I argue for three claims: (1) For Berkeley, given his view of extension, Euclidean (classical) geometry must be empirically false; a view famously explicit in his early Notebooks (NB). (2) The method of selective attention for the purpose of representative generalization, as presented in the Introduction to The Principles of Human Knowledge (PI),Footnote 1 plays no significant role in generalizing proof results in classical, i.e., Euclidean, geometry. (3) That, in practice, Berkeley must have considered classical geometry a useful fiction; i.e., that strictly speaking, the fundamental terms of classical geometry, “point,” “line,” “plane,” etc., lack reference.Footnote 2
In this regard I make a distinction between abstraction as Berkeley envisioned it in PI, and idealization. I contend as a consequence that idealization, for the purpose of classical proof, automatically serves the goal of representative generalization without need to selectively attend to diagram properties beyond, of course, taking a perceived or constructed figure to be, qua perceived, as triangular, square, circular, etc.
Empirical Geometry
In entry A 770 of (NB) (1709) Berkeley writes:
Qu: whether geometry may not properly reckon’d among the Mixt Mathematics. Arithmetic and Algebra being the only abstracted pure i.e. entirely nominal. Geometry being an application of these to points.Footnote 3
Berkeley’s well-known answer in the Notebooks is “yes.” The object of geometry, unlike arithmetic or algebra, is sensible extension composed of sensible minima. Therefore, many if not all theorems of Euclidean or classical geometry, and therefore the postulates, are false. Some examples: The ratio between circumference and diameter of a circle is the same for all circles (NB B340). Line segments are infinitely divisible. (B26). The diagonal of a square isn’t commensurable with its sidesFootnote 4 (NB B258).
This is well known, but as commentators note, it’s risky to take the early Notebooks as authoritative about Berkeley’s ultimate views. Douglas Jesseph and Zoltan Szabo, for example believe Berkeley’s later view of geometry in The Principles demonstrates a significant shift in his thinking about geometry.Footnote 5 Certainly there’s an important change in emphasis. However, textual and formal considerations suggest Berkeley never relinquished his view that sensible extension is composed of sensible minima. In the New Theory of Vision (NTV), Berkeley claims that both segments of visible and tangible extension are composed of minimaFootnote 6 (See also NTV 80-83). In PHK 123, Berkeley questions whether classical geometry requires line segments to be infinitely divisible, suggesting that Euclidean geometry might work with line segments considered to have a finite number of points; a project which, while not pursued, suggests he still believes extensive segments are composed of minima.
Formal considerations, perhaps more important, also dictate taking sensible finite extension to be non-continuous. I take it as an a priori truth accepted by both Berkeley and later Hume that sensed line segments are composed of sensible atoms. This is a point about phenomenology. The alternative—that as a segment phenomenally diminishes, there will be for every putative minimum one appearing smaller—would be unintelligible to Berkeley. Jesseph raises the interesting problem that if Berkeley accepts: (1) that there is nothing hidden in what we immediately see, and (2) line segments appear continuous, then we should accept (3) finite line segments are continuous. Jesseph is I think correct to the extent we can’t immediately visually or tactually perceive segments as composed of minima. Then, per impossible, we would perceive boundaries between minima that are less than a minimum. But this is consistent with there being a last sensible atom as a segment visually diminishes, and a first as one comes into view.Footnote 7 In NTV 80, Berkeley notes since the minimum visible cannot by its nature be distinguished into parts, it must be the same for every “creature” with vision. He believes this is a necessary truth. In NTV 81, Berkeley writes: “Now for any object to contain several distinct visible parts, and at the same time be a minimum visible, is a manifest contradiction” (Berkeley’s italics). I think this is compatible with a finite visual length being composed of minima though of necessity looking continuous.Footnote 8 See also NTV 83, where Berkeley, remarking on the “perfections” of the “visive faculty,” mentions two; “first that of comprehending in one view a greater number of visible points; Secondly, of being able to view them all equally and at once, with the utmost clearness and distinction.” This might suggest (problematically for Berkeley), as Jesseph notes, that we should see each minimum as bounded by other minima, that is, see the boundaries, which, of course, we can’t do. But again Berkeley likely means we see clearly and distinctly all the minima, though not as joined minima. There is nothing in the visual content not seen.
