ThesisPDF Available

Models as epistemic artefacts: Toward a non-representationalist account of scientific representation

kansi_polkuina.FH10 Mon Nov 07 22:39:12 2005 Page 1
Tarja Knuuttila
Models as Epistemic Artefacts:
Toward a Non-Representationalist
Account of
Scientific Representation
Philosophical Studies from the University of Helsinki 8
Tarja Knuuttila
Models as Epistemic Artefacts
What are the origins of the epistemological difficulties
concerning representation? How have philosophers of
science studying models and representation reacted to these
difficulties? How else might scientific models be approached
if not representationally? In addressing such questions,
Models as Epistemic Artefacts: Toward a Non-Representationalist
Account of Scientific Representation seeks to reinvigorate the
philosophical discussion of models and scientific repre-
sentation by proposing a new, artefactual approach to
models that loosens their epistemic value from representation
and ascribes it instead to their materially embodied con-
straints and interactive enablings. It also proposes that the
very problem of representation should be released from its
representationalist underpinnings. After a contextualising
introductory essay, the first two original research articles
deal with more general issues concerning representation,
while the remaining four articles examine representation
in the context of modelling. The articles also take part in
the current discussion of representation in science and
technology studies, semiotics and cognitive science.
Filosofisia tutkimuksia Helsingin yliopistostailosofisia tutkimuksia Helsingin yliopistosta
ilosofisia tutkimuksia Helsingin yliopistostailosofisia tutkimuksia Helsingin yliopistosta
ilosofisia tutkimuksia Helsingin yliopistosta
Filosofiska studier från Helsingfors universitetFilosofiska studier från Helsingfors universitet
Filosofiska studier från Helsingfors universitetFilosofiska studier från Helsingfors universitet
Filosofiska studier från Helsingfors universitet
Philosophical Studies from the University of HelsinkiPhilosophical Studies from the University of Helsinki
Philosophical Studies from the University of HelsinkiPhilosophical Studies from the University of Helsinki
Philosophical Studies from the University of Helsinki
Department of Philosophy
Department of Social and Moral Philosophy
P.O. Box 9 (Siltavuorenpenger 20 A)
00014 University of Helsinki
Marjaana Kopperi
Sami Pihlström
Panu Raatikainen
Petri Ylikoski
Bernt Österman
ISSN 1458-8331
ISBN 952-10-2797-5 (paperback)
ISBN 952-10-2798-3 (PDF)
Edita Prima Oy
Vantaa 2005
Models as Epistemic Artefacts:
Toward a Non-Representationalist Account of
Scientific Representation
Tarja Knuuttila
This study seeks to situate the philosophical discussion on models
and scientific representation within the larger context of questioning
representation that is taking place in other fields, especially in science
and technology studies. It addresses four related questions: (i) What
kinds of different reactions there have been to the puzzle of
representation? (ii) From where do the seeming epistemological
difficulties concerning representation stem? (iii) How can
representation be approached in a non-representationalist way? (iv)
What kinds of things are models, and how do they give us
A new artefactual approach to models is advanced that loosens the
epistemic value of models from representation and ascribes it instead
to their materially embodied constraints and interactive enablings.
The thesis draws four additional major conclusions: (1) Our
understanding of modelling should not be reduced to models
representing some external target systems. Apart from being
representative things models are typically also productive things
whose workability and experimentability is crucial for their epistemic
value. (2) Representation should be approached both from the use and
production points of view. (3) From the use point of view
representation appears as a two-fold phenomenon that is based both
on the medium-specific affordances of the material sign-vehicle and
on the intentional process of relating the sign-vehicle to the
represented. (4) From the production point of view a major portion of
the work of representation that is taking place in sciences
concentrates on what is already represented and modelled somehow.
Looking at representation from this angle stresses the methods,
ingredients and various representative devices that are needed in
producing models.
The study consists of a contextualising introductory essay and six
original research papers. The first two articles deal with more general
issues concerning representation. The next four articles study
representation in the context of modelling. Common to all of them is
the idea that models can be seen as epistemic artefacts. The articles
refer to this idea in discussing the interrelated questions of
representation, modelling, and cognition.
Preface ...................................................................................................... 7
List of original publications ....................................................... 10
Part 1. Summary
1. Introduction ........................................................................................ 12
1.1 Background ................................................................................... 12
1.2 Aims and plan .............................................................................. 16
1.3 A note on the method ................................................................... 18
2. Representation and its discontents .................................................. 21
2.1 The concept of representation ..................................................... 21
2.2 Representationalism .................................................................... 2 3
2.3 Against representation?............................................................... 27
Renouncing representation.............................................................. 27
Deconstructing representation ........................................................ 29
Reconstructing representation ......................................................... 31
3. Scientific models in the philosophy of science ............................ 37
3.1 Syntactic and semantic views on models .................................. 37
3.2 A practice-oriented approach to models ................................... 40
3.3 Models as representations........................................................... 42
3.4 Models as epistemic artefacts ..................................................... 48
4. An overview of the original articles ................................................. 50
5. Conclusions .......................................................................................... 68
Bibliography ............................................................................................ 71
Part 2. Original articles
My fascination with the question of representation has been the red
thread running through my travels in different disciplines, and it has
finally led me to the completion of this doctoral dissertation in
philosophy. While taking my first Master’s degree in economics, I
started to wonder how seemingly very austere and unrealistic
economic theories and models are able give us knowledge about our
complex economic life, what kind of knowledge that is, and,
ultimately, what science is all about. In the philosophy courses given
in the Helsinki School of Economics I came to understand that these
problems were philosophical in nature. I wish to express my gratitude
to Professor Uskali Mäki for this insight, for his teaching and for the
scholarly example he has set over the years.
I probably would never have started to work on this doctoral thesis
had I not studied semiotics at the end of 1990’s. The circle of
semioticians that surrounded Professor Eero Tarasti at the University
of Helsinki was international, liberal and yet academically ambitious,
and this spirit can, I think, be attributed directly to Eero Tarasti
himself. Over the years I have participated in various projects and
events under his leadership, and I am grateful for his considerate
guidance and assistance.
In the Center for Activity Theory and Developmental Work
Research I found the same kind of multidisciplinary and active milieu
that I had experienced in semiotics. I was fortunate to be a doctoral
student in the Center during the years when it was among those units
at the University of Helsinki that had been nominated by the Academy
of Finland as Centers of Excellence. Many foreign guest lecturers and
research fellows visited the centre and Professor Yrjö Engeström led
all of this activity in a very convincing and expansive manner. I wish
to thank all the personnel of the Center for Activity Theory and
Developmental Work Research, especially Ritva Engeström, Jaakko
Virkkunen, Leena Harjula-Jalonen and Heli Kaatrakoski.
I have been blessed with three supervisors, who all have
contributed to this dissertation in different ways. The long and
8Tarja Knuuttila
enthusiastic discussions I have had with Professor Reijo Miettinen
about the importance of artefacts have obviously had a great effect
on how I conceive of models. Moreover, my participation in the
remarkable research group led by Professor Miettinen has taught me
that science is above all a collaborative and collective effort.
Professor Ilkka Niiniluoto supervised my thesis before becoming
Rector of the University of Helsinki. I am grateful for his help and
encouragement, above all for his open-minded attitude to
philosophy. Professor Matti Sintonen has guided me through the
difficult (and admittedly nervewracking) phase of finishing this
thesis. He has been supportive and always on hand to answer my
questions. Moreover, I owe him a lot for his insights into the history
of philosophy.
Professor Fred Karlsson, Professor Timo Honkela and Docent Atro
Voutilainen have helped me substantially with the case studies on
language technology and neural network modelling. In addition, both
Honkela and Voutilainen co-authored one of the articles I am
submitting as a part of this thesis. Erika Mattila has been my helpful
partner in our common endeavour to understand modelling better,
loyal both in good and bad days—and always ready to work even
My special thanks go to the following colleagues and friends with
whom I have collaborated in recent years and from whose advice
and generosity I have benefited: Kristian Bankov, Marcel Boumans,
Sampsa Hyysalo, Hanna Johansson, Timo Kaitaro, Marja-Liisa
Kakkuri-Knuuttila, Jonna Kangasoja, Sakari Katajamäki, Hannele
Kerosuo, Janne Lehenkari, Aki Petteri Lehtinen, Juha Leminen,
Johannes Lenhard, Andrea Loettgers, Endla Lõhkivi, Martina Merz,
Sami Pihlström, Veikko Rantala, Kristina Rolin, Max Ryynänen, Jussi
Silvonen, Juha Tuunainen, Marja Väätäinen and Petri Ylikoski. I also
want to thank Henry Fullenwider and Marjatta Zenkowicz warmly for
English-language revision of some of the manuscripts, and Tinde
Päivärinta for the layout of this book.
I am grateful to Professor Mary S. Morgan and Professor Mauricio
Suárez, whose work I greatly appreciate, for agreeing to pre-examine
my thesis and comment on it.
Scholarships from the Finnish Cultural Foundation and the
Research Funds of the University of Helsinki made it possible for me to
finish this thesis. I have received travel grants from the Chancellor of
the University of Helsinki, the Finnish Konkordia Fund and the
Philosophy of Science Association (NSF Travel Grant).
Finally, it is ultimately people who make academic life a
worthwhile endeavour. More often than not, colleagues become
friends and sometimes they become more than friends. I am so happy
that I have been able to share so many things with Susanna Välimäki
and Harri Veivo. What funny, witty and caring companions you have
As for family and friends, I wish to thank one and all (some of
whom have already been mentioned above). And I am ever grateful
to my parents Raili and Petteri Knuuttila, my husband Panu
Savolainen, my nephew Jan, my daughter Laura and son Sauli for all
their love and care!
I dedicate this book to my beloved sister Petra Anneli (1962-1999),
who would have loved to see me making it.
Helsinki, October 2005
Tarja Knuuttila
10 Tarja Knuuttila
List of original publications
1) Knuuttila, Tarja (2003), “Is Representation Really in Crisis?”,
Semiotica 143: 95-111.
2) Knuuttila, Tarja (2002), “Signing for Reflexivity:
Constructionist Rhetorics and Its Reflexive Critique in Science
and Technology Studies”, Forum Qualitative Sozialforschung /
Forum: Qualitative Social Research, [On-line Journal], 3(3).
Available at:
3-02/3-02knuuttila-e.htm (52 paragraphs).
3) Knuuttila, Tarja and Atro Voutilainen (2003), “A Parser as an
Epistemic Artifact: A Material View on Models”, Philosophy of
Science 70 (Proceedings): 1484–1495.
4) Knuuttila, Tarja (in press), “Models, Representation, and
Mediation”, Philosophy of Science 72 (Proceedings), 2005.
5) Knuuttila, Tarja (in press), “From Representation to Production:
Parsers and Parsing in Language Technology”, in Johannes
Lenhard, Günther Küppers and Terry Shinn (eds.), Simulation:
Pragmatic Constructions of Reality. Sociology of the Sciences
Yearbook. New York: Springer, 2006.
6) Knuuttila, Tarja and Timo Honkela (in press), “Questioning
External and Internal Representation: The Case of Scientific
Models”, in Lorenzo Magnani (ed.), Computing, Philosophy,
and Cognition. London: King’s College Publishing, 2005.
Part 1. Summary
12 Tarja Knuuttila
1. Introduction
1.1 Background
If there is any theme that unites the heterogeneous postmodern
discussions in the fields of philosophy, humanities and cultural
theory, then representation is certainly a good candidate. Different
“postist“ movements such as poststructuralism, postcolonial theory
and certain currents of feminism have questioned representation in
several ways. The main thrust of these critiques has been that our
cultural representations are not to be trusted since there is no way for
us to ascertain whether they (re)present their objects—or reality—
truthfully. Moreover, this scepticism has concerned even scientific
representations which especially in the field of science and
technology studies have been “unmasked“ as more or less contingent,
and thus questionable, products of their times. The outcome of the
recent discussion has been that there is something wrong with our
received notion of representation, which conceives of knowledge as
accurate representation of the independently existing world. In this
sense the crisis of representation is not a new phenomenon, but can be
dated back at least to the beginning of the 20th century, where it can
be found for instance in abstract art in general, dadaism and cubism.
When it comes to science, already both John Dewey and Martin
Heidegger criticised the idea of knowledge as that of spectating the
world as a picture,1 a theme that was picked up by Richard Rorty in
his seminal and controversial Philosophy and the Mirror of Nature
(1980). This book can be regarded as a successful manoeuvre in
bringing the notion of representation to the centre of the debate in
analytical philosophy as well.
Neither was representation of central interest in the philosophy of
science before the 1980’s. 2 The term representation was not often used,
and when it was, it was neither thematised nor questioned. In the
1980’s the situation started to change largely due to the semantic
approach to models. As the semantic approach loosened itself from
the linguistic paradigm of the received view and began to conceive of
theories as extra-linguistic entities, as families of (theoretical) models,
1 Cf. Dewey’s critique of the “spectator theory of knowledge” in Dewey
(1929), and Heidegger’s famous essay “The Age of the World Picture” in
Heidegger (1977).
the question then became how these entities were linked to the world.
Unlike with propositions and sentences, such terms as “true” and
“false” did not seem to be apt in dealing with the relationship
between models and their target systems. “Representation” was more
This turn away from truth to representation also implied for many
a definite change in how science was perceived. As Woodward has
noted: “The notion of [adequate] representation is a more general idea
than the notion of a statement’s being ‘true’, with representation
having to do with a qualitative notion of ‘fit’ between a model and
world—a notion that admits of degrees” (2002, 380).3 For some, the
notion of representation has even served as a way to circumvent the
question of realism altogether (see e.g. Frigg 2003, 12).
Now, twenty years later, a lively discussion concerning scientific
representation is in full sway but it is largely still conducted in
relation to scientific models and more recently, simulations (e.g.
Winsberg 1999). This is actually quite remarkable. It suggests, among
other things, how internalistic4 this discussion has been. Indeed,
Callender and Cohen have recently claimed that the specific problem
2 Of course there were notable exceptions among the analytical
philosophers, such as Sellars (1968) and Rosenberg (1974). However, it was
typical for the so-called linguistic turn of the analytical philosophy to use
such terms as meaning, sentence and proposition, rather than
representation (see Rorty [ed.] 1992). Nevertheless, it can be claimed, as
Rorty (1980) in fact does, that in the background of this linguistically
oriented philosophy there still loomed the old problem of representation,
which motivated both the empiricists and idealists in their battle against
scepticism. See Popkin (1980) on the constitutive role of the problem of
scepticism in Western philosophical thought.
3 Pace Woodward, the notion of truth can also be made to admit of degrees:
sentences or statements can be classified according to their closeness to
truth, that is according to their degree of truthlikeness or verisimilitude (see
e.g. Niiniluoto 1999).
4 By “internalistic” I do not refer to the internalism/externalism debate in
epistemology (or in the philosophy of mind) but to the discussion in the
sociology of scientific knowledge on whether or not science should be
approached as a conceptual and rational activity only. To the extent that
philosophers have tended to take this view on science, it has led them to be
internalistic also in the sense of rather staying inside their own discipline.
The relatively recent “return” to naturalism has changed this attitude quite
a bit (see the section 1.3).
14 Tarja Knuuttila
of scientific representation is a non-problem and that the discussion
of scientific representation has neglected “the philosophical work on
representation in general” (2005, 2). I think that there is a grain of truth
in their accusation even though it misconstrues much of the work
done in the burgeoning field of scientific representation. For one thing,
the philosophical writers on scientific representation have (pace
Callender and Cohen) considered the question of scientific
representation in a larger framework: for instance several authors
have related their discussion of representation in science to
representation in art (e.g. Suárez 1999, French 2003, Frigg 2003).
Ronald Giere’s work on scientific representation has drawn resources
from cognitive science throughout the years.
Having admitted this, it nevertheless seems to me that the present
discourse on scientific representation has remained a rather solitary
enterprise. It has been first and foremost interested in how (and by
virtue of what) models represent reality. That science is a hugely
complex and historically layered artefactual and representative
enterprise has been largely ignored in this discussion, which has
mainly been focused on ready-made models isolated from the
practices of their production.
Another peculiar character of the philosophical discussion of
scientific representation, also due to its internalistic character, is the
way it has succeeded in remaining nearly totally untouched by the
critical discussions concerning the notion of representation in the
other fields of inquiry. The discussants have agreed that models are
representative entities—otherwise, it has been supposed, they cannot
give us knowledge—even though no common understanding of what
representation involves has emerged. To be sure, already in 1983 Ian
Hacking in his Representing and Intervening proposed shifting the
focus of the philosophy of science from representing to intervening.
This was an interesting move indeed—and far ahead of its time—
taking into account that philosophers of science did not generally
understand scientific endeavour in terms of representation in those
By the time Hacking was writing the turn to intervening was
something that had already taken place in science and technology
studies (STS), where the so-called laboratory studies aiming at
accounting for how scientific facts were produced in scientific work
were accumulating (Latour and Woolgar [1979]1986, Knorr Cetina
1981, Lynch 1985a). But the question of representation kept on coming
back. It refused to be repressed—if only because the constructivists
presented their laboratory stories as descriptions of what “actually”
took place in scientific “practice”. This led not only to discussions of
the problem of reflexivity (Woolgar [ed.] 1988) but also to some
insightful studies on how representations are constructed with the
help of diverse means and procedures (Lynch and Woolgar 1990),
how they function as “working places” (Amann and Knorr Cetina
1990) and as “immutable mobiles” that “draw together” things and
diverse activities (Latour 1990).
Another source from which the philosophical work on scientific
representation could have taken nourishment, but in fact has not done
so (with such notable exceptions as Ronald Giere and Paul Thagard)
is cognitive science and the philosophy of mind. Especially in
cognitive science some fundamental representationalist assumptions
concerning the interrelationship between cognition and
representation have been called into question by researchers studying
the importance of environment and artefacts for what can be called
“embodied, situated and distributed cognition” (Clark 1999; see also
Varela et al. 1991, Hutchins 1995, Clark 1997). These studies have in
part contributed to the recent discussion of the notion of mental
representation, where different standpoints vis-à-vis representation
have been taken, ranging from adopting less metaphysically
impregnated words and expressions such as “tools for thinking”
instead of “representation “ (Dennett 2000) to arguing for a need to
“rehabilitate representation” (Smith in Clapin [ed.] 2002).
An interesting feature of the above-mentioned discussions in the
field of STS and cognitive science is that, irrespective of their
scepticism concerning our representationalist heritage, they have
actually resulted in fresh approaches to representation—leading also
to radical reconstructions of what should count as knowledge and
cognition. Nothing like that, I claim, has yet happened in the
philosophy of science. Even so, I think that at least two features of the
discussion of scientific representation should make this discussion
rewarding also from a more general point of view. Firstly, the recent
work on scientific representation has focused on specifying and
evaluating the relative merits of different dyadic and triadic accounts
of representation (e.g. Giere 2004, Suárez 2003 and 2004, Frigg 2002
and 2003, Bailer-Jones 2003). The relevance of this kind of theoretical
work is not limited to scientific representation only. Secondly,
philosophers are now seriously interested in studying the functioning
16 Tarja Knuuttila
of specific models in scientific practice. The volume Models as
Mediators, edited by Mary S. Morgan and Margaret Morrison (1999),
can be seen as an epitome of this kind of work. In the introductory
articles of that book they lay the basis for a research programme for
studying models from the point of view of scientific practice
(Morrison and Morgan 1999a, 1999b). Morrison and Morgan’s
approach to models as mediating instruments provides a potential
bridge between philosophical theorising and the more practice-
oriented approach of STS.
The almost non-existent dialogue between the aforementioned
different discussions of representation in general and scientific
representation in particular set the double agenda of this thesis: On
the one hand I situate the philosophical discussion of scientific
representation in the larger context of questioning representation that
has taken place in other fields, especially in STS. On the other hand I
take part in the discussion of models and representation in the
philosophy of science. The red thread that ties these interests together
is the material and artefactual approach to representation that I am
both developing and arguing for in this dissertation. What is special
about this approach is that it gives a non-representationalist account
of scientific representation.
1.2 Aims and plan
As already hinted by the double agenda adopted, the aims of this
study range from the more general to the more specific. At the most
general level, the goal of this study is to investigate the philosophical
or, more exactly, the epistemological problem of representation that
seems to underlie the recent critiques of representation. Thus, I first
study the various different reactions to the puzzle of representation.
Secondly, on this basis I attempt to diagnose the causes of the seeming
epistemological difficulties concerning representation. The hypothesis
guiding this study is that the so-called problem of representation: i.e.
“how is it possible for one thing to represent something else?” (Crane
1995, 9), is due to our representationalist heritage and that
representation need not be understood in a representationalist way.
This leads me, thirdly, to set as my next goal the developing of a non-
representationalist approach to scientific representation. Since the
discussion of scientific representation has so far taken place in the
context of modelling, my aim of developing an alternative approach to
scientific representation entails, fourthly, answering the question of
what kinds of things scientific models are and how they give us
The underlying assumption of this study is that scientific
representation is a very good place to study the puzzles of
representation and representationalism. This is due to science being
perhaps the one human activity that is most critically dependent on
the representationalist notion of representation. We are accustomed to
thinking that our scientific theories and models refer to an
independently existing reality outside of them and that they can at
best be considered as truthful, or accurate, descriptions of it. On the
other hand, however, sciences typically study unobservable entities
with the most sophisticated investigative machinery created by
humans so far, which makes the question of scientific representation
especially challenging. Last but not least, to inquire into
representation is to inquire into the nature of knowledge as well. The
epistemic value of models has traditionally been attributed to their
representative dimension. Thus studying how scientific models give
us knowledge—and what this knowledge is like—helps us to
understand more generally in which ways representation, knowledge
and, ultimately, cognition are intertwined.
That this dissertation consists of individual articles has some
consequences for how its aims are tackled. Since articles typically
focus on relatively narrow questions, part of the background of these
more specific topics is provided by the present summary (Part 1).
Hence to contextualise the individual articles and to give them a
common raison d’être, I discuss first what is meant by the notions of
representation and representationalism and how these concepts are
related to one another. Then I take up the phenomenon of the crisis of
representation from the point of view of scientific representation.5 I
study the diverse ways in which certain recent authors have sought to
settle the problem of representation and ask whether or not we should
renounce the concept of representation altogether or rather strive to
interpret it in a different way. As I do not think that it is possible—or
5 Despite the institutional differences and different goals of art and science,
the question of representation provides a common point of view from
which the specific features of art and science can be fruitfully compared
and examined (see e.g. Baigrie, ed. 1996 and Jones and Galison, eds. 1998).
18 Tarja Knuuttila
even desirable—to renounce or deconstruct the notion of
representation entirely, my discussion of models is an attempt to
show how the problem of representation could be dealt with in a
positive way—yet taking into account the recent critique of
I suggest that in trying to approach representation in a fresh way
one should resist the traditional temptation to ground the
representative power of actual representative artefacts in primary
immaterial and ideal representations (e.g. abstract structures, “mental
models” or concepts). Instead, the actual processes of scientific
representing should be analysed from the epistemic point of view. This
means focusing on the production and use of materially embodied
representations. Along these lines I approach models, which have
generally been regarded as representations, and advance an
alternative conception of models as epistemic artefacts.
The main points of this thesis can be summarised as follows: i)
models are human-made artefacts which are used to interact with the
world rather than merely to represent it; ii) thus instead of being
abstract theoretical constructions they are better conceived of as
entities that are materialised in some media; iii) the epistemic value of
models accrues importantly from their material dimension, which
explains why models have many other epistemic functions besides
that of representing the world; iv) the representational function of
models should not be approached in “representationalist” terms; v)
representation is best understood as activity that relies both on the
medium-specific affordances of the material sign-vehicle and the
intentional process of relating the sign-vehicle to its object.6
1.3 A note on the method
Being careful examiners of other scientists’ methods and
presuppositions, philosophers of science were remarkably silent on
their own method until recently when the discussions of naturalism
(see e.g. Giere 1988 and 1999; Kitcher 1992, Downes 1993) and the
disunity of science (Dupré 1993, Galison and Stump 1996) swept over
the field. As to the discussion of naturalism Philip Kitcher (1992)
argues that philosophy of science is slowly (re)turning to the
6 I am grateful to Mauricio Suárez in helping me to explicate the main
points of my thesis.
naturalist epistemology. It is based on rejecting either one or both
“Post-Fregean” assumptions of how to pursue epistemology.
According to them, epistemology should be both an apsychological and
an a priori investigation—“knowledge is to be given a ‘logical
analysis’” (57). These assumptions were constitutive for analytical
philosophy for decades. Accordingly, “for at least a period,
philosophers could be confident of their professional standing,
priding themselves of a method—the method of conceptual analysis—
which they, and they alone, were trained to use“ (54).
On the level of doing philosophical research, the naturalist
epistemology means that we should take seriously the results of
scientific research. According to Giere “any conclusions one reaches
about the nature of science are subject to criticism based on theoretical,
historical, psychological, or social investigations into particular
scientific practices” (1999, 5; see also Giere 1988, 8-10). Thus as a
result of adopting the naturalist epistemology my work is also
multidisciplinary. In tackling the problem of representation I have
drawn resources from semiotics, STS, activity theory (see Engeström
and Escalante 1995, Engeström et al. 1999, Miettinen 2001) and
distributed cognition. What is common to these approaches is that
they, in one way or another, concentrate on mediation provided by
signs or tools in explaining human activity and cognition.
Furthermore, several of the articles of this study participate also in the
philosophically inclined discussions in the aforementioned fields.
Last but not least, my approach to the problem of representation is
uncompromisingly naturalist in the sense that I do not want to posit
any fundamental representations behind those things that we actually
use in scientific practices in order to represent.
Provided that we accept the results of empirical science as part of
philosophical reasoning, should we then stop at that? Is there a place
for empirical study in philosophical argumentation? I think that there
is, if only because a lot of research done in the philosophy of science
proceeds by presenting cases from specific disciplines, taking
historical data into account as well. Since I approach representation
and modelling from the point of view of scientific practice, I have felt
a need to get some grasp of the practices themselves. Thus in one of
the studies concerning modelling (“From Representation to
Production: Parsers and Parsing in Language Technology”) I present
a case of modelling in language technology in which I use interviews
of the researchers working in the field. This is partly because research
20 Tarja Knuuttila
on the procedures and phases of modelling is largely non-existent,
and knowledge of it remains largely at the tacit participant-level.
My empirical data consists of 24 transcribed semi-structured
interviews that I conducted in two phases between the years 2000 and
2002 and in spring 2004 with researchers and developers of language
technology who have either been doing language technological
research in the Research Unit for Multilingual Language Technology
in the Department of General Linguistics in University of Helsinki, or
have been otherwise affiliated to the group. The key researchers and
developers were interviewed more than once for a relatively long
period (from 1 ½ hours to 3 hours). The written material on which the
case study is based consists of publications of the researchers
interviewed, other literature on computational linguistics and
language technology, and some reports and documents concerning
the Research Unit for Multilingual Language Technology and
language technology in Finland in general. I have also benefited from
casual discussions with the language technologists, attended some
courses on language technology and taken part in language
technological events and activities. Moreover, I have been observing a
language-technological project funded by National Technological
Agency of Finland (TEKES), in which researchers from the Department
of General Linguistics have participated. Some of the researchers
interviewed also commented on the earlier drafts of my papers on
language technology, and Study 3 “A Parser as an Epistemic Artefact:
A Material View on Models” was co-authored by Atro Voutilainen, a
member of that team.7
As for the claims of the disunity of science, it is already widely
acknowledged that physics may not provide an appropriate model for
approaching other sciences. Yet, in discussions of modelling the
theoretical models of physics are typically taken as representatives of
models in general. In this study I have instead used models from the
fields of computational linguistics and artificial intelligence as
examples. One of the purposes for choosing them has been to show the
limitations of the traditional notion of models as representations of
their target systems. As the traditional notion fits inadequately these
models, I wish to articulate an approach to models that would suit
them better. I do not claim that the computational models I have
chosen can serve as representatives of models in general—if only
because no model can. Nevertheless, the importance of computational
models and simulations in science is constantly increasing and in fact
creating new fields of inquiry. On the other hand, I would like to
suggest that the notion of models as epistemic artefacts is useful as a
general approach to models, as a way of viewing them from another
angle than has been customary.
2. Representation and its discontents
2.1 The concept of representation
The word representation comes from Latin raepresentare “to make
present or manifest or to present again” (Pitkin 1967, 241). It was
almost exclusively applied to inanimate objects, the political idea of
human beings representing each other appeared only after the
fourteenth century (ibid.). In her etymological study of the concept of
representation Pitkin describes the meaning of raepresentare in the
following way:
It can mean to make them [inanimate objects] literally present, bring
them into someone’s presence, accordingly it also comes to mean
appearing in court in answer to summons, literally making oneself
present. It can also mean the making present of an abstraction through
or in an object, as when a virtue seems embodied in the image of a
certain face. And it can mean the substitution of one object for another,
instead of the other (Pitkin 1967, 240).
The modern usage of representation developed from the latter
meaning of the aforementioned quotation (Volli 2003).8 Thus, for
instance, Prendergast (2000) discriminates between two basic
meanings of the term: between the older “re-presentation” and the
7 It appears to me that since scientific inquiry is nowadays so specialised
and heterogeneous, the philosophers and other researchers who are
studying science should work collaboratively, or at least in constant
dialogue, with the scientists whose work and discipline they are examining
(unless they themselves have some experience of the field of inquiry in
question). The ethnographical approach to scientific work, once so popular
among STS researchers, was successful in making scientific work and its
material dimension visible. However, I do not think that “anthropological
observation” per se could help us get any more than a superficial
understanding of the content and methods of different disciplines.
IntroductionRepresentation and its Discontents
22 Tarja Knuuttila
more recent “standing for”. Instead of striving to produce the illusion
of presence, to re-present, the representative relation of standing for is
that of substitution, of substituting something absent with something
present. The substitution can take the form of simulacrum, but it is a
form of representation as making present (in the older sense of
representation) only if it produces an illusion of presence by virtue of
being an accurate replica of the real thing.
As distinguished from “re-presentation”, representation as
“standing for” is not to be confused with the thing itself. Pitkin notes
how representation as “standing for” seems to require a certain
distance or difference as well as resemblance and correspondence
(1967, 68). Representation as “standing for” is typically approached
through the metaphors of portrait, map or mirror: what they have in
common is that they are all renderings of an “original” in a medium
different from it. Thus the function of representation as “standing for”
is to bring knowledge: it “consists of the presence of something from
which we can draw accurate conclusions about the represented,
gather information about the represented because it is in relevant
ways like the represented” (Pitkin 1967, 81). When applied to political
sphere this idea of representation provided a justification for
representative democracy. The substituting representatives for the
whole people were like a copy, a second-best approximation of direct
democracy, which would have been the ideal system according to the
Anglo-American democratic ideology (ibid., 84).
As for my focus on the epistemic value of models, I find Pitkin’s
evaluation of representation as “standing for” in the political realm
insightful: “…[this] view of representation … does not allow for an
activity of representing. … It has no room for any kind of representing
as acting for, or on behalf of, others; which means that in the political
realm it has no room for the creative activities of the representative
legislature, the forging of consensus, the formulating a policy” (90). In
fact, I am going to claim in this thesis that treating models as
8 There have been, of course, many other conceptions of representation,
but representation as “standing for” seems to be the most common and
prevailing one (see e.g. Peirce CP 2.273; Palmer 1978, 262; Crimmins 1991,
791). Even though typical for the modern period, it originates from the
scholastic aliquid stat pro aliquo. However, Nöth (1995) claims that medieval
philosophers used more often another formulation, namely supponit aliquid
pro aliquo, which means “something serves in place of something else” (84).
predominantly representative entities—as surrogates that stand for—
ignores their material and interactive side, from which their heuristic,
mediating and many other epistemic capabilities arise.
2.2 Representationalism
Representation as “standing for“ seems to be embedded in
representationalism. According to representationalism, the sensing and
knowing mind cannot have direct acquaintance with its objects. It can
approach them only through ideas, which are assumed to represent
them. Thus knowledge is conceived of as an assemblage of
representations that reproduce accurately, i.e. stand truthfully for,
what is outside the mind (or, after the “linguistic turn”, outside
linguistic description—or other external representations) (see e.g.
Rorty 1980, 3-6; Gutting 2001, 336). The crucial difficulty of
representationalist theory is that the mind “supposes” that its ideas
represent something else but it has no access to this something else
except via another idea.
The conception of a mind as container of ideas is most commonly
associated with Locke. He was criticised already by other empiricists,
but he nonetheless succeeded in articulating the way the early modern
age thought about perception and knowledge. In his Les mots et les
choses (1966) Michel Foucault claims that this age, the “Classical Age”
as he calls it, was that of representation par excellence. It strove for a
universal method of analysis that would perfectly order
representations and signs to mirror the real orderings. Language
coincided with thought, it was the transparent system of signs that
represented the representation, that is, the ideas. Moreover, it was
assumed that language by its very nature made successful
representation possible, which led Foucault to characterise “classical
language” as “the common discourse of representation and things”
(1970, 311). Thus the Classical age relied on the ability of language to
represent the world as it is, but as a consequence man as representer
was left out of the “picture”. This is what Foucault aims to show with
his lengthy analysis of Velásquez’s painting Las Meninas, to which the
first chapter of Les mots et les choses (1966) is dedicated. Foucault takes
Las Meninas to be a perfect image of Classical representation, which
makes the different aspects of representation visible, yet leaving them
dispersed because there is no “unifying” subject. The mirror behind
the painter shows what he sees and what he is in the process of
Representation and its Discontents
24 Tarja Knuuttila
depicting on his canvas, the ruling couple Philip IV of Spain and his
wife Mariana. Yet neither the painting itself nor the act of painting are
shown in Las Meninas: the painter is standing back from his work; if he
had been in the process of painting, he would have disappeared
behind the canvas. Moreover, as spectators, we realize that the royal
couple is in fact standing in our place—everything else is depicted
except the subject of representation and the act of representing.
As the Modern age no longer treated discourse as a perfectible and
transparent medium, the representational capacity of man became
available as a distinct object of knowledge. Foucault remarks how
Kant and the age that followed him did not complain about man’s
limited capacity to know how things are. They rather turned it into an
advantage, a condition of all factual, empirical knowledge.
Epistemology, as an attempt to explain and justify how things in
general can be given to representation, was born. Foucault writes:
The Kantian critique…questions representation, not in accordance with
the endless movement that proceeds from the simple element to its
possible combinations, but on the basis of its rightful limits. Thus it
sanctions for the first time that event in the European culture which
coincides with the end of eighteenth century: the withdrawal of
knowledge and thought outside the sphere of representation. That
space is brought into question in its foundation, its origin, and its
limits: and by this very fact, the unlimited field of representation,
which Classical thought had established…now appears as a
metaphysics (Foucault 1970, 242).
Interestingly, whereas for Foucault Kant broke away from the
representationalist paradigm, in the alternative philosophical story of
representationalism provided by Richard Rorty, Kant more or less
perfected it and thus laid down the lines for the representationalist
tradition of (analytical) philosophy to come.9 In his Philosophy and the
Mirror of Nature (1980) Rorty claims that two assumptions have
dominated the representationalist tradition of Western philosophy.
