Publications

Abstract

Participants' eye movements were monitored as they heard sentences and saw four pictured objects on a computer screen. Participants were instructed to click on the object mentioned in the sentence. There were more transitory fixations to pictures representing monosyllabic words (e.g. ham) when the first syllable of the target word (e.g. hamster) had been replaced by a recording of the monosyllabic word than when it came from a different recording of the target word. This demonstrates that a phonemically identical sequence can contain cues that modulate its lexical interpretation. This effect was governed by the duration of the sequence, rather than by its origin (i.e. which type of word it came from). The longer the sequence, the more monosyllabic-word interpretations it generated. We argue that cues to lexical-embedding disambiguation, such as segmental lengthening, result from the realization of a prosodic boundary that often but not always follows monosyllabic words, and that lexical candidates whose word boundaries are aligned with prosodic boundaries are favored in the word-recognition process.

Abstract

Eye movements were monitored as participants followed spoken instructions to manipulate one of four objects pictured on a computer screen. Target words occurred in utterance-medial (e.g., Put the cap next to the square) or utterance-final position (e.g., Now click on the cap). Displays consisted of the target picture (e.g., a cap), a monosyllabic competitor picture (e.g., a cat), a polysyllabic competitor picture (e.g., a captain) and a distractor (e.g., a beaker). The relative proportion of fixations to the two types of competitor pictures changed as a function of the position of the target word in the utterance, demonstrating that lexical competition is modulated by prosodically conditioned phonetic variation.

Abstract

Several theorists have proposed that distinctions are needed between different positive emotional states, and that these discriminations may be particularly useful in the domain of vocal signals (Ekman, 1992b, Cognition and Emotion, 6, 169–200; Scherer, 1986, Psychological Bulletin, 99, 143–165). We report an investigation into the hypothesis that positive basic emotions have distinct vocal expressions (Ekman, 1992b, Cognition and Emotion, 6, 169–200). Non-verbal vocalisations are used that map onto five putative positive emotions: Achievement/Triumph, Amusement, Contentment, Sensual Pleasure, and Relief. Data from categorisation and rating tasks indicate that each vocal expression is accurately categorised and consistently rated as expressing the intended emotion. This pattern is replicated across two language groups. These data, we conclude, provide evidence for the existence of robustly recognisable expressions of distinct positive emotions.

Abstract

The difficulty of the task of segmenting a speech signal into its words is immediately clear when listening to a foreign language; it is much harder to segment the signal into its words, since the words of the language are unknown. Infants are faced with the same task when learning their first language. This study provides a better understanding of the task that infants face while learning their native language. We employed an automatic algorithm on the task of speech segmentation without prior knowledge of the labels of the phonemes. An analysis of the boundaries erroneously placed inside a phoneme showed that the algorithm consistently placed additional boundaries in phonemes in which acoustic changes occur. These acoustic changes may be as great as the transition from the closure to the burst of a plosive or as subtle as the formant transitions in low or back vowels. Moreover, we found that glottal vibration may attenuate the relevance of acoustic changes within obstruents. An interesting question for further research is how infants learn to overcome the natural tendency to segment these ‘dynamic’ phonemes.

Abstract

This letter evaluates potential benefits of combining human speech recognition ~HSR! and automatic speech recognition by building a joint model of an automatic phone recognizer ~APR! and a computational model of HSR, viz., Shortlist @Norris, Cognition 52, 189–234 ~1994!#. Experiments based on ‘‘real-life’’ speech highlight critical limitations posed by some of the simplifying assumptions made in models of human speech recognition. These limitations could be overcome by avoiding hard phone decisions at the output side of the APR, and by using a match between the input and the internal lexicon that flexibly copes with deviations from canonical phonemic representations.

Abstract

Articulatory feature (AF) modelling of speech has received a considerable amount of attention in automatic speech recognition research. Although termed ‘articulatory’, previous definitions make certain assumptions that are invalid, for instance, that articulators ‘hop’ from one fixed position to the next. In this paper, we studied two methods, based on support vector classification (SVC) and regression (SVR), in which the articulation continuum is modelled without being restricted to using discrete AF value classes. A comparison with a baseline system trained on quantised values of the articulation continuum showed that both SVC and SVR outperform the baseline for two of the three investigated AFs, with improvements up to 5.6% absolute.

