DRAWING THEORY
Visual Languages In Science
“A depiction
is never just an illustration… To understand a visualisation is … to inquire into … its principles of exclusion
and inclusion, to detect the roles that it makes available, to understand the
way in which they are distributed, and to decode the hierarchies and
differences that it naturalises.” (Fyfe & Law,1)
“[We scientists] view our pictures only as ancillary illustrations of what we defend by words. Few scientists would view an image itself as intrinsically ideological in content.” (Gould,28)
Martin Rudwick’s 1976 paper, The Emergence of a Visual Language for Geological Science 1760-1840, introduced the concept of the visual language into the realm of science studies:
[geological] prints are a means of communication, and as such the new modes of representation required new modes of perception by those who looked at them. Furthermore, the relation between the object depicted and its visual representation was never straightforward, however ‘realistic’ the intentions of the illustrator: artistic representation is always a visual language, which has to be learned and which changes over time. (Rudwick, 1976,151)
Since then, philosophers and historians of science have latched on to Rudwick’s discussion of material and artistic influences on the development of visual languages, and have become comfortable in their use of the term to talk about systems of scientific representation. However, many of these discussions about visual languages in science leave out a central discussion of theory-laden seeing and representing. This is because Rudwick’s term, visual language, is often examined from what could be called a visual rhetorical perspective – i.e. what information the image is supposed to communicate and how it relays that information – ignoring a more primary discussion of why the “relation between the object depicted and its visual representation was never straightforward”.
The concept of a visual language was first discussed in Ernst Gombrich’s Art and Illusion: A Study in the Psychology of Pictorial Representation (1976, n.7), a text which examines not only what visual vocabularies are and how they are used but, more importantly, how they structure the observer’s perception and representation of their subject. The move of Gombrich’s representational terminology into science studies requires an examination of those theories of seeing and representing in science, particularly in the realm of theory-influenced observation and representation. For, as Gombrich’s vocabulary transfers to the domain of science studies, it will become clear that particular scientific languages affect and restrict not only what scientists can and do represent, but also what they -- and therefore others -- can see.
This paper will first outline Gombrich’s original theory of seeing and representing, then will attempt to synthesise his account with those of three important philosophers of science, and will finally discuss historical examples of the theory-laden visual languages in practice.
GOMBRICH ON PICTORIAL REPRESENTATION
Ernst Gombrich’s Art and Illusion (1956) deals with the limitations of an artistic vocabulary and the associated issues of the artist’s perception and representation of the world around him. Because an artist tries to render the three dimensional world in a specific, often two-dimensional medium, he must have recourse to some kind of visual language to represent what he sees. This visual language is like a ‘code’: contrasting pigments, lines, and shades on canvas articulate and encode perceived relationships between elements in the world and communicate these relationships to the viewer:
What a painter inquires into is not the nature of the physical world but the nature of our reactions to it. He is not concerned with causes but with the mechanisms of certain effects … to explore the capacity of our minds to register relationships rather than individual elements. (Gombrich,49)
The visual language is not entirely shaped by the choice of medium, it is also constructed in response to the relationships that require representation. But how to sort the vast range of visual stimulation into categorical relationships? This is the role of the mental set, which sorts experience into expected and pre-formed categories – in Gombrich’s terms, schemata. It should be noted that schemata themselves are simply a categorization (Gombrich,74) – such as ‘church’, ‘city’, ‘animal’, or ‘man’ – and are “the starting point of the artist’s vocabulary” (Gombrich,183).
Once experience is sorted by the mental set, then the artist employs a visual language to articulate the schematised relationships between perceived objects in the visual field, making minor corrections to the schema in use to conform to individual instances. So, when an artist begins his work,
[he] begins not with his visual impression but with his idea or concept… The individual visual information, those distinctive features…, are entered, as it were, upon a pre-existing blank or formulary. And, as often happens with blanks, if they have no provisions for certain kinds of information we consider essential, it is just too bad for the information… And just as the lawyer or the statistician could plead that he could never get hold of the individual case without some sort of framework provided by his forms or blanks, so the artist could argue that it makes no sense to look at a motif unless one has learned how to classify and catch it within the network of a schematic form. (Gombrich,73)
The resulting image is “not a faithful record of a visual experience
but the faithful construction of a relational model” (Gombrich,90). Thus, a mental set has a significant effect
on what is ‘representable’: if the schemata do not correspond well to
experience or the visual language in use is too rigid, it will be increasingly
difficult to render a ‘realistic’ representation.