Selective Attention and Representative Generalization
Berkeley comments at least once in the Notebooks about selective attention, writing, “Considering length without breadth is considering any length be the Breadth what it will” (NB A722). In PI, Berkeley considers geometrical proof in the context of discussing how, in a world of particulars, language can possess general terms. Commenting on the bisection of a line, he writes: “I believe we shall acknowledge that an idea when considered in itself is particular, becomes general by being made to represent or stand for all other particular ideas of the same sort” (PI 12 my italics).
Similarly in PI 15:
Thus, when I demonstrate any proposition concerning triangles it is to be supposed that I have in view the universal idea of a triangle, which ought not to be understood as if I can frame an idea which was neither equilateral nor scaleon, nor equicrural, but only that the particular triangle I consider, whether of this or that sort it matters not, does equally stand for and represent all rectilinear triangles whatsoever, and is in that sense universal. All of which seems very plain and not to include any difficulty in it (my italics).Footnote 9
In PI 16 Berkeley writes:
To which I answer that, although the idea I have in view whilst I make the demonstration be, for instance, that of an isosceles rectangular triangle whose sides are of a determinate length, I may nevertheless be certain it extends to all other rectilinear triangles of what sort of bigness soever. And that because neither the right angle, nor the equality, nor determinate length of the sides are at all concerned in the demonstration. . . .And here it must be acknowledged that a man may consider a figure merely as triangular, without attending to the particular qualities of the angles, or the relations of the sides. So far he may abstract: but this will never prove, that he can frame an abstract general inconsistent idea of a triangle. (my italics, [also NB A723, PHK 126]
Ideas (excluding ideas of reflection and imagination) are, for Berkeley, sensible objects. Thus, opposing Locke, Berkeley takes the abstract general idea of a triangle to be inconsistent, having to be simultaneously equilateral and scalene.Footnote 10 The method of selective attention allegedly avoids this. We focus on a property of a constructed triangle, say, its triangular character, needed to prove a theorem; e.g., that the sum of its angles equals two right angles. Since just that property of the diagram is involved the theorem allegedly applies to all triangles.
However, though we can pay attention to the linearity of the drawn sides of a triangle, ignoring their width and depth, we can’t observe that those lines conform to the postulates, axioms, and definitions of classical geometry. We can’t perceive of constructed triangles that their boundaries are infinitely divisible though that’s arguably implied by the postulates.Footnote 11 Although in The Elements it’s not part of the definition of a straight line—“a line which lies evenly with the points on itself” (Heath 153)—that line segments are continuous, that appears implied by postulate 2, a construction postulate, which states that [one can] “produce a finite straight line continuously in a straight line”Footnote 12 (Heath 196). If continuousness implies denseness—that between any two segment points there exists a point between them, then Euclidean straights are not composed of minima. Moreover, as Heath notes, later commentators on Euclid’s “proposition (theorem) 10—“to bisect a given finite straight line”—thought infinite divisibility was either a presupposition of classical geometry, or a consequence of the ability to construct incommensurable lengths (Heath 268). For Berkeley all sensible finite lengths, being composed of minima, would be commensurable. In NB B262 he reminds himself to consider whether the “incommensurability of diagonal and side” [of a square] assumes a unit be “divisible ad infinitum.”
