Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

8.1 Introduction

In her paper of 1990 entitled Epistémologie et didactique, Michèle Artigue reflected on 10 years of practice within the French mathematics educationFootnote 1 community, while stressing the need for epistemology for the working researcher. First, she underlined the need for epistemological awareness as an experience for the researcher, enabling distance between the researcher and their personal mathematical culture; second, she pointed out that some knowledge of the history of mathematics was of a key component of didactical research, either to understand the historical development of a mathematical concept, or to understand the shaping of mathematics as a ruled cultural activity.

This chapter will, to a large extent, directly echo the original Artigue paper, starting with a common take on “epistemology”. Of course, the word has many meanings: rather than the noun “epistemology”, which seems to denote a well defined research field, we will use the adjective “epistemological” to denote the endeavour of deriving insight from knowledge/awareness of the history of mathematics that is relevant from a ME research perspective.Footnote 2 Also, in spite of the fact that it discussed several well-known and important papers from the 1980s, the Artigue paper was not a literature review. Rather, the literaturereferenced by Artigue was discussed in detail from a methodological perspective, focusing on the use of concepts from MER and science education research such as “epistemological obstacle” and “conception”.

This chapter will also focus on issues of method, although with a shift of emphasis. Instead of focusing directly on didactical concepts, we will mainly discuss research practices at the intersection of two autonomous fields of knowledge: MER on the one hand, and the history of mathematics on the other; in this context, “history of mathematics” will denote the outcome of the work of historians of mathematic.

It seems to us that since Artigue’s 1990 paper, the interactions between the two research communities have become less intense; certainly not because the need for epistemological inquiry which Artigue clearly spelled out has faded, but probably from the conjunction of two factors. First, a growing professionalisation of the two research communities occurred. Second, the theory of didactical situations assigned a role to the history of mathematics within a didactical theoretical framework, making it easier for the ME researcher to engage with history. The multiplication of theoretical frameworks, and the decline of interest in concepts such as “epistemological obstacle” probably made it less clear, in particular for early-career ME researchers, why and how interactions with the historical community could benefit them.

We will first endeavour to describe the structural differences between the two fields of research: MER and the history of mathematics (HM). We will then use a methodological viewpoint to analyse several classical and more recent works in MER, in order to document the range of possible interactions and stress fruitful leads. Our range of examples will cover recent works in physics education research to the extent that they bring to light complementary connections between the histories of science and didactics.

8.2 Two Autonomous Fields of Research

8.2.1 Structural Differences

MER and HM are two different and autonomous disciplines: each has its own empirical field of investigation, its own set of legitimate questions, its own way of validating claims, its own reference works, etc. That fact may be self-evident; however, we feel it should be taken into account in order to pave the way for fruitful collaborations. It is also a fact that ME researchers and historians of mathematics have often been speaking at cross purposes: when historians of mathematics read what ME researchers say about the history of mathematics, the typical reaction goes: “this is not history, but a sketchy reconstruction of history framed within a-historical categories; what really happened is really much more complicated than that, you know …”; which ME researchers are usually fully willing to acknowledge while wondering why historians would deny them the right to make heuristic use of the HM, usually in a preliminary phase to their main investigation. For them, learning about history (which is one of the things historians do) is a means to learn something from history (which is not what historians do). Reciprocally, ME researchers may sometimes be surprised by the lack of theoretical frameworks in the work of historians, since such frameworks provide the main tools for describing and analysing specific issues, and enable researchers to integrate their particular study within a growing and soundly-structured body of knowledge about the learning of mathematics in educational contexts. Even though some historians occasionally borrow concepts from some theoretical frameworks,Footnote 3 they usually feel they have no use for theoretical frameworks from MER, because they don’t use theoretical frameworks at all!

The purpose of this chapter is not to claim that these common misunderstandings are only the result of the relative isolation of the two communities, and that they would soon fade if both groups of researchers decided to work together with an open mind. Quite the contrary, we think these misunderstandings point to differences which are structural, and our purpose is to sketch ways of living with this fact.

Since the intended audience of this chapter is that of ME researchers, we would like to briefly describe some elements of the work of historians. Of course, our approach is descriptive and not normative.

MER and HM have at least this in common: contrary to what research mathematicians do, the object of their investigation is not mathematics, and this object is not studied primarily mathematically. Footnote 4 Rather, both historians and ME researchers study how agents engage with mathematics, in a context which can be described; mathematics is necessary to make sense of this engagement and this context, but cannot possibly be the only background tool.

