Leuninger 1993, 1996).

1.2 Etiology Despite differences between the speech production models described above, the locus of TOTs is similar. A lexical node in the semantic system (lemma) becomes activated, but there is insufficient activation transmitted to its associated phonological nodes (lexeme) to enable the word's retrieval. What causes this breakdown between lemma and lexeme retrieval? Two main hypotheses have been suggested: (1) The blocking hypothesis, where TOT states result from the presence of interlopers (Jones, 1987; Jones & Langford, 1987), and (2) the incomplete activation hypothesis, where TOTs result from insufficient activation of the to-beretrieved word (Meyer & Bock, 1992). In the blocking hypothesis, another word can come to mind that prevents the intended word from being retrieved, resulting in a TOT. Interlopers can result from partial phonological information that is shared between the target and another word, creating competition. For example, given the cue Aida Composer, participants may recover the initial letter V, the final letter I, and know that it is a short Italian name and could then recall Vivaldi instead of Verdi (Maril, Wagner, & Schacter, 2001, p. 658). One type of incomplete activation hypothesis is the transmission deficit hypothesis (TDH; Burke, MacKay, Worthley, & Wade, 1991; MacKay & Burke, 1990), a corollary of node structure theory (MacKay,

Nina Jeanette Hofferberth-Sauer and Lise Abrams

47

1987), which arranges nodes (conceptual representations) into a hierarchical network of multilevel systems, including semantic, syntactic, and phonological systems. Activation of nodes (and subsequent retrieval of the information it contains) is dependent on the most-primed-wins principle: When nodes in the same level receive simultaneous priming, only the node receiving the most priming can be activated. The TDH defines the locus of incomplete activation as a failure to fully transmit activation from the lexical to the phonological nodes as a result of weakened connections between these nodes. The frequent availability of partial target word information in a TOT state supports this view. Connections are thought to weaken as a function of frequency of use (words not used frequently are more susceptible to TOTs), recency of use (less recently-used words are more likely to have TOTs), and the normal aging process (more TOTs occur with increasing age).

1.3 Eliciting and measuring TOTs In the 1930s, Wenzl (1932; 1936) and Woodworth (1934) collected naturally-occurring TOTs in diary studies. About 30 years later, the first systematic laboratory study was undertaken by Brown and McNeill (1966). They read out definition-like questions of low-frequency words to elicit TOTs, and participants had to write down the information they could retrieve while experiencing a TOT. The results showed that participants often had access to partial information of the target word, such as letters in the word, the number of syllables, and words with similar sound and/or similar meaning. Brown and McNeill (1966) also found that the degree of orthographic overlap between words given by the participants and the targets was highest for the beginning of the word, intermediate for the ending, and lowest for the middle part, resulting in a U-shaped serial position function (p. 330f.). Furthermore, participants not only provided the initial letter of the target word but occasionally more than one letter, e.g., ex (for extort) or con (for convene). Brown and McNeill (1966) did not make reference to the significance of producing the first syllable in these cases, but they suggested that this could arise because “some letter (or phoneme) sequences are stored as single entries having been ‘chunked’ by long experience“ (p. 331). Many studies attempt to induce TOTs by presenting a specific definition of a target word, such as “A navigational instrument used in measuring angular distances, especially the altitude of sun, moon, and stars at sea" for sextant (Brown & McNeill, 1966, p. 333). If participants are able to retrieve the target, they say it aloud or write it down. When

