"In this book, Richard Kern explores how technology matters to language and the ways in which we use it. Kern reveals how material, social and individual resources interact in the design of textual meaning, and how that interaction plays out across contexts of communication, different situations of technological mediation, and different moments in time."
Date: Thu, 2 Jun 2005 13:55:21 +0100 From: Maarika Traat Subject: Information Structure: Theoretical and Empirical Aspects
EDITOR: Steube, Anita TITLE: Information Structure SUBTITLE: Theoretical and Empirical Aspects SERIES: Language, Context & Cognition 1 PUBLISHER: Mouton de Gruyter YEAR: 2004 ISBN: 3110179342
Maarika Traat, School of Informatics, The University of Edinburgh
[Note: The symbols ~, A, and E are used for logical negation, universal quantification and existential quantification respectively. -- Eds.]
The book is a collection of fourteen papers. The papers report on the state of the art work on different aspects of information structure. The papers vary greatly as far as the significance of their contribution and clarity of presentation are concerned. The first eight papers are theoretical in nature, while the last six report on empirical data from production and perception experiments. The majority of the papers analyse information structural phenomena in German, but some other languages also receive coverage: Hungarian, Czech, Polish, Russian, Vedic. Both word-order related phenomena and prosodic phenomena are examined. The majority of the papers were contributed by the members of the Research Group "Linguistic Foundations of Cognitive Science: Linguistic and Conceptual Knowledge" at the University of Leipzig, but papers were also invited from researchers in the field elsewhere (Prague, Potsdam, Budapest).
In the following section we are going to give a brief summary of each of the fourteen papers in the order they appeared in the book. Finally, we are going to present a short evaluation of the book as a whole.
"Degrees of Contrast and the Topic-Focus Articulation" by Eva Hajičová and Petr Sgall studies the phenomenon of contrast and degrees of its intensity as determined by the type of contrast. However, it is left unspecified what the authors mean precisely by 'intensity': whether we are speaking about some prosodic metrics or somebody's perception. It is more likely to be the latter, since the authors use the words 'contrast is felt even stronger', but it still remains unclear whether the decisions are based on the authors' personal intuitions or on the results of some perception study.
The study is set in the Praguian Functional Generative Description (FGD) framework, which includes a description of the topic-focus articulation (TFA). The first part of the paper deals with TFA. The authors define it as a relation of 'aboutness' where focus holds about its topic. The authors also include a section about annotating Prague Dependency Treebank with TFA. However, no further clarification is given how this goal is achieved with the rather vague definition of TFA.
The rest of the paper discusses contrast in TFA. The means of expressing contrast in Czech include strong pronominal forms and rising stress; the latter, however, is also exploited towards other ends (e.g. 'open continuation'). Focus has been viewed as a choice from a set of alternatives (Rooth 1985), and as such always be contrastive. German linguists often distinguish between contrastive and non- contrastive focus, though. The authors of the present paper propose that there are two degrees of contrast present in the focus in German, signalled by the structure of the language. In Czech the contrast in focus does not vary in its intensity. This is not the case with contrast in topic, or more precisely on context bound (CB) items, however. The authors suggest that the intensity of contrast on a CB item depends on how narrow the focus is, whether the set of alternatives is explicitly mentioned or implicit, and the range of the set of alternatives.
In "Information Structure and Modular Grammar", Anita Steube, Kai Alter and Andreas Späth present a model of grammar set in the generative framework, that consists of Lexicon, Semantic Form, S- Structure, and Phonological Form, in which information structure is realised at each level of grammar. They view information structure as a pragmatic phenomenon, which is part of the cognitive model. The pre-structuring of information takes place before entering into the grammar model. The lexicon serves as the interface between the conceptual system and the grammar. When the concepts are mapped onto lexical items their representations are enriched with information structural features.
The authors mostly discuss information structure in terms of background and focus, although they also define topic and comment. The default organisation of information in German is that background elements precede focus elements. The focus domain starts to the right of the attitudinal adverbials. However, focus domain and focus of a sentence are not identical: constituents base-generated in the focus domain can move out of the domain. There are two kinds of movement: information structure motivated and purely syntactically motivated movement. If the syntactically motivated movement takes focus constituents out of the focus domain, they need to be explicitly marked by the focus feature [+F] in the S-Structure, which later gets realised in prosody. Similarly, the background elements remaining in the focus domain need to be marked by [-F]. When the S-Structure is mapped on Phonological Form the constituents marked with [-F] are de-accented, and the ones with [+F] are accented. In what follows the authors use their model to analyse German categoric and correction sentences, and sentences with the Bridge Contour.
