(Saha, 2018)
You can also download and read or share a .pdf of the complete text of this essay by scrolling down to the bottom of this post and clicking on the Download icon.
The Integration Illusion: Why Calculus’s Greatest Trick is a House of Cards
1. Introduction
Calculus textbooks sell integration as the noble inverse of differentiation: a yin-yang harmony that lets us reverse-engineer areas, volumes, and accumulated change from instantaneous rates. “If f′(x) = g(x), then ∫g(x) dx = f(x) + C,” they intone with priestly certainty. Strip away the varnish, however, and integration collapses into a fragile, context-bound sleight-of-hand. The plurality of “integration theories”—Riemann, Lebesgue, Henstock-Kurzweil, Daniell, and more—is not progress toward a grand unified truth; it is a graveyard of patched-up failures, each papering over the cracks of the last. From a skeptical lens, integration is not discovery: it’s damage control.
2. The Fundamental Theorem of Calculus: A Lie We Tell Freshmen
The FTC is the jewel in the crown of calculus: if f is continuous on [a,b] and F(x) = ∫ₐˣ f(t) dt, then F′(x) = f(x). Beautiful—if f is continuous. But most functions are not. The Dirichlet function (1 at rationals, 0 at irrationals) is discontinuous everywhere yet not Riemann-integrable on [0,1] (Riemann, 1854), but Lebesgue integrable with value 0 with integral 0. We just integrated a function, via Lebesgue, that never settles down. The FTC fails spectacularly for Thomae’s function, for the Cantor distribution, or for any pathological beast. Differentiation and integration are not true inverses: differentiate a smooth function, integrate it back, and you recover the original only up to a constant—if everything is nice. Change domains, add discontinuities, and the round-trip breaks. The FTC is a photocopier that sometimes eats the margins.
3. Riemann: A 19th-Century Kludge
Bernhard Riemann’s 1854 partition-and-squeeze method is elegant for teaching but comically restrictive (Riemann, 1854). It fails on the Dirichlet function (integrable, but not by Riemann’s own criterion for most partitions) and on any bounded function with “too many” discontinuities. Riemann integration is like measuring a coastline with a ruler: it works for smooth beaches but collapses on fractals. The integral exists only if upper and lower sums agree—a condition so fragile that most bounded functions on [0,1] are not Riemann-integrable. Yet we teach it first because it is drawable. Lebesgue had to invent measure theory just to rescue the concept.
4. Lebesgue: Throwing Out the Baby to Save the Bathwater
Henri Lebesgue inverted Riemann’s approach in 1902, partitioning the range rather than the domain and bolting on measure theory (Lebesgue, 1902). The payoff: vastly more integrable functions and powerful convergence theorems, dominated convergence, monotone convergence, Fatou’s lemma (Royden and Fitzpatrick, 2010). The cost: a sledgehammer of σ-algebras, measurable sets, and measure spaces that obscures geometric intuition. The Lebesgue integral of the Dirichlet function is 0, but no classical antiderivative exists. It still fails on non-measurable Vitali sets, courtesy of the Axiom of Choice. Lebesgue integration is less a theory of area than a statistical averaging trick. It tells you the “average height” of a function but says nothing about its path.
5. Henstock-Kurzweil: The Integral That Ate Everything—and Spat Out Usability
Independently discovered in the 1950s–60s, the HK (gauge) integral extends Riemann’s method with variable partition sizes (“gauges”: see Henstock, 1963; Kurzweil, 1957). It integrates every derivative—no exceptions. If F′(x) = f(x) pointwise, then ∫ₐᵇ f = F(b) − F(a). Sounds perfect. But it is non-absolute (∫|f| may not exist even if ∫f does), computationally useless, and ignored by physicists and engineers. HK is the ultimate skeptic’s integral: it works everywhere at the cost of being uniquely unhelpful—like a universal solvent that dissolves the beaker.
6. The Antiderivative Myth
Of the functions we can differentiate, only a tiny sliver of them have elementary antiderivatives:
• e^(−x²) → error function erf(x) • √(sin x) → elliptic integrals • (sin x)/ x → Si(x)
The Risch algorithm (1969) proves decidability for elementary functions but fails in practice beyond exam problems (Risch, 1969). Numerical quadrature (Simpson, Gauss, Monte Carlo) is the real workhorse because exact integration is a fantasy.
7. The Philosophical Punchline: Plurality Proves Indeterminacy
The coexistence of Riemann, Lebesgue, HK, Daniell (1918), Pettis, Bochner, and gauge integrals, is not a historical accident; it is conceptual underdetermination. No theory dominates across all desiderata:
| Desideratum | Riemann | Lebesgue | HK |
| Geometric intuition | ✓ | ✗ | ✗ |
| Generality (functions) | ✗ | ✓ | ✓✓ |
| Convergence theorems | ✗ | ✓✓ | ✗ |
| Computational tractability | ✓ | ✗ | ✗ |
| Elementary machinery | ✓ | ✗ | ✓ |
No meta-principle ranks these trade-offs. “Naturalness” is context-dependent: Riemann feels natural in calculus classrooms, Lebesgue in measure-theoretic probability, HK in differential-equation recovery problems. Category theory reveals structural relationships but cannot reduce the plurality to a single canonical integral (Fremlin, 1974). Axiomatization fails for the same reason—every axiom system underdetermines the concept (McShane, 1973; Berberian, 1965).
This is not mere technical diversity. It is family-resemblance pluralism à la Wittgenstein (Wittgenstein, 1953): integration theories share overlapping properties—linearity, positivity, some fundamental theorem—but no single essence. The concept of “integration” is a loose cluster, not a Platonic object with determinate extension.
8. Applied Math’s Quiet Rebellion
In physics and engineering, the theoretical circus is irrelevant. Numerical quadrature, Monte Carlo, and Fourier transforms converge under weak conditions and scale to high dimensions. These methods treat integrals as expected values or algebraic manipulations, not as metaphysical truths. The “theory” of integration is a historical artifact—like phlogiston.
9. The Skeptical Conclusion: Integration is a Useful Fiction
Integration works because:
- Physical systems are usually smooth enough.
- We engineer problems to fit polynomials and exponentials.
- When it fails, we redefine success (distributional derivatives, Sobolev spaces).
But push beyond the textbook, and the cracks show. The plurality of integration theories is not convergence on truth—it is divergence from a broken starting point. Riemann was a stopgap. Lebesgue was a patch. Henstock-Kurzweil was a desperate overreach.
True skepticism: the integral does not exist in the world. It is a model—and like all models, it is wrong, but some versions are useful. The real question is not: “Which integration theory is correct?,” but instead: “When does pretending the integral exists stop being worth the cognitive load?”
Calculus integration is like democracy: deeply flawed, occasionally dangerous, and the best we’ve got—until mathematics hands us a better abstraction.
REFERENCES
(Berberian, 1965). Berberian, S.K. Measure and Integration. London: Macmillan.
(Daniell, 1918). Daniell, P.J. “A General Form of Integral.” Annals of Mathematics 19, 4: 279–294.
(Fremlin, 1974). Fremlin, D.H. Topological Riesz Spaces and Measure Theory. Cambridge: Cambridge Univ. Press.
(Henstock, 1963). Henstock, R. Theory of Integration. London: Butterworths.
(Kurzweill, 1957). Kurzweil, J. “Generalized Ordinary Differential Equations and Continuous Dependence on a Parameter.” Czechoslovak Mathematical Journal 7, 82: 418–449.
(Lebesgue, 1902). Lebesgue, H. « Intégrale, longueur, aire. » Annali di Matematica Pura ed Applicata 7, 1: 231–359.
(McShane, 1973). McShane, E.J. “A Unified Theory of Integration.” American Mathematical Monthly 80, 4: 349–359.
(Riemann, 1854). Riemann, B. Über die Darstellbarkeit einer Function durch eine trigonometrische Reihe. Abhandlungen der Königlichen Gesellschaft der Wissenschaften zu Göttingen. 13: 87–132.
(Risch, 1969). Risch, R.H. “The Problem of Integration in Finite Terms.” Transactions of the American Mathematical Society 139: 167–189.
(Royden and Fitzpatrick, 2010). Royden, H.L. and Fitzpatrick, P.M. Real Analysis. 4th edn., Hoboken NJ: Prentice Hall.
(Saha, 2018). Saha, S. “Wonders of Integral Calculus.” Medium. 29 July. Available online at URL = <https://medium.com/cosmus-mathematicus/wonders-of-integral-calculus-d1ba9abf4050>.
(Wittgenstein, 1953). Wittgenstein, L. Philosophical Investigations. Trans. G.E.M. Anscombe. Oxford: Blackwell.
Against Professional Philosophy is a sub-project of the online mega-project Philosophy Without Borders, which is home-based on Patreon here.
Please consider becoming a patron!