Beyond this common agent-based approach, dissimilarities become striking: ME researchers study learners, while historians of mathematics tend to focus on experts.Footnote 5 ME researchers have direct access to the living agents they study, which means empirical data can be gathered, hypotheses can be put to the test in finely-tuned conditions, and cognitive processes can be investigated. Historians of mathematics have indirect access to the agents they study, and it is part of their field to attempt to assess what biases exist (for example, critique of sources, and careful methodological reflection on corpus delineation). Historians of mathematics have to deal with events that happened once, but that can be understood, compared, and to some extent, fit into narrativesFootnote 6; MER has an experimental side to it, and can aim for invariants and reproducibility.Footnote 7

The fact that historians depend heavily on the availability of sources and do not explicitly rely on theoretical frameworks does not imply that their work is purely descriptive and erudite. To use Kuhn’s phrase, historians solve puzzles, just as any researcher does, whatever their field. We would like to illustrate this agent-based, puzzle-solving approach from three different angles.

8.2.2 Echoing Questionnaires

First, let us mention the kind of questions that historians aim to tackle. A very general and context-free list of questions can be found, for instance, in Catherine Goldstein’s (1999, 187–188) methodological paperFootnote 8:

At a given period in time, what were the networks, the social groups, the institutions, the organizations where people practiced mathematics or engaged with mathematics? Who were mathematicians? In what conditions did they live; in what conditions did they carry out mathematical work? How were they educated and trained? What did they learn?

Why did they work in mathematics, in what preferred domain? What did this domain mean to them? (…) Where did mathematicians find problems to be solved? What were the form and origins of these problems? Why was some result considered as very important, or of lesser importance? According to which criteria? What was considered to be a solution to a problem? What had to be proven, and what did not require a proof (tacitly or explicitly)? Who decided so? When was a proof accepted or rejected? When was an explicit construction deemed indispensable, optional or altogether irrelevant?

When, where and how mathematics were written? Who wrote, and for whom? For instance, were new results taught, were they printed, were they applied? What got transmitted? To whom was it transmitted, in which material and intellectual conditions?

What changed and what remained fixed (and according to what scale, to which criteria)?

This list strikingly echoes the list of questions which Guy Brousseau considered to be meaningful for MER when he attempted to derive didactically-relevant insight from a study of the history of mathematics. When discussing Georges Glaeser’s paper of 1981 on the epistemological obstacles relative to negative numbers, Brousseau summed up Glaeser’s approach, and then pointed to what he would consider to be the more relevant questions:

This formulation shows what failed Diophantus or Stevin, seen from our time and our current system. We thus spot some knowledge or possibility which failed 16th century authors and prevented them from giving the “right” solution or the proper formulation. But this formulationFootnote 9 hides the necessity to understand by what means people tackled the problems which would have required the handling of isolated negative quantities. Were such problems investigated? How were they solved? (…) What we now see as a difficulty, how was it considered at the time? Why did this “state of knowledge” seem adequate; relative to what set of questions was it reasonably efficient? What were the advantages of this “refusal” to handle isolated negative quantities, or what drawbacks did it help avoid? Was this state stable? Why were the attempts at changing it doomed to fail, at that time? Maybe until some new conditions emerge and, some “side” work be done, but which one? These questions are necessary for an in-depth understanding of the construction of knowledge [pour entrer dans l’intimité de la construction de la connaissance] (…).Footnote 10

In both lists, we can see that a focus on agency does not mean that the object of study is a freely creative cognitive agent. Quite the contrary: agents are born in a world which preexists and constrainsFootnote 11 their actions. When it comes to mathematical activity, constraints come from a great variety of sources, ranging from the material environment (a Chinese abacus is not an electronic calculator) to epistemic values (such as rigour, generality, simplicity, accuracy, applicability) and epistemic categories (such as definition, justification, proof, example, algorithm, analysis/synthesis, principle). The historical contingency of these constraints does not imply that they have no a-historical components, be they mathematical properties (a rule such as “minus times minus equals minus” is not compatible with distributivity of × over +) or semiotic properties (an algebraic shorthand with no parentheses—such as Cardano’s—has different properties from Bombelli’s). Making historical sense of how actors engage with mathematics involves understanding how they act within a given set of constraints, what meaning they give to their actions, and in what respect these actions alter the system of constraints.