The paper discusses important aspects of German information structure. However, it would have profited immensely by having an introduction which would provide the reader with guidance through the paper, and a conclusion summarising the points the authors themselves consider most important. Without these important components the paper feels unfocussed.
"Negative Descriptions of Events: Semantic and Conceptual Aspects of Sentence Negation and its Relevance for Information Structure" by Andreas Späth and Martin Trautwein discusses the relation between the scope of negation and the information structure of a sentence. In the surface form of a sentence, negation has the striking role of dividing the individual information units of the sentence and marking their function. Only the part on the right of the negation (i.e. focus) is subject to truth-conditional evaluation, the part of the sentence that precedes the negation is presupposed. Hence a determiner phrase's (DP) position before the negation in the sentence makes it presuppositional. This helps to explain why Slavic languages do not need a definite article, which is generally viewed as the source of presupposition, to mark their DPs as specific.
The authors argue that due to the negation's fixed position before the focus of the sentence, external negation as used in Propositional Logic is not appropriate for describing the truth conditions of natural languages. Moreover we can only speak of the truth of an event in a specific spatio-temporal domain. The authors examine three approaches to logically representing negative events: negative event quantification (~Ee[e INST p]), negative instantiation (Ee[~(e INST p)]) and negative propositional condition (Ee[e INST ~p]). Even though the latter performs best of the above three as an approximation to reality, it is still not adequate since it does not take into account the semantic and syntactic constraints on sentence structure. In the section following this discussion the authors show how syntax determines the way the semantics of negative sentences is composed.
Beáta Gyuris' "Two Types of Contrastive Topics?" starts with a hypothesis that there are two kinds of contrastive topic DPs in Hungarian: the ones that are only licensed by the very same DP in the previous discourse and the ones that can also be licensed by other DPs. The author presents a number of Hungarian question-answer pairs that seem to reflect this tendency. Moreover, this tendency seems to be related to the monotonicity of the DPs: the DPs with monotone increasing determiners seem to be able to be used in a much wider range of contexts than the ones with monotone decreasing and non-monotonic determiners.
At first the author analyses the data in the framework proposed by Büring (1997). It turns out that his mechanism cannot account for all the Hungarian examples of syntactically possible, but uninterpretable sentences with contrastive topic DPs. In what follows, the author resorts to Kadmon's (2001) theory on discourse congruence to explain why certain sentences with contrastive topic DPs cannot be used as answers to certain questions. Kadmon proposed a restriction that the topic value of the answer declarative sentence with a contrastive topic corresponds to the focus value of the last question under discussion (QUD). Gyuris supplements Kadmon's restriction with two more restrictions: i) in order for a declarative sentence with a contrastive topic to be accepted as an appropriate answer to a question, this overt question and the last QUD have to be sub-questions of the same 'superquestion', and ii) in case the overt question and the last QUD are not identical, the answer declarative must not entail a complete answer to the overt question. The paper rejects the initial hypothesis about the twofold division of DPs in Hungarian, and proposes that the the data presented can be accounted for by the above three restrictions.
In "Information Structure -- Two-dimensionally Explicated" the author, Ingolf Max, presents a two-dimensional representation for sentence semantics. The first meaning dimension represents the proper ordinary meaning of the sentence, which is given by the matrix of the sentence semantic form (SF) in first order logic. The second meaning dimension relates to the information structure of the sentence: it represents the logical form of the background of the sentence, according to background/focus partitioning. In the case of categorical sentences, the second dimension is a sub-conjunction of the conjunction in the proper SF of the sentence on the first dimension. In thetic sentences any tautology can serve as the background.
Max introduces a new kind of presupposition-preserving negation (~^) in the form of a special reduction operator that works on both dimensions. Using this negation and the relations of 'necessitation' and 'entailment' the author can define important semantic notions of 'presupposition' and 'assertion'.
Using his two-dimensional representation, Max models various linguistic phenomena: categorical sentences, thetic sentences, generalization, hat contour and correction sentences. In his approach the scope inversion accompanying the hat contour can be modelled in a straightforward manner: inverting the direct order of AE and ~^ is all that is needed.