We might say that Euclidean theorems could be true of some but not all figures. For example, the Pythagorean theorem applies to right triangles with certain sets of triples, for example, sides of 3, 4, and 5 minima. However, the deeper question is why someone believes the Pythagorean theorem true in any particular case. Presumably she would refer to proposition 47 in a text of Euclid (Heath, 349). [or a translation of the time.]Footnote 13 But of course there would be no way of knowing whether the construction used in proving proposition 47 had the requisite number of minima to be a Pythagorean triple. We would rightly say that’s irrelevant to the proof, but again that returns us to the question of what if anything the demonstration is about.
Jesseph and Szabo do recognize that proof results can’t strictly apply to diagrams used in a demonstration.Footnote 14 Indeed Principles PHK 126, as Szabo points out, illustrates the difficulty Berkeley would have in thinking proof results apply to actual constructions. Berkeley first reminds the reader that he has explained in PI 15 what he means by “universal ideas” with respect to “theorems and demonstrations:” “that the particular triangle I consider, whether of this or that sort it matters not, does equally stand for and represent all rectilinear triangles whatsoever, and it is in that sense universal” (my italics). Presumably the quantifier’s scope in the italicized phrase includes the diagram (a specific rectilinear triangle) in the proof. PI 15 and 16 (above) support this presumption.Footnote 15
In PHK 126, however, Berkeley gives “universal” a limited extension; a demonstration refers only to those figures where a needed construction, e.g., bisecting a line segment, is empirically possible. Berkeley claims that the actual size of a segment in a diagram—that it is an inch long—though said to contain ten thousand parts,” is “indifferent to the demonstration.” He writes rather that the inch line is “universal in its signification in the sense that it ‘represents innumerable lines greater than itself’ in which may be distinguished ten thousand parts or more, though there may be not an inch in it” [my emphasis]. The difficulty is that whereas the discussion in PI asks us to ignore the actual dimensions of figures used for proofs, PHK 126 makes the size of drawn segments relevant to what figures theorems refer to. Szabo notes this issue about quantification; on the one hand, Berkeley seems to claim that a line segment used in demonstrations can represent all segments, while on the other hand it apparently represents only segments where a division is practically possible. Szabo correctly writes that if we have to check “whether the proof of the theorem can be applied to a particular idea,” we have in fact no standard of generalization.Footnote 16
The following perhaps exemplifies Szabo’s point. Suppose a geometer proves the sum of the angle theorem for a constructed obtuse triangle. How could she know the theorem applies to a constructed acute triangle? The ordinary (and Berkeleian) reply is that in the proof angle size plays no role. Angle size is indeed irrelevant in Euclid’s proof but not simply because it plays no role, though that’s true, but because the conclusion isn’t strictly true of any sensible triangle.Footnote 17
Idealization vs. Abstraction
Idealizations, to borrow a phrase from Michael Weisberg, are “intentional fictions.”Footnote 18 My claim here is that Berkeley would as a matter of course take all of classical (pure) geometry to be an intentional fiction; the points, lines, planes, etc., related by the postulates are, strictly speaking, referentially empty.Footnote 19 As mentioned, Szabo correctly notes the difficulties Berkeley has in thinking both that geometry was about sensible extension and that there could still be a standard of generality for geometrical proof. He suggests one solution would be to deny that Berkeleian ideas are in fact the subjects of proof. Szabo writes: “first of all the possibility that [classical] geometry does not have objects has not been discussed at all.”Footnote 20 But although Berkeley doesn’t explicitly discuss whether classical geometry literally has objects, some admittedly brief comments from De Motu show he found referentially empty general terms useful in mechanics and geometry.
And just as geometers for the sake of their art make use of many devices which they themselves cannot describe nor find in the nature of things, even so the mechanician makes use of certain abstract and general terms, imagining in bodies force, action, attraction, solicitation, etc. which are of first utility for theories and formulations, as also for computations about motion, even if in the truth of things, and in bodies actually existing, they would be looked for in vain, just like geometers’ fictions made by mathematical abstraction (DM 39) (my italics).
The phrase “made by mathematical abstraction” is interesting, for although Berkeley evidently thinks such abstraction legitimate, it isn’t a process of selectively attending to a real property, say, the color, of a perceived object, which then can represent that color on the surface of other objects. In DM 17, referring to “impressed forces,” Berkeley writes:
Force, gravity, attraction, and terms of this sort are useful for reasonings and reckonings about motion and bodies in motion, but not for understanding the simple nature of motion itself or for indicating so many distinct qualities. As for attraction, it was certainly introduced by Newton, not as a true, physical quality, but only as a mathematical hypothesis (Berkeley’s emphasis).Footnote 21
Similarly, although we can’t selectively attend to Euclidean properties of perceived figures if there are no such properties, or, perhaps equally sufficient, if we can’t in principle discern whether a figure is Euclidean, we take it as a matter of useful convention that the figure satisfies the postulates. The convention isn’t arbitrary but rather idealizes appearances. Idealizing appearances here means that we consider various real “straights,” for example, plumb lines, or lines constructed with a straight edge, to have Euclidean properties, that is, to conform to the Euclidean postulates; for example, no intersecting straight lines enclose a space. In fact Berkeley must have viewed in this way—as idealizations or geometrical fictions—Newton’s figures in the Principia, Euclid’s (or a translator’s) diagrams in The Elements, when Berkeley studied geometry, and his own diagrams in The Analyst. Idealizations are neither sensible objects nor Platonic Forms, (something Berkeley certainly rejected), but ways we decide to treat sensible objects, say, geometrical diagrams, either for theoretical reasons (e.g., doing proofs) or for practical concerns (e.g., carpentry, architecture).
Jesseph suggests that in The Analyst, where Berkeley uses classical geometry to criticize Newton and Leibniz’s calculus, he rejected his earlier critique of Euclidean geometry in the Notebooks. Jesseph quotes the following:
It hath been an old remark that Geometry is an excellent Logic. And It must be owned that when the Definitions are clear; when the Postulata cannot be refused, nor the Axioms denied; when the distinct Contemplation and Comparison of Figures, their Properties are derived, by a perpetual well-connected chain of consequences, the Objects being still kept in view, and the attention ever fixed on them; there is acquired a habit of Reasoning, close and exact and methodical: which habit strengthens the Mind, and being transferred to other Subjects is of general use in the inquiry after Truth. But how far this is the case of our Geometrical Analysts, it may be worth while to considerFootnote 22 (my italics).
In my view, the best way to think of the phrase “when the Postulata cannot be refused” is that within a certain perceptual range some postulates—e.g., that in a plane two oblique straight lines intersect at only one point—appear self-evident.Footnote 23 Hume later claimed this postulate is perceived as true only within a limited visual expanse; a point Berkeley perhaps, as a point about observation, would have agreed with.Footnote 24
However, Berkeley, for demonstrative purposes, would idealize the sensible construction as Euclidean, as he would have needed auxiliary lines or circles permitted by the construction postulates. Idealization—conceived here—involves no special act of imagination, and certainly not abstraction as articulated in PI; rather the boundaries of constructed polygons, for example, are simply treated as Euclidean straights; satisfying the classical postulates.Footnote 25 Berkeley accepted Euclidean straights (representing light rays) as useful fictions in geometrical opticsFootnote 26 (NTV 13, 14). In dynamics he accepted the parallelogram of forces as a fictional but useful device for computing resultant forces (DM 18). Moreover, though not mentioning the case, he likely would have accepted as idealizations the frictionless surfaces and perfect spheres Galileo assumes in formulating the law of free fall.Footnote 27
Of course unlike the parallelogram of forces, diagrams in proofs are real figures; for Berkeley, bits of sensible extension. But as idealized—i.e., treated as satisfying the postulates—they are no more real than the perfect spheres and frictionless planes of Galileo. And as Galileo, to make use of classical geometry, introduced idealizing assumptions, Berkeley must have taken all of Euclidean geometry itself to be a useful fiction. Apropos here is a section from a “Dialogue” of Leibniz (1677) between A (presumably Leibniz), and an interlocutor B, about geometric constructions. B notes the importance of “contemplating constructed figures accurately.”