8.2.3 An Example

Let us now flesh out this notion of agent-based approach—this focus on mathematical agency—from another angle. We will use the diagram below (see Fig. 8.1) to illustrate several methodological points.

Fig. 8.1
figure 1

First diagram from Descartes’ Geometry

The very same diagram (Fig. 8.1) appears in two of the most influential works in the history of mathematics: Euclid’s Elements and Descartes’ La Géométrie. One could argue that not only the diagram is the same, but also the mathematical content is the same; however, the parts these diagrams play in both works are strikingly different.

In Euclid’s Elements Footnote 12 (ca. 300 BCE), this diagram comes with proposition 14 of book II, a proposition which solves the following construction problem: to construct a square equal (in area) to a given rectangle. If the sides of the rectangle are equal (in length) to FG and GH, then the perpendicular IG is the side of the sought-after square, which Euclid proves using proposition 47 of book I (which we call Pythagoras’ theoremFootnote 13). At the end of book I, a series of propositions established that, for any given polygon, a rectangle with the same area could be constructed (with ruler and compass only), hence proposition 14 provides the final positive solution to the problem of quadrature of polygons (i.e., to transform any polygonal area into a square). In turn, this fact implies that—at least for polygons—area is a well behaved magnitude: areas can be compared (since square areas can) and added (since the Pythagorean construction provides a means to add square areas). On this basis, a modern reader would conclude that a theory of measure is possible for polygonal areas; the modern reader also knows that this requires the set of real numbers. Euclid was well aware of the fact that the theory of well-behaved geometrical magnitude (even line-segments, for which comparison and addition are straightforward) requires more than natural numbers and their ratios. The solution he presented in book V is a number-free solution, based on the notion of ratio of magnitudes and not of measure. The positive result of II.14 also points to open questions in the theory of magnitudes, in particular the extension of the theory beyond the case of polygons (the case of the circle being of prime importance).

The same diagram (Fig. 8.1) appears on the second page of Descartes’ La Geométrie (1637Footnote 14). Along with another diagram (Fig. 8.2), he aims to define operations on segmentsFootnote 15; operations for which he would use the same names as for arithmetical operations. In Fig. 8.2, if AB denotes a unit segment, then BE will be called the “product” of segments BC and BD.

Fig. 8.2
figure 2

Second diagram from Descartes’ Geometry

In Fig. 8.1, if FG is the unit segment, then IG will be called the “square root” of GH. Descartes then adds that he would not only use the same names as those of arithmetical operations, but that he would also resort to the same signs as in algebra: letters for segments (known or unknown), and symbols such as × and √ for the above mentioned constructions. The project was to use the means of algebra (rewriting rules, elimination in simultaneous equations, identification in polynomial equalities, method of indeterminate coefficients) to capture and analyse geometrical relations between segments; among such relations, those expressed by one equation in two unknowns capture plane curves.Footnote 16

This specific Cartesian project is quite different both from Euclid’s, and from what we call either algebra or coordinate geometry. In the Elements, proposition II.14 solved an area problem; in terms of magnitudes, considering two line-segments could lead either to a new segment (by concatenation, which can be seen as a form of addition), or to an area (that of a rectangle, which can be seen as a form of multiplication), or to a ratio (which is not a geometrical entity, but not a number either). On the contrary, Descartes uses elementary constructions (with an ad hoc unit segment) to define operations such as “times”, “divide” or “root” as internal operations within the domain of segments; this enables him to make free use of algebraic symbolism while warranting geometrical interpretability.Footnote 17 This system, however, involves no global coordinate system; it does not even involve coordinates, if by coordinates we mean (real) numbers, since no such numbers play any part in the system.Footnote 18

The fact that Descartes’ system is an algebra of segments has other far-reaching consequences. Let us mention one of general epistemological importance. At first, when we read in La Géometrie (Descartes 1954, p. 303) that the solution of equation \( {z^{ 2} = az + b^{2} } \) (z being unknown, a and b known) can be expressed by

$$ {z = \frac{ 1}{ 2}a + \sqrt {\frac{ 1}{ 4}aa + bb} }, $$

we feel we are on familiar ground. However, we need to recall that this formula is not a symbolic summary for a list of arithmetical operation on numbers, but is a symbolic summary for a geometrical construction program; a ruler-and-compass construction program which involves two concatenations, two multiplications (Fig. 8.2), the construction of a “square-root” segment (Fig. 8.1), and three midpoints. This, in turn, means that the algebraic manipulation of formulae and equations deals with the transformation and comparison of geometric construction programs. Here, the comparison with Al-Khwarizmi (ca 820 CE) is striking:

Roots and numbers equal to squares; for instance, if you say: three roots and four in numbers are equal to one square.