In "Topic Constraints in the German Middlefield" Brigitta Haftka takes a close look at the word order in the middlefield of German categorical sentences. This is the position for contextually known information. Haftka distinguishes two kinds of topics: 'proper topic'/'aboutness topic'/'theme' on the one hand and 'anaphoric topics' on the other. 'Proper topic' is what the rest of the sentence is predicated about. Anaphoric topics are known background elements that are stored in the narrow short-term memory of the speaker. Proper topic is moved to the top of the sentence, while anaphoric topics fill the so- called Wackernagel position, which follows the proper topic and precedes the sentence adverbial position.
The study shows that the order of anaphoric topics in Wackernagel position is highly grammaticalised. Haftka formulates seven constraints that model how the word order is determined in the topicality middlefield. The main body of the paper is followed by an appendix where the author presents an optimality theoretic ranking scheme for the above constraints.
"Contrastive Word Stress in Vedic Endo- and Exocentric compounds" by Rosemarie Lühr is an extensive study of stress patterns in different classes of Vedic compounds. The basic stressing rules previously proposed for Vedic compounds leave a lot of counter-examples unexplained. Lühr shows that stress shifts in Vedic compounds are brought along by the need to mark contrast i) between 'substantive' and 'adjective' parts of speech ii) compound internally between the parts of speech of a constituent of the compound and the compound as a whole. The author uses optimality theoretic approach to explain the stress shifts.
The paper would have benefited immensely if instead of overwhelming the reader with a multitude of examples, the author would have used more of the space for explanatory purposes. Several central notions lack a clear definition, e.g. 'bahuvrihis', which refers to a type of Vedic compound.
In "Towards a Scalar Notion of Information Structural Markedness" Thomas Weskott proposes to view information structural (IS) markedness in the light of contextual requirement. The more requirements a given IS-variant of a sentence has to the context it is being uttered in, the more marked it is. The author makes three assumptions that serve to make decisions about information structural complexity of sentences: i) the basic word order in German sentences is SVO, ii) the default phrasal stress falls on the most deeply embedded constituent, and iii) the complexity of an IS-variant depends on the two dimensions of information structural bracketing: Topic- Comment-Structure (TCS) and Focus-Background-Structure (FBS). The last assumption implies that the bigger the overlap is between topic and background on the one hand, and between comment and focus on the other hand, the less complex the information structure of the given sentence is. Violation of any of the three assumptions adds to the complexity of the IS of the given sentence.
Thetic sentences are the least complex sentences from the IS point of view, since they place no restrictions to their context, and can be uttered 'out-of-the-blue'. Categorical sentences have varying degrees of complexity depending on how many of the above assumptions they violate.
1a) [Der Kellner]T/B [beleidigte den GAST]C/F. 1b) [Den GAST]T/F [beleidigte der Kellner]C/B. (T(opic), C(omment), B(ackground), F(ocus))
According to Weskott's theory the above two sentences represent the two opposite ends of the spectrum of information structural markedness: while the first sentence (1a) follows all the assumptions, the second one (1b) violates all three of them. Hence, the information structure of the second sentence is maximally complex. Weskott shows that his approach about the scale of IS markedness and its relation to the utterance context fits in nicely with empirical data about sentence processing difficulties.
"Prosody in Dialogues and Single Sentences: How Prosody Can Influence Speech Perception" by Claudia Hruska and Kai Alter studies the role of intonational focus. They describe three different perception experiments. The audio data used represented three different focus conditions: neutral focus (2a), broad focus (2b), and narrow focus (2c). Neutral focus requires no preceding context, and is purely syntactically determined, while both broad and narrow focus need to be embedded in context. This context was provided by preceding the sentence containing the given focus type by an appropriate question.
2a) [Peter vershpricht Anna zu arbeiten](IPh) und das Büro zu putzen. 2b) Q: Was verspricht Peter Anna zu tun? A: Peter vershpricht Anna zu (ARBEITEN und das BÜRO zu putzen) (F).
2c) Q: Wem verspricht Peter zu arbeiten und das Büro zu putzen? A: Peter vershpricht (ANNA)(F) zu arbeiten und das Büro zu putzen.
The first perception experiment explored the role of intonational focus in dialogues, which in the present case consisted of question-answer (QA) pairs. In addition to appropriate QA pairs, other pairs were constructed where the focus in the answer did not match the question. In the second and third perception experiment the subjects were presented with single sentences that were either neutrally focussed or contained a broad or a narrow focus.