48

Resolving Tip-of-the-Tongue States with Syllable Cues

participants cannot retrieve the target, they are asked to indicate if they do not know the word or are having a TOT, where the difference between these responses is defined in terms of the imminence of recall (Brown & McNeill, 1966, p. 325). Ways of defining TOTs have varied dramatically across studies. Some researchers define TOTs as equivalent to long or extended retrieval, i.e., a word that does not come to mind immediately must be a TOT. For example, Hamberger and Seidel (2003) defined a TOT as taking at least two seconds, using Goodglass, Theurkauf, and Wingfield’s (1984) assertion that normal retrieval consists of two stages: automatic (up to 1.5s) and effortful (after 1.5s). Vigliocco et al. (1997) defined TOTs in terms of the availability of partial information, where TOTs were only counted as such when participants could provide partial information. One criticism of TOT research is that it is difficult to measure TOT states objectively. In most lab studies, TOTs are measured subjectively, i.e., a participant indicates having a TOT, and the experimenter has to assume that participants have accurate access to and reporting of their metacognitions. Furthermore, TOT incidence can be influenced by various factors unrelated to the words themselves, such as the wording of instructions. When participants are instructed that the target words are difficult (“95% of students had great difficulty answering them”), participants indicate DON'T KNOW more often than when the target words are indicated as easy (“95% of students had little difficulty answering them”), which increases participants' likelihood of reporting a TOT state (Harley & Bown, 1998). In some studies, clarifications are added to make sure that participants don’t report a TOT simply because the target word is from a familiar area in which the word should be known (Ravizza, 2003). It can be difficult to compare TOT studies because they do not use the same instructions or target words, and they also differ in the design. As noted above, it makes an immense difference how to define TOTs and what to count as a TOT. Furthermore, TOT incidence can vary broadly both across items and participants, and create a “fragmentary data problem” (Brown & McNeill, p. 328). In many studies, there are participants who never experience TOTs, which can result in a “hidden subject-selection bias” (Brown, 2012, p. 54) because participants who did not report any TOTs are excluded from the analyses. Similarly, a small number of words can be overrepresented in TOT analyses given that not all words elicit TOTs (Brown 2012, p. 54f.). Providing information about target words, their definitions, and TOT rate per target word can help other researchers. For example, Abrams, Trunk, and Margolin (2007) summarized

Nina Jeanette Hofferberth-Sauer and Lise Abrams

49

TOT incidence for 163 target words, collected in five experiments (Abrams, White, & Eitel, 2003; Abrams & Rodriguez, 2005; Robinson, Abrams, & Bahrick, 2004; White & Abrams, 2002).

1.4 TOT resolution TOT resolution can be achieved through external search strategies (such as looking up the word online, in a dictionary, or by asking someone) or through internal strategies (such as mentally going through the alphabet or generating similar words). Laboratory studies often involve cueing procedures with a cue-target relationship that is not readily obvious, yet efficient to boost activation of the target word and assist TOT resolution. The terminology in the literature sometimes differs between the terms primes and cues. We use the term cue here as defined by Brown (2012), where “primes precede and cues follow the initiation of a target word search procedure” (p. 67). Cueing designs can be used to either manipulate TOT incidence (the probability of a TOT occurring) or influence TOT resolution (retrieving the intended word after reporting a TOT). Some studies have shown that phonologically-related cue words (in comparison to semantically-related cue words) increase TOT incidence, when the cue word was presented immediately after the definition (Jones, 1989; Jones & Langford, 1987; Maylor, 1990; but see Meyer & Bock, 1992; and Perfect & Hanley, 1992, for critiques on Jones’ stimuli). In contrast, the effect of phonological cues on TOT resolution is generally facilitatory. Phonological cues boost activation to the target words during TOTs relative to unrelated words (Abrams et al., 2003; Abrams & Rodriguez, 2005; Abrams, Trunk, & Merrill, 2007; Burke, MacKay, Worthley, & Wade, 1991; Farrell & Abrams, 2011; Harley & Bown 1998; Heine, Ober, & Shenaut, 1999; James & Burke 2000; Meyer & Bock, 1992; Rastle & Burke, 1996; White & Abrams, 2002). A phonologically-related cue is defined differently in these studies in terms of the degree of phonological overlap, including phonological neighbours, orthographic or phonological syllables within cue words, overlapping initial sound and/or letter, and shared number of syllables or stress pattern. For example, the phonological cues used by Meyer and Bock (1992) were described as having “the same initial sound and letter, the same number of syllables, and the same stress pattern as did the corresponding targets but were unrelated to them in meaning“ (p. 717). However, the items themselves varied widely, with some cues sharing only the initial sound and letter, while others shared the entire first syllable. Furthermore, while the majority of TOT studies did not control

50

Resolving Tip-of-the-Tongue States with Syllable Cues

the grammatical class of the cue, Abrams and Rodriguez (2005) demonstrated that the likelihood of phonological cues increasing TOT resolution depends on whether the target and cue shared grammatical class (see below). Thus, more precisely defining phonological cues will make it easier to compare findings across studies, and we focus on a definition where phonological overlap includes the initial syllable of the target word.