This claim is
supported by several examples in the text.
When Gombrich compares a Chinese and a British painting of the same
scene at Derwentwater, it is easy to see how a visual language influences
representation, as “the relatively rigid vocabulary of the Chinese tradition
acts as a selective screen which 
admits only the features for which schemata exist” (Gombrich,85).
But the most memorable of his examples are those animals dessiné al vif. In their depictions of elephants, lions (fig.
1), porcupines, rhinoceri (fig. 2), and whales, artists consistently
resorted to the use of pre-existing schemata, apparently incapable of
representing what was actually in front of their eyes in spite of their claims
to “truthfulness” (Gombrich,78-9). This is not surprising: the artist must
start from a mental set which categorises experience into elements the artist
knows how to represent. However, the
result of a “mental set which makes the artist look for certain aspects in the
scene” (Gombrich,85) can be an image which we would consider most un-life-like
and even “half-invented” (Gombrich,81).
Gombrich believes the issue to be not one of learning to see but learning to discriminate (Gombrich,172), cautioning: “it is dangerous to confuse the way a figure is drawn with the way it is seen”. But because the mental set categorises experience for the artist and leaves out those details which the schema or visual language are limited in representing, we cannot expect these ‘preconceived prejudices’(Gombrich,82) to refrain from intervention in the artist’s perception as well as his representation of the world. If this were not so, why should “you ha[ve] first to learn and practice how to draw ‘a man’ before you were even allowed in the life class” (Gombrich,157)? Why would, after studying schemata for chubby children, Gombrich state, “I never thought they could exist, but all of a sudden I saw such children everywhere” (Gombrich,168)?
What really happens when an artist or a scientist approaches must approach experience with a preconception of what he will see? The treatment of expectation, observation, and representation put forward by three major authors in science studies sheds new light on Gombrich’s theories, problematising the use of a visual language in science.
SEEING AND REPRESENTING IN SCIENCE STUDIES
Sorting Phenomena: Hanson And Kuhn
“There is more to seeing than meets the eyeball” (p.7) declares Norwood Russell Hanson in his 1958 book’s opening chapter entitled “Observation”. In his ensuing discussion of ‘theory-laden observation’, Hanson alerted philosophers of science to the importance of observation and visual representation in science, and the need for a proper analysis of these topics. For philosophers of science such as Hanson and, later, Thomas Kuhn, “theories and interpretations are ‘there’ in the seeing from the outset” (Hanson,10), and the boundaries between ‘seeing’, ‘perceiving’ and ‘representing’ are blurred.
Hanson places the act of ‘seeing’ in the very organization or categorization of the visual experience, asserting relationships between phenomena, and developing a “way in which elements are appreciated” (Hanson,13). For Hanson, this process is reliant upon some sort of context:
The context is part of the illustration itself [and is] ‘built into’ thinking, imagining, and picturing. We are set to appreciate the visual aspect of things in certain ways. Elements in our experience do not cluster at random. (Hanson,15)
Observation, for Hanson, is not only coloured from the start by theory and expectation due to prior knowledge, it is further influenced by what we know how to express, what language is available to us. The function of language here is significant. Language is not only used to express what we know, but also to ‘set’ our visual sensations: “how else could they be appreciated in terms of what we know?” (Hanson,26) A language is therefore highly theory-laden in itself, as it articulates those relationships which the theory-context detects or finds significant and then affects what relationships can be articulated. A visual language is similarly laden with contextual preconceptions, this time because it articulates the relationships which the mental set sees fit to discriminate.
These concepts are taken up in Thomas Kuhn’s Structure of Scientific Revolutions, in his description of “normal science”. This discipline operates under a paradigm which not only has considerable predictive power, but which also ‘sorts’ experience, providing categories for perception and articulating the relationships between phenomena. Phenomena must be forced into a “preformed and relatively inflexible box that the paradigm supplies” (Kuhn,24), and because of the large role expectation plays in observation under a paradigm, a scientist operating under a paradigm may not even see an anomaly when it arises. So a paradigm controls scientific perception both by confining the range of observation to those areas which the paradigm has pre-identified, and by sorting all experience into categories – whether it fits or not.
The Kuhnian paradigm, therefore, operates like the ‘mental set’ described by Gombrich. Both provide categories for the artist or the scientist to sort his/her perceptions of the world, thus prescribing what is observed and what can be represented. Under a paradigm, normal science is directed towards “the articulation of those phenomena and theories that the paradigm already supplies” (Kuhn,24); under a mental set, while a schema may be corrected for individual representations, the actual categories into which experience is sifted also do not change. Just as paradigmatic procedures “restrict the phenomenological field accessible for scientific investigation” (Kuhn,60), the visual language with which an artist articulates a schema restricts the represent-able phenomenological field. And finally, when a paradigm or a mental set is fundamentally altered, it is not the artist or scientist’s interpretation of an objective observation which changes, but the very ‘seeing’ of the phenomenon itself (Kuhn,120). In spite of Gombrich’s Popperian sympathies, both Kuhn and Gombrich talk about perception in a surprisingly similar way: as governed by a meta-theory which categorises experience and mediates what can be perceived[1].
For these philosophers of science, the context, paradigm or theory is absolutely primary to the seeing process: it provides categories for the scientist to sort their perceptions of the world and not only prescribes what is ‘seen’ in the first place, but also what can be represented. This is why it is necessary for scientists to be well trained in the theory behind their work before they are taught how to practice in their field: they must ‘learn how to see’. A visitor to a laboratory “must learn some physics before he can see what the physicist sees [before the context will] throw into relief those features of the objects before him which the physicist sees as indicating resistance” (Hanson,17). In any case, “‘seeing’ in science is a profoundly ‘theory-laden’ undertaking[, as] [o]bservation of x is shaped by prior knowledge of x” (Hanson,19), or at least an understanding or expectation of how x should fit into a pre-existing paradigm. It is interesting that, just like the student in the physics laboratory, Gombrich’s artist-in-training must attend art school to acquire a mental set, to learn how to sort what he sees into schemata and learn a visual language to represent what he has learned to see.
In his Representing and Intervening, Ian Hacking declares:
Human beings are representers. Not homo faber, I say, but homo depictor. People make representations. (Hacking,132)
By ‘representations’, Hacking means not only artistic likenesses but also theoretical models and “complicated speculations which attempt to represent our world” (Hacking,133-4). Representations, he claims, are primary, after which a ‘reality’ is constructed to which the representation can be compared:
Reality is an anthropomorphic creation… It will be protested that reality, or the world, was there before any representation or human language. Of course. But conceptualizing it as reality is secondary. First there is this human thing, the making of representations. Then there was the judging of representations as real or unreal, true or false, faithful or unfaithful. Finally comes the world, not first but second, third or fourth.” (Hacking,136)
This concept is complicated when Hacking introduces that of ‘style’, without which there is no representation. These styles “grow with representation as materials are worked, and craftspeople produce artifacts …” (Hacking,137). For each different style, there is a different way of representing, there is therefore a different corresponding concept of reality.
Hacking’s claim that different styles of representation conceptualise different realities is crucial to understanding the problem of the visual language. To bring the metaphor back to the visual arts, ‘styles’ can be seen as individual mental sets and their corresponding collections of schemata or languages. If each style represents differently, each also constructs reality differently. As an example, Balla, Cezanne and Picasso each present a different visual language aimed at capturing different aspects of reality: colour play on the retina, speed and motion, or multiple dimensions. As each artist’s visual language aims to represent different things, each correspondingly sorts the world differently into categories and relationships. In Hanson’s terms, each ‘sees’ the world differently; in Hacking’s terms, each presents a different reality. This is why Gould later argues that “a complex shift in ideas is epitomised by an alteration in pictures” (Gould,26): visual languages not only concretise different relationships and articulate particular paradigms, they also articulate the representer’s concept of reality.
If there are different realities articulated by different representations, which representation should we trust? In extreme readings of the Kuhnian incommensurability problem, there are “no criteria for saying which representation of reality is the best” (Hacking,144). This concept had been articulated by Duhem earlier: “there is no truth of the matter – there are only better or worse systems of representation, and there might well be inconsistent but equally good images of mechanics” (qtd. in Hacking,143). The existence of multiple systems of representation can only call into question the ‘real’ nature of reality (Hacking,139): it cannot distinguish the best representational system to choose. This may not present a problem in the visual arts, where different systems of representation are employed by different artists. But the scientist’s purpose, Hacking argues, is to look for the best representational model: that which places elements in meaningful relationships to each other, which best builds or fits a paradigm, which best represents reality. The scientist who selects a particular visual language does so because she feels it best provides a system of representation for the reality established by her paradigmatic theory.