Besides the judgment of the subjects about the appropriatenesss of the sentences, electroencephalogram (EEG) was continuously recorded, and event-related brain potentials were measured. In the dialogues' experiment the subjects were very good at differentiating between the matching and non-matching question-answer pairs (98%) . The ERP data showed that inappropriate intonation impairs comprehension. De-accentuation of new information causes bigger problems than having superfluous accents in the sentence. The ERP recordings showed that listeners concentrate their attention at the sentence positions where new information is expected. In the single sentence experiments neutral focus was preferred. The acceptance rate for sentences with narrow focus was especially low (8%).
"On the Independence of Information Structural Processing from Prosody" by Ulrike Toepel and Kai Alter continues in the same vein as the previous paper in the book. This time the perceptual studies are about whether the subjects are able to distinguish between a narrow new focus accent and a contrastive focus accent in German sentences. The authors carry out two perception experiments. They use the same data in both experiments: 44 three sentence dialogues where the third sentence is the critical one. Besides the appropriate focus accent in appropriate context they create two inappropriate conditions, changing the focus of the third sentence so that a contrastive accent appears in a new focus context and a new focus accent appears in a contrastive context. The only difference between the experiments was that in the first experiment the subjects were asked a content question based on each dialogue, while in the second experiment they had to explicitly pay attention to the prosody of the dialogues and assess its appropriateness. During the experiments the electroencephalogram (EEG) was continuously recorded.
Based on ERP (event-related brain potential) data, the authors conclude that subjects are sensitive to the subtle differences that the new and the contrastive focus accent exhibit. Similarly to the previous paper, the authors found that over-specification of prosodic information (a contrastive focus accent in a new focus context) can be dealt with without bigger problems, while prosodic under-specification (a new focus accent in a contrastive focus context) causes processing difficulties.
"The Prosodic Pattern of Contrastive Accent in Russian" by Grit Mehlhorn describes the results from a production experiment and three perception experiments that aimed at establishing whether contrastive focus accent represented a different category from the new focus accent in Russian.
In the production experiment the subjects had to read experimental sentences set in appropriate context. The collected data revealed that contrastive focus accents were characterised by a higher pitch than the new focus accents. The length of the syllable carrying the main accent was considerably increased in the case of the contrastive focus accent. In contrast to sentences with a new focus accent, the ones with a contrastive accent have strongly centred contours.
In the three perception experiments the subjects were asked to determine the position of the accent, characterise the prominence of the accent on a five-grade scale, and to characterise the pattern of the perceived accent in terms of rises and falls. The results showed that the subjects found it easier to correctly determine the location of contrastive focus accents (98.8%) than new focus accents (54.6%). The average grade of prominence the subjects assigned to contrastive accents (4.42) was considerably higher than that assigned to new focus accents (2.24). Even though there was not perfect agreement among the subjects about the exact pattern of neither new focus nor contrastive accents in Russian, they systematically assigned a different pattern to contrastive accents as opposed to new focus accents. The author concludes that the experimental data proves that contrastive and new focus accents represent different categories in Russian.
In "Focus Structure and the Processing of Word Order Variations in German", the authors, Britta Stolterfoht and Markus Bader study the focus structural effects in processing German sentences with scrambled word order. The basic assumption is that in order for focus to project (i.e. produce wide focus reading) the constituent carrying the nuclear accent has to be in its base position and in the sister position of the verbal head (Haider & Rosengren 2002). If the nuclear accent falls on a moved constituent, only narrow focus reading is possible.
The preferred word order in German is subject before object (SO). However, it is possible to scramble the object before the subject (OS). Sometimes the syntactic function of the DPs can be locally or globally ambiguous. The sentences in Examples 3a and 3b are ambiguous in respect to their word order until the reader reads the finite verb. Then the number information (singular or plural) of the verb disambiguates the syntactic functions of the preceding DPs.
3a) Maria hat behauptet, dass [die Tante die Nichten begrüßt hat](F). 3b) Maria hat behauptet, dass die Tante(i) [die Nichten](F) t(i) begrüßt haben.
According to the above assumption, at this point the reader, besides performing a syntactic re-analysis, should also perform a focus structural revision. The authors test this hypothesis in two reading experiments. They measure the event-related brain potentials (ERPs) of the subjects. The experimental data supports the hypothesis, since in the case of scrambled sentences, an additional negative ERP effect was observed. The authors interpret this as the result of focus structural revision.