2. Cuing studies 2.1 Studies with syllable cues in word lists The general methodology of these cueing studies involves presentation of general knowledge questions or definitions whose answers correspond to a target word. Participants are asked to indicate whether they know the target (and to produce it if so), do not know it, or are having a TOT. After TOT responses, a list of words is presented one at a time, where either all of the words are unrelated to the target, or some are phonologically related by containing specific syllables of the target word. Participants are asked to read each word and sometimes make some type of judgment about it. After the word list has been presented, the original TOT-inducing question is shown again to see if the target word can now be retrieved. Cueing is measured as the difference in TOT resolution following a list containing phonologically-related words relative to a list with solely unrelated words. Following all questions, a multiple-choice recognition test is given for unresolved TOT questions so that only "correct" TOTs, i.e., target words that participants correctly recognize, are included in analysis. James and Burke (2000, Experiment 2) were the first to use syllables to cue TOT resolution, showing that both younger (M = 19 years) and older adults (M = 72 years) were more likely to resolve TOTs when phonologically-related words that cumulatively contained the target's syllables were presented during a TOT, relative to unrelated words. For example, when having a TOT for the target abdicate, reading a list that contained the words indigent, abstract, truncate, tradition, and locate (syllables italicized here for emphasis) increased TOT resolution. White and Abrams (2002; see also Abrams et al., 2003, Experiment 2) further highlighted the significance of the syllable unit by showing that the locus of phonological cueing effects on TOT resolution is the target's first syllable. Using a list where three of the words contained only one of the target's syllables, either the first (ab), middle (di), or last syllable (cate), they found that both younger (M = 20 years) and young-old adults (M = 67 years) experienced greater TOT resolution following lists containing first-

Figure 1. Experimental procedure used in White and Abrams (2002).

Nina Jeanette Hofferberth-Sauer and Lise Abrams 51

52

Resolving Tip-of-the-Tongue States with Syllable Cues

syllable cues than unrelated lists, whereas resolution was unaffected by middle- and last-syllable cues.3 See Figure 1 for an illustration of the procedure. An even older group, old-old adults, M = 77, did not show firstsyllable cueing. Abrams et al. (2003) demonstrated that the benefits to younger adults' TOT resolution were unique to a full first syllable, as cues containing the target's initial letter or initial phoneme (but not syllable) had no effect on TOT resolution. Abrams and Rodriguez (2005; see also Abrams, Trunk, & Merrill, 2007) qualified the previously demonstrated first-syllable cueing effect on TOT resolution in younger adults by illustrating the relevance of a phonological cue's grammatical class. Targets (e.g., rosary, a noun) were cued with a list, where one of the words contained the target's first syllable and either shared (robot, a noun) or did not share the target's grammatical class (robust, an adjective). TOT resolution was increased following lists containing a phonological cue in a different grammatical class relative to a list of unrelated words, replicating previous findings of a first-syllable cueing effect. However, phonologically-related cues in the same grammatical class had no effect on TOT resolution. Abrams et al. (2007) replicated these findings and extended the finding of first-syllable cueing on TOT resolution to young-old adults (M = 69 years), who showed slightly reduced first-syllable cueing from phonological cues in a different grammatical class compared to younger adults. In contrast, old-old adults (M = 80 years) did not show significant first-syllable cueing from phonological cues in a different grammatical class and instead showed an inhibitory effect on TOT resolution following cues in the same grammatical class. Together, these findings support models of speech production where phonologically-related words transmit activation to the weakened phonological representations that are thought to cause TOTs, increasing the likelihood of retrieving the target. Access to the initial syllable is necessary for TOT resolution to occur, as the initial phoneme does not transmit sufficient activation to enable target retrieval. However, several variables, specifically grammatical class and aging, mediate the relationship between phonological cueing and resolution of TOT states. Grammatical class functions to activate appropriate candidates for production, and phonological cues can only facilitate TOT resolution when the cue itself is not a potential competitor for retrieval. The normal aging process exacerbates the weakening of phonological representations so that 3

Word lists were also shown after DON’T KNOW and KNOW responses to minimize participants' awareness of the relation between the phonological cues and targets.

Nina Jeanette Hofferberth-Sauer and Lise Abrams

53

cues that help to resolve TOTs at earlier ages do not help (and can even hurt retrieval) in the later stages of aging, suggesting that additional phonological input may be necessary to resolve TOTs as we age.

2.2 Studies with syllable cues in pseudowords Pseudoword cues offer a vehicle for presenting a TOT target’s first syllable in the appropriate position independent of embedding that phonology in an existing lexical representation. Abrams, White, Merrill, and Hausler (2007) investigated whether phonological pseudoword cues, defined as multisyllabic, pronounceable strings of letters that were not real English words and contained the target's first syllable, could help to resolve TOT states. Cues were presented auditorily among a list containing three words and three pseudowords, to which participants made a lexical decision judgment (said 'yes' or 'no' as to whether the item was a real English word). In Experiment 1, phonological cues were assigned a suffix unassociated with a specific grammatical class. The results showed significant first-syllable cueing, where TOT resolution following lists with a phonologically-related pseudoword was greater than resolution following lists containing an unrelated pseudoword. In Experiment 2, phonological cues were given a suffix that was commonly associated with a particular grammatical class, which was either the same as or different from the target's part of speech. As seen previously with word cues, the results showed that more TOTs were resolved following a pseudoword cue in a different grammatical class relative to an unrelated pseudoword, demonstrating significant cueing of TOT resolution. In contrast, TOT resolution following a pseudoword cue in the same grammatical class and an unrelated pseudoword was equivalent. Pureza, Soares, and Comesana (2012) elicited TOTs in European Portuguese speakers using a picture naming task, after which a list of 14 words and pseudowords were visually presented for lexical decisions. Four of the pseudowords were pseudohomophone cues, containing either the first syllable, last syllable, or no syllable of the target word in order to manipulate the syllabic position. After the lexical decision task, the TOTinducing picture was shown again, and resolution was assessed. Although a different methodology from previous studies, the results demonstrated a significant syllabic pseudohomophone cueing effect on TOT resolution. Following TOT responses, seeing pseudohomophone cues increased target retrieval, especially for longer words (four-syllable words in comparison to two- and three-syllable words). The authors also report a posteriori analysis, which “seems to show that the positional syllable frequency has

54

Resolving Tip-of-the-Tongue States with Syllable Cues

an effect especially for the first position” (p. 13), as was found by White and Abrams (2002) and Abrams et al. (2003). These results with pseudoword phonological cues demonstrate that while encountering the initial syllable during a TOT is essential to resolving the TOT as shown in previous research, the syllable does not need to be embedded within a real word. When pseudowords possess sufficient phonological and appropriate grammatical information, they are able to influence the subsequent activation of the target’s phonology for resolving TOTs, suggesting that the basis of phonological cueing of TOT resolution at least begins from sublexical components. Furthermore, firstsyllable phonological cueing of TOT resolution is not limited to English, showing that it also occurs in languages with well-defined syllable boundaries, such as European Portuguese.

2.3 Studies with syllable cues presented individually In this chapter, we present new data using a reaction-time (RT) experiment to determine whether the correct first syllable of the target word facilitates TOT resolution and whether a first syllable different from the target's first syllable but with matched frequency inhibits TOT resolution (HofferberthSauer, in preparation). In contrast to the studies of Abrams et al. (2003), the cue consisted only of the first syllable alone and was not embedded within a word that shared the same first syllable as the target. Furthermore, RTs, which have to our knowledge never previously been reported in TOT research with phonological cueing, allow for a more objective measure of TOTs, one that is not dependent on participants' self-report. Forty-eight under- and postgraduates between 21 and 35 years (30 female, 18 male, M = 29.5 years, SD = 3.7) from Heinrich-Heine-University Dusseldorf participated in this study. They were native speakers of German and were paid for their services. 2.3.1. Procedure The definitions (in German) were presented on a computer screen using the program Presentation. The definitions were selected from two pilot studies that were conducted to determine the definitions that induced more TOT states and had a higher rate of name agreement, in order to minimize the occurrence of TOTs for a word different from the intended target (Hofferberth, 2012). Out of a pool of 353 definitions, 240 definitions were presented to induce TOTs. The frequencies of the German target words as well as the frequencies of the first syllables of these nouns were taken from

Nina Jeanette Hofferberth-Sauer and Lise Abrams

55

Figure 2. Experimental procedure used in Hofferberth (2012). While the experiment was conducted in German, an example translated into English is shown here for illustration purposes and for easier comparison with other experiments conducted in English.

the DLEX database. The frequencies of the syllables were needed in order to match the correct and incorrect syllables. The correct syllable is the one that fits the target word, whereas the incorrect syllable was taken from another target word in the experiment with matched frequency. Participants read definitions on a computer screen and pushed a button to indicate KNOW, DON’T KNOW or, TOT, respectively. When in a TOT state, a written cue was presented for 25 seconds that was either the correct first syllable of the target word, an incorrect syllable with matched frequency as the fitting syllable, or a neutral baseline condition consisting

56

Resolving Tip-of-the-Tongue States with Syllable Cues

of a row of Xs, "xxx". The experimental procedure is illustrated in Figure 2. For example, the definition for the target meteorite was “A mass of stone or metal that has reached the earth from outer space”. When the participant indicated to be in a TOT state, the correct first syllable me, the incorrect first syllable sa, or the control condition xxx appeared on the screen. If the TOT state was not resolved within 25 seconds, a question appeared on the screen as a general manipulation check to assess whether participants were correctly reporting TOTs. Participants were asked “Is the following word the word you were looking for?”, together with a word that was the target word 50% of the time and a word semantically but not phonologically related to the target word the other 50% of the time. No participant answered the question incorrectly (incorrect endorsement and incorrect rejection) more that 60% of the time (M = 48.2%, SD = 11.7%). 2.3.2. Results 2.3.2.1 Reaction Times The RTs differed between KNOW, DON’T KNOW, and TOT, F(2, 94) = 41.9, p < .001, as seen in figure 3. The RTs to report KNOW (M = 5265 ms, SD = 1085 ms) were faster than DON’T KNOW responses (M = 5923 ms, SD = 1311 ms), t(47) = 4.49, p < .01, which in turn were faster than TOT responses (M = 6474 ms, SD = 1122 ms), t(47) = 4.34, p < .01. With respect to the influence of a cue, RTs to report KNOW after the presentation of one of the cues in the TOT state were faster with the correct cue (M = 4055 ms, SD = 1689 ms) in comparison to the incorrect cue (M = 7807 ms, SD = 3984 ms), t(47) = 6.92, p < .01, or the control condition (M = 7559 ms, SD = 2582 ms), t(46) = 10.42, p < .01, with no difference between the latter two conditions, t(46) =.40, p = .69. 2.3.2.2 TOT Incidence and Resolution Rates The TOT rate was 20.2% (= 2,326 TOTs out of overall 11,520 stimuli). After the cue was presented, 862 (37.1%) were accurately resolved TOTs where the answer was consistent with the target word, 354 (15.2%) were inaccurately resolved TOTs where the answer differed from the target word, and 1,110 (47.7%) of the TOTs remained unresolved after 25 seconds.

Nina Jeanette Hofferberth-Sauer and Lise Abrams

57

ȗȗȗ

ͺͲͲͲ

ȗȗȗ

͹ͲͲͲ

ȗȗȗ

͸ͲͲͲ ͷͲͲͲ ͶͲͲͲ ͵ͲͲͲ ʹͲͲͲ ͳͲͲͲ Ͳ ‘™

‘Ʈ–‘™



Figure 3. Reaction times to definition. X-axis: Response type, y-axis: Mean RTs (in ms). Error bars represent +/- 1 SD; *** = p < .001

2.3.2.3 Effects of Cue on TOT Resolution4 The number of accurate TOT resolutions differed between the three types of cues, F(2, 94) = 260.6, p < .001. When given the correct first syllable, TOTs were accurately resolved more often (M = 73.5%, SD = 18.6%) in comparison to the control condition (M = 24.3%, SD = 16.4%), t(47) = 16.4, p < .001, which had more resolved TOTs than an incorrect syllable (M = 16.0%, SD = 13.6%), t(47) = 3.7, p = .001 (see Figure 4).

4

The items on which participants experienced TOTs could not be determined ahead of time. So, the number of TOTs per person differed, and it was impossible to give each participant an equal number of the three cue types (correct syllable, incorrect syllable, no cue). To adjust for these differing numbers of cues, a percentage of TOT resolution was computed per participant, where the number of resolutions was divided by the number of TOTs that occurred in each cue condition.

Resolving Tip-of-the-Tongue States with Syllable Cues

58

100

***

90

***

80

***

70 60 50 40 30 20 10 0 Correct cue

Incorrect cue

Control (xxx)

Figure 4. Accurate TOT resolutions. X-axis: Cue type, y-axis: Mean accurate TOT resolution (in %) Error bars represent +/- 1 SD; *** = p < .001

The number of inaccurate TOT resolutions also differed between the three types of cues, F(2, 94) = 17.6, p < .001. Fewer TOTs were inaccurately resolved when given the correct first syllable (M = 9.2%, SD = 7.8%) in comparison to an incorrect syllable (M = 15.8%, SD = 14.4%), t(47) = 3.6, p = .001, which in turn had fewer inaccurate TOT resolutions than the control condition (M = 20.2%, SD = 14.4%), t(47) = 2.4, p = .023 (see Figure 5).

Nina Jeanette Hofferberth-Sauer and Lise Abrams

100

59

***

90

***

80

*

70 60 50 40 30 20 10 0 Correct cue

Incorrect cue

Control (xxx)

Figure 5. Inaccurate TOT resolutions. X-axis: Cue type, y-axis: Mean inaccurate TOT resolution (in %) Error bars represent +/- 1 SD; *** = p < .001, * = p