Procedure: halve the number of roots, you get one and one half; multiply it by itself, you get two and one quarter; add four, you get six and one quarter; work out the root, which is two and one half; add half the number of roots – that is one and one half – you get four, which is the root of the square; and the square is sixteen. (Rashed 2007, p. 106). (Trans. RC).

With its purely rhetorical algebra and its use of generic examples, Al-Khwarizmi’s text may look less familiar than Descartes’ formula. However, it presents a bona fide list of operations which enables one to solve an equation, in a numerical context. In the rhetorical context, algorithms are easy to express, but not so easy to compare, transform and calculate upon.Footnote 19 One of the properties of Descartes’ system is that its symbolic algebra allows for calculation to operate on algorithms; the fact that the basic steps of the algorithms involved are ruler and compass constructions and not numerical operations is irrelevant, and testifies to the meta-level function of symbolic algebra.

This interplay between the familiar and the not-so-familiar (yet understandable) may feel disorienting at first, but this disorienting effect is positive, as Artigue stressed. It has a critical function, helping the researcher to distance him or herself from his/her own mathematical cultureFootnote 20; and a heuristic function, suggesting new viewpoints on seemingly familiar notions, for instance, the role of symbolism in algebra, or the role of real numbers in geometry (as measures and as coordinates). At least two other functions can be mentioned. First, it helps identify problems to which there are no straightforward answers, for instance, what should we consider to be the geometrical analogue of numerical multiplication, at least for one-dimensional objects? In particular, should the analogue of the product be one-dimensional or two-dimensional? A long series of different–yet mathematically sound—constructions provides different answers to this question, including dimension-changing solutions (going down with the dot product, or up with the exterior product). Secondly, it helps question the notion of identity. It could be argued that, from a purely mathematical point of view, Euclid and Descartes rely on the same content associated with Fig. 8.1; this probably makes sense, but is not necessarily very helpful, either to the historian or to the ME researcher. Indeed, researchers in both fields aim to analyse how content depends on, for instance, semiotic resources, or intended use.

To conclude this example from Euclid-Descartes, we would like to explain why we chose such an example. On the one hand, the example is relatively small scale; we did not need to include it in any large-scale narrative on the “stages” in the history of geometry for this sketchy comparison to serve the four functions listed above of epistemological inquiry. On an even smaller scale, the comparison with a short passage of Al-Khwarizmi could play a relevant part even with no background “big picture” on the history of algebra, or even on the Kitāb al-Jabr wa-al-muqābala. On the other hand, to compare the uses of the same diagram required that its role in the whole structure of the works (the Elements and La Géométrie) be analysed. It requires some knowledge of history to make sense of highly sophisticated but largely forgotten theoretical constructs such as the classical theory of ratios or the 17th century research program of construction of equations. This knowledge cannot derive from a quick look at short extracts from the original sources, and probably not even from a lengthy examination of the complete books; here, we depend on secondary sources and the work of professional historians such as Bernard Vitrac (Euclide 19902001) for Euclid and Bos (2001) for Descartes.

8.2.4 Solving Puzzles, Crafting Puzzles

We would now like to illustrate the puzzle-solving side of research in HM, by giving short descriptions of a selection of recent works in that field. We also wish to illustrate, with six examples, the difference between a naïve question and a research question. Indeed, it is natural to begin with a naïve question, and curiosity is the first driver in all fields of inquiry; however, delineating a specific question, on the basis of available documents and of the state of research (historiography) is a key stage when practicing research-level historical investigation. Since this sample is not a survey, we can also proceed chronologically.

Example 1

Contrasting rational reconstructions: In Proust (2012), the author studies the algorithm displayed in paleo-Babylonian tablets when working out the reciprocals of large numbers in the sexagesimal system. The clay tablets display instances of calculations, but no general descriptions of the method (much less any justifications), which is why historians endeavour to come up with reconstructions of the algorithm. A pioneer in the history of Babylonian mathematics, Otto Neugebauer (1899–1990) reconstructed an algorithm on the basis of a few tablets- an algorithm which required that additions be used along the way. However, in the floating point sexagesimal number system, and in the purely numerical context of these tablets, addition is not possible (whereas products and reciprocals make perfect sense). On the basis of a much larger sample of tablets, Proust reconstructed a different algorithm, one which is fully compatible with a floating point arithmetic.