"Intonational Patterns in Contrast and Concession" by Carla Umbach, Ina Mleinek, Christine Lehmann, Thomas Weskott, Kai Alter, Anita Steube describes an experiment that was conducted to verify Lang's (2001) hypothesis that in German the contrastive and the concessive readings of but/aber-sentences are reflected in their distinct prosody.
The experiment was a speech production task where the subjects had to read but/aber-sentences in their respective contrastive and concessive contexts. The phonological analysis of results did not reveal any systematic difference in the F0-contours of the two readings. A further statistical analysis was performed to uncover any hidden tendencies, but it also failed to reveal a systematic difference between the two readings.
The authors admit that there is a big inter- and intra-subject variation in the data, that does not allow them to provide a proof to Lang's hypothesis. However, at the same time they avoid disproving the hypothesis, saying that other uncontrolled factors may have been present during the experiment. They also pronounce the possibility that the distinction in intonational patterns proposed by Lang could still emerge in the case of a larger sample size of subjects.
In "Prosody in Contrast. Prosodic Distinction of Contrast and Correction Readings of Polish Adversative Coordinate Structures" Dorothee Fehrmann examines the prosody of Polish adversative coordinate structures, where the first conjunct contains a negative marker (see Example 4).
4) Piotr nie ma samochodu, ale motocykl. Peter S-Neg has car Conj motorbike
Depending on the context the sentence appears in, such coordinate structures can be interpreted as contrast or correction. Fehrmann investigates the question of whether these two readings are distinguished by their prosody in Polish. In order to do that she conducted a production experiment, where subjects had to read lexically and syntactically identical adversative coordinate constructions embedded in context to give them either a contrast or correction reading.
The resulting intonation contours exhibited a lot of variability, and therefore Fehrmann concludes that it is not obligatory in Polish to differentiate between contrast or correction readings of adversative structures by prosodic means. However, she does identify a frequent intonation contour, that she calls 'the default-IC', with which correction constructions are correlated more often than contrast constructions. The default-IC is a single intonational phrase that covers the whole coordinate construction. In the case of contrast it is more likely that there is a sentence internal intonational boundary between the two conjuncts. The author concludes that rather than being determined by the conceptual interpretation type, the intonational patterns of adversative coordinate constructions depend on the particular information structure of the conjuncts.
The book is a diverse body of papers, all of which study some aspect of information structure. The papers vary greatly as far as the significance of their contribution and clarity of presentation are concerned. As a downside, they exhibit the same vagueness of definition so characteristic of the field of information structure, while basing the theory mostly on syntactically simple and short examples. Some of the papers suffer from structural deficits, such as the lack of introduction and conclusion. In other papers central terminology is used without providing a clear definition.
There are also some minor editorial issues, which nevertheless can cause confusion. Besides some typos, in the first paper there are some formatting problems where example text gets mixed up with the text of the main body of the paper (page 10), and on pages 154 and 155 seven lines of text are printed twice.
All in all, several interesting ideas concerning information structure are put forward in the collection. The book does a great service to the heterogeneous field of information structure already merely by bringing together between the same covers a body of papers from a number of researchers in the field. This is hopefully a step forward towards a more unified treatment of information structure over different languages and schools of thought.
Büring, D. (1997): The Meaning of Topic and Focus. The 59th Street Bridge Accent. London, New York: Routlege.
Haider, H. and I. Rosengren (2002): Scrambling -- Non-triggered Chain Formation in OV-languages. MS. Salzburg University and Lund University.
Kadmon, N. (2001): Formal Pragmatics. Malden, MA, Oxford: Blackwell.
Lang, E. (2001): Kontrastiv vs. implikativ: Interpretationseffekte intonatorischer Distinktionen bei Koordination. In: A. Steube and C. Umbach (eds), 113-138.
Rooth, M. (1985): Association with focus. PhD dissertation. University of Massachusetts, Amherst.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Maarika Traat is a PhD student at the University of Edinburgh, UK. She is working under the supervision of Mark Steedman and Johan Bos. Her doctoral study focuses on developing a semantic representation with information structure compatible with first order logic, and embedding this semantics in a categorial grammar formalism. Her other current research interests are the syntax and semantics of English cleft constructions, and calculating presuppositions.