One is the Kantian, according to which human knowledge should be
understood as a special relation between the objective knowledge
substrate offered by the world and the active cognitive abilities of the
subject. The other assumption is the platonic belief that there is a way
to portray things that can reach them as they are. However, also for
Rorty the representationalist tradition has roots in the beginning of the
modern age. Rorty sees representationalism as a reaction to the
emergence of the new natural sciences, which led to the relocation of
the intentional and phenomenological in the realm of the mental. Thus
philosophy turned to inspect the ideas in the mind. The rationalists
and the empiricists alike tried to establish a secure foundation for
knowledge, and that foundation was to be found from inside of the
mind. The earlier distinction between appearance and reality was
replaced by that between inner and outer and the question became
“How can I escape the veil of ideas?” (1980, 160). Kant provided this
an elegant answer, according to Rorty, by changing the empiricist
question of psychophysiological mechanisms into a discussion of the
legitimacy of science itself (1982, 145). He took up an unresolved
scientific problem—the relation of sensations to their objects—and
turned it into a question about the possibility of knowledge, which
question was to be resolved in the sphere of representation. Thus
Rorty holds Kant responsible for the foundational pretensions of
philosophy: “the Kantian picture of concepts and intuitions getting
together to produce knowledge is needed to give sense to the idea of
‘theory of knowledge’ as a special philosophical discipline” (1980,
It is important to bear in mind that Rorty’s anti-
representationalism is not only an attack against the notion of
representation as “mirroring”, corresponding or resembling but even
more importantly against what he calls “privileged representations”
9 Rorty seems not to pay enough attention to the fact that for Kant the
activity of thinking is not mimetic but constructive (Rusterholz 2003, 53).
Kant can be considered one of the founding fathers of constructivism. As
Matti Sintonen (1995, 1996) has argued, the gist of Kant’s transcendental
method and of the entire “Copernican Revolution” was in acknowledging
the active contribution of the mind. The transcendental method in
epistemology requires not just that one examines the objects of putative
knowledge but also that one focuses on the conceptual tools used in
knowledge acquisition as well as on what one must do to obtain
knowledge. Nor does this constructivism confine to the representation of
the external world. Jaakko Hintikka (1973, 1974) has maintained that there
is a close relationship between Kant’s transcendentalism and his theories of
space, time, and mathematics. On these themes, see Sintonen (2005,
forthcoming). Also Sismondo (1996) and Hacking (1999) consider Kant to
be the forefather of constructivism, Sismondo even dubs the whole STS
constructivism the “Neo-Kantian science”.
Representation and its Discontents
26 Tarja Knuuttila
and their privileging of philosophy.10 This clearly emerges from the
following short resume of his antirepresentationalist story:
To describe this development as a linear sequence is of course
simplistic, but perhaps it helps to think of the original dominating
metaphor as being that of having our beliefs determined by being
brought face to face with the object of belief (the geometrical figure
which proves the theorem, for example). The next stage is to think that
to understand how to know better is to understand how to improve the
quasi-visual faculty, the Mirror of Nature, and thus to think of
knowledge as an assemblage of accurate representations. Then comes
the idea that the way to have accurate representations is to find, within
the Mirror, a special privileged class of representations so compelling
that their accuracy cannot be doubted. The privileged representations
will be the foundations of knowledge… Philosophy as epistemology
will be the search for the immutable structures within which
knowledge, life and culture must be contained—structures set by the
privileged representations, which it studies. (Rorty 1980, 163).
Rorty’s reconstruction of the history of Western philosophy is not only
provocative but also revealing. It is indeed not difficult to find
privileged representations in which philosophers have sought to
ground their endeavour: ideas, concepts, logical forms, Husserlian
“essences”… and structures, which have featured importantly in the
discussions of models and representation.11
10 According to Rorty, the turn to language did not entail any notable
change for the representationalist tradition of philosophy, since analytical
philosophy attempted to formulate many traditional epistemological
problems by linguistic means. In Why Does Language Matter to Philosophy?
(1975) Hacking expressed a similar view. He calls the early modern age the
“heyday of ideas”, whereas he considers himself to be writing in the
“heyday of sentences”. Hacking notes how language was not an interesting
problem during the heyday of ideas, which was instead interested in
“mental discourse”. Hacking nevertheless claims that “the structure of the
recent philosophical problem situation... is formally identical to the
seventeenth century one, but the content of that structure is different”
(1975, 167). The problem of the “interface between the knower and the
known” has remained much the same, but public discourse has replaced
mental discourse and sentences have replaced ideas.
2.3 Against representation?
The remarkable thing about the “postmodern” discussions of
representation and representationalism is that they have been
conducted nearly exclusively in negative terms. The concept of
representation has been found lacking both in theory and practice.
Theoretically, it seems that our Kantian heritage has left us captives of
the sphere of representation. Either representation cannot reach the
things in themselves, providing only a deficient substitute for them, or
then it creates an effect of reality that it strives to capture, but in vain
(because there is nothing to capture—according to the most adamant
poststructuralists). Hence it is hardly astonishing that the question of
representation has been a subject of constant epistemological concern
and even horror (an expression by Steve Woolgar). Moreover, at a
practical level, given the diversity of the representations we use and
the complexity of our practices of representation, it seems clear that the
general concept of representation as “standing for“ does not help us
much in explaining what kinds of things representations are and how
they are supposed to represent.
In this situation many have felt that something needs to be done—
but what? At least three different positions vis-à-vis the problematics
of representation have been taken. Some wish to renounce
representation and simply evade the problem as unproductive and
unfruitful. Then, instead of putting the question of representation
aside, one can attempt to deconstruct the notion—to show that in
talking about representation we are actually talking about different
ways of rendering, referring to, denoting, indicating, etc. that do not
share any common core that entitles us to call them representations.
Finally, one can seek to reconstruct our notion of representation in
such a way that it pays due attention to the criticism.
Renouncing representation
In Philosophy and the Mirror of Nature Rorty (1980) concludes that it is
time to break with the received epistemological framework and to stop
11 On the other hand this predilection for privileged representations has
been typical of the cognitive sciences as well, where such hypothetical
entities such as concepts, symbolic structures, mental models, prototypes,
schemes, etc. have been ascribed to our minds to explain our cognitive
Representation and its Discontents
28 Tarja Knuuttila
thinking of knowledge in terms of representing accurately that what is
outside the mind and, subsequently, language. Actually, we should
stop thinking about language in terms of representation entirely:
We must get the visual, and in particular the mirroring, metaphors out
of our speech altogether. To do that we have to understand speech not
only as not the externalizing of inner representations, but as not a
representation at all. We have to drop the notion of correspondence for
sentences as well as for thoughts and see sentences as connected with
other sentences rather than with the world. We have to see the term
“corresponds to how the things are“ as an automatic compliment paid
to successful normal discourse rather than as a relation to be studied
and aspired throughout the rest of the discourse. (371-72, italics mine)
As noted above, Rorty is predominantly interested in targeting
philosophy as a discipline that takes upon itself the task of
legitimising the knowledge claims of the other areas of culture.
Without the notion of knowledge as accuracy of representation and
the consequent privileged representations, the claims that
“philosophy should consist of ‘conceptual analysis’ or ‘phenom-
enological analysis’ or ‘explication of meanings’ or examination of
‘the logic of our language’…would not have made sense” (1980, 12).
Rorty has actually been rather uninterested in science.12 However,
claims similar to his have been presented also in the context of
science. Ian Hacking and Andrew Pickering, among others, have
contested the centrality of representation for our notion of what
science is about. Hacking (1983) aims to turn from truth and
representation to experimentation and manipulation:
Realism and anti-realism scurry about, trying to latch on to something
in the nature of representation that will vanquish the other. There is
nothing there. That is why I turn from representing to intervening.
Pickering (1995), for his part, contrasts “representational and
performative idioms of thinking about science”:
12 Hacking notes, rightly I think, that “Rorty’s version of pragmatism is yet
another language-based philosophy, which regards all our life as a matter
of conversation” (1983, 63).
The representational idiom casts science as, above all, an activity that
seeks to represent nature, to produce knowledge that maps, mirrors,
or corresponds to how the world really is. In so doing, it precipitates a
characteristic set of fears about the adequacy of scientific representation
that constitute the familiar philosophical problematics of realism and
objectivity… (5)
…My suggestion is that we should see science (and of course,
technology) as a continuation and extension of this business of coping
with material agency. And, further, we should see machines as central to
how scientists do this. (7)
In the face of these claims I wonder whether it is desirable, or even
possible, to do without the notion of representation? Is renouncing
representation just a question of talking differently, and consequently
even thinking differently—using perhaps the proposed notions of
conversation (Rorty), intervention (Hacking) or machines (Pickering)
instead? But are not scientific intervening, testing and producing
effects, or “coping with the material agency“ activities involving
complex representative artefacts and skills? If so, then what could be
meant by the need to get rid of representation or to contrast it with an
intervening or a performative idiom? I think that this former question
is especially justified in the context of models, which occupy an
interesting middle ground between representation and experimen-
tation. What seems to motivate the aforementioned attempts to get rid
of the notion of representation for good is that continuing to use the
concept would eventually evoke the representationalist conception of
science. I do not see that this needs to be the case, as the various
attempts to reconstruct or rehabilitate representation show.
Deconstructing representation
Instead of evading the question of representation, scholars in the field
of science and technology studies have attempted to tackle it head-on.
Inspired by ethnomethodology they have gone into “the field” to
observe how scientists “actually” go about representing. Their studies
deconstruct scientific representation into complex processes making
use of various “documents“ or “inscriptions“. Representations
become things that are worked upon, being ultimately “rich
depositories of ‘social’ actions“ (Lynch and Woolgar 1990, 5). In
examining these “representative documents”, the studies in question
begin by inquiring: “What do the participants, in this case, treat as
Representation and its Discontents
30 Tarja Knuuttila
representation?” (ibid., 11), rather than asking what is meant by
What follow from this approach are studies that meticulously
follow the “assembly line”, the processes of constructing scientific
representations. From this point of view scientific representation
appears as a subtle “dialectic of gain and loss“ (Latour 1995). It is not
just a question of reduction or simplifying. Some methods of
representation further fragment, upgrade and define the specimen in
order to reveal its details, whereas others add visual features for the
purposes of clarifying, extending, identifying, etc. Often the aim of
scientific representation is to mould the scientific object so that it can
assume a mathematically analysable form or to be more easily
described and displayed using different textual devices (see e.g.
Latour 1990, Lynch 1985b and 1990). Scientific representation widens
in these studies into an expanded process of circulating and
arranging diverse extracts, “tissue cultures”, photographic traces,
diagrams, chart recordings, verbal accounts and so on into a serial
order. The epistemological problem concerning the relation of
scientific representations to the reality surrounding us is due,
according to this view, to our forgetfulness of these expanded material
and social processes behind the finished representations. “Through
successive stages [sciences] link us to an aligned, transformed,
constructed world” (Latour 1999; 79, see also Latour and Woolgar
[1979]1986, 69).
I find this approach insightful and corrective of the received view
on representation in its stress on the complicated process of
producing scientific representations. Yet something seems to be
missing from these studies. This has been expressed nicely by Ronald
Giere, who entitles his review of them tellingly: “No representation
without representation” (Giere 1994). Despite all the careful studies
on how scientists go about representing, nothing is actually said
about what possibly makes the inscriptions examined representative—
except of course for the overall claim that the representation and
represented both emerge and merge in the same material process of
scientific work.
It seems that these studies are playing a double game. On the one
hand they proceed as if they had excluded any consideration of the
epistemological question of representation and were instead interested
in the representative practices as social phenomena only. As Michael
Lynch puts it: “For the sociological purposes, the ‘real object’ is the
representation in hand, e.g. the visual display, and not the invisible
phenomenon or abstract relationship out there” (1990, 154). On the
other hand these studies nevertheless have a distinct epistemological
aim to “explode“ the supposed homogeneous conception of
representation in order to make room for the “deeds performed, when
those [representational] items are embedded in action“ (Lynch 1994,
146). This is in line with ethnomethodology, which has inspired much
research in STS particularly with its disdain of theoretical discourse
(In Study 2 I examine this double agenda of STS in some detail).
Now there is something contradictory about this way of
proceeding. Firstly, in an effort to deconstruct the notion of
representation the protagonists rely on what is commonly taken as
representation already in their choice of case studies. Rather than
exploding the notion of representation these cases reveal what a
complicated phenomenon scientific representation is. Secondly, in
order to challenge what they take to be the philosophical view on
representation, these studies claim that they show us what really goes
on in scientific representation. But then the question is whether
ethnomethodologists and other STS scholars themselves are relying on
a rather traditional notion of representation—the very same notion
they set out to demolish.
Reconstructing representation
Why is representation such a bad word for so many? It appears that
this reaction results from our inherited representationalist conception
of knowledge, which ties knowledge firmly to representation and for
which representation is a static relation between that which
represents (statements of language, ideas in the mind, abstract
structures, etc.) and that which is represented (“reality“, the “world“,
some physical system, and so on). Whether the representation is true
of its intended object is a question of observing a correspondence
(analysed most often as isomorphism or similarity) between the
representation and that which is represented. Moreover, according to
this view, that which is represented, i.e. reality, consists of a fixed
totality of mind- and representation-independent objects. I do agree
with the critics of representation that this view of representation
deserves to be set aside. However, I do not think that this conclusion
should lead us to banish the notion of representation (or the
representative idiom) altogether—if only because the question of
Representation and its Discontents
32 Tarja Knuuttila
representation seems to stay with us however uncomfortable we are
with it. The aforementioned struggles of STS scholars with
representation provide good examples of this.
On the other hand the notion of experimental manipulability,
which Hacking (and to a certain extent Pickering) professes, does not
really succeed in its task of replacing representational realism. In the
first place, it fits well only those sciences that allow for experimen-
tation, and thus it is not particularly suitable for the social sciences. To
illustrate, let us take the case of economics. Mäki (1996) argues that
whereas the existence of the entities to which the theoretical terms
refer is problematical for the natural sciences, the problems of
economics are different. The argument from experimental
manipulability cannot easily be applied to economics. Apart from the
difficulty of conducting experiments in economics, experimental
manipulability is not needed for ontological realism in economics
because we are not uncertain to which entities many theoretical terms
of economics (such as “consumer” or “firm”) refer. Furthermore, as I
have already noted, efforts to avoid or renounce representation do not
work even in the experimental sciences except on the level of general
declarations. This is because as soon as we start to inquire how
scientist intervene, we find ourselves engaged in complex represen-
tative processes involving specialised artefacts that record, chart,
trace, etc.
Since representation cannot so easily be dismissed, one possibility
of tackling the problem of representation is to loosen it from its
representationalist grip by reconstructing it. This effort is actually
being made on several fronts, both in philosophy and in other fields of
study. In regard to my line of argumentation I find three approaches
especially relevant: the work on embodied and distributed cognition
in cognitive science, the recent pragmatic approaches on models and
scientific representation in the philosophy of science, and the
theorizing and historical studies of scientific objects in STS. One way
to see how these very different approaches contribute to a non-
representationalist conception of representation is to see how each of
them attacks one of the following three basic assumptions of
1) Knowledge consists of a collection of (internal) representations
2) These representations correspond accurately to the bits and
pieces of reality
3) Reality is already made up: it consists of a fixed totality of
representation-independent objects
Recent work in the field of embodied and distributed cognition has
questioned the first assumption by concentrating on what has been
called “cognition in the wild” (Hutchins 1995). This economical and
embodied cognition uses external scaffolding, environmental clues
and cheap tricks in its cognitive tasks instead of creating complete,
internal representations of the world. The argument is that the human
brain evolved originally to coordinate the body, which made cognition
action-oriented rather than reflective. Instead of one single central
processor controlling all the cognitive activities, evolution preferred a
solution with many, more specialized processors (see e.g. Varela et al.
1991, Clark 1997 and 2003). On this basis it is possible to claim that
our cognition is distributed between individuals and artefacts
(Hutchins 1995) and that it is also largely skill-based and tool-using.
Thus for instance Dennett stresses the importance of “florid
representing”, which depends on the “objectification” of certain skill-
based contents into tools suitable for exercising other skills (Dennett
2000). Latour takes up this theme of the cognitive value of creating
objects about objects when he argues that the possibility of
superimposing, reshuffling and recombining signs and inscriptions
can engender totally new phenomena (Latour 1990). Giere (2002a) has
approached scientific cognition as distributed, claiming that “most
models in science, even in classical mechanics, are too complex to be
fully realised as mental models” (10).13 The recent, active discussion of
expertise and the tacit dimension of knowledge (see e.g. Dreyfus and
Dreyfus 1986, Collins and Evans 2002, and Study 5 of this
dissertation) is partly related to these developments.
Representation and its Discontents
13 An important precursor of this line of thinking was Lev Vygotsky, who
already in the 1920’s suggested that the development of a child’s higher
psychological functions is a result of the internalisation of social forms of
action mediated by signs and tools (Vygotsky 1978). Cognition is thus not
only embodied but also cultural. Interestingly, we do seem to use
internalised public tools, such as language, when thinking and trying to
solve cognitive problems “in our heads”.
34 Tarja Knuuttila
Whereas the first approach concentrates predominantly on the
nature of knowledge and cognition, the second line of attacking
representationalism focuses directly on the relation of representation
to what is represented. These studies criticise the assumption that
representation is a dyadic relation of correspondence between the
representative vehicle and its target. In the field of the philosophy of
science a host of pragmatic approaches to representation have recently
been proposed (Suárez 2004, Giere 2004, Bailer-Jones 2003, Frigg
2003). They all stress the importance of the use and users for
representation. Consequently, what is common to these approaches is
that they focus on the intentional activity of representation users and
deny that the relation of representation to what is represented can be
based only on the respective properties of the representative vehicle
and its target object (see Suárez 2004). Instead of asking “how do
models represent the world?” one should rather ask “how models are
used to represent the world” (Giere 2004). In arguing for three- or even
four-place relation between representation and the represented, these
pragmatic approaches in fact follow the tradition set up by C. S. Peirce,
for whom the sign relation was irreducibly a triadic relationship
including the sign (or “representamen”), the object and the
interpretant. Furthermore, Peirce understood sign activity, semiosis, in
terms of inferentiality and mediation, both of which have been
discussed in recent literature on models and representation.14
Finally, representationalism can be challenged by denying that it
makes sense to think about the world as consisting of representation-
independent objects. This is because “objects” are not available to us
except through representation. From a thoroughly constructivist point
of view, talking about representation should not pose any problem,
since representations are freed from representing the world as it is.
Quite the contrary, we construct scientific, artistic and other objects by
way of representing them. This theme has been insightfully explored
by the work on ”epistemic things” or, alternatively, “epistemic
objects” by Hans-Jörg Rheinberger (1997) and Karin Knorr Cetina
(1999, 2001). In their work, conventional scientific entities appear as
open-ended epistemic things that are actually effects of manifold and
interwoven representative practices. For neither Rheinberger nor Knorr
Cetina does this open-endness mean relativism, since scientists are
not free to construct objects as they wish—representative practices
constitute part of “experimental systems” (Rheinberger) and are
constantly “in the process of being materially defined” (Knorr Cetina
2001, 181). The implication is that the material world somehow resists
scientists’ attempts to mould it, the “the things strike back”, as Latour
(2000) has put it. 15 In scientific practice, this constraint is actively
sought, which is most evident in the case of experimentation.16
However, I will argue later in the individual studies of this
dissertation that this is also done in modelling, even though there also
the constraint itself is constructed and not just the setting in which it
is aspired after, as in experimentation. If scientific objects are treated
as constructions, then their historical trajectories become possible
topics of their own. Thus Lorraine Daston claims in her introduction
to the collection of essays Biographies of Scientific Objects that scientific
objects can be both real and historical; that is, their existence is
relative: “new scientific objects pour forth, and old ones fade away”
(2000, 5). That this way of thinking seems strange to most
philosophers is due exactly to the fact that they tend to think that the
scientific objects were always there waiting to be found.17 But then the
history of science does not offer much support for this thought.
14 On Peirce’s later stress on mediation, see Bergman (2004, ch. 4). Eco
(1984, ch. 1) and Merrell (1995, ch. 1) have in turn stressed the inferential
character of Peirce’s conception of the sign.
15 The “objectualists” appear to claim that reality is both a construction and
a constraint, which is of course confusing. It appears to me that behind their
way of seeing things there is actually a distinction between reality and the
real. This distinction, which can be traced to the work of French
psychoanalyst Jacques Lacan, somehow captures the flavour of the
poststructuralist theorising of the ineffable. (French poststructuralist theory
has had an important influence on both Knorr Cetina’s and Rheinberger’s
views on epistemic things). Reality is an effect of the coherence of our
representative practices: it is, as Hacking has aptly put it, a second-order
concept that follows from our practice of representation. “The world has an
excellent place, even if not the first one… It was found by conceptualising
the real as an attribute of representations“ (1983, 136). Thus reality is, from
the very outset, a fictive construction which is continuously questioned by
the real. The real is the unrepresentable ground of the subject’s own being
and that of the world beyond it (see Boothby 2001, 12). Thus the real
constrains the way reality can be constructed and maintained (see also
Ragland-Sullivan 1996, 192).
16 For instance Ludwik Fleck claims that the “general aim of the
intellectual work” is to create a “maximum thought constraint”: “The
work of the research scientist means that in the complex confusion and
chaos he faces, he must distinguish that which obeys his will from that
which arises spontaneously and resists it” (1979, 95).
Representation and its Discontents
36 Tarja Knuuttila
I find all these different strategies of shattering the assumptions
underpinning representationalism important and have tried to
accommodate them in my work on scientific models. Through the
notion of models as epistemic artefacts I have tried, firstly, to detach the
epistemic (and cognitive) value of models from their representational
status. Secondly, as artefacts, models are used in scientific practice
many ways, one of which is representation. I am not contesting
representation itself but rather the way it is conventionally
understood—thus I side with forces of the pragmatists of
representation. Thirdly, my view on models as first and foremost
artefacts rather than representations implies that I do not think that
there is any pre-packaged collection of objects available for us to
represent. Rather, it seems to me that models help us to imagine what
there could be, as Morgan (2004) and Frigg (2003, ch. 5) have pointed
out. With the notion of the epistemic artefact I have nevertheless
wanted to show how importantly tied our imagination is to materiality
and to the established ways of using various representative media.
Moreover, the artefactual account of models easily lends itself to the
consideration of the processes by which models and their subsequent
productive properties are built—an aspect that has been stressed by
the STS studies on representation (see above).
Next I will turn into a discussion of scientific models. To ground
the idea of models as epistemic artefacts I will provide a short
overview on how models have been treated in the philosophy of
science and how the question of representation has arisen in that
17 The idea that our objects of knowledge are not independent of our
knowing dates back at least to Kant, see the footnote 10. Several
philosophers have also attacked the notion of the “ready-made world” (see
e.g. Putnam 1982, Tuomela 1985). What distinguishes these views from the
recent interest in scientific objects is that, unlike the philosophical
discussion, the study of scientific objects does not understand construction
predominantly in conceptual and linguistic terms but rather as social and
material activity.
3. Scientific models in the philosophy of science
The discussion of models in the philosophy of science has
heterogeneous beginnings, testifying to a variety of theoretical, formal,
and practical aspirations that appear to have different and even
conflicting goals (Bailer-Jones 1999, 32). Thus in addition to
approaches that have predominantly been interested in the pragmatic
and cognitive role of models in scientific enterprise, attempts have
been made to establish, within a formal framework, what scientific
models are. The syntactic view of models, once the “received view“,
and the semantic approach to models, the prevailing model-theoretic
approach, are both attempts of this kind.
3.1 Syntactic and semantic views on models
According to the syntactic view, a model is designed to give an
interpretation of an uninterpreted formalism or calculus. Thus,
according to Ernest Nagel, for example, a model is an interpretation of
“the abstract calculus which supplies some flesh for the skeletal
structure in terms of more or less familiar conceptual or visualizable
materials“ (1961, 90). For the proponents of the syntactic view a
scientific theory was such an uninterpreted or partially interpreted
formalism, a purely syntactic structure consisting of a set of axioms.
To interpret a theory was to specify a model for it, which makes all the
axioms of the theory true (or false). Consequently, a model for a theory
T could be defined as a set of true propositions with the same formal
structure or calculus as T (96).
The semantic conception of models contested this “linguistic“ view
of models by replacing the syntactic formulation of the theory with the
theory’s models, which are non-linguistic entities. In this view,
theories are not assemblages of propositions or statements, but
“extralinguistic entities, which may be described or characterized by a
number of different linguistic formulations“ (Suppe 1977, 221). These
extralinguistic entities—models—are taken to be structures that are
defined either by the use of set-theoretical predicates (e.g. Suppes 1961,
da Costa and French 1990) or by the use of suitable mathematical
language (e.g. van Fraassen 1980). The emergence of the semantic
conception dates back to the 1960’s with impulses both from
mathematics and computer science (see Suppe 1989, Prologue).
Scientific Models in the Philosophy of Science
38 Tarja Knuuttila
Of the semantic approaches to models (and theories) perhaps the
best-known are those of van Fraassen (1980) and Giere (1988). Their
approaches differ from each other in the degree of their abstractness
and in the ways they utilise aspects of the semantic approach to
accommodate their divergent standpoints to the empiricism-realism
debate (French and Ladyman 1999, 104).
Van Fraassen, the (constructive) empiricist, advances a “new
picture of theories“:
To present a theory is to specify a family of structures, its models; and
secondly, to specify a certain part of those models (the empirical
substructures) as candidates for the direct representation of observable
phenomena. The structures which can be described in experimental and
measurement reports we call appearances: the theory is empirically
adequate if it has some model such that all appearances are isomorphic
to empirical substructures of that model (1980, 64).
Ronald Giere, a (constructive) realist, denies that the relation between
a model and a real system should be isomorphic. Giere (1988)
develops his account of models on the basis of classical mechanics as
presented in advanced textbooks. He proposes that, to take an
example, the “linear oscillator“ referred to in mechanics texts is not a
single model with different specific versions but a cluster of models of
varying degrees of specificity (80). Thus Giere finds in the standard
textbooks “a population of models consisting of related families of
models“ (82). The models as such are not true or false with respect to
the world; the role of the theory is rather to claim a “good fit“ between
the models and some important types of real systems. Consequently,
Giere suggests that a theory is comprised of two elements: (1) a
population of models, and (2) various hypotheses linking those
models with systems in the real world (85).
Giere argues that the relationship between models and the world is
not primarily that of truth, correspondence or isomorphism, but
similarity. He is not very worried about the vagueness of the notion of
similarity, since in his opinion the cognitive sciences are
accumulating evidence that “human cognition and perception operate
on the basis of some sort of similarity metric“ (81). Moreover, Giere
claims that even the links between the models are rather relations of
similarity than logical connections. As a consequence a scientific
theory turns out not to be a well-defined entity. Nothing in the
structure of any model itself could determine whether it belongs to a
given family of models or not. It is up to the scientific community to
judge whether the resemblance is sufficient.18
The semantic conception replaced the “received view”, becoming
itself a received view.19 Frederic Suppe (1989) states confidently that
“the Semantic Conception of Theories today probably is the
philosophical analysis of the nature of theories most widely held
among the philosophers of science; it frequently is used to analyse and
treat other philosophical problems” (3). That a conception of theories
should provide us a prevailing approach to models seems already at
the outset somewhat paradoxical. It tends to downplay models in
respect to theories, as has been pointed out by Morrison and Morgan:
The semantic view claims that models, rather than theory, occupy a
centre stage, yet most, if not all of the models discussed within the
framework fall under the category ‘models of theory’ or ‘theoretical
models’… Viewing models strictly in terms of their relationships to
theory draws our attention away from the process of constructing
models and manipulating them, both of which are crucial in gaining
information about the world, theories and the model itself. (1999a, 7-
Consequently, Morrison and Morgan propose that we should
investigate the models actually used in science to understand what
kind of entities scientific models are and how they function. In this
18 Giere’s views of models have remained generally similar over the
years, but they show signs of changing. In Giere (1999) he claims that
models are “humanly constructed abstract entities” (168). It seems to me
that this idea does not fit very well with the notion of distributed
cognition, a topic that he has shown a good deal of interest recently (2002a,
2002b). The idea that models are abstract entities is better suited to the
idea of “mental models” that was still popular in cognitive science some
years ago (see Giere [ed.] 1992). But in order for cognition to be
distributed, models must be materially existing things of some kind—if
they are to function as central tools of science and mediate between people
and other artefacts.
19 German structuralism (see e.g. Balzer, Moulines and Sneed 1987) also
has a model-based approach to scientific theories. They are however more
concerned with the architectonic of sciences than with individual models.
Recently, several researchers have developed the so-called “partial
structures” view, which limits the isomorphic relations between the model
and its target system (see e.g. Bueno 1997, and French and Ladyman 1999).
Scientific Models in the Philosophy of Science
40 Tarja Knuuttila
they are following a long tradition of approaching the epistemic value
of models from the point of view of scientific practice.
3.2 A practice-oriented approach to models
It can in fact be claimed that the very discussion of models was
originally motivated by practice-oriented considerations—even the
proponents of the semantic view understood themselves as providing
a more realistic picture of theories (see van Fraassen 1980, 64). Yet the
practice-oriented approach has usually involved taking into account
different aspects of the actual making of science. In the 1960’s issues of
theory reconstruction and theory change as well as of scientific
discovery prompted many philosophers to start to study models
(Bailer-Jones 1999, 31). Various writers including Achinstein (1968),
Black (1962), Hesse (1966) and Hutten (1954) likened models to
analogies and metaphors in their attempt to understand how models
function in scientific reasoning. Moreover, both Max Black and Peter
Achinstein created typologies of models in an effort to give a more
complete account of the variety of models used in science.
It is of interest that both Black and Achinstein started their
discussion of models by considering three-dimensional physical
objects, which Black thought were the “standard cases“ of models in
a literal sense of the word. He called them scale models, covering “all
the likenesses of material objects, or systems, or processes, whether
real or imaginary, that preserve relative proportions“ (1962, 220).
When we make scale models, Black points out, our purpose is to
reproduce in a relatively manipulable or accessible embodiment,
selected features of the “original“. “We try to bring the remote and the
unknown to our own level of middle-sized existence. There is,
however something self-defeating in this aim, since … we are forced to
replace a living tissue by some inadequate substitute, and a sheer
change of size may upset the balance of factors in the original“ (221).
Thus, “inferences from a scale model to an original are intrinsically
precarious and in need of supplementary validation and correction“
(ibid.). Achinstein, too, paid attention to the manipulability or
“workability“ of physical models (which he called representational
models). According to him “representational models, although used
in all the sciences, are particularly central in engineering. Instead of
investigating an object directly, the engineer may construct a
representation of it, which can be studied more readily” (1968, 209).
Morrison and Morgan’s conception of models as mediators that
function as investigative instruments amounts to a return to this
tradition, in that they also stress the epistemic importance of building
and manipulating models.
The notion of models as mediators builds importantly on the work
of Nancy Cartwright (1983). In arguing that the fundamental laws of
physics do not describe the regularities that exist in nature,
Cartwright reverts to models. There is a gap between the general
theoretical principles of physics and the messiness and complexity of
data which phenomenological laws in turn strive to capture. It is the
task of models to bridge that gap: “The route from the theory to reality
is from theory to model, and then from model to the phenomenological
law. The phenomenological laws are indeed true of the objects of
reality—or might be; but the fundamental laws are true only of objects
in the model” (4). For a model to function as a bridge between theory
and data, a model has to include some genuine properties of the
objects modelled. But in addition to that, models contain properties of
convenience and fiction (15). These features are needed to bring the
objects modelled into the confines of the theory. Building models is
pragmatic activity, “adjustments are made where literal correctness
does not matter very much in order to get correct effects where we
want to get them; and very often…one distortion is put right by
another” (140).
The similarities between Cartwright’s account of models as
bridges and that of Morrison and Morgan’s as mediators are clear. In
both accounts models occupy the middle space between the theory
and the world (or data), thus linking them. What is more, both
accounts stress how “additional elements” are brought into models.
This is exactly what makes models able to connect the different realms
but it is also what makes them “at least” partly autonomous, a point
that Morrison and Morgan especially stress. What I find particularly
fresh and rewarding in Morrison and Morgan’s approach is precisely
their emphasis on the epistemic importance of the independence of
models and their mediating capabilities. I think that their approach
can actually be used as a starting point for developing a non-
representationalist account of models (granting that anything like this
need not have been their original intention). However, before turning
to that, I wish to examine the idea that models are, first and foremost,
Scientific Models in the Philosophy of Science
42 Tarja Knuuttila
3.3 Models as representations
As already noted, a certain understanding of representation is part
and parcel of the semantic approach, since the structures specified by
models are posited as possible representations of either the observable
phenomena or, even more ambitiously, the underlying structures of the
real target systems. This relation of representation between a model
and its target is understood in terms of isomorphism or partial
isomorphism. Thus, according to the semantic view, the structure
specified by a model represents its target system if they are structurally
isomorphic. By isomorphism I refer to a kind of mapping that can be
established between the two that preserves the relations among their
elements. Isomorphism can be defined formally, which is one of the
charms and motivating forces behind the semantic theory. Not all
protagonists of the semantic theory agree with the formulation
presented above. For instance Giere (1999) thinks that it represents a
“carryover from an older picture of science”:
The clearest expression of the old picture is to be found in the
philosophy of mathematics and logic, and in formal semantics. The
idea is that the structure of reality mirrors the structure of set-theory.
Reality is conceived of as consisting of discrete objects, sets of discrete
objects, sets of sets of objects, sets of ordered pairs of objects, and so
on. True statements are those that describe objects as belonging to the
sets to which they in fact belong. A complete science would be the
conjunction of all and only the true statements of this set-theoretically
structured reality. (78)
In terms of the argument of this thesis it is evident that the above-
mentioned picture of science painted by Giere is thoroughly
representationalist: it depicts representation in terms of a
correspondence created by isomorphism, supposes that reality
consists of a fixed set of discrete objects and posits that there is a
privileged set of representations, structures that are, to use an
expression of Rorty “automatically and intrinsically accurate” (1980,
Overall, the critics of the semantic approach have not worried very
much about its representationalist view of science. Instead they have
limited their critique to the tendency of the semantic approach to
reduce representation to isomorphism (or similarity). The critics have
argued, among other things, that denoting a symmetrical, reflexive
and transitive relation, isomorphism does not satisfy the formal and
other criteria we might want to affirm concerning representation (see
e.g. Suárez 2003, Frigg 2003). I find these critiques both thorough and
conclusive, which is the reason this issue does not concern me here. I
rather turn to the alternative accounts of models that have been
presented by the critics of the semantic approach.