Example 2

Re-problematising familiar practices: In his now classic work, The Shaping of Deduction in Greek Mathematics, Netz (1999) attempted to re-historicise the endeavours of the Greek mathematicians of the classical and Hellenistic periods, with the aim of helping us to question many things we take for granted. This is difficult for at least two reasons: first, some of the basic elements of practice displayed in these texts—in particular, the practice of discussing lettered diagrams using only explicit axioms and formerly established results—is so familiar to us that we cannot imagine how a few men strove to establish this specific cultural form based on the background of other cultural activities; and second, because we feel we know that mathematics was a central intellectual activity in these periods, as many texts from Plato and Aristotle seem to indicate. Netz established that this was not the case, and discussed why Plato and Aristotle distorted our perception of historical realities. In a stimulating review of this erudite book, Latour (2008) emphasised the extent to which it echoed central methodological trends in the social history of science.

Example 3

Describing a reception as a form of hybridation: The question of the circulation of mathematics between different cultural areas—and not only different periods—is also a central field of investigation. For instance, in a paper published in 1996, Karine Chemla discussed the introduction of “western” mathematics in 17th century China by Jesuit missionaries. It was usually thought that, in this period, the indigenous Chinese tradition of mathematics was to a large extent forgotten in China, and that western mathematics had been adopted passively. Actually, studying the Chinese sources leads to a more nuanced picture. In particular, when Jesuit Matteo Ricci and Chinese scholar Li Zhizao collaborated to write a treatise of arithmetic based on Clavius’ Epitome arithmeticae, they ended up with much more than a translation: Li added many elements from the indigenous tradition, in particular the fangcheng algorithm to solve simultaneous linear equations.Footnote 21 This work of synthesis did not stir interest in the West; in China, however, the introduction of western mathematics revived scholarly interest in classical Chinese mathematics and triggered comparative studies of both traditions.

Example 4

Renewing our understanding of a crucial step in the development of mathematics: Our image of the beginning of calculus was significantly altered in the 1990s by the publication of a hitherto little-studied Leibnizian manuscript (Knobloch 1993). Until then, our understanding of the calculus according to Leibniz was based on miscellaneous short texts (tracts, letters), in which few attempts at justification were given. It was generally thought that even though Leibniz claimed that his calculus could be justified by the rigorous methods of the Ancients, he actually relied on infinitesimals (which, he admitted, were only “useful fictions”). With the De quadratura arithmetica, which was written before the invention of his calculus, Leibniz wrote a long treatise, in a deductive style which both emulated and improved the exhaustion proof-scheme in a way that, were it to be reformulated in a symbolic and numerical context, would be closer to the current ε-δ proofs than those of the Ancients. Moreover, the term “fiction” was already used in this context, to denote abbreviations for calculations dealing only with finite quantities.

Example 5

Unravelling a forgotten branch of mathematical analysis: It is well-known that for the founders of calculus, the prime goal was the study of curves defined by ordinary differential equations, in a geometrical or physical context. Pen and paper, and formulaic solutions were not the whole story, as was demonstrated by the deep and original work of Dominique Tournès (following the work of Henk Bos). Tournès (2003) studied the intense work on graphical methods and graphing devices carried out from the very beginning (for example, Leibniz, Newton, Jean Bernoulli, Euler), up until the advent of digital instruments in the second half of the 20th century. This work brings to light a great wealth of largely forgotten mathematical ideas and techniques, shows the continuity between the algebraic research program on the “construction of equations” (as in Descartes) and the late-17th and 18th century researches on ODEs, and documents the deep connections between the most theoretical considerations on the one hand, and the demand for approximation methods (be they graphical, mechanical or numerical) in the engineering communities on the other hand. Since 2003, Tournès furthered his work on integration instruments,Footnote 22 and headed a collective research program on the mathematical and professional contexts of numerical calculation.Footnote 23

Example 6

Studying the emergence of a meta level articulation: In the didactics of analysis, it is customary to distinguish between point-wise, local and global properties of functions. The distinction between “local” and “global” is now widely accepted in the scholarly mathematical world, