Former Department of Developmental and Comparative Psychology
Max Planck Institute for Evolutionary Anthropology
Deutscher Platz 6
Language is perhaps the most important uniquely human skill, and is therefore a central focus of the department. We concentrated on attempting to explain the processes by which children develop language as a function of their socio-cognitive skills in interaction with mature language users, and how they build later and more complex linguistic forms on previously established ways of saying things.
Research is conducted both in Leipzig and at the Max Planck Child Study Centre in Manchester, England, using naturalistic, experimental, and modeling methodologies. A very important aspect of our work has been the collection of ‘dense databases’ of the speech of children and their interlocutors. Most corpus-based child language captures, we estimate, about 2% of a child’s waking day. This can lead to major problems of sampling depending on the area under investigation. We estimate that our corpora have increased the sampling to between 7-15% depending on the exact recording regime.
As our work on human social cognition has established, children bring already quite sophisticated socio-cognitive skills to language learning. Children can distinguish what is given and new for their communicative partner and they can specify by pointing. However they have to be able to use these skills to understand the communicative intent of speakers and to learn the conventional ways of using language to inform.
We are investigating when children can successfully mark ‘given’ and ‘new’ in their utterances, and how this is influenced by shared visual and/or verbal perspectives with their communicative partner. In a number of studies we have shown that, although two-year-olds do adjust the information they provide as a function of both visual and discourse perspectives, it is not clear whether this is really adjusted to the listener, nor how children deal with coordinating visual context and prior discourse. We look at these factors by seeing under what circumstances children will use a pronoun instead of a fully descriptive noun phrase and by seeing what they mention or leave unmentioned in their replies to questions.
In studies of word learning, we have shown the intimate interrelation between pragmatic cues to reference and more semantic cues such as lexical contrast. We are investigating how children learn names for objects as a function of the relative strengths of pointing, prosodic stress patterns, newness and givenness of the referent, and the granularity with which it is described. We have an ongoing series of experiments to investigate children’s development of the use of determiners (e.g. a versus the) and adjectives to specify objects. More complex structures for identifying speakers’ reference to entities other than objects are also being studied, for instance, different types of questions, children’s understanding of words like ‘again’ (nochmal) and ‘too’ (auch), and relative clauses. In a training study we have shown that children can be trained to provide more informative referring expressions, but they are much more successful if they were speakers rather than overhearers - suggesting that using language provides the best basis for learning it.
Children have to learn how their language ‘works’ without being explicitly told. A usage-based theoretical framework suggests that children do this by first learning ‘pieces of language’ with which to communicate. Once children have learned a number of these, they will start to break them down into productive patterns with slots for either words or inflections, depending on the language. As this process continues, children will start to identify the separate parts of what they are hearing and assign meaning to them. Thus children can be highly productive with language from very early on. We investigate the scope of that productivity and how it changes with development.
In studies where we ‘trace back’ English-speaking children’s novel utterances to utterances they have previously heard or said, we have shown that at two years, their utterances are often exact repeats of something they have said before or simple but productive ‘slot-and-frame’ patterns where productivity largely consists in a single operation of filling an ’object word’ into a slot in a schema. With development, the complexity of the noun phrases placed in these slots changes as does the semantic range of the slots. We have also ‘reversed’ this study by extracting a lexically-specific grammar from each child corpus and then seeing how well it can predict the novel utterances that the child produces. We found that the grammars produced for the 2-year-old data were computationally much less complex than those produced for the 3-year-old data. We also found differences between the 2-year-old and 3-year-old grammars in the extent to which the addition of a ‘noun’ or ‘verb’ category improved performance: only the ‘noun’ category made the grammar more successful at two years whereas by three years, the addition of a verb category also improved performance. We are also conducting a series of studies of the mixed utterances of German-English bilinguals: these provide a major challenge for almost all theories in that the child is unlikely to have heard them before. We are investigating the nature of the underlying representation that could produce such mixed utterances.
A central feature of our studies is an analysis of the speech that children are actually hearing. In almost all cases, we have found close, though sometimes complex, relationships with children’s own language development. We are extending previous research on the context of children’s language learning to other communicative contexts and other languages. In a comparative study of Russian, German and English Child Directed Speech (with Sabine Stoll in Linguistics), we have shown, first, that, as we have previously found for English, the initial strings of Russian and German utterances addressed to children are very highly repetitive but that, second, typological differences between the languages have an effect on this repetitiveness, though not to the extent that might be predicted by a purely grammatical approach to language structure. Also with Sabine Stoll, we are conducting a study of the communicative context in which children from a traditional, rural society (in Eastern Nepal) learn to talk.
Despite the fact that by far the greatest volume of research on language development has been done on English, it is not very typical of the languages of the world, as it has very rigid syntactic work order and little inflectional morphology. We are interested in seeing how children learning languages with much more complex inflectional systems work out how to use these systems. Our major focus has been on the Polish system of noun declensions, which shows a number of different inflectional forms per noun, substantial allomorphy, and a complex set of factors governing the choice of endings. In a series of experiments using novel nouns (some with Ewa Dąbrowska) we have shown that there is clear evidence of developmental progression. For instance, 3.5-year-old children are capable of generalising a particular case across genders, but, on the other hand, even 4-year-olds are nowhere near ceiling in transforming a noun presented in one case to its appropriate form in a different case. We are also investigating (with Sabine Stoll) children’s learning of even more complex morphology in a richly polysynthetic language.
Different languages provide different clues to how to identify agents and patients in sentences. In a series of experiments investigating children’s comprehension of transitive syntax in English, German, French and Cantonese, we have shown there is a developmental progression from understanding the prototype for a particular language to being able to use individual cues when they do not occur in prototypical utterances (e.g. with novel verbs, with inanimate subjects, or with case marking conflicting with word order). Our results show a close match with the availability and reliability of different cues in the speech that the children hear. We are now conducting a similar study in Polish. The idea of a ‘prototype’ with a variety of cues working together is being used to study the development of children’s understanding of the ditransitive, passive and relative clause constructions of English and Cantonese. We have shown strong prototype effects with relative clauses and complementation (for both English and German): early complex sentences are based on earlier learned simple constructions (in the case of relative clauses) and highly frequent verbs (in the case of complementation). A second series of studies suggests that the input has strong effects both in terms of affecting error rates and in providing discourse support in the early stages of the production of some complement structures. We are also investigating the contribution of prosodic structure to children’s early comprehension of constructions – an entirely neglected area and one that we feel may be critical in helping children to identify particular constructions.
More complex syntax allows speakers to highlight topics or to move them into the background, to situate events in complex time frames, to convey states of mind and to manipulate aspects of the Gricean ‘cooperative’ contract. We have the most complete diary in existence of one child’s development of complex syntax (the Braunwald diary) and are using this for a microanalytic study of the development of different constructions. An analysis of this child’s earliest finite complement structures shows that each matrix clause is used for a particular interpersonal function (I wish X, You know what X) and followed by some type of proposition, but the latter is not fully coordinated either syntactically or semantically with the former. We are now experimentally investigating children’s understanding of this coordination and of the full range of meanings possible in the matrix.