Interestingly, the critics of the semantic approach nevertheless
share with them the same presupposition that the main task of models
is to represent the world. In arguing against the semantic view of
models, some of them have explicitly made representation the crucial
property of models (see e.g. Frigg 2003, 33; Hughes 1997, and Suárez
1999), which is presumably something that most philosophers
generally agree in. A characteristic statement of this linkage is that
given recently by Paul Teller (2001):
I take the stand that, in principle, anything can be a model, and that
what makes a thing into a model is the fact that it is regarded or used
as a representation of something by the model users. Thus in saying
what a model is the weight is shifted to the problem of understanding
the nature of representation. I do not begin to have a workable
account of representation, so what is accomplished by this move? The
point is that when people demand a general account of models, an
account, which will tell us when something is a model, their demand
can be heard as a demand for those intrinsic features of an object which
make it a model. But there are no such features. We make something
into a model by determining to use it to represent.
While I do agree with Teller that it is our achievement that models
represent, I would rather not propose that we make a thing into a
model by deciding to use it as a representation of some aspect of the
world. In my opinion there are several reasons why we should resist
binding models too firmly together with representation:
1) Needless stumbling over the “puzzle of representation”. It is perhaps no
wonder that the proponents of the semantic view have not presented
any explicit analyses dealing with scientific representation, since the
semantic approach takes the question of “in virtue of what do models
represent” to be solved by reverting to isomorphism.20 The problem of
representation arises when, on the one hand, models are regarded as
being primarily representational entities and, on the other hand, it is
granted that constructing models involves idealisations,
Scientific Models in the Philosophy of Science
44 Tarja Knuuttila
approximations, conventions, fictions, and the like. Thus the question
becomes, as Callender and Cohen have aptly put it: “How can
[models] represent, if they, well, misrepresent?” (2005, 5). The urgency
of this question is, of course, directly linked to the representationalist
conception of knowledge. It dissipates once we stop linking the
epistemic value of models to their being more or less accurate
representations of the independently existing, real target systems.
What is more, then we are not in such a pressing need of giving an all-
around explanation for the possibility of representation. Rather,
whether or not a model represents something will be judged case by
case in view of its specific goals and with the help of other
information at hand—and general philosophical intuitions
concerning what makes a model a representation are largely
redundant in this task. Perhaps the best a purely philosophical
analysis can do is to offer a deflationary account of representation, as
Suárez (2004) has suggested.
2) Many scientific models cannot be considered as clear-cut representations of
any specific external systems. In the case of many scientific models we do
not know what exactly they are supposed to represent. This is
especially striking for instance in artificial intelligence, where new
kind of reality is being created. For instance, in his study of early
synthetic brain models Asaro (2005) claims that “it is hard to conceive
of just what entities these models were supposed to be modelling”
(55), granting that these models were models of mental functions,
whose ontological status is questionable as opposed for instance to
those of behaviour or neurons. It seems to be more true of scientific
practice that rather than functioning as straightforward
representations of some “real” systems, models often present us some
tentative mechanisms, processes or solutions that can then function as
a basis for various inferences. On many occasions models are used
primarily as demonstrations, exemplifications, proofs of existence,
test-beds, etc. Interestingly, even bad models and errors give us
knowledge, which runs counter to the idea that knowledge is based on
20 This is of course a somewhat simplified claim. Both van Fraassen and
Giere have been stressing the pragmatic nature of representation in their
recent writings (Giere 2004, van Fraassen 2004). But then I do not know
whether Giere can still be classified among the adherents of the semantic
representing things rightly (in relevant respects and aspects). That
this should be so is occasionally reflected in the way authors
interested in scientific practice write about representation. For
instance, when discussing representation, Morrison and Morgan
(1999b) use such expressions as “rendering” and “translation”. Their
choice of these words seems to result from their uneasiness with the
traditional notion of representation as that of “standing for” (though
they refer to “mirroring”). “Rendering” and “translation” are
commonly-used words in STS.21 Both terms refer to the
accomplishments that are brought about by the participants’ actions.
From this perspective, representation becomes a complicated
productive and interpretative process, which is one of the points I
make in Study 5.
3) Conceiving of models as representations accepts the cognitive challenge of
modelling in the wrong way. It is as if we already knew enough of the
world to be able to represent the real objects waiting there to be
represented. The representational approach also seems to assume that
we already know how to represent the objects to be represented and
have the appropriate means at hand. But if this were the case we
would not need to represent the object in the first place for other than
perhaps didactic and practical purposes (for those who do not know
what we know and for the purposes of communicating and
remembering). Thus representationalism approaches science from the
point of view of finished science. I suggest that we model the
phenomena because we usually do not know enough about those
“systems”. Thus one has to conjecture what the different features of
the phenomena are like, how they are organised and how they
function. This conjecturing has to happen through and with the help
of certain already familiar representative means (formal languages,
diagrams, three-dimensional objects, modelling methods, etc.), the use
21 “Rendering” is used by ethnomethodologists, who thereby dissociate
themselves from the representationalist paradigm (see Garfinkel 2002, ch.
3). “Translation”, in turn, has been made popular by the actor-network
theory. Latour (1999) defines it in the following way: “Instead of opposing
words and the world, science studies, by its insistence on practice has
multiplied the intermediary terms that focus on the transformations so
typical of the sciences… In its linguistic and material connotations,
[translation] refers to all the displacements through other actors whose
mediation is indispensable for any action to occur” (311).
Scientific Models in the Philosophy of Science
46 Tarja Knuuttila
of which is by no means a minor part of the problem. As I argue in
Study 5, the problem of representation does not concern only what is
being represented, but also the ways to do it. Moreover, the means of
representation are themselves conducive to certain kinds of solutions
and problems—which is due to the affordances of the medium used.22
4) Models have many other distinctive uses apart from representation.
Because of this, it is misleading to reduce modelling to representation.
Firstly, models typically occupy an intermediate space between
representing and experimenting. They share with representation the
property that their medium of expression is different from that of their
object of reference. On the other hand they share with experimentation
the fact that one can try out different possibilities with them. For
instance, it is possible to study how the “structures [of the model]
behave when the parts of the model are put together and when we
vary certain things in the model” (Morgan 2002, 220). Another typical
feature of modelling is the way we “model” entities and processes of
new domains with the help of already existing theoretical tools,
computational templates and methods borrowed from another more
established domains. One might call this procedure “representing
as”, but typically in this context the expression used is to “model on”;
e.g. “many sciences are modelled on mechanics”. This is related to the
cognitive value of modelling: since the phenomena we are interested
in seem often quite impenetrable to us, we can probe them by applying
to them modelling methods and templates that have proven to be
successful elsewhere. This is linked to the third point I want to make.
Regarding models as representations have led many to argue that
models are intended for specific phenomena (see Frigg 2003, 33;
Suárez 1999). This seems to follow from the idea that models are,
inherently, “models of” something. But models are as importantly
“models for” something, as Fox Keller (2000) has pointed out. They
are not just representative entities, but often also productive ones. I do
agree that when a model represents, the model is intended for a
22 A nice example of this is provided by Morgan (2004), where she
discusses the mathematisation of economics, how in the beginning of the
mathematisation of economics both “the process of making
representations” and “the representing relation” remained partly opaque
for the early developers. Moreover, Morgan notes how the
mathematisation of economics changed the way economists understand
and depict the economy.
certain phenomenon. But it seems to me to be more true of scientific
practice that models are characteristically entities that can be used to
represent many different kinds of things. The defenders of the
“models of” view can certainly try to account for this by converting
the material object representing different target phenomena into
different models. This seems to me a legitimate philosophical move,
yet it loses sight of one important characteristic of modelling: the
detachability and the consequent re-applicability of certain already
rather stabilised representative apparatuses to diverse kinds of
problems and data.
I think that the basic problem of the representational approach in
general is that it treats science as if it were already finished, or at least
quite ready.23 That this has not worried philosophers very much over
the years is for the most part due to their preoccupation with what can
be called as “the rational reconstruction of science”.24 However, if
philosophers are, as I think they should be, interested in the very
practices through which knowledge is being created, then the
representational approach to models proves to be limiting. This is
somewhat paradoxically acknowledged by the recent pragmatist
accounts of scientific representation by Giere (2004) and Suárez
(2002a, 2004) which both make it clear that nothing very substantive
can be said about the relationship of representation in general. In
these accounts representation is contextualised to the users’, or
agents’, activity of representing (Giere) or inferring (Suárez) in view of
their specific aims and goals.25 Thus, on my interpretation, these
accounts actually point out the need for an approach that, instead of
Scientific Models in the Philosophy of Science
23 Here I speak rather of “representational” instead of “representionalist”,
since those authors, like Teller, who consider models as representations do
not underwrite all the tenets of representationalism (as semantic conception
does, in my opinion).
24 This attitude started to change, of course, already in the early 1960’s due
to the so-called “historicist turn” with philosophers such as Stephen
Toulmin, Thomas Kuhn and Norwood Russell Hanson.
25 Giere’s four-place definition of scientific representation and Suárez’s
inferential conception of scientific representation are presented in Study 4
of this dissertation. This discussion is, however, based on the papers they
presented at the 2002 Biennial Meeting of the Philosophy of Science
Association (Giere 2002c, Suárez 2002b) and not on their final published
symposia papers (Giere 2004, Suárez 2004).
48 Tarja Knuuttila
focusing only on representation provided by ready-made models,
takes a wider perspective on the epistemic value of models—an
approach that takes seriously the epistemic importance of building
and interacting with models.
3.4 Models as epistemic artefacts
It seems to me that Morrison and Morgan’s framework of models as
mediators provides an alternative to considering models as
representations. They explicitly recognise the many tasks of models, of
which representation is just one. As I will treat their approach in the
individual articles of this dissertation, I will not go into its details
here, but rather rehearse what I take as the most important
characteristics of their programme, viz. the independence and
mediating characteristics that they ascribe to models. Nevertheless, it
appears to me that Morrison and Morgan do not quite cash out the
radical potential of their programme. In order to turn the models-as-
mediators approach into a true alternative to the prevailing models-as-
representations approach, one needs to free models from the theory-
data framework still present in Morrison and Morgan (1999b) and to
interpret models materially, thus granting them an individual status
as epistemic artefacts.
Provided that the word mediation can be understood in a very
wide sense covering complex semiotic processes, including the
activities of various actors, the mediation between theory and data (or
world) that Morrison and Morgan concentrate on is a rather limited
version of mediation. Morrison and Morgan point out that it is the
(relative) independence of models that makes them able to mediate. I
agree, but would like to add that it is in turn the material dimension,
and not just “additional elements”, that makes models able to
mediate. It is the material dimension of objects that gives them the
robustness necessary to maintain their identity across different sites.26
But making this move means also leaving the conceptual and ideal
world of philosophy and entering into the social and material world
of human actors, where material objects, usually human-made
artefacts, draw together numerous activities and different actors.
From the scientific practice point of view, it is easy to see that scientific
models are such things. This is especially evident in computer
modelling, where models more often than not act as boundary objects
(Star and Griesemer 1989) between different groups of scientists from
different disciplines.27
Now, apart from enabling the mediation between different actors,
materiality is also important for the epistemic functioning of models.
Morrison and Morgan attribute a large part of the epistemic
significance of models to the processes of their construction and
manipulation instead of focusing unilaterally on representation. As
they stress that we learn from models by building and manipulating
them, their approach in fact implies that there needs to be something
more concrete than just conceptual ideas to work on. For us to learn
from a model, it has to be a self-contained thing with which we can
interact and work. The epistemic enablings of models are due to the
way in which their materially embodied constraints intersect with
their intentional uses. Consequently, I suggest that models are
intentionally constructed and materially embodied things, epistemic
artefacts, the constraints of which are characteristically turned into
affordances for epistemic purposes. A fuller formulation of this idea is
presented in Study 3.
I suggest that conceiving of models as epistemic artefacts provides
a fresh outlook on models and their epistemic value. It attributes the
epistemic value of models to their epistemic productivity rather than
reducing it to the representation of some pre-existing natural or social
systems. It also provides a new angle from which to approach the
characteristics of models and representation. Namely, approaching
models as artefacts makes it natural to ask how and of what elements
they are made. This directs our attention to the processes of
representation that underlie “ready-made” models. From this point of
view representation appears as a complicated process that makes use
of various representative means to convey something that has already
been theoretically rendered or otherwise prepared into another
medium or form. With this in mind we can turn to the original articles,
on which the main burden of proof of this dissertation rests.
26 By “material” I mean something that is concrete and corporeal,
occupying space and time and able to interact with other things and
human beings. For a discussion of the ontology of material objects in the
social world, see Harré (2002).
27 See Mattila (2005) for an analysis of the interdisciplinary work of
modelling the transmission of infectious disease.
Scientific Models in the Philosophy of Science
50 Tarja Knuuttila
4. An overview of the original articles
The articles are ordered according to the time of their writing. This
seemed to be the most natural way to organise them in view of both
their content and the emergence of the ideas presented in them: from
them one can trace the development of the artefactual approach to
scientific representation that I am arguing for in this thesis. As already
mentioned in the Introduction, these articles participate in the specific
discussions of scientific representation in various fields: philosophy of
science, science and technology studies, semiotics and cognitive
science. Consequently, in the abstracts of the articles below I shall, in
addition to presenting their main arguments and results, also
contextualise them and show how they relate to each other. The first
two articles deal with more general issues concerning representation.
The first focuses on the so-called crisis of representation in the
humanities, and the second deals with the problem of reflexivity in
the field of science and technology studies. The next four articles
study representation in the context of modelling. Common to all of
them is the idea of models as epistemic artefacts. They use this concept
in discussing the interrelated questions of representation, modelling,
and cognition.
Study 1:
Knuuttila, Tarja
Is Representation Really in Crisis?”
Semiotica 143 (2003), 95-111.
“Is Representation Really in Crisis?” is written mainly for
semioticians and literary critics. Accordingly it assumes some
background knowledge of semiotics. It claims that there is nothing
very “postmodern“ or recent about the crisis of representation. The
“crisis” has mainly dressed a traditional epistemological problem in
more up-to-date garb. The most important point of the article is to
show, however, how the work done on scientific representation in STS
is relevant for the discussions of the crisis of representation. I argue
that the excesses of postmodernist discourse depend on a too
idealistic and traditional understanding of representation, an
understanding for which the work being done in STS offers a good
antidote. The theme of representation as artefact-using activity, which
is central to this thesis, appears already in this article.
The article opens by distinguishing between two types of claims
concerning the crisis of representation: an ontological and an
epistemological version of it. The ontological version takes for granted
that the crisis has been brought about by the expansion of mass
communication, which has made our modern life world increasingly
packed with representations and virtual artefacts of all sorts. It is
supposed that rather than being in contact with reality, we are thus
increasingly dealing with representations of it. On this basis it has
then been claimed that reality proper is in fact receding—without us
even noticing.
The epistemological version, in turn, questions the relationship
between our representations and reality. The problem is how our
ideas, words or other signs are able to “correspond” to real objects
when they seem to be very different kinds of things. The linguistic turn
of the last century externalised the sphere of representation, yet the
problem of representation stayed much the same. A recent externalist
version of the problem of the empiricists can be found for instance in
Umberto Eco’s semiotics (e.g. 1984). According to his notion of
An Overview of the Original Articles
52 Tarja Knuuttila
encyclopedia, a sign cannot refer to an object but is simply interpreted
by another sign. Thus the sphere of representation starts to look like a
prison from which there is no access to reality proper.
It appears to me, however, that the proponents of the crisis of
representation have not paid enough attention to the technological
efficiency of our mediated lifestyle, which critically depends on our
reliance on man-made representations of diverse sorts. I argue in “Is
Representation Really in Crisis?” that postmodern discussions of
representation have more often than not, forgotten that
representations are themselves concrete, material things that have real
effects. To make the point I present some exemplary work on
representation done in the field of STS. Bruno Latour and Steve
Woolgar’s Laboratory life (1986[1979]) has shown how any scientific
representation is a result of many laborious translations accomplished
using diverse and often complicated inscription devices. The paradox
of the process of representation is that once the final representation is
reached, the process of making it is excluded in the discussions about
what it means. Representations are treated as if they were direct
“signatures of the phenomena under study“, but as a by-product of
this procedure the problem of representation emerges. I also discuss
the work of Michael Lynch, who has attempted to show how in
scientific representation the artificial properties introduced by the
representational devices merge with the natural object and thus make
it understandable in the first place.
Now, the problem is how to interpret these constructivist insights.
I claim that they indicate a way out of the crisis of representation;
however totally different conclusions from constructivism have also
been drawn. For many, constructivism has meant relativism28
(witness for instance the unhappy phenomenon of the so-called
“science wars”). Reading Richard Rorty and Hilary Putnam I try to
diagnose why being both a constructivist and a realist has been such
an uneasy position for many philosophers. This seems to be at least
partly due to the common predisposition of philosophers to treat
construction as if it were a matter of conceptual and linguistic activity
only. The situation becomes different once it is realised that our
representative practices and their products take part in complicated
28 This is of course one important reason why the constructivist
standpoints presented in STS have created so much heated discussion (see
e.g. Hacking 1999; Gross and Levitt 1994).
social and material activities with their own specific ends. These
worldly activities, in which the representations themselves are
produced and used, establish the link between them and those things,
the processes and phenomena they are about. Moreover, this is
actually a process of co-construction: through the interrelated
activities of theoretical reflection, representation and experimentation
scientific objects come into being. This kind of constructivist outlook
leads to relativism only if we accept the tenets of representationalism,
according to which the world consists of some fixed totality of
representation-independent objects and expect that in order to give us
knowledge our representations have to depict these objects truthfully
or accurately.
An Overview of the Original Articles
54 Tarja Knuuttila
Study 2:
Knuuttila, Tarja
“Signing for Reflexivity: Constructionist
Rhetorics and Its Reflexive Critique in Science
and Technology Studies”
Forum Qualitative Sozialforschung / Forum: Qualitative Social
Research, (2002), [On-line Journal], 3(3). Available at: http://
(52 paragraphs).
As, paradoxically, many constructivists29 themselves also understood
realism in representationalist terms, the accusations of relativism that
have been levelled at constructivists are at least partly justified. The
second article inspects what happens when one takes into account the
constructedness of our representations, but is still attached to the
ideas of metaphysical realism and the accompanying correspondence
theory of truth. I claim that it was exactly this combination that led
“reflexivists” to claim that scientific discourse is fundamentally
flawed in its attempt to create an illusion of objectivity. Since the
1970’s these kinds of reflexive pronouncements were launched
especially in the fields of anthropology, sociology and social studies
of science. The declarations were followed by attemps to write
scientific articles so that they would more obviously display their own
undeniable artificiality.
The second article studies the discussion of reflexivity in STS. The
STS discussion offers an especially good place to study reflexivity
since reflexive critique was ambitiously practised there on at least in
three interrelated levels. Firstly, the STS reflexivists pointed out the
inherent reflexivity of STS itself: its aim of studying scientific study
scientifically.30 Secondly, though being in principle favourable to the
emerging constructivist programme, the reflexivists nevertheless
29 In this study I have used “constructionism” instead of “constructivism”
in line with Hacking (1999). In the rest of this thesis I have used the STS
participants’ term “constructivism” instead.
30 This critique applies especially to the “strong programme” of the so-
called Edinburgh school, which still opted for causal explanation.
criticised the constructivists for forgetting the constructedness of their
own accounts. From the reflexivists’ point of view the constructivists’
effort to describe what really goes on in science was dubious taking
into account their own constructivist doctrines. Thus, thirdly, the
reflexivists proposed that one should renew scientific writing by
adopting literary devices that would make explicit the
constructedness of any scientific account.
In regard to the aforementioned claims, “Signing for Reflexivity:
Constructionist Rhetorics and Its Reflexive Critique in Science and
Technology Studies” examines what kind of a problem reflexivity in
STS is, and whether the new ways of writing proposed by the
reflexivists constitute an appropriate reaction to the problems of
scientific representation. The article attempts to show that the
artificiality and conventionality of our representations lead to
epistemological problems only if we assume that in order to give us
knowledge of the world, our representations have to be more or less
accurate or truthful reflections of it. If this is not the case, the rationale
for writing reflexively vanishes. In fact, if we admit that any scientific
representation is also a purposeful construction, then the question
becomes in what ways representations satisfy their intended scientific
uses best. Literary devices borrowed from artistic discourses may
entirely miss the mark when it is a matter of conveying some features
or results as clearly and convincingly as possible. In fact, the
reflexivists’ own discourse testifies to this. Their critiques of
constructivism were written in the form of conventional critiques.
Only after the point was thus successfully brought home did they start
to experiment with reflexive writing.
Nothwithstanding their fruitless struggle with (literary)
representation, the reflexivists made a genuine contribution in
pointing out the self-refuting tendencies of constructivist rhetorics.
Namely, the assertion that all knowledge is local, situated and socially
accomplished contingent achievement seems to be either trivially true
or leads to reflexive paradoxes. The question is why these kinds of
claims were then repeated over and over again in STS—despite the
reflexivist critique presented. I suggest that STS authors were, more or
less consciously, practising deconstruction: they were attempting to
contest the hierarchies invested in traditional distinctions such as
global/local, theory/praxis, or general/specific. This work need not
challenge the epistemic value of science per se, but if it gives the
An Overview of the Original Articles
56 Tarja Knuuttila
appearance of doing so, it leads to a position that easily turns against
the STS scholars themselves. A good example of this has recently been
presented by Lynch and Cole (2002). They analyse the difficulties
encountered by one of the authors, who, being known for his STS
connections, had to account for his conception of science before being
accepted to give expert testimony in a legal case.31
31 “Signing for Reflexivity: Constructionist Rhetorics and Its Reflexive
Critique in Science and Technology Studies” was already in press when this
interesting paper was presented in the EASST 2002 Conference in York,
which is the reason I do not refer to it in the Study 2.
Study 3:
Knuuttila, Tarja and Atro Voutilainen
“A Parser as an Epistemic Artefact: A Material View
on Models”
Philosophy of Science 70 (Proceedings), (2003), 1484–1495.
In opposition to the widespread belief (also exemplified by the
reflexivist struggles) that the artificial features of scientific
representations somehow question their objectivity and epistemic
status, this article claims that artefact building does bring us
knowledge in various ways. The article takes part in the philosophical
discussion of scientific models by proposing a new approach to
models. I suggest that models can be conceived of as epistemic artefacts,
the epistemic value of which derives from their being intentionally
constructed and purposefully constrained material things. In addition
to stressing the cognitive importance of artefactuality, this approach
also departs from tradition by relating the epistemic value of models to
their material dimension. The reigning semantic approach to models
treats them as abstract theoretical entities for which only structural
“two- or three-dimensional” representation matters. This view on
models has as its goal to present a unifying, formal account of models,
which, despite its proponents claims to the contrary, actually has
problems in accommodating the very diversity of models in scientific
practice. One may suppose, however, that the very plurality of models
in scientific practice is not inconsequential, thus leading one to ask
how these diverse things function in science.
The article takes as its starting point Morrison and Morgan’s
(1999b) practice-oriented conception of models as investigative
instruments and develops it in the direction suggested by Marcel
Boumans (1999) in the very same collection of articles. Whereas for
Morrison and Morgan models are (partly) independent mediators
between theory and data, Boumans loosens models from their
subservience to the theory-data-framework by treating them as
constructed entities made up of various ingredients. Inspired by this
line of work I however take one step further and argue that models
used in science are actually epistemic artefacts, whose epistemicity
An Overview of the Original Articles
58 Tarja Knuuttila
importantly accrues from their intentionality32 and materiality. The
intentionality and materiality of epistemic artefacts are coupled
through the notion of affordance due to James J. Gibson’s ecological
theory of perception (1979). What is remarkable about this notion is
that it cuts across the dichotomy between the subjective and objective
stressing the complementarity between the environment and the
organism. Thus affordances are on the one hand based on the objective
material properties of the environment and on the other hand on its
consequences for the specific organism. 33 Applied to humans and
artefacts this means that the materiality of an artefact constrains the
uses to which it can be put. I argue that this element of constraint is in
fact conducive to scientific reasoning, if devised in a skilled and
purposeful way, which is where intentionality comes into the picture.
The notion of an epistemic artefact is applied to a language-
technological artefact, a parser. The description of the parser is written
by my co-author Atro Voutilainen, a language technologist. A parser
is interesting model, in that its very status as a model is questionable
from the traditional semantic point of view. As a working program, it
cannot easily be rendered as an abstract structure, given that it seems
to be more like a thing, or an instrument, that is primarily appreciated
for what it produces: a morpho-syntactic analysis of written text.
However, approaching a parser as an epistemic artefact discloses its
affinity to various other things that scientists call models. In the case
of a parser the constraints built into the language description are
made operative by implementing it as a computer program.
Consequently, the notion of models as epistemic artefacts makes room
also for considering computer programs as models.
Finally, I discuss the implications of the artefactual approach
developed for the question of representation. The article ends by
taking up many themes concerning models and representation that
will be more fully developed in the next two articles. Thus I argue that
models as epistemic artefacts give knowledge in many other ways
than just via direct representative links, being typically used as both
objects and tools of inquiry. Moreover, I point out how the different
roles of a model can be closely linked. Consequently, the instrumental
success of a parser, i.e. its producing reliably what we expect from it,
makes it also an interesting object to study in our effort to understand
language and cognition.
Study 4:
Knuuttila, Tarja
“Models, Representation, and Mediation”
(in press) Philosophy of Science 72 (Proceedings), (2005).
This article continues and expands the discussion of models and
representation started by the previous article. It relates the notion of
models as epistemic artefacts to the discussion on models and
representation in the philosophy of science. The article opens by
observing that even though philosophers of science typically ground
the knowledge-bringing aspects of models in representation, they have
widely different opinions of what constitutes representation. This of
course is bound to awaken suspicions of whether philosophers
actually do agree as to what the epistemic value of models depends on.
Consequently, I examine the recent discussion of models and
representation, where a definite change from more traditional dyadic
structuralist approaches to triadic pragmatist ones can be discerned.
From a brief discussion on the shortcomings of the structuralist
approach I move on to the pragmatic approaches advanced by
Daniela Bailer-Jones (2003), Ronald Giere (2002c) and Mauricio
Suárez (2002b).
Of the pragmatic accounts presented, I find Bailer-Jones’s idea of
approaching the question of representation through the propositions
entailed by models somewhat misguided. It loses sight of the diverse
media that models make use of, a consideration that seems to be
important for understanding the specific scientific value of models as
representations. Moreover, the talk about propositions in the context
of models and representation appears to be somewhat paradoxical,
given that the interest in models in the philosophy of science has been
motivated by a desire to break away from the language-orientedness of
32 Artefacts can be claimed to be intentional in the very general sense of
being directed towards some goal or thing, which intentionality is
bestowed upon them by human activity.
33 For an interpretation of affordances in terms of material dispositions, see
Harré (2002) and Scarantino (2003).
An Overview of the Original Articles
60 Tarja Knuuttila
the so-called “received view” (see e.g. van Fraassen 1980, 44;
Woodward 2002, 279-380; Frigg 2003, 10-11). In contrast, both the
similarity account of representation by Ronald Giere and the
inferential account by Mauricio Suárez seem to give interesting and
partly complementary approaches to representation. Thus, whereas
Giere stresses the properties of the representative vehicles (i.e. they
should be similar to their targets in specified ways), Suárez focuses on
the inferential activities of “competent and informed” users of
What is interesting about these pragmatic approaches is that they
agree that no more than a minimalist account of representation can be
given, since representation is essentially an accomplishment of the
users of the representation. Users make the model to represent that
which it is modelling, relying, of course, on the properties of the
model. Thus it may seem that the model becomes a model first when it
is used as such, i.e. representing something with its help (see the
discussion above). However, it appears to me that taking the
pragmatic approach seriously actually implies that the link between
the models and representation is looser than is usually supposed. I
suggest that, rather than being representations, models should be
approached as epistemic artefacts, as self-contained artificial objects
that can be used in scientific endeavour in a multitude of ways. This
makes room for the observation that from the point of view of scientific
practice the representative links to reality provided by models are
often more indirect, hypothetical and preliminary than we have
thought. Moreover, the very same models can be used to represent
different kinds of things. Consequently, in “Models, Representation,
and Mediation” I end up articulating a two-fold approach to
representation. My main idea is to approach representation as a dual
phenomenon, comprised of both a material sign-vehicle and an
intentional, three-place relation of representation that connects the
sign-vehicle to whatever is being represented.
This two-fold approach makes it possible to study scientific
representation from two different angles. On the one hand one should
study the artefacts, the models, themselves: their specific constraints
and affordances, as well as the media, methods and previous
knowledge they rely on. On the other hand one should study how the
representative relation between the sign-vehicle and its object is
achieved. Besides, I find the distinction between representation as a
sign-vehicle and as a relation illuminating because the word
“representation” is frequently used in both of these senses, causing
them easily to coalesce into another. This is in fact what happens in
dyadic accounts of representation, for example in the structuralist
approach, which assumes that representative vehicles are able to
create the relation of representation by themselves, due to their
inherent properties.
An Overview of the Original Articles
62 Tarja Knuuttila
Study 5:
Knuuttila, Tarja
“From Representation to Production: Parsers and
Parsing in Language Technology”
(in press) in Johannes Lenhard, Günther Küppers
and Terry Shinn (eds.)
Simulation: Pragmatic Constructions of Reality.
Sociology of the Sciences Yearbook. New York: Springer, (2006).
In Study 4 I claimed that one of the reasons to treat models as
epistemic artefacts is to loosen their epistemic value from their
representative function. Consequently, the task of this article is to
show what other epistemic roles besides representation a scientific
model can have. I proceed by presenting a case study of the Constraint
Grammar Parser, a computational model of a syntax whose primary
goal is to give a morpho-syntactic analysis of a running text. In Study
3 the principles of building a slightly different kind of parser, the
Finite-State Constraint Grammar, were presented. The parser was
introduced by a linguist, who has himself been fabricating it. Here my
goal is somewhat different: as I attempt to paint a down-to-earth
picture of the place of a parser in language technological research, the
study is based on the publications of a group of scientists making
parsers, as well as on interviews conducted with them.
As a language technological artefact the parser can be claimed to
imitate human performance in the sense that its task is to give same
kind of analyses to a text that a human linguist can. Thus it is a
simulation both in the sense of being a stand-in for a human linguist
and being a computer program that executes the grammatical
constraints of a specially devised language description. Yet, even
though the task of a parser is to simulate the “output” of a human
linguist, it cannot be taken as a realistic representation of the linguistic
competence of a linguist. The parser thus presents us a clear case of
being a model for something instead of being a model of something (Fox
Keller 2000). It is, rather, a new kind of thing in itself than a
representation of something that exists already. If the epistemic value
of models were tied to representation, the parser would not appear to
be a very interesting epistemic thing in the first place. One might even
question its status as a model, as already suggested in “A Parser as
an Epistemic Artefact: A Material View on Models“. In my opinion,
this does not show that there is something wrong with the parser as an
example of a model but rather with the philosophical outlook on
models. As I have already argued, it is unclear in the case of many
scientific models what exactly they are models of. Thus I would not
make being a representation a criterion for what being a model
amounts to, which has also the additional problem of how then to
distinguish models from the other scientific representations which we
would not prefer to characterise as models. The answer to this
question is, as I have already claimed and will claim in this article
especially, that we usually take as models such purposefully
fabricated things that we can work, interact and experiment with—
and it seems that computer simulations even radicalise this aspect of
The uneasy status of models as representations has of course been
noticed by some philosophers of science and thus there has been some
discussion of models as tools in addition to that of models and
representation (see e.g. Cartwright, Shomar and Suárez 1995, Klein
2001). Indeed, it is a typical move in the general discussions concern-
ing representation and representationalism that the things we take to
be representations are claimed to be tools rather than representations.
Thus, for instance, Rorty’s critique of representationalism asserts that
linguistic expressions, and consequently language itself, should be
considered as tools. Considering representative artefacts as tools
offers a convenient way for him to circumvent the variously
intertwined notions of representation, realism and truth. The
Constraint Grammar Parser I consider in this article is obviously a tool,
but I argue that it is other things, too. It can be also considered as a
worthy research object in its own right and as an inferential device.
The fact that parsers are also research objects of computational
linguistics and language technology serves to underline the fact that
models need not be treated just as surrogates for reality proper. In its
role as an inferential device, the parser comes closest to functioning like
a representation of some kind. However, since it cannot be taken to
offer any realistic representation of human linguistic competence, it
gives indirect evidence of it instead.
An Overview of the Original Articles
64 Tarja Knuuttila
Last but not least, I have tried to make the work of representation
characteristic of modelling visible in my treatment of the parser
construction. This is because the discussions of models and
representation have been by and large neglectful of the way models
are themselves products of various representative procedures. As they
are results of the work of representation, they present us something,
although it is uncertain how that something is supposed to represent
something else in the external world. Looking at the way models are
built offers also new insights into representation. Here representation
appears as a mediative activity that conveys something that is already
known or assumed, transforming it with the help of diverse
representative means. Consequently, the cognitive problem of
modelling concerns not only what is being represented but also how
that can be done. Thus in addition to depicting the multifacetedness of
parsers as epistemic artefacts, the article “From Representation to
Production: Parsers and Parsing in Language Technology” tries to
convey the flavour of the laborious work of representation that is
inherent in modelling and of the expertise that is born in the very
Study 6:
Knuuttila, Tarja and Timo Honkela
“Questioning External and Internal Representation:
The Case of Scientific Models”
(in press) in Lorenzo Magnani (ed.),
Computing, Philosophy, and Cognition.
London: King’s College Publishing, (2005).
One way to put that what I have attempted to say in the previous
articles on models and representation is that the things we treat as
representations do not always act representationally and their
epistemic value need not derive from their ability to represent. Models
are a case in point. I have argued that they give us knowledge in many
ways and that they can be made to bear on our knowledge of the
world even if they cannot be taken as straightforward representations
of external phenomena. The rationale for treating models as epistemic
artefacts lies exactly in the potential of this approach to explain the
epistemic value of models without recurring to their representative
function. The target of my criticism is not representation itself, but
representationalism, which in the case of models has such a firm grip
on us that it has seemed nearly beyond doubt that models give us
knowledge because they succeed in representing something external to
themselves. This has led, as we have seen, to claims that to be a model
is to represent. In my opinion this eventually leads us to ask why we
believe that to know is to be in command of a representation that
depicts something that already exists rightly, in relevant respects or
Why is representation so important? An answer already hinted at
is provided by the representationalist theory of knowledge, which
seems to be deeply ingrained in our epistemological thinking. Thus,
for instance, the author of a recent monograph on models and
representation does not hesitate to claim that: “Science aims at giving
us empirical knowledge. If models are to serve this purpose, they must
be representational. … They can instruct us about the nature of reality
only if we assume that (at least some of) the model’s aspects have
counterparts in the world. Hence in order to be a source of knowledge
An Overview of the Original Articles
66 Tarja Knuuttila
models must be representational.” (Frigg 2003, 19). As already noted
earlier, the roots of representationalism lie in the idea that the external
world is not directly presented to the consciousness, but only through
mental content that can be regarded as a collection of internal repre-
sentations of some kind. Thus knowledge is identified with internal
representation, which poses the question of the relationship between
internal and external representation.
In order to know how internal and external representation are
linked to each other, one would need to know what internal
representations are like and how they are connected to the external
world (including external representations). Unfortunately, there is
even less agreement concerning internal representation than external
representation. Whereas we are empirically acquainted with the
external representations we produce and use, the same is not true of
internal representation. There is no agreement as to what kind of
representations our minds house. Such being the case, the article
“Questioning External and Internal Representation: The Case of
Scientific Models” approaches the question of internal representation
by examining what can be inferred about our cognition from our
external practices of representation. Study 6 examines in particular a
special kind of model, the self-organising map (SOM), which is
approached as a multifunctional epistemic artefact. Timo Honkela, a
computer scientist experienced in building and studying SOMs, has
characterised them in this article. What gives a special twist to the
case of the SOM in the context of internal and external representation
is that it can also be used, among other things, as a model of the mind.
Thus it can be conceived of as an external representation, which
conveys us knowledge of our supposed internal representations.
However, even in this capacity the SOM, rather than being a
representation of our brains, functions as a device with which some
interesting results concerning the adaptive cognitive processes can be
shown. Consequently, it rather demonstrates and exemplifies than
really represents.
As for the question of internal and external representation, we
argue that the idea of models as epistemic artefacts finds support in
recent trends in cognitive science according to which cognition is
distributed between embodied humans and their environment. This
human environment is largely artefactual, and thus the distinctively
human kind of knowledge can be regarded as a product of our
artefactual practices (including language). Thus both external and
internal representations should be understood rather performatively
than representationally, as oriented towards action, and as parts of
complicated causal, cultural and inferential webs. This of course sets
limits on how well the notion of representation itself can be expected
to explain cognitive and epistemic processes. Our conclusion is that
the explanatory power of the representationalist paradigm is
diminishing as a consequence of such developments as distributed
cognition, which ultimately relativises the strict division between the
internal and external representation.
An Overview of the Original Articles
68 Tarja Knuuttila
5. Conclusions
It seems that the crisis of representation emerges from the uneasy
encounter of our representationalist epistemological heritage with the
observation that our representations are artificial and conventional
constructs. The conception of knowledge of the modern era has been
representationalist in the sense that correspondence with the world
“as it is” has provided the criterion for whether a representation is
true or not. From this point of view any traces of artifice must make the
representation seem less truthful and more contrived. Yet, as the
representations we produce and use cannot be but artificial,
suspicions concerning their “objectivity” start to abound. How can
our obviously constructed and media-specific representations really
stand for reality in the sense of depicting some aspects of it accurately?
The answer, I suggest, is not to doubt the artificiality of
representation but rather our representationalist convictions. As I
have argued in this dissertation, our representationalist heritage ties
together several different ideas in assuming that knowledge consists
of a collection of representations that correspond accurately to reality,
which is comprised of a fixed totality of representation-independent
objects. What I have attempted to do here is to join others in untying
the knots that are characteristic of representationalism: the conflation
of knowledge and representation, representation and correspondence,
and finally representation and (metaphysical) realism. Even though I
have been inspired and motivated by these big questions my work has
remained largely in the sphere of modelling and scientific
In this dissertation I have developed a practice-oriented artefactual
approach to models that loosens the epistemic value of models from
representation (understood in the representationalist sense). Treating
models as epistemic artefacts attributes their epistemic value to the
interplay of their material and intentional dimensions, which is due to
their being both purposefully constrained and materially defined, yet
interpretatively open things. What is new about my approach is its
stress on the materiality and artefactuality of models—properties that
have been of secondary importance for the philosophical tradition,
which values the abstract, theoretical and conceptual. From my
naturalist point of view, we do not however represent with the help of
structures or conceptual ideas alone. These ideas they have to be built
into, and with the help of, specific media, which partly determines
that what can be conveyed through models and representations.
As to the question of modelling and scientific representation, this
dissertation draws the following conclusions:
Our understanding of modelling should not be restricted to the
view that models represent some external target systems accurately.
Apart from being representative things, models are typically also
productive things whose workability and experimentability are
crucial for their epistemic value. Models can function not only as tools
and inference generators, but also as research objects in their own
right. In the capacity of inference generators models can be used as
representations. In scientific practice, however, they also function as
exemplifications, proofs of existence, demonstrations and test-beds.
Thus conceiving of models as representations loses the sight of many
of their distinctive properties. What is more, it gets the cognitive
challenge of modelling the wrong way, as it assumes that we already
knew what our relevant target systems were and had the appropriate
means at hand to represent them.
Representation can be approached both from the use and the
production points of view. Philosophical analyses of scientific
representation have so far concentrated rather one-sidedly on the use
of ready-made models. From this starting point, I have argued that
representation should be approached as a two-fold phenomenon that
is based both on the medium-specific affordances of the material sign-
vehicle and on the intentional process of relating the sign-vehicle to
whatever it is that is being represented. The fact that a sign-vehicle is
a materially constructed historical artefact leads us to consider the
complex culturally constructed artefactual chains through which our
knowledge of the world is actually mediated. As no sign-vehicle is
representative of anything in and of itself, the intentional process of
relating it to its object is needed for the representation to become
Treating models as artefacts makes their production processes
visible. Seen from this angle a large part of the work of representation
that is taking place in sciences is conveying into another form that
what is already represented and modelled somehow. Looking at
representation from this point of view stresses the methods,
ingredients and various representative devices that are needed in
producing models. The production point of view on representation
70 Tarja Knuuttila
seems to me an important complement to the use point of view. It
shows how any ready-made model is already a complex
representative achievement in itself and not an isolated theoretical
entity. I think that this has a certain sobering effect: one needs not be
puzzled about what connects the model and its supposed target
system since the model is from the very outset a result of various
procedures of connecting. Thus representing lies indeed at the heart of
the modelling enterprise, but not where it has been conventionally
taken to be.
Achinstein, Peter (1968), Concepts of Science. Baltimore: Johns Hopkins Press.
Amann, Klaus and Karin Knorr Cetina (1990), “The Fixation of (Visual) Evi-
dence”, in Michael Lynch and Steve Woolgar (eds.), Representation in Sci-
entific Practice. Cambridge, MA: MIT Press, 85-122.
Asaro, Peter (2005), On the Origins of the Synthetic Mind: Working Models,
Mechanisms, and Simulations. PhD. Dissertation. Urbana: University of Il-
linois at Urbana-Champaign.
Baigrie, Brian S. (ed.) (1996), Picturing Knowledge: Historical and Philosophical
Problems Concerning the Use of Art in Science. Toronto: Toronto University
Bailer-Jones, Daniela M. (1999), “Tracing the Development of Models in the
Philosophy of Science”, in Lorenzo Magnani, Nancy J. Nersessian and
Paul Thagard (eds.), Model-Based Reasoning in Scientific Discovery. New
York: Kluwer, 23-40.
Bailer-Jones, Daniela (2003), “When Scientific Models Represent”, Interna-
tional Studies in the Philosophy of Science 17: 59-74.
Balzer, Wolfgang, C. Ulises Moulines and Joseph D. Sneed (1987), An Archi-
tectonic for Science. The Structuralist Program. Dordrecht: Riedel.
Bergman, Mats (2004), Fields of Signification: Explorations in Charles S. Peirce’s
Theory of Signs. PhD. Dissertation. Philosophical Studies from the
University of Helsinki 6. Helsinki: University of Helsinki.
Black, Max (1962), Models and Metaphors. Ithaca: Cornell University Press.
Boothby, Richard (2001), Freud as Philosopher: Metapsychology After Lacan.
New York and London: Routledge.
Boumans, Marcel (1999), “Built-In Justification”, in Mary S. Morgan and
Margaret Morrison (eds.), Models as Mediators. Perspectives on Natural and
Social Science. Cambridge: Cambridge University Press, 66-96.
Bueno, Otavio (1997), “Empirical Adequacy: A Partial Structure Approach”,
Studies in the History and Philosophy of Science 28: 585-610.
Callender, Graig and Jonathan Cohen (2005), “There Is No Special Problem
About Scientific Representation”, a paper submitted for review.
Available at:
Cartwright, Nancy (1983), How the Laws of Physics Lie? Oxford: Clarendon
Cartwright, Nancy, Towfic Shomar and Mauricio Suárez (1995), “The Tool
Box of Science”, in L. Nowack (ed.), Poznan Studies in the Philosophy of Sci-
ence. Amsterdam: Rodopi, 137-149.
Clapin, Hugh (ed.) (2002), Philosophy of Mental Representation. Oxford: Oxford
University Press.
Clark, Andy (1997), Being There: Putting the Brain, Body, and World Together
Again. Cambridge, MA: MIT Press.
72 Tarja Knuuttila
Clark, Andy (1999), “Embodied, situated, and distributed cognition”, in Wil-
liam Bechtel and George Graham (eds.), A Companion to Cognitive Science,
Malden, MA: Blackwell Publishing, 506-517.
Clark, Andy (2003), Natural-Born Cyborgs. Oxford: Oxford University Press.
Collins, H.M. and Robert Evans (2002), “The Third Wave of Science Studies:
Studies of Expertise and Experience”, Social Studies of Science 32(2): 235-
Crane, Tim (1995), The Mechanical Mind: A Philosophical Introduction to Minds,
Machines and Mental Representation. Harmondsworth: Penguin Books.
Crimmins, Mark (1991), “Representation”, in Hans Burkhardt and Barry
Smith (eds.), Handbook of Metaphysics and Ontology, Vol. 2. Munich,
Philosophia, 791-793.
da Costa, Newton C. A. and Steven French (1990), “The Model-Theoretic
Approach in the Philosophy of Science”, Philosophy of Science 57: 248-265.
Daston, Lorraine (ed.) (2000), Biographies of Scientific Objects. Chicago: Chi-
cago University Press.
Dennett, Daniel (2000), “Making Tools for Thinking”, in Dan Sperber (ed.),
Metarepresentations: A Multidisciplinary Perspective. Oxford: Oxford Uni-
versity Press, 17-29.
Dewey, John (1929), The Quest for Certainty. A Study of the Relation of Knowl-
edge and Action. London: George Allen and Unwin.
Downes, Stephen M. (1993), “Socializing Naturalized Philosophy of Sci-
ence”, Philosophy of Science 60: 452-468.
Dreyfus, Hubert L. and Stuart E. Dreyfus (1986), Mind over Machine. New
York: Macmillan.
Dupré, John (1993), The Disorder of Things. Metaphysical Foundations of the
Disunity of Science. Cambridge: Harvard University Press.
Eco, Umberto (1984), Semiotics and the Philosophy of Language. Bloomington:
Indiana University Press.
Engeström, Yrjö and Virginia Escalante (1995), “Mundane Tool or Object of
Affection? The Rise and Fall of Postal Buddy, in Bonnie A. Nardi (ed.),
Context and Consciousness: Activity Theory and Human-Computer Interaction.
Cambridge: MIT Press, 265-290.
Engeström, Yrjö, Reijo Miettinen and Raija-Leena Punamäki (eds.) (1999),
Perspectives on Activity Theory. New York: Cambridge University Press.
Fleck, Ludwik (1979), Genesis and Development of a Scientific Fact. Chicago: The
University of Chicago Press.
Foucault, Michel (1966), Les mots et les choses. Paris: Gallimard.
Foucault, Michel (1970), The Order of Things. London, Tavistock and New
York: Pantheon.
Fox Keller, Evelyn (2000), “Models Of and Models For: Theory and Practice
in Contemporary Biology”, Philosophy of Science 67 (Proceedings): S72-
French, Steven (2003), “A Model-Theoretic Account of Representation (Or, I
Don’t Know Much About Art...but I Know It Involves Isomorphism”,
Philosophy of Science (Proceedings) 70: 1472-1483.
French, Steven and James Ladyman (1999), “Reinflating the Semantic Ap-
proach”, International Studies in the Philosophy of Science 13 (2): 103-121.
Frigg, Roman (2002), Models and Representation: Why Structures Are Not
Enough. Measurement in Physics and Economics Discussion Paper Series.
London: London School of Economics.
Frigg, Roman (2003), Re-presenting Scientific Representation. PhD. Disserta-
tion. London: London School of Economics.
Galison, Peter and David J. Stump (1996), The Disunity of Science: Boundaries,
Contexts, and Power. Stanford: Stanford University Press.
Garfinkel, Harold (2002), Ethnomethodology’s Program: Working Out Durkheim’s
Aphorism. Edited and introduced by Anne Warfield Rawls. Lanham, MD:
Rowman and Littlefield.
Gibson, James J. (1979), The Ecological Approach to Visual Perception. Boston:
Houghton Mifflin.
Giere, Ronald, N. (1988), Explaining Science: A Cognitive Approach. Chicago
and London: The University of Chicago Press.
Giere, Ronald N. (ed.) (1992), Cognitive Models of Science. Minnesota Studies
in the Philosophy of Science, Vol. XV. Minneapolis: University of Min-
nesota Press.
Giere, Ronald N. (1994), “No Representation without Representation”, Biol-
ogy and Philosophy 9: 113-120.
Giere, Ronald N. (1999), Science without Laws. Chicago and London: The
University of Chicago Press.
Giere, Ronald N. (2002a), “Scientific Cognition as Distributed Cognition”, in
Peter Carruthers, Stephen Stich and Michael Siegal (eds.), Cognitive Bases
of Science. Cambridge: Cambridge University Press, 285-299.
Giere, Ronald N. (2002b), “Models as Parts of Distributed Cognitive Systems”,
in Lorenzo Magnani and Nancy J. Nersessian (eds.), Model Based
Reasoning: Science, Technology, Values. Dordrech: Kluwer, 227-241.
Giere, Ronald N. (2002c), “How Models Are Used to Represent Physical Re-
ality”, a paper presented at PSA 2002. Available at: http://philsci-
Giere, Ronald N. (2004), “How Models Are Used to Represent Reality”, Phi-
losophy of Science (Symposia) 71: 742-752.
Gross, Paul R. and Norman Levitt (1994), Higher Superstition: The Academic
Left and Its Quarrels With Science. Baltimore: Johns Hopkins University
Gutting, Gary (2001), French Philosophy in the Twentieth Century. Cambridge:
Cambridge University Press.
74 Tarja Knuuttila
Hacking, Ian (1975), Why Does Language Matter to Philosophy? Cambridge:
Cambridge University Press.
Hacking, Ian (1983), Representing and Intervening. Cambridge: Cambridge
University Press.
Hacking, Ian (1999), The Social Construction of What? Cambridge, MA:
Harvard University Press.
Harré, Rom (2002), “Material Objects in Social Worlds”, Theory, Culture &
Society 19(5/6): 23-33.
Heidegger, Martin (1977), The Question Concerning Technology and Other
Essays, translated by William Lovitt. New York: Harper & Row.
Hesse, Mary (1966), Models and Analogies in Science. Milwaukee: University of
Notre Dame Press.
Hintikka, Jaakko (1973), Logic, Language-Games and Information. Kantian
Themes in the Philosophy of Logic. Oxford: The Clarendon Press.
Hintikka, Jaakko (1974), Knowledge and the Known. Historical Perspectives in
Epistemology. Dordrecht-Holland/Boston-U.S.A: D. Reidel.
Hughes, R. I. G. (1997), “Models and Representation”, Philosophy of Science
64: S325-S336.
Hutchins, Edwin (1995), Cognition in the Wild. Cambridge, MA: MIT Press.
Hutten, E. H. (1954), “The Role of Models in Physics”, British Journal for the
Philosophy of Science 4: 284-301.
Jones, Caroline and Peter Galison (eds.) (1998), Picturing Science, Producing
Art. New York: Routledge.
Kitcher, Philip (1992), “The Naturalists Return”, Philosophical Review 101: 53-
Klein, Ursula (2001), Experiments, Models, Paper Tools. Cultures of Organic
Chemistry in the Nineteenth Century. Stanford: Stanford University Press.
Knorr Cetina, Karin (1981), The Manufacture of Knowledge. An Essay on the
Constructivist and Contextual Nature of Science. Oxford: Pergamon Press.
Knorr Cetina, Karin (1999), Epistemic Cultures: How the Sciences Make Knowl-
edge. Cambridge, MA: Harvard University Press.
Knorr Cetina, Karin (2001), “Objectual Practice”, in Theodore R. Schatzki,
Karin Knorr Cetina and Eike von Savigny (eds.), The Practice Turn in
Contemporary Theory. London and New York: Routledge, 175-188.
Latour, Bruno (1990), “Drawing Things Together”, in Michael Lynch and
Steve Woolgar (eds.), Representation in Scientific Practice. Cambridge, MA:
MIT Press, 19-68.
Latour, Bruno (1994), “On Technical Mediation-Philosophy, Sociology, Ge-
nealogy”, Common Knowledge 3: 29-64.
Latour, Bruno (1995), “The ‘Pédofil’ of Boa Vista: A Photo-Philosophical
Montage”, Common Knowledge V4, N1: 144-187.
Latour, Bruno (1999), Pandora’s Hope: Essays on the Reality of Science Studies.
Cambridge, MA and London: Harvard University Press.
Latour, Bruno (2000), “When Things Strike Back-A Possible Contribution of
Science Studies to the Social Sciences”, British Journal of Sociology 51: 107-
Latour, Bruno and Steve Woolgar ([1979]1986), Laboratory Life: The Construc-
tion of Scientific Facts. 2nd ed. Princeton: Princeton University Press.
Lynch, Michael (1985a), Art and Artifact in Laboratory Science. A Study of Shop
Work and Shop Talk in a Research Laboratory. London: Routledge and
Lynch, Michael (1985b), “Discipline and the Material Form of Image: An
Analysis of Scientific Visibility”, Social Studies of Science 15: 37-66.
Lynch, Michael (1990), “The Externalized Retina: Selection and Mathemati-
zation in the Visual Documentation of Objects in Life Sciences”, in
Michael Lynch and Steve Woolgar (eds.), Representation in Scientific
Practice. Cambridge, MA: MIT Press, 153-186.
Lynch, Michael (1994), “Representation is Overrated: Some Critical Re-
marks About the Use of the Concept of Representation in Science
Studies”, Configurations: A Journal of Literature, Science and Technology 2:
Lynch, Michael and Simon A. Cole (2002), “STS on Trial: Dilemmas of Ex-
pertise”, a paper presented at EASST 2002 Conference, York.
Lynch, Michael and Steve Woolgar (1990), “Introduction: Sociological Ori-
entations to Representational Practice in Science”, in Michael Lynch and
Steve Woolgar (eds.), Representation in Scientific Practice. Cambridge, MA:
MIT Press, 1-18.
Mattila, Erika (2005), “Interdisciplinarity ‘In the Making’: Modelling Infec-
tious Diseases”, forthcoming in Perspectives on Science 13 (4).
Merrell, Floyd (1995), Peirce’s Semiotics Now: A Primer. Toronto: Canadian
Scholars’ Press.
Miettinen, Reijo (2001), “Artifact Mediation in Dewey and in Cultural-His-
torical Activity Theory”, Mind, Culture, and Activity 8: 297-308.
Morgan, Mary S. (2002), “Experiments without Material Intervention: Model
Experiments, Virtual Experiments, and Virtually Experiments”, in Hans
Radder (ed.), Philosophy of Scientific Experimentation. Pittsburgh: Univer-
sity of Pittsburgh Press, 216-235.
Morgan, Mary S. (2004), “Imagination and Imaging in Model Building”, Phi-
losophy of Science (Symposia) 71: 753-766.
Morgan, Mary S. and Margaret Morrison (eds.) (1999), Models as Mediators.
Perspectives on Natural and Social Science. Cambridge: Cambridge Univer-
sity Press.
Morrison, Margaret and Mary S. Morgan (1999a), “Introduction”, in Mary S.
Morgan and Margaret Morrison (eds.), Models as Mediators. Perspectives on
Natural and Social Science. Cambridge: Cambridge University Press, 1-9.
76 Tarja Knuuttila
Morrison, Margaret and Mary S. Morgan (1999b), “Models as Mediating In-
struments”, in Mary S. Morgan and Margaret Morrison (eds.), Models as
Mediators. Perspectives on Natural and Social Science. Cambridge: Cam-
bridge University Press, 10-37.
Mäki, Uskali (1996), “Scientific Realism and Some Pecularities of Econom-
ics”, in R.S. Cohen, R. Hilpinen and Qiu Renzong (eds.), Realism and Anti-
Realism in the Philosophy of Science. Dordrecht: Kluwer, 427-447.
Nagel, Ernest (1961), The Structure of Science. New York: Harcourt, Brace &
Niiniluoto, Ilkka (1999), Critical Scientific Realism. Oxford: Clarendon.
Nöth, Winfried (1995), Handbook of Semiotics. Bloomington: Indiana Univer-
sity Press.
Palmer, Stephen E. (1978), “Fundamental Aspects of Cognitive Representa-
tion”, in Eleanor Rosch and Barbara B. Lloyd (eds.), Cognition and Catego-
rization. Hillsdale NJ: Erlbaum, 259-303.
Peirce, Charles S. (1931-1958) [CP], Collected Papers of Charles Sanders Peirce.
Charles Hartshorne, Paul Weiss, and A.W. Burks (eds.), Vols. 1-8.
Cambridge MA: Harvard University Press.
Pickering, Andrew (1995), The Mangle of Practice: Time, Agency, and Science.
Chicago: The University of Chicago Press.
Pitkin, Hanna F. (1967), The Concept of Representation, Berkeley: University of
California Press.
Popkin, Richard H. (1980), “Scepticism and Anti-Scepticism in the Latter Part
of the Eighteenth Century”, in The High Road to Pyrrhonism (edited by Ri-
chard A. Watson and James E. Force). San Diego: Austin Hill Press.
Prendergast, Christopher (2000), The Triangle of Representation, New York:
Columbia University Press.
Putnam, Hilary (1982), “Why There Isn’t a Ready-Made World”, Synthese
Ragland-Sullivan, Ellie (1986), Jacques Lacan and the Philosophy of Psychoanaly-
sis. London, Canberra: Croom Helm.
Rheinberger, Hans-Jörg (1997), Toward a History of Epistemic Things. Stanford:
Stanford University Press.
Rorty, Richard (1980), Philosophy and the Mirror of Nature. Oxford: Basil
Rorty, Richard (1982), Consequences of Pragmatism. Brighton: Harvester Press.
Rorty, Richard (ed.) (1992), The Linguistic Turn. Essays in Philosophical Method.
Second edition with two retrospective essays. Chicago and London: The
University of Chicago Press.
Rosenberg, Jay F. (1974), Linguistic Representation. Dordrecht: D. Reidel.
Rusterholz, Peter (2003), “On the Crisis of Representation”, Semiotica 143: 53-
Scarantino, Andrea (2003), “Affordances Explained”, Philosophy of Science 70:
Sellars, Wilfrid (1968), Science and Metaphysics. Variations on Kantian Themes.
London: Routledge and Kegan Paul.
Sintonen, Matti (1995), “Knowing and Making: Kantian Themes in
Hintikka’s Philosophy”. Grazer Philosophische Studien 49: 121-134.
Sintonen, Matti (1996), “The Come-back of Constructivism”. Revue Roumaine
de Philosophie 40 (3-4): 217-225.
Sintonen, Matti (2005), “From the Logic of Questions to the Logic of In-
quiry”, forthcoming in Randall E. Auxier and Lewis E. Hahn (eds.), The
Philosophy of Jaakko Hintikka. Library of Living Philosophers. Illinois:
Sismondo, Sergio (1996), Science without Myth: On Constructions, Reality and
Social Knowledge. Albany, NY: SUNY Press.
Star, Susan Leigh and James R. Griesemer (1989), “Institutional Ecology,
‘Translations’ and Boundary Objects: Amateurs and Professionals in
Berkeley’s Museum of Vertebrate Zoology, 1907-39”, Social Studies of Sci-
ence 19: 387-420.
Suárez, Mauricio (1999), “Theories, Models, and Representations”, in
Lorenzo Magnani, Nancy J. Nersessian and Paul Thagard (eds.), Model-
Based Reasoning in Scientific Discovery. New York: Kluwer, 75-83.
Suárez, Mauricio (2002a), Pragmatics of Scientific Representation. Discussion
paper 66/02. Centre for the Philosophy of Natural and Social Science.
London: London School of Economics.
Suárez, Mauricio (2002b), “An Inferential Conception of Scientific Represen-
tation”, a paper presented at PSA 2002. Available at: http://philsci-
Suárez, Mauricio (2003), “Scientific Representation: Against Similarity and
Isomorphism”, International Studies in the Philosophy of Science 17: 225-244.
Suárez, Mauricio (2004), “An Inferential Conception of Scientific Represen-
tation”, Philosophy of Science (Symposia) 71: 767-779.
Suppe, Frederick (1977), The Structure of Scientific Theories, 2nd ed. [1st ed,
1974]. Urbana: University of Illinois Press.
Suppe, Frederick (1989), The Semantic Conception of Theories and Scientific
Realism. Urbana: University of Illinois Press.
Suppes, Patrick (1961), “A Comparison of the Meaning and Uses of Models
in Mathematics and the Empirical Sciences”, in Hans Freudenthal (ed.),
The Concept and the Role of the Model in Mathematics and the Natural and So-
cial Sciences. Dordrecht: Reidel, 163-177.
Teller, Paul (2001), “Twilight of the Perfect Model Model”, Erkenntnis 55:
Tuomela, Raimo (1985), Science, Action and Reality. Dordrecht: D. Reidel.
van Fraassen, Bas (1980), The Scientific Image. Oxford: Oxford University
78 Tarja Knuuttila
van Fraassen, Bas (2004), “Science as Representation: Flouting the Criteria”,
Philosophy of Science (Symposia) 71: 794-804.
Varela, Francisco, E. Thompson and Eleanor Rosch (1991), The Embodied
Mind. Cambridge, MA: MIT Press.
Volli, Ugo (2003), “Crisis of Representation, Crisis of Representational
Semiotics?” Semiotica 143: 25-33.
Vygotsky, L. S. (1978), Mind in Society. The Development of Higher Psychological
Processes. Eds. Michael Cole et al. Cambridge, MA: Harvard University
Winsberg, Eric (1999), “Sanctioning Models: The Epistemology of Simula-
tion”, Science in Context 12: 275-292.
Woodward, Jim (2002), “A Book Review on Ronald Giere, Science without
Laws”, Philosophy of Science 69: 379-406.
Woolgar, Steve (ed.) (1988), Knowledge and Reflexivity. New Frontiers in the
Sociology of Knowledge. Beverly Hills: Sage.
... If this is right, the GARD model from the origin of life studies shows us that exploratory models can have very indefinite, or indeterminate, connections to the actual world; this indeterminacy is where the epistemic value of this model lies. Indeterminacy is valuable because it provides motivation and an interpretive structure, both of which prompt new lines 1A number of authors discuss models without targets and how such models relate to various accounts of modeling is intensely complex (Frigg and Nguyen, 2016;Gelfert, 2018;Massimi, 2018;Massimi, 2019;Knuuttila, 2005;French, 2003;French, 2014;Giere, 2010;Poznic, 2016;Suárez, 2003;Contessa, 2011) of research and a context for evaluating new evidence. Scientists can use indeterminate models in this way because of the uncertainty over the target: were it known whether there was a target, the model would not have the same epistemic value. ...
... There is another way of thinking about models that places less emphasis on modality and the target phenomenon: the artifactual approach (Knuuttila, 2010;Knuuttila, 2011;Knuuttila and Loettgers, 2013a;Knuuttila and Loettgers, 2013b;Knuuttila and Loettgers, 2016;Knuuttila, 2017;Knuuttila and Koskinen, 2020;Knuuttila, 2005). This approach is fruitful and makes important advances over the imaginary approach, particularly in terms of ontology and in terms of the manipulatability of models. ...
The aim of this article is to use a model from the origin of life studies to provide some depth and detail to our understanding of exploratory models by suggesting that some of these models should be understood as indeterminate. Models that are indeterminate are a type of exploratory model and therefore have extensive potential and can prompt new lines of research. They are distinctive in that, given the current state of scientific understanding, we cannot specify how and where the model will be useful in understanding the natural world: in this case, the origin of life on Earth. The purpose of introducing indeterminacy is to emphasize the epistemic uncertainty associated with modeling, a feature of this practice that has been under emphasized in the literature in favor of attempts to understand the more specific epistemic successes afforded by models.
... No campo de Educação em Ciências, pesquisadores vêm concentrando esforços para melhor compreender o papel de professores e de representações usadas em salas de aula e a aprendizagem de estudantes em interação com esses professores e, também, fazendo uso dessas representações (Evagorou et al., 2015). Contudo, pesquisas sobre representações nesse campo têm focado principalmente na influência de representações como representação da realidade (Morrison & Morgan, 1999), ao invés de representações como objetos de pensamento para a construção de sentidos (Knuuttila, 2005). ...
Full-text available
A noção de mediação é central na psicologia sócio-histórica, como originalmente adotada por Vigotski e ampliada por Wertsch com a apresentação da sua Teoria da Ação Mediada. Algumas propriedades da ação mediada são usadas neste artigo como princípio organizador ao analisarmos relatos, em uma perspectiva histórico/processual, de seis professores de Química do Ensino Superior sobre o uso do recurso educacional usualmente reconhecido como modelos moleculares em seus percursos profissionais. Focamos principalmente nos processos mediados por modelos moleculares, nas mudanças que ocorreram nesses processos e no reconhecimento de limitações/possibilidades de seus usos nas suas aulas, à luz da Teoria da Ação Mediada. Como principais resultados, apontamos algumas transformações das ações dos professores, a partir de adaptações, substituição e/ou abandono de modelos moleculares na prática docente e alterações em suas habilidades em usá-los, muitas vezes motivadas pelo reconhecimento de limitações/possibilidades desse recurso no ensino. A partir disso, apresentamos contribuições e implicações desta pesquisa para a área de Educação em Ciências.
Full-text available
Ausgangspunkt und Zielsetzung des Beitrags ist die Notwendigkeit eines medienpädagogischen Perspektivwechsels vom Einsatz von Medien im Unterricht zur Kultur der Digitalität, bzw. kollektiven Praktiken. Theoretischer Bezugspunkt ist ein Verständnis der Digitalisierung aller Lebensbereiche unter Bezugnahme auf praktiktheoretische Überlegungen, bzw. Theorien sozialer Praktik. Durch den Perspektivwechsel ergeben sich mindestens die drei im Weiteren näher ausgeführten Fragestellungen/ Forschungsschwerpunkte: (1) Das Individuelle und das Kollektive: Wie bedingen individuelle und kollektive Medienpraktiken einander? (2) Digitalisierung und Bildungsinstitutionen: Die handlungstheoretische Weltinterpretation bildungs- politischer Akteure und die Idee der Generalisierung von Praktiken trifft mit Wucht auf die Kontingenz und Immanenz medialer Praktiken (3) Mensch und Algorithmen: Wie verhält sich Mensch und Algorithmus sowie Struktur und Praxis zueinander? (4) Bildung als gestaltende Auseinandersetzung mit Unbestimmtheit: Welche Implikationen ergeben sich für den (Medien-)Bildungsbegriff. Der Beitrag endet mit der Feststellung, dass Medienpädagogik unter dieser Perspektive und mit dem Bezug auf Materialität, Unbestimmtheit und Kollektivität zentral für die weitere Exploration und Untersuchung allgemein pädagogischer Fragen wird.
Full-text available
In the last decades, science literacy has been progressively emphasised as an aim of science education. However, the science education literature does not present a consensus about both the meaning of science literacy. By studying such literature, we identified some of these aspects that may also be relevant for promoting students’ critical education. These are decision-making, critical thinking, critical positioning, and self-awareness-raising. Based on this review of the literature, our main goal has been that of discussing how these aspects could be inter-related and favour science literacy in the context of nature of science teaching. This does not mean coming up with a new ready definition for science literacy, but rather enhancing the discussion and supporting its understanding as something more complex, resulting from an intricate interaction of many components and that may be constantly promoted and developed in science lessons. Such a perspective can help teachers’ perceptions that science literacy of their students is an attainable aim, as well as encourage them to foster it.
Full-text available
A atividade visa promover o desenvolvimento, por parte dos estudantes, do pensamento crítico sobre alguns aspectos de NdC a partir da história de Marie Curie. Para isto, propomos o uso de um caso histórico, uma vez que tal abordagem de ensino nos possibilita promover uma reflexão sobre alguns aspectos de NdC a partir de um contexto científico que envolve parte da história de Marie Curie. Nesta atividade, os estudantes podem refletir sobre alguns aspectos de NdC, principalmente os históricos e antropológicos (influência cultural) cuja discussão não é favorecida no capítulo anterior e, consequentemente, ampliar suas visões sobre Ciências.
Full-text available
Esta atividade visa promover a mobilização e/ou desenvolvimento, por parte dos estudantes, do pensamento crítico sobre e com alguns aspectos de NdC a partir da análise da tragédia ambiental que ocorreu em Brumadinho/MG. Para isto, propomos o uso de um caso contemporâneo, uma vez que tal abordagem requer o engajamento dos estudantes na discussão de uma situação-problema que envolve um contexto sociocientífico. Portanto, ao retomar tal tragédia ambiental, por meio de um caso contemporâneo, os estudantes podem refletir sobre alguns aspectos de NdC, principalmente os econômicos e filosóficos (ética), e aplicá-los na análise do caso. Isto porque o caso histórico (apresentado no capítulo 3), diferentemente deste que é de natureza argumentativa e proporciona um ambiente favorável para isto, não possibilita esta última ação (aplicação de alguns aspectos de NdC para se posicionar criticamente e/ou tomar consciência de outros aspectos, situações-problema e/ou problemas) e, consequentemente, ampliar suas visões sobre Ciências.
Full-text available
Esta sequência de atividades tem como objetivo introduzir aspectos de Natureza da Ciência (NdC) de maneira contextualizada e explícita em contextos de ensino para estudantes do Ensino Médio. Propomos três atividades que envolvem etapas diferentes de uma intervenção, sendo elas: (i) Atividade 1, etapa inicial, que visa favorecer a explicitação das ideias iniciais de estudantes sobre Ciências; (ii) Atividade 2, etapa principal, que objetiva introduzir aspectos de NdC de maneira contextualizada e explícita; e (iii) Atividade 3, etapa final, que visa identificar se, e em que extensão, houve uma ampliação da visão de estudantes sobre Ciências.
Full-text available
Esta sequência visa promover a tomada de consciência por parte dos estudantes com relação à existência de alguns aspectos de NdC a partir da vivência de práticas científicas e epistêmicas análogas às que cientistas vivenciam ao produzir conhecimentos científicos. Para isto, propomos o uso de atividades de modelagem, uma vez que esta abordagem de ensino engloba uma série de outras práticas científicas como as relacionadas à investigação, experimentação, argumentação, entre outras, utilizadas por cientistas para produzir conhecimentos científicos.
Biological brains are first and foremost the control systems for biological bodies. Biological bodies move and act in rich real‐world surroundings. These apparently mundane facts are amongst the main driving forces behind a growing movement within cognitive science – a movement that seeks to reorient the scientific study of mind so as to better accommodate the roles of embodiment and environmental embedding.