A: True, but we must recognize that these figures must also be regarded as characters, [symbols] for the circle described on paper is not a true circle and need not be; it is enough that we take it for a circle.
B: Nevertheless it has a certain similarity to the circle, and this is surely not arbitrary.
A: Granted; therefore figures are the most useful of charactersFootnote 28 (my italics).
The significant point is that taking Euclidean geometry to idealize appearances (within a certain range) solves the problem of representative generalization without need for acts of selective attention.Footnote 29 The “standard of generalization,” (Szabo) in doing a proof is built into one’s conception of the diagram. In treating, as opposed to recognizing, a construction with straight edge and compass as a Euclidean isosceles triangle, theorems deduced (its base angles are equal) ipso facto apply to other observed or constructed figures taken to be Euclidean isosceles triangles.
Selective attention of course plays some role here. We might want to know the criteria used for selecting a particular or constructed figure as Euclidean. However we solve that problem, it’s one distinct from the alleged role of selective attention in doing proofs. For example, with straight edge and compass I describe a triangle on paper, in order to prove the sum of the angle theorem. I’ve already decided to consider both the drawn boundaries as Euclidean straights, as well as the additional required line constructed through the vertex parallel to the base. I’ve assumed, via the fifth postulate, that one and only one such parallel exists. The problem of representational generalization is solved. That is, the “sort” or “kind” I’m dealing with—a Euclidean triangle—is established by making these assumptions, which is in fact to idealize the figure. The theorem proven then is true of any other Euclidean triangle. If I simply attend to the sensible “triangular character,” of the figure, the Berkeleian idea, then the proof doesn’t get started. Again idealizations are neither Berkeleian ideas nor Platonic forms. Conceived operationally they are not sensible objects at all, but rather involve conventions about how to treat certain sensible objects. And if we get very odd results in applying this geometry to the world we refine our measurement procedures, or—as with Einstein—ultimately change the geometry we apply to the world; a possibility likely not considered in the early 18h century.Footnote 30
Brief (Speculative) Post-script: Berkeley, Formalism and Geometry.
I think Berkeley might have found congenial, had they been around, the ideas of David Hilbert (1899). For Hilbert, basic Euclidean terms, “point,” “line,” “plane,” etc. lack extra-systemic reference, but are implicitly defined by their relations in the postulates.Footnote 31 In particular they don’t necessarily refer to space.Footnote 32 Berkeley’s query in NB A770 (quoted above) whether geometry is applying the formal (“nominal”) systems of arithmetic and algebra to “points,” is suggestive here. But it seems unlikely he would identify Euclidean geometry simply with its axiomatic structure. If we take the Analyst2 passage above seriously, then the “postulata” of geometry must at least seem self-evident about what we see (or perhaps touch).Footnote 33
There have been good discussions of Berkeley’s “formalist” approach with respect to arithmetic and algebra.Footnote 34 Robert Baum notes the ambiguity in discussions of formalism, between (1) when the non-grammatical terms in a system lack any denotation, and (2) when such terms denote, but can be manipulated according to rules with no attention paid to their referents. Berkeley, as Baum observes, generally accepts the latter interpretation for arithmetic. (PHK 120-122, Alciphron VII 12,13), but does mention what he calls “the algebraic mark which, denotes the root of a negative square, hath its use in logistic operations, although it be impossible to form an idea of any such quantity” (Alciphron, VII, 14, my italics).
However, neither Baum, nor Jesseph discuss how Berkeley might find “formalism,” in the sense of an uninterpreted calculus, like algebra (a set of marks manipulable through rules), as a way of envisioning classical geometry. The historical problem was developing geometry’s axiomatic structure in words or symbols but dispensing with diagrams except, as with Hilbert, as aids in grasping the formalism.
