21st ANNUAL WORKSHOP, 4-6 July 2004
Genova (Genoa), Italia,

at the University:
Università di Genova, Facoltà di Lettere e Filosofia, via Balbi, 2

For each lecture  45 minutes are available, including discussion.

Sunday 4 July, 18.00 hrs, Welcome party, Ristorante Amadeus,
via P.E. Bensa, 40c (Piazza della Nunziata), Genova

Monday 5 July

10.00 – 12.00, Morning session

12.00 – 14.30, Lunch

14.30 – 18.00, Afternoon session

Tuesday 6 July

9.00 – 12.00 Morning session

12.00 – 14.30, Lunch

14.30 – 18.00, Afternoon session

Address of the scientific meetings:

Università di Genova,
Facoltà di Lettere e Filosofia
via Balbi, 2, Genova,
Tel. (+39) 010 209 5781


21st Annual Workshop, Genova, Italy, 4-6 July 2004

Abstracts of papers to be presented, in alphabetical order on the surname of the first author. The symbol @ has been replaced by ‘at’ to reduce the chance that E-mail addresses can be found on the webpage by search engines for spammail.

Jörg Becker

Universitaet der Bundeswehr, Munich, Germany

Constructive constructivism

According to Piaget, consciousness means having a model of the world and of oneself. We look into the microprocess of creating such a model, based on variability, irreversibiliy, and selectivity. Such a model of a cognitive system may be used for developing desires which serve for making decisions at bifurcation points ("chance management").

Address: Universitaet der Bundeswehr Muenchen

Fakultaet fuer Elektrotechnik und Informationstechnik

Institut fuer Physik, 85577 Neubiberg, Germany

E-mail: Joerg.Becker at

Margherita Benzi

University of Genoa

Causal Inference and Simpson’s Paradox

   If we admit that positive (negative) causes raise (lower) the probability of their effects, then positive and negative associations between properties in observed data can be interpeted as indicators of causal connections between the associated properties. Yet there are cases in which the statistical connection between two properties does not correspond to the causal connection. One of the most interesting cases is provided by the so called "Simpson's paradox", where the data structure is such that the 'sign' of the association between a pair of properties can consistently be inverted in each subpopulation when the population is partitioned (e.g., a pharmacological treatment can be associated with a higher recovery rate for treated patients, but the two subpopulations of treated female patents and treated male patients can each have lower recovery rates when compared with untreated female patients and untreated male patients).

  In this paper I recall and discuss some of the main philosophical problems that Simpson's paradox raises, focussing on the debate on theories of probabilistic causality (according to which causes can be reduced to probabilities), and on theories of causal inference.

Address: Margherita Benzi

Dipartimento di Filosofia - Universita' di Genova

via Balbi 4, 16126 Genova , Italy

E-mail: Benzi at

Anna M. Borghi

University of Bologna

Do Object Concepts Incorporate Motor Information?

   Traditional views of conceptual organization see concepts as being made of  propositional symbols related arbitrarily to their referents. This implies  the existence of a process by which sensorimotor experience is translated  into amodal symbols. Recently many authors have shown that limitations of  these views by proposing that concepts are, rather, grounded in  sensorimotor activity (Barsalou, 1999; Harnad, 1990; Thelen & Smith, 1994).  Neuroimaging studies studies support the hypothesis that some kinds of  motor information are directly elicited by concept-nouns (Chao & Martin, 2001). The idea that concepts are grounded in sensorimotor activity and  action-based is compatible with two claims. According to the first claim,  concepts can be conceived of directly as patterns of potential action  (Glenberg, 1997), thus allowing us to respond quickly to environmental  stimuli.  According to the second claim, concepts are being made of  "perceptual symbols" (Barsalou, 1999). These symbols are modal, i.e. that  they are represented in the same perceptual systems that gave rise to them.  Storing information in terms of perceptual symbols guarantees flexibility,  i.e. it allows us to interact with objects in different ways depending on  our current goals, on the current perspective and on the current context. In the presentation I will discuss evidence in support of both claims. Most  specifically, I will report experimental evidence obtained with  compatibility tasks, feature production tasks and verification tasks, as  well as with connectionist Artificial Life simulations. The results show that motor information is incorporated directly into  concepts for simple interaction with their referents, particularly with  manipulable objects. However, when it comes to performing complex  goal-oriented actions with complex objects we may access more general  perceptual and situational information and use it in a flexible manner. I will show that this is true both in presence of objects and when object  concepts are referred to by words,  i.e. that not only the visual  representation of objects incorporates motor information (Tucker & Ellis,  1998; Borghi, Di Ferdinando & Parisi, in preparation), but that also  concept-nouns elicit motor information (Borghi, 2004; Borghi, Glenberg &  Kaschak, in press).

Address: Anna M. Borghi, Department of Psychology, University of Bologna

Viale Berti Pichat, 5, 40127 Bologna, Italy,

Email: borghi at  Webpage:  HYPERLINK ""

G.J. Dalenoort

University of Groningen, The Netherlands

The a priori Role of Theories and Models for Understanding Cognition

   Many scientists and theoreticians of science have emphasized the essential  role of the apriori notions on which we base our theories and models. This  is true for all natural sciences, but it is especially important in the  sciences concerned with mind and brain, where a large diversity of opinions  can still exist on such basic notions as: what is an appropriate explanation, what can we explain at all, what is the role of experiments, and what is the  role of theories and theoretical notions.

  These ideas stem from the concern I feel on the direction research, funding,  and, acceptation of papers for conferences and journals are taking. The growth of the dominance of what may be called the 'empirical stance' is alarming. There is little concern of the role our a priori notions have on the design of  experiments, on the interpretation of observations and measurements.

  In physics there is a common basis for the way we argue, and for what counts as an argument. Although physicists may differ in their opinions of the  appropriateness of a given model or assumption, there is no fundamental difference on the consequences of such assumptions. How different is this in disciplines concerned with cognitive science, such as psychology, linguistics, the neurosciences,  philosophy.

  In the paper some examples will be given of the role of a priori theoretical assumptions, and subsequently an analysis will be given of some controversial issues in cognitive science, and how we can avoid useless controversies.

Address: Dept. of Psychology (E&A), University of Groningen

Grote Kruisstraat 2/1, 9712 TS  Groningen, The Netherlands

E-mail:  g.j.dalenoort at 

Peter Eberle

University of Linz, Austria

User Modelling in the Context of Distributed Cognition

   The presentation addresses distributed cognition and its use as a framework for modelling characteristics of users of interactive technologies. The proposed approach for static and dynamic user modelling shows how symbolic representations might serve as reference model to dynamic user representations. Besides the interplay between symbolic and sub-symbolic knowledge processing the emergence of novel parameters (constructs) in designing user interfaces will be discussed. Emerging parameters leverage implicit user needs and make them transparent through explicit representation. They allow for dynamic adaptation of user interfaces, in particular when evolving through actual interaction of users with the software system.

Address: Department of Business Information Systems

Communications Engineering, University of Linz

Freistädterstraße 315, A-4040 Linz, Austria,     E-mail: Peter.Eberle at

Marcello Frixione

Universita' di Salerno

Can Reference be Naturalized?

   This abstract deals with the possibility of giving a naturalistic account of 'reference' of natural language. Can reference be naturalized? I shall argue that  this type of question admits a positive as well as a negative answer:

a) In a stronger sense, the possibility of giving a naturalistic account of the relation of reference is unlikely because it seems to involve a full acceptance of the commonsense ontology within a naturalistic view of the world. Let us suppose that reference is a two-place relation R, such that R(e, r) is true if and only if e is a linguistic expression and r is the reference of e. If we assume that relation R can be naturalized (i.e., that R can figure within a naturalistic theory of some kind), then we must assume that all the entities of the commonsense discourse (i.e., all the entities natural language expressions can be about) can be accommodated with a naturalistic view of the world. Forms of this argument have been proposed by Bloomfield, Chomsky, and Fodor (in the solipsistic phase of his thought).

b) In a weaker sense, it is likely that that reference can be naturalised, in the sense of naturalising the referential abilities of the speakers. In other words, it seems plausible that a naturalistic account can be given of everything happens in the heads of the  speakers when they use natural-language expressions to refer to the world.

Dipartimento di Scienze della Comunicazione,

Universita' di Salerno

via Ponte Don Melillo, I-84084 Fisciano (Salerno), Italy

E-mail: frix at

Omar Gelo

University of Ulm, Germany

Cognitive Linguistics and Metaphors: Toward an Account of Meaning Construction and Extension

   Meaning construction and extension is since a long time at the centre of scientific debate within cognitive sciences, and the role of analogical-metaphorical processes within such processes received in the last years a renewed attention. It has been observed how abstract reasoning and language are based on our perceptual-motoric schemas and pre-verbal concepts (Scholnick & Cookson, 1984). Moreover, in most of our everyday life situations we use our experience to structure new informations, in a way that cannot be fully explained be formal/propositional information processing (Carbonell & Minton, 1991). Metaphorical-analogical processes are a primary cognitive device through which knowledge is organized and creatively modified (Gentner & Wollf, 2000).

  Cognitive linguistics (Lakoff & Johnson, 1980) has provided a theory of metaphorical meaning and language that fits the recent models of mental architecture and function developed in the last years within the cognitive sciences (Bucci, 1997; Edelman, 1987; Dennet, 1991). "Conceptual organization" (generically mental and specifically linguistic) is analyzed in relation to "world experience" (bodily-perceptual-motoric aspects and extralinguistic ones), and metaphorical processes have a key role within the dynamics between those different levels. Metaphorical language is not just an alternative to a literal description of reality, but reflects at a linguistic level cognitive modalities of information processing. Metaphorical processing allows to use already structured, concrete and familiar experience and knowledge domains to create new ones, more abstract and less familiar (Carbonell & Minton, 1991; Lakoff & Johnson, 1980). Their function is that of using perceptual-motoric experience to structure abstract concepts (Johnson, 1987), accessing and expressing of emotional states (Ortony & Fainsilber, 1987) and facilitating cognitive restructuration (Gentner & Wollf, 2000).

  Providing some of the most recent computational models of metaphor, the present contribution intends to expose how a cognitive view of metaphor can contribute to a better comprehension of how knowledge develops, organizes and modifies itself.

Address: University of Ulm

Department of Psychosomatic Medicine and Psychotherapy

Section of Informatic in Psychotherapy

Am Hoechstraess, 8 , 89081 Ulm - Germany

E-mail: omargelo at

Alberto Greco

University of Genoa

A Meta-Theoretic System for Constructing Correspondences in

Cognitive Science

   The most usual issue arising in cognitive science meetings is that a multidisciplinary cooperation should be welcome, but that we really don't know how to accomplish it. I argue that the problem is that we lack a meta-theoretic framework or guidelines about how to sort and match different perspectives. This is a consequence of the fact that the dominant unifying concept in cognitive science has been "information": such notion worked as a common umbrella, useful for connecting and consolidating perspectives, but became reductionistic because every approach could be translated into others only using the computational language.

   Since the computational metaphor was the only available tool, so far, the easiest way of attempting the construction of a meta-theoretic framework for cognitive sciences was to distinguish between low-level and high-level processes and to try to establish a path linking them, either bottom-up or top-down. (As a bottom-up example, one may try to see how meaningful conscious thoughts arise from some sensorial input; reversing it the top-down way, obviously, one could see how symbolic processes might direct sensory ones).

   The problem is that different disciplines investigating different levels speak very different languages. The reason is that they segment in different ways the commonsense reality: in other words, they actually have different objects. Using the concept of information, sometimes, one can succeed in translating different explanations into a common language: e.g. neural processes can be considered as raw data upon which higher-level processing is performed. But in this case the contribution of single cognitive disciplines to the overall explanation is poor, because the specificity of the language, say, of neurons, or qualia, and so on, is lost into the new computational account.

   A different solution is to try to establish correspondences between different accounts. The fact that each cognitive discipline describes and explains the same commonsense "fact" from a peculiar point of view, using specialized predicates, should not be lead us to consider different accounts as excluding each other or incompatible. Finding correspondences is a step further, going beyond a simple catalog of different descriptions of facts, because it can provide a sort of a meta-theoretic map for the translation of one description into the other in the context of specific empirical data (particular tasks).

   I shall elaborate a detailed proposal about how such a map can be accomplished. My suggestion is that, if we want a meta-theoretical perspective not reducing different views to a single account (like the computational one), then some general dimensions should be defined, sufficiently general to cover all cognitive phenomena. The general dimensions here proposed are: state, event, flow, flow-chain.

   Definitions. The basic element in the system I am proposing is "state". There are different descriptions associated with a commonsense fact considered as a snapshot in a single time. Neural language may refer to the state of a neuron, or of a brain area; behavioral language may refer to a glandular secretion, or to a muscle or body movement; phenomenal language may refer to a state of consciousness, and so on. (For brevity, we shall elliptically speak of "states" instead of "descriptions of states"). States are following one another. More precisely, we can identify another state just when the former state changes. We define "event" such a change of state. In a more coarse grain, we call "flow" a sequence of events in time. When we consider different flows in parallel at the same time interval, we have a "flow-chain".(In all such terms, the word "description" is always implied).

   A typical flow-chain about cognitive phenomena includes several flows: physical, behavioral, sensorial, consciousness, etc. Since a clear-cut distinction between first-person and third-person descriptions is convenient, I shall refer to the "flow-reader" as to the system that reads flows, by identifying states and events in the system: the same individual or cognitive system or some other individual or CS. The description of an event includes the description of an initial state, of a transformation or change, of a subsequent state. A state  change may also be defined in terms of information (in a never-changing system there would be more entropy than information) and so our terminology is more general but compatible with the computational approach. In identifying states and events, a flow-reader operates as a categorical system (to identify it must first discriminate differences).

   Different cognitive disciplines provide us with collections of descriptions, a chain-flow of events constructed upon the same commonsense fact. The aim of a meta-theoretic approach is to find links and correspondences between them. Links between events may be causal or simply correlational. Then two kinds of links may be constructed in our chain-flow system: horizontal (links between events in the same flow) an vertical (links between events in different flows). Vertical links (multidisciplinary cooperation) can add explicative power to the overall model, since some links, that would only be correlational inside a single flow, might become causal when moving to different flows. Examples of analyses of specific tasks will be provided and some consequences for communication between disciplines will be discussed.

Address: Psychology Division, Department of Anthropological Sciences

University of Genoa

DISA (Dip. di Scienze Antropologiche) Department

Laboratory of Psychology and Cognitive Science

Universita' degli Studi - Via Balbi, 4; 16126  Genova , Italy

E-mail: greco at    Webpage:

Ton Jörg

University of Utrecht, The Netherlands

Minds in Evolution - An Evolutionary Perspective on Learning and Knowledge Formation through Human Interaction

   In my presentation the topic will be what has been called 'the new science' by Vygotsky. In his  view there was a serious crisis in psychology which one should deal with (Vygotsky, 1926/1997).  One of his main points was that we should first have a kind of 'theory of the crisis' before we could  be able to 'solve the crisis' (Ibid.). It is my intention to show that Vygotsky had not been very  successful in formulating such a theory of the crisis. That's why the crisis is still there, as has been  notified, among others, by the editors of his work "The Historical Meaning of the Crisis in  Psychology' (Rieber & Wollock, 1997). In the field of psychology and in education educational  research some of the scholars speak about the crisis quite openly (Westland, 1978; Cohen &  McLaughlin, 1993; Egan, 1997; Burkhardt & Schoenfeld, 2003).

  Now that his collected work finally has been translated in a more adequate (extended) form in six  volumes (1987, 1997, 1999), it may be time to rethink some of his main thoughts about the field  of (educational) psychology. 

  One of the main goals of Vygotsky he has not worked out the way he had been thinking of, is the  goal of humanizing determinism as a fundamental part of his new science (Vygotsky, 1997, Vol. 4,  p. 58-59). In my view there still is not a new method and a new methodology for the new science  of psychology he advocated with so much passion. It is no accident that there still is no adequate  theory on interaction in the field of education and cognitive science. Bates et al. (1999) formulated  this problem quite eloquently into the question: "What will a good theory of interaction look like  when it arrives?" (p. 590) It seems that we are still the victims of old habits of thought, unable to  escape them, f.i. the way we conceive of determinism (see e.g. Mainzer, 2004, p. 37). We are so  much used to what Starobinski describes as the Calculable, that we have very great problems of  dealing with the Incalculable, that goes beyond control and prediction (Starobinski, 2003; see also  Mainzer, 2004).