Abstract

This paper addresses the problem of recognizing a vocabulary of over 50,000 city names in a telephone access spoken dialogue system. We adopt a two-stage framework in which only major cities are represented in the first stage lexicon. We rely on an unknown word model encoded as a phone loop to detect OOV city names (referred to as ‘rare city’ names). We use SpeM, a tool that can extract words and word-initial cohorts from phone graphs from a large fallback lexicon, to provide an N-best list of promising city name hypotheses on the basis of the phone graph corresponding to the OOV. This N-best list is then inserted into the second stage lexicon for a subsequent recognition pass. Experiments were conducted on a set of spontaneous telephone-quality utterances; each containing one rare city name. It appeared that SpeM was able to include nearly 75% of the correct city names in an N-best hypothesis list of 3000 city names. With the names found by SpeM to extend the lexicon of the second stage recognizer, a word accuracy of 77.3% could be obtained. The best one-stage system yielded a word accuracy of 72.6%. The absolute number of correctly recognized rare city names almost doubled, from 62 for the best one-stage system to 102 for the best two-stage system. However, even the best two-stage system recognized only about one-third of the rare city names retrieved by SpeM. The paper discusses ways for improving the overall performance in the context of an application.

Abstract

Humans are able to recognise a word before its acoustic realisation is complete. This in contrast to conventional automatic speech recognition (ASR) systems, which compute the likelihood of a number of hypothesised word sequences, and identify the words that were recognised on the basis of a trace back of the hypothesis with the highest eventual score, in order to maximise efficiency and performance. In the present paper, we present an ASR system, SpeM, based on principles known from the field of human word recognition that is able to model the human capability of ‘early recognition’ by computing word activation scores (based on negative log likelihood scores) during the speech recognition process. Experiments on 1463 polysyllabic words in 885 utterances showed that 64.0% (936) of these polysyllabic words were recognised correctly at the end of the utterance. For 81.1% of the 936 correctly recognised polysyllabic words the local word activation allowed us to identify the word before its last phone was available, and 64.1% of those words were already identified one phone after their lexical uniqueness point. We investigated two types of predictors for deciding whether a word is considered as recognised before the end of its acoustic realisation. The first type is related to the absolute and relative values of the word activation, which trade false acceptances for false rejections. The second type of predictor is related to the number of phones of the word that have already been processed and the number of phones that remain until the end of the word. The results showed that SpeM’s performance increases if the amount of acoustic evidence in support of a word increases and the risk of future mismatches decreases.

Abstract

In this paper, we present an automatic speech recognition (ASR) system based on the combination of an automatic phone recogniser and a computational model of human speech recognition – SpeM – that is capable of computing ‘word activations’ during the recognition process, in addition to doing normal speech recognition, a task in which conventional ASR architectures only provide output after the end of an utterance. We explain the notion of word activation and show that it can be used for ‘early recognition’, i.e. recognising a word before the end of the word is available. Our ASR system was tested on 992 continuous speech utterances, each containing at least one target word: a city name of at least two syllables. The results show that early recognition was obtained for 72.8% of the target words that were recognised correctly. Also, it is shown that word activation can be used as an effective confidence measure.

Abstract

We have recently developed a new model of human speech recognition, based on automatic speech recognition techniques [1]. The present paper has two goals. First, we show that the new model performs well in the recognition of lexically ambiguous input. These demonstrations suggest that the model is able to operate in the same optimal way as human listeners. Second, we discuss how to relate the behaviour of a recogniser, designed to discover the optimum path through a word lattice, to data from human listening experiments. We argue that this requires a metric that combines both path-based and word-based measures of recognition performance. The combined metric varies continuously as the input speech signal unfolds over time.

Abstract

The fields of human speech recognition (HSR) and automatic speech recognition (ASR) both investigate parts of the speech recognition process and have word recognition as their central issue. Although the research fields appear closely related, their aims and research methods are quite different. Despite these differences there is, however, lately a growing interest in possible cross-fertilisation. Researchers from both ASR and HSR are realising the potential benefit of looking at the research field on the other side of the ‘gap’. In this paper, we provide an overview of past and present efforts to link human and automatic speech recognition research and present an overview of the literature describing the performance difference between machines and human listeners. The focus of the paper is on the mutual benefits to be derived from establishing closer collaborations and knowledge interchange between ASR and HSR. The paper ends with an argument for more and closer collaborations between researchers of ASR and HSR to further improve research in both fields.