The concept of a visual language which Rudwick incorporates into science studies, then, does in fact put “key theoretical concepts into visual form, by creating a style of visual reasoning, and by establishing a boundary that defines research problems and projects and central or peripheral” (Pang). Just as Gombrich bridges the theoretical and the representational through the visual language, encapsulating relationships as determined by a mental set, the scientific visual language makes the theoretical relationships determined by a paradigm, visual.[2] Thus a scientific visual language reflects the scientist’s concept of reality and, as a visualisation of paradigmatic relationships, it constricts the scientist’s perception of reality as well.
‘SEEING’ WITH A SCIENTIFIC VISUAL LANGUAGE
Uses And Limitations Of A Visual Language
When artists employ a visual language, they are articulating the discriminations made by a pre-existing mental set, the meta-theory of perception which sorts experience into categories of relevance and representability. When scientists employs a visual language, they are also implicitly engaging in a theory-laden undertaking. A pre-existing paradigm provides a mental set, discriminating which experiences should and can be represented. Choice of visual language makes concrete the relationships between perceived phenomena, determining what the scientist is able to express and, therefore, to a large extent what he is able to see. Further, radically different languages can be seen as depicting different realities. In short, the visual vocabulary available to the scientist reflects and supports the paradigmatic theory, influences what can be represented, and restricts what is actually perceived in and as the real world. Several examples from the history of science can be cited in which the use of a particular visual language influences what is depicted and what is seen – or not seen.
Like Gombrich’s al vif artists discussed above, the artists
in Rudwick’s Scenes from Deep Time (1992) resort to an available visual
vocabulary to represent their pre-historic scenes. Martin, Unger, and Hawkins
employ pre-existing schemata to “reconstruct… animals from disarticulated
skeletons” (Gould,23), representing the pre-Adamic creature as a composite
monster of floppy dog-ears, fish-like scales, or elephantine trunks (fig.
3), and eventually develop a specialised visual language and set of schemata
for imaging prehistoric scenes. These images illustrate clearly the extent to
which “many of our pictures are incarnations of concepts masquerading as
neutral descriptions of nature.
These [images] are the most potent sources of conformity, since ideas passing as descriptions lead us to equate the tentative with the unambiguously factual. Suggestions for the organization of thought are transformed to established patterns in nature. Guesses and hunches become things.” (Gould,28)
In this case, the visual language’s ‘realisation’ of the theoretical lends a sense of reality to otherwise necessarily tentative hypotheses: i.e. the existence of trunks, scales, or floppy ears. And although some of Rudwick’s pictures may seem laughable today, these early representations have certainly shaped our modern image of the dinosaur.
Another use of a visual language in the history of science is seen in Galileo’s revolutionary discovery of lunar topography. Although Galileo’s telescope certainly gave him a better view of the moon, his experience as a student of optics and as a member of the Accademia del Disegno gave him a distinct advantage: he possessed the pre-requisite mental set and visual language to not only ‘see’ craters on the moon, but also to represent what he saw:
Galileo’s [moon sketching] clearly shows how well he understood the way that sunlight differentiates a concave hollow from a convex protrusion…Because Galileo had already practiced such perspective-projection exercises during his youthful studies of geometry and optics, he was able to recognize instantly that the mysterious markings on the magnified moon were not caused by alabaster-like marbling as the traditional Aristotelians maintained, but by the sun illuminating an irregular, opaque lunar plain… (Lynch & Edgerton,104-5)
This
revolutionary vision is called into sharper relief with the comparison of
Galileo’s images to those of Thomas Harriott, an English contemporary. It
becomes clear that the images in Siderius Nuncius would not have been
possible without the pre-
existence of a mental set which enabled Galileo to see what
retinal images he was presented with in terms of topographical
relationships. Further, because “[l]inear
perspective and chiaroscuro shadow-rendering had been the stuff of
European art and printed-book illustration already for a century and a half”
(Lynch & Edgerton,105), Galileo could use a commonly understood visual
language to effectively communicate the new way of seeing the moon to his
contemporaries, profoundly influencing Harriott’s subsequent drawings of the
same object (see fig. 4). Some authors even maintain that Galileo
manipulated his drawings, distorting his visual language “in the service of his
verbal argument that [the moon’s] surface is rugged like that of the earth”
(Winkler & Helden,217). In this case, a particular and existing mental set
and associated visual language were needed in order to see and communicate a
new reality.
But most refined scientific visual languages have little room for
novelty. Consider, for example, Bohr’s atomic model (fig. 5). In this
visual language, an articulation of relationships perceived by a scientific
paradigm, a nucleus is drawn of protons and neutrons, surrounded by the correct
number of orbiting electrons in perfectly circular rings. The language is
highly tailored to represent precisely and only what is of concern to atomic
physicists working under ‘normal science’; there is, for example, no room for
neutrinos or x-rays. However, a visual language so refined can only be
inflexible, as it represents only what relationships the theory has determined,
what particles exist in that reality. There is no room for observations which
do not ‘fit the box’. As the
theoretical paradigm encountered crisis and a shift was made towards quantum
mechanics, physicists were presented with the significant problem not only of
representing elements such as quarks, electron clouds or probability
relationships, but also – necessarily -- of seeing them in the first place.
If the use of a visual language guarantees an essentially theory-laden enterprise, what possibilities exist for objective accounts? Although the process of seeing and representing can never truly be free from the lens of theory, the description of astronomic image processing by Michael Lynch and Samuel Edgerton may create the possibility of freeing the eye from its theory-laden glance. This is because, with the aid of complex computer graphic tools, the astronomers which Lynch and Edgerton observe can easily manipulate photographs, converting data “into processed images in order to more easily “see the physics” (Lynch & Edgerton,111):
The machinery simulates what the astronomer would have us see as the ‘natural’ object… The programming is also used to sharpen the scattered configurations of light to pack the figure back into what the object is supposed to look like. (Lynch & Edgerton,120-21)
The astronomer can utilise any one of a series of visual languages to convey his data, “selecting colors, composing textures, framing features, imposing scales, and establishing gestalt coherences” employed by that particular language[3] to communicate the relationships which he finds relevant. For example, an image of a double-tailed comet was ‘teased out’ through the manipulation of colour tones and contrasts, highlighting and communicating a ‘discovery’ (Lynch & Edgerton,116) which would not be visible or representable in another visual language. Although, through the use of such visual-language filters, an astronomer can easily impose his theory onto an observation, a simple click of a button allows the image to be viewed through the filter of another visual language, another mental set – another paradigm.
Certainly, each of these visual languages ‘comes with’ a particular theory inextricably attached. However, the availability of multiple visual languages which isolate and articulate many different possible relationships may bring the possibility of a minor degree of autonomy from the paradigm in a scientific community. As Helen Longino argues,
…the greater the number of different points of view included in a given community, the more likely it is that its scientific practice will be objective… points of view cannot simply be allowed expression but must have an impact on what is ultimately thought to be the case... (Longino,80)
It is possible for visual languages to provide these required different points of view, and a dialogue between these languages may eventually “block the influence of subjective preference at the level of background beliefs” (Longino,73): the process Longino finds so crucial to scientific objectivity. Hacking himself states that “as systems of representation multiply, we become sceptics and form the idea of mere appearance” (Hacking,141-42) as something distinct from reality. If the ‘monopoly’ of a visual language in a scientific discipline can severely limit the possibilities for perception, the ability to switch between multiple languages at the click of a mouse – each language theory-laden to be sure, but each articulating a different theory – may create a greater possibility for ‘objective seeing’ in science.
Ultimately, the problem with visual languages is that they concretise the unknown, the invisible or the theoretical in a way that makes theory seem like reality. Bohr’s atomic model described what the atom was ‘really like’ for scientists in the same way that the mechanical dinosaurs in Jurassic Park turned a theoretical view of Deep Time into what dinosaurs were ‘really like’. In these cases, it is important to remember the distinction Hacking makes between representation and realities: the representation must remain distinct from whatever ‘reality’ is believed to exist objectively. Gombrich makes a similar claim as to the ‘truth’ of images:
If all art is conceptual, the issue is rather simple. For concepts, like pictures, cannot be true or false. They can only be more or less useful for the formation of descriptions. (Gombrich,89)
A representation is simply the articulation of a particular reality, a particular ‘seeing’ of the world through a paradigmatic mental set. It may have claim to the viewer’s reality, but it should have no claim to a transcendent ‘reality’. Hopefully, just as “those who understand the notation will derive no false information from the drawing” (Gombrich,90), those who understand the limitations and theory-laden nature of visual languages in science will know not to read a representation as ‘the way things really are’. Perhaps the ease with which images are manipulated on a computer will have an impact on the general understanding of scientific image making: at the very least, Lynch and Egerton’s examination of astronomical image processing demonstrates the many languages into which the same image can be ‘translated’ instantly, and this multiplicity of representations in the public domain may increase awareness and critical thinking on the part of the viewer.
Martin Rudwick opens The Emergence of a Visual Language for Geological Science with the claim that it is “worth studying the historical development of the visual language of geology not only for the way in which it gradually enabled the concepts of a new science to be more adequately expressed, but also as a reflection of the growth of a self-conscious community of geological scientists” (Rudwick,1976,151). But our analysis of the original Gombrichian description of the visual language from Hanson, Kuhn, and Hacking’s points of view has enabled the scientific visual language to grow beyond this description. Considering the above discussion, the scientific visual language is worth studying in its historical context for the way in which it expresses a paradigm-informed reality, and the way in which it enables or disables the scientist’s ability to ‘see’ and represent their world.
BIBLIOGRAPHY
Bull, Malcolm. Scheming Schemata. British Journal of Aesthetics, Vol.34, No.3, 1994, 207-217.
Fyfe, Gordon & John Law. “Introduction: on the invisibility of the visual”. Picturing Power : Visual Depiction and Social Relations. Fyfe & Law, eds. London: Routledge, 1988. 1-14.
Gombrich, E.H. Art and Illusion: A study in the Psychology of Pictorial Representation. The A.W. Mellon Lectures in the Fine Arts, 1956. Princeton: Bollingen Series, 1960.
Gould, Stephen Jay. “The iconography of expectation”. In: Wonderful Life: The Burgess Shale and the Nature of History. New York: W.W. Norton, 1989. 23-52
Hacking, Ian. Representing and Intervening. Cambridge: Cambridge University Press, 1983.
Hacking, Ian. Style for Philosophers. Studies in the History and Philosophy of Science, Vol. 23 No. 1, 1992. 1-20.
Hanson, Norwood Russell. Patterns of Discovery: An inquiry into the conceptual foundations of science. Cambridge: Cambridge University Press, 1958.
Kuhn, T.S. The Structure of Scientific Revolutions. 2nd Ed. Chicago: University of Chicago Press, 1962.
Latour, Bruno. Drawing Things Together. 19-68.
Longino, Helen E. “Values and Objectivity.” Science as Social Knowledge: Values and Objectivity in Scientific Inquiry. Princeton: Princeton University Press, 1990.62-82.
Lynch, Michael and Samuel Edgerton. “Abstract Painting and Astronomical Image Processing.” The Elusive Synthesis: Aesthetics and Science. A.I. Tauber ed. Netherlands: Kluwer Academic Publishers, 1996. 103-24.
Nersessian, Nancy J. “How Do Scientists Think? Capturing the Dynamics of Conceptual Change in Science.” Cognitive Models of Science. Ronald N. Giere, ed. Vol XV, Minnnesota Studies in the Philosophy of Science. Minneapolis: University of Minnesota Press, 1992.
Pang, Alex Soojung-Kim. “Visual Representation and Post-Constructivist History of Science”. Historical Studies in the Physical and Biological Sciences. Vol. 27, 1997. 139-71.
Rudwick, Martin J.S. “The Emergence of a Visual Language for Geological Science 1760-1840”. Historia Scientiarum. Vol. xiv, 1976. 149-95.
Rudwick, Martin J.S. Scenes From Deep Time. Chicago: University of Chicago Press, 1992.
Winkler, Mary & Albert Van Helden. “Representing the Heavens: Galileo and Visual Astronomy.” ISIS. Vol. 83, 1992. 194-217.
Wittgenstein, Ludwig. On Certainty. 1969. Ed. Anscombe & Wright. Trans. Denis Paul and G. Ansombe. New York: Harper Torchbooks, 1972.