  According to Vygotsky (1997), we should take the whole of personality as a subject in human  psychology and the research of it: "The development of personality and the development of  reaction of the personality are essentially two aspects of one and the same process." (Vol. 4, pp.  59; emphasis added) It is this statement which can be taken as a starting point for theorizing on  the processes of human interaction, thereby taking generativity as "the (simultaneous) cause and  effect of individualization." (Sassone, 1996, p. 519) By bringing man forward to the centre  determinism can be described, then, as humanized. (see Vygotsky, 1997, Vol. 4, p. 58-59)   We may, now, formulate the goal of a theory of learning through human interaction as 'a causal  generative theory of being through becoming.' (cf. Bhaskar, 1993) The whole person is taken to be  involved in the processes going on. The learning which takes place may be defined as a form of  Generative Learning (GL). It is not who is doing the interaction we are focusing on but what makes  the process of interaction generative as a process and in its effects on the participants? This goal  could be rephrased as aiming at understanding the cognitive dynamics of knowledge formation  (see Raven & Krohn, 2000, p. lv; cf. Bechtel, 1998). The implication of it may be described as a  kind of 'liquefying rigidity' by which "To be" is replaced by "To become" (see Zilsel, 2000, on the  significance of Darwin's theory of evolution).  The learning taking place may be defined as a process of change and reciprocal transformation, of  metamorphoses in behaviour (Vygotsky, 1978; see also Mainzer, 2004, p. 19). It is the patterns of  selectivity active in both the intra- and inter-mental processes of involution and evolution,  operating together, with their dialectic dynamics of opposing forces (of growth) which may  'produce' those phenomena (Vygotsky, 1997, vol. 4; Starobinski, 2003; see also Thom, 1975; and  Mainzer, 2004). It is these very phenomena which goes beyond prediction and control, but not  beyond our imagination! Their description may, however, be more closely linked to reality than we  may expect at first sight (Mainzer, 2004, p. 325, p. 407; see also Wolfram, 2002). A new kind of  language for the description of the dialectic dynamics of interaction may be needed for doing so  (see Starobinski, 2003, p. 221; cf. Thom, 1975; Bhaskar, 1993; and Mainzer, 2004).

Bates, E., Elman, J.L., Johnson, M.H., Karmiloff-Smith, A., and Plunkett, K. (1998). Innateness and  emergentism. In W. Bechtel & G. Graham (Eds.) A Companion to Cognitive Science, pp.  590-601. Malden (MA): Blackwell Publishers.

Bechtel, W. (1998) Representations and cognitive explanations: Assessing the dy-namicist's  challenge in cognitive science. Cognitive Science, vol. 22, (3), 295-317.

Bhaskar, R. (1993). Dialectic. The pulse of freedom. London: Verso.

Cohen, D.& McLaughlin, M. (Eds.) (1993). Teaching for understanding. Challenges for  policy and practice. San Francisco: Jossey-Bass Publishers.

Egan, K. (1997). The educated mind. How cognitive tools shape our understanding. Chicago: The  university of Chicago Press.

Mainzer, K. (2004). Thinking in complexity. The computational dynamics of matter, mind,  and mankind. Berlin: Springer.

Raven, W. Krohn, and R. S. Cohen (Eds.), Boston Studies in the philosophy of science. Vol. 200.  Dordrecht: Kluwer Academic Publishers.

Rieber, R.W. & Wollock, A.S. (1997). Prologue. Vygotsky's  "Crisis" and its Meaning Today. In R.W. Rieber & A.S. Wollock (Eds.), The collected works of L.S. Vygotsky. Vol. 3, Problems of the theory and history of psychology. New York.: Plenum Press.

Sassone, L. A. (1996). Philosophy across the curriculum: A democratic Nietzschean  pedagogy. Educational Theory, Fall 1996, vol. 46, nr. 4, pp. 511-524.

Starobinski, J. (2003). Action and reaction. The life and adventures of a couple. New York:  Zone books.

Thom, R. (1975). Structural stability and morphogenesis. An outline of a general theory of  models. Reading (MA): W.A. Benjamin inc.  Vygotsky, L. (1978). Mind in Society. Cambridge (MA): Harvard University Press.

Vygotsky, L. (1987). Collected works, Vol. 1. Problems of General Psychology. In R.W. Rieber &  A.S. Carton (Eds.). New York.: Plenum Press. Vygotsky, L. (1997). Collected works, Vol. 3. Problems of the theory and history of  psychology.  In R.W. Rieber & A.S. Wollock (Eds.). New York.: Plenum Press.

Vygotsky, L. (1997). Collected works, Vol. 4. The history of the development of higher mental  functions.  In R.W. Rieber (Ed.). New York.: Plenum Press.

Vygotsky, L.S. (1997). Educational psychology. Boca Raton: St Lucie Press. USA.

Westland, G. (1978). Current Crisis of Psychology. London: Heinemann.

Wolfram, S. (2002). A New Kind of Science. Wolfram Media Inc.

Zilsel, E. (2000). The social origins of modern science. In D. Raven, W. Krohn, and R. S.  Cohen (Eds.), Boston Studies in the philosophy of science. Vol. 200. Dordrecht: Kluwer Academic Publishers.

Address: IVLOS, University of Utrecht,  P.O. Box 80137,

3508 TC  Utrecht, The Netherlands

E-mail: t.jorg at

Carlo Penco

Universita' di Genova

Context and Prototypes

   In the paper I assume the need of a coordination between subsymbolic and symbolic tools in treating  knowledge representation and reasoning. The talk aims to clarify the notion of "context" in the  treatment of reasoning and semantics. To look for a clarification, I will start with two contrasting  theories: the "objective" theory of contexts (Kaplan), where context is a set of features of the world,  and the "cognitive" theory of context (McCarthy), where context is the cognitive background of a  speaker or agent in respect to a situation. These two kinds of theories of contexts can be partially  compatible with the classification given by John Perry who distinguishes among pre-semantic,  semantic and post-semantic context. The point of the talk, which develops ideas presented in a previous  paper on "Three alternatives on context", is to show some shortcoming of the cognitive theory of  context and to give some general lines for an answer. 

Address:  Carlo Penco

Dipartimento di Filosofia - Universita' di Genova

via Balbi 4, 16126 Genova +39-010-2099795

E-mail: penco at   Webpage: 

Thomas Riga[1], Angelo Cangelosi[1], Alberto Greco[2]

1, University of Plymouth, U.K.; 2, University of Genoa, Italy

Teaching Robots by Using Imitation and Natural Language

   The development of increasingly intelligent and autonomous robots (and simulated agents) requires the design of efficient human-robot interaction systems. This includes approaches based on imitation (Billard, 2000; Dautenhahn, 2003), instruction-based learning (Lauria et al., 2000), and linguistic communication (Roy et al., 2003; Steels & Kaplan, 1999). A new approach that combines imitation and natural language is proposed here. We developed an integrated system of two simulated robots in a virtual reality world. On this platform we studied how perceived actions are matched with correspondent self-generated actions and how these actions are memorized and successively reproduced autonomously. We addressed the question of how agents learn to perform basic actions on objects by using simple imitative mechanisms like mimicking. We taught the robots more complex behaviours using a natural language interface. In a first simulation an imitator agent learned to perform actions when receiving a linguistic description of them. Furthermore it learned to give a linguistic description of actions performed by a teacher agent. Descriptions contained a verb, indicating the action, and a noun, referring to the object involved in the action. Brain imaging studies (Pulvermueller, 2003; Cappa & Perani, 2003) reveal that language processing activates different brain areas for verbs and nouns. We analyzed the internal neural network structure using synthetic brain imaging techniques (Arbib & al., 2000) and compared the results to experimental brain imaging data: results showed that verb processing recruits neurons in the area responsible for motor program execution and noun processing involved the area responsible for object recognition. These results are in line with the existing experimental data. They support the view that language integrates with neural structures responsible for action execution and observation (Glenberg & Kaschak, 2002; Rizzolatti & Arbib, 1998; Steels, 2000). During the second simulation agents learned to execute basic actions by mimicking them, while simultaneously learning words corresponding to these actions. Furthermore they learned combined actions by receiving linguistic descriptions of them, containing these previously acquired words. The agents merged basic actions into a composite action following exclusively linguistic descriptions. They did this by transferring the grounding of the symbols referring to basic actions to the symbol indicating the combined action (Greco, Riga & Cangelosi, 2003). The third simulation was similar to the second, but instead of teaching combined actions through a natural language interface we taught sequences of actions. After having acquired the basic actions together with the words referring to them, the agents received descriptions of macro-actions consisting of a sequence of basic actions. When presented with the word referring to the macro-action, they correctly produced the sequence without ever having seen it being performed. The results of the last two simulations emphasize on language as a means through which new behaviours can be acquired quickly and effortlessly, building on experience accumulated by previous generations of agents. They highlight the importance of cultural transmission in cognitive development. The proposed model is a first step towards the design of learning systems for robot programming through demonstration. The long-term goal is to develop a framework for training robots both through imitation and by using a natural language interface.

Arbib, M.A., Billard, A., Iacoboni M., & Oztop E. (2000). Synthetic brain imaging: grasping, mirror neurons and imitation. Neural Networks, 13, 975-997.

Billard, A. (2000). Learning motor skills by imitation: a biologically inspired robotic model. Cybernetics & Systems, 32, 1-2, 155-193.

Cappa, S.F., & Perani, D.  (2003). The neural correlates of noun and verbal processing.  Journal of Neurolinguistics, 16 (2-3), 183-189.

Dautenhahn, K.(2003). Playing and Learning with Robots. In: The Future of Learning, Eds. M. Tokoro and L. Steels, IOS Press, pp. 163-177.

Glenberg A., & Kaschak M. (2002). Grounding language in action. Psychonomic Bulletin & Review, 9 (3) , 558-565.

Greco A., Riga T. & Cangelosi A. (2003), The acquisition of new categories through grounded symbols: An extended connectionist model. Joint ICANN/ICONIP 2003 Conference, Turkey, June 2003.

Lauria S., Bugmann G., Kyriacou T., & Klein E. (2002). Mobile robot programming using natural language. Robotics and Autonomous Systems, 38 (3-4): 171-181.

Pulvermueller F. (2003). The neuroscience of language. On brain circuits of words and serial order. Cambridge University Press.

Rizzolatti, G., & Arbib, M. (1998). Language within our grasp. Trends in Neuroscience, 21, 188-194.

Roy, D., Hsiao, K., & Mavridis, N. (in press). Mental Imagery for Conversational Robots. IEEE Transactions on Systems, Man, and Cybernetics.

Steels, L. (2000). Mirror neurons and the action theory of language origins. Proceedings of Architectures of the Mind, Architectures of the Brain.

Steels, L., & Kaplan, F. (2000). AIBO's first words: The social learning of language and meaning. Evolution of Communication, 4(1).


Thomas Riga & Angelo Cangelosi,

Adaptive Behaviour & Cognition Research Group University of Plymouth, U.K. {thomas.riga, angelo.cangelosi} at

A. Greco: Psychology Division, Department of Anthropological Sciences

University of Genoa

DISA (Dip. di Scienze Antropologiche) Department

Laboratory of Psychology and Cognitive Science

Universita' degli Studi - Via Balbi, 4; 16126  Genova , Italy

E-mail: greco at    Webpage:

Laura Salmon

University of Genoa

What about a Translation Device?

A Theoretical Proposal on Human Linguistic Translation Processes.

   In spite of an old interest on translation products (texts) dating back to the ancient times, only in the last two decades academic scholars started developing new scientific proposals on translation processes. Although in the first half of the past Century machine translation was one of the most significant aims of Artificial Intelligence (AI), it became evident that a basic translation theory is needed in order to state the very problem of human versus machine language processing. In the present paper, a basic Theory of human Translation Processes (TTP) is presented. TTP is based on the most relevant psycho- and neurolinguistic data and on the assumption that humans did evolve an innate Translation Device, acting most of all at the preconscious level of the human holistic intelligence. As in the case of language processing, interlanguage processing (translation switching) is due to a complex and refined interaction of  (A) the cognitive, emotional, computational (statistical) faculties with (B) the data acquired during the life experience by a person belonging to a linguocultural group (i.e., individual and shared encyclopaedia). TTP is an attempt to introduce Translation Theory as a scientific field in a new epistemological view.

Address: Laura Salmon

Facolta' di Lingue e Letterature straniere

Dipartimento di Scienze della Comunicazione Linguistica e Culturale (DISCLIC)

P.zza S.Sabina, 16126 Genova

E-mail: Salmon at