Abstract

In this paper, we present a novel computational model of human speech recognition – called SpeM – based on the theory underlying Shortlist. We will show that SpeM, in combination with an automatic phone recogniser (APR), is able to simulate the human speech recognition process from the acoustic signal to the ultimate recognition of words. This joint model takes an acoustic speech file as input and calculates the activation flows of candidate words on the basis of the degree of fit of the candidate words with the input. Experiments showed that SpeM outperforms Shortlist on the recognition of ‘real-life’ input. Furthermore, SpeM performs only slightly worse than an off-the-shelf full-blown automatic speech recogniser in which all words are equally probable, while it provides a transparent computationally elegant paradigm for modelling word activations in human word recognition.

Abstract

The ultimate goal of our research is to develop a computational model of human speech recognition that is able to capture the effects of fine-grained acoustic variation on speech recognition behaviour. As part of this work we are investigating automatic feature classifiers that are able to create reliable and accurate transcriptions of the articulatory behaviour encoded in the acoustic speech signal. In the experiments reported here, we analysed the classification results from support vector machines (SVMs) and multilayer perceptrons (MLPs). MLPs have been widely and successfully used for the task of multi-value articulatory feature classification, while (to the best of our knowledge) SVMs have not. This paper compares the performance of the two classifiers and analyses the results in order to better understand the articulatory representations. It was found that the SVMs outperformed the MLPs for five out of the seven articulatory feature classes we investigated while using only 8.8–44.2% of the training material used for training the MLPs. The structure in the misclassifications of the SVMs and MLPs suggested that there might be a mismatch between the characteristics of the classification systems and the characteristics of the description of the AF values themselves. The analyses showed that some of the misclassified features are inherently confusable given the acoustic space. We concluded that in order to come to a feature set that can be used for a reliable and accurate automatic description of the speech signal; it could be beneficial to move away from quantised representations.

Abstract

Good teachers know their students, and exploit this knowledge to adapt or optimise their instruction. Traditional teachers know their students because they interact with them face-to-face in classroom or one-to-one tutoring sessions. In these settings, they can build student models, i.e., by exploiting the multi-faceted nature of human-human communication. In distance-learning contexts, teacher and student have to cope with the lack of such direct interaction, and this must have detrimental effects for both teacher and student. In a past study we have analysed teacher requirements for tracking student actions in computer-mediated settings. Given the results of this study, we have devised and implemented a tool that allows teachers to keep track of their learners'interaction in e-learning systems. We present the tool's functionality and user interfaces, and an evaluation of its usability.

Abstract

During language production and comprehension, information about a word's syntactic properties is sometimes needed. While the decision about the grammatical gender of a word requires access to syntactic knowledge, it has also been hypothesized that semantic (i.e., biological gender) or phonological information (i.e., sound regularities) may influence this decision. Event-related potentials (ERPs) were measured while native speakers of German processed written words that were or were not semantically and/or phonologically marked for gender. Behavioral and ERP results showed that participants were faster in making a gender decision when words were semantically and/or phonologically gender marked than when this was not the case, although the phonological effects were less clear. In conclusion, our data provide evidence that even though participants performed a grammatical gender decision, this task can be influenced by semantic and phonological factors.

Abstract

This study investigated the time course of phonological encoding during speech production planning. Previous research has shown that conceptual/semantic information precedes syntactic information in the planning of speech production and that syntactic information is available earlier than phonological information. Here, we studied the relative time courses of the two different processes within phonological encoding, i.e. metrical encoding and syllabification. According to one prominent theory of language production, metrical encoding involves the retrieval of the stress pattern of a word, while syllabification is carried out to construct the syllabic structure of a word. However, the relative timing of these two processes is underspecified in the theory. We employed an implicit picture naming task and recorded event-related brain potentials to obtain fine-grained temporal information about metrical encoding and syllabification. Results revealed that both tasks generated effects that fall within the time window of phonological encoding. However, there was no timing difference between the two effects, suggesting that they occur approximately at the same time.

Abstract

In this study, we investigated grammatical feature selection during noun phrase production in German and Dutch. More specifically, we studied the conditions under which different grammatical genders select either the same or different determiners or suffixes. Pictures of one or two objects paired with a gender-congruent or a gender-incongruent distractor word were presented. Participants named the pictures using a singular or plural noun phrase with the appropriate determiner and/or adjective in German or Dutch. Significant effects of gender congruency were only obtained in the singular condition where the selection of determiners is governed by the target’s gender, but not in the plural condition where the determiner is identical for all genders. When different suffixes were to be selected in the gender-incongruent condition, no gender congruency effect was obtained. The results suggest that the so-called gender congruency effect is really a determiner congruency effect. The overall pattern of results is interpreted as indicating that grammatical feature selection is an automatic consequence of lexical node selection and therefore not subject to interference from other grammatical features. This implies that lexical node and grammatical feature selection operate with distinct principles.

Abstract

This study investigated the encoding of metrical information during speech production in Dutch. In Experiment 1, participants were asked to judge whether bisyllabic picture names had initial or final stress. Results showed significantly faster decision times for initially stressed targets (e.g., LEpel 'spoon') than for targets with final stress (e.g., liBEL 'dragon fly'; capital letters indicate stressed syllables) and revealed that the monitoring latencies are not a function of the picture naming or object recognition latencies to the same pictures. Experiments 2 and 3 replicated the outcome of the first experiment with bi- and trisyllabic picture names. These results demonstrate that metrical information of words is encoded rightward incrementally during phonological encoding in speech production. The results of these experiments are in line with Levelt's model of phonological encoding.

Abstract

Speakers can use pronouns when their conceptual referents are accessible from the preceding discourse, as in 'The flower is red. It turns blue'. Theories of language production agree that in order to produce a noun semantic, syntactic, and phonological information must be accessed. However, little is known about lexical access to pronouns. In this paper, we propose a model of pronoun access in German. Since the forms of German pronouns depend on the grammatical gender of the nouns they replace, the model claims that speakers must access the syntactic representation of the replaced noun (its lemma) to select a pronoun. In two experiments using the lexical decision during naming paradigm [Levelt, W.J.M., Schriefers, H., Vorberg, D., Meyer, A.S., Pechmann, T., Havinga, J., 1991a. The time course of lexical access in speech production: a study of picture naming. Psychological Review 98, 122-142], we investigated whether lemma access automatically entails the activation of the corresponding word form or whether a word form is only activated when the noun itself is produced, but not when it is replaced by a pronoun. Experiment 1 showed that during pronoun production the phonological form of the replaced noun is activated. Experiment 2 demonstrated that this phonological activation was not a residual of the use of the noun in the preceding sentence. Thus, when a pronoun is produced, the lemma and the phonological form of the replaced noun become reactivated.

Abstract

Recent research in speech production has shown that syntactic structure is reflected in segmental phonology--the application of certain phonological rules of English (e.g., palatalization and alveolar flapping) is inhibited across phrase boundaries. We examined whether such segmental effects can be used in speech perception as cues to syntactic structure, and the relation between the use of these segmental features as syntactic markers in production and perception. Speakers of American English (a dialect in which the above segmental effects occur) could indeed use the segmental cues in syntax perception; speakers of British English (in which the effects do not occur) were unable to make use of them, while speakers of British English who were long-term residents of the United States showed intermediate performance.

Abstract

We have used the genome-wide marker genotypes from Genetic Analysis Workshop 15 Problem 2 to explore joint evidence for genetic linkage to rheumatoid arthritis across several samples. The data consisted of four high-density genome scans on samples selected for rheumatoid arthritis. We cleaned the data, removed intermarker linkage disequilibrium, and assembled the samples onto a common genetic map using genome sequence positions as a reference for map interpolation. The individual studies were combined first at the genotype level (mega-analysis) prior to a multipoint linkage analysis on the combined sample, and second using the genome scan meta-analysis method after linkage analysis of each sample. The two approaches were compared, and give strong support to the HLA locus on chromosome 6 as a susceptibility locus. Other regions of interest include loci on chromosomes 11, 2, and 12.

Abstract

The basic idea behind this task is to find out how languages encode basic shape distinctions such as dimensionality, axial geometry, relative size, etc. More specifically, we want to find out (i) which formal means are used cross linguistically to encode basic shape distinctions, and (ii) which are the semantic distinctions that are made in this domain. In languages with many shape-classifiers, these distinctions are encoded (at least partially) in classifiers. In other languages, positional verbs, descriptive modifiers, such as “flat”, “round”, or nouns such as “cube”, “ball”, etc. might be the preferred means. In this context, we also want to investigate what other “grammatical work” shapeencoding expressions possibly do in a given language, e.g. unitization of mass nouns, or anaphoric uses of shape-encoding classifiers, etc. This task further seeks to determine the role of shape-related parameters which underlie the design of objects in the semantics of the system under investigation.

Abstract

Based on the conversation analysts’ insights into the various forms of third person reference in English, this paper first presents the inventory of forms Kilivila, the Austronesian language of the Trobriand Islanders of Papua New Guinea, offers its speakers for making such references. To illustrate such references to third persons in talk-in-interaction in Kilivila, a case study on gossiping is presented in the second part of the paper. This case study shows that ambiguous anaphoric references to two first mentioned third persons turn out to not only exceed and even violate the frame of a clearly defined situational-intentional variety of Kilivila that is constituted by the genre “gossip”, but also that these references are extremely dangerous for speakers in the Trobriand Islanders’ society. I illustrate how this culturally dangerous situation escalates and how other participants of the group of gossiping men try to “repair” this violation of the frame of a culturally defined and metalinguistically labelled “way of speaking”. The paper ends with some general remarks on how the understanding of forms of person reference in a language is dependent on the culture specific context in which they are produced.

Abstract

One of the central aims of the "Cognitive Anthropology Research Group" (since 1998 the "Department of Language and Cognition of the MPI for Psycholinguistics") is to research the relationship between language, culture and cognition and the conceptualization of space in various languages and cultures. Ever since its foundation in 1991 the group has been developing methods to elicit cross-culturally and cross-linguistically comparable data for this research project. After a brief summary of the central considerations that served as guidelines for the developing of these elicitation devices, this paper first presents a broad selection of the "space games" developed and used for data elicitation in the groups' various fieldsites so far. The paper then discusses the advantages and shortcomings of these data elicitation devices. Finally, it is argued that methodologists developing such devices find themselves in a position somewhere between Scylla and Charybdis - at least, if they take the requirement seriously that the elicited data should be comparable not only cross-culturally but also cross-linguistically.

Abstract

Visiting the Trobriand Islands is advertised as being the highlight of a trip for tourists to Papua New Guinea who want, and can afford, to experience this 'ultimate adventure' with 'expeditionary cruises aboard the luxurious Melanesian Discoverer. The advertisements also promise that the tourists can 'meet the friendly people' and 'observe their unique culture, dances, and art'. During my research in Kaibola and Nuwebila, two neighbouring villages on the northern tip of Kiriwina Island, I studied and analysed the encounters of tourists with Trobriand Islanders, who sing and dance for the Europeans. The analyses of the islanders' tourist performances are based on Erving Goffman's now classic study The Presentation of Self in Everyday Life, which was first published in 1959. In this study Goffmann analyses the structures of social encounters from the perspective of the dramatic performance. The situational context within which the encounter between tourists and Trobriand Islanders takes place frames the tourists as the audience and the Trobriand Islanders as a team of performers. The inherent structure of the parts of the overall performance presented in the two villages can be summarized - within the framework of Goffman's approach - in analogy with the structure of drama. We find parts that constitute the 'exposition', the 'complication', and the 'resolution' of a drama; we even observe an equivalent to the importance of the 'Second Act Curtain' in modern drama theory. Deeper analyses of this encounter show that the motives of the performers and their 'art of impression management' are to control the impression their audience receives in this encounter situation. This analysis reveals that the Trobriand Islanders sell their customers the expected images of what Malinowski (1929) once termed the '...Life of Savages in North-Western Melanesia' in a staged 'illusion'. With the conscious realization of the part they as performers play in this encounter, the Trobriand Islanders are in a position that is superior to that of their audience. Their merchandise or commodity is 'not real', as it is sold 'out of its true cultural context'. It is staged - and thus cannot be taken by any customer whatsoever because it (re)presents just an 'illusion'. The Trobriand Islanders know that neither they nor the core aspects of their culture will suffer any damage within a tourist encounter that is defined by the structure and the kind of their performance. Their pride and self-confidence enable them to bring their superior position into play in their dealings with tourists. With their indigenous humour, they even use this encounter for ridiculing their visitors. It turns out that the encounter is another manifestation of the Trobriand Islanders' self-consciousness, self-confidence, and pride with which they manage to protect core aspects of their cultural identity, while at the same time using and 'selling' parts of their culture as a kind of commodity to tourists.

Abstract

Arbeitet man als muttersprachlicher Sprecher des Deutschen mit Corpora gesprochener oder geschriebener deutscher Sprache, dann reflektiert man in aller Regel nur selten über die Vielzahl von kulturspezifischen Informationen, die in solchen Texten kodifiziert sind – vor allem, wenn es sich bei diesen Daten um Texte aus der Gegenwart handelt. In den meisten Fällen hat man nämlich keinerlei Probleme mit dem in den Daten präsupponierten und als allgemein bekannt erachteten Hintergrundswissen. Betrachtet man dagegen Daten in Corpora, die andere – vor allem nicht-indoeuropäische – Sprachen dokumentieren, dann wird einem schnell bewußt, wieviel an kulturspezifischem Wissen nötig ist, um diese Daten adäquat zu verstehen. In meinem Vortrag illustriere ich diese Beobachtung an einem Beispiel aus meinem Corpus des Kilivila, der austronesischen Sprache der Trobriand-Insulaner von Papua-Neuguinea. Anhand eines kurzen Auschnitts einer insgesamt etwa 26 Minuten dauernden Dokumentation, worüber und wie sechs Trobriander miteinander tratschen und klatschen, zeige ich, was ein Hörer oder Leser eines solchen kurzen Daten-Ausschnitts wissen muß, um nicht nur dem Gespräch überhaupt folgen zu können, sondern auch um zu verstehen, was dabei abläuft und wieso ein auf den ersten Blick absolut alltägliches Gespräch plötzlich für einen Trobriander ungeheuer an Brisanz und Bedeutung gewinnt. Vor dem Hintergrund dieses Beispiels weise ich dann zum Schluß meines Beitrags darauf hin, wie unbedingt nötig und erforderlich es ist, in allen Corpora bei der Erschließung und Kommentierung von Datenmaterialien durch sogenannte Metadaten solche kulturspezifischen Informationen explizit zu machen.

Abstract

One of the many results of the "Space" research project conducted at the MPI for Psycholinguistics is that there are three "Frames of spatial Reference" (FoRs), the relative, the intrinsic and the absolute FoR. Cross-linguistic research showed that speakers who prefer one FoR in verbal spatial references rely on a comparable coding system for memorizing spatial configurations and for making inferences with respect to these spatial configurations in non-verbal problem solving. Moreover, research results also revealed that in some languages these verbal FoRs also influence gestural behavior. These results document the close interrelationship between language, culture and cognition in the domain "Space". The proper description of these interrelationships in the spatial domain requires language and culture specific ontologies.

Abstract

This handbook chapter summarizes some of the problems of nominal classification in language, presents and illustrates the various systems or techniques of nominal classification, and points out why nominal classification is one of the most interesting topics in Cognitive Linguistics.

Abstract

This project aims to investigate how speakers of various languages in indigenous cultures verbally reason about moral issues. The ways in which a solution for a moral problem is found, phrased and justified will be taken as the basis for researching reasoning processes that manifest themselves verbally in the speakers’ arguments put forward to solve a number of moral problems which will be presented to them in the form of unfinished story plots or scenarios that ask for a solution. The plots chosen attempt to present common problems in human society and human behaviour. They should function to elicit moral discussion and/or moral arguments in groups of consultants of at least three persons.

Abstract

A corpus of digitized speech errors is used to compare the prosody of correction patterns for word-level vs. sound-level errors. Results for both peak F0 and perceived prosodic markedness confirm that speakers are more likely to mark corrections of word-level errors than corrections of sound-level errors, and that errors ambiguous between word-level and soundlevel (such as boat for moat) show correction patterns like those for sound level errors. This finding increases the plausibility of the claim that word-sound-ambiguous errors arise at the same level of processing as sound errors that do not form words.

Abstract

Three experiments examined whether infants recognise functors in phrases, and whether their representations of functors are phonetically well specified. Eight- and 13- month-old English infants heard monosyllabic lexical words preceded by real functors (e.g., the, his) versus nonsense functors (e.g., kuh); the latter were minimally modified segmentally (but not prosodically) from real functors. Lexical words were constant across conditions; thus recognition of functors would appear as longer listening time to sequences with real functors. Eightmonth- olds' listening times to sequences with real versus nonsense functors did not significantly differ, suggesting that they did not recognise real functors, or functor representations lacked phonetic specification. However, 13-month-olds listened significantly longer to sequences with real functors. Thus, somewhere between 8 and 13 months of age infants learn familiar functors and represent them with segmental detail. We propose that accumulated frequency of functors in input in general passes a critical threshold during this time.

Abstract

Incremental sentence generation imposes special constraints on the representation of the grammar and the design of the formulator (the module which is responsible for constructing the syntactic and morphological structure). In the model of natural speech production presented here, a formalism called Segment Grammar is used for the representation of linguistic knowledge. We give a definition of this formalism and present a formulator design which relies on it. Next, we present an object- oriented implementation of Segment Grammar. Finally, we compare Segment Grammar with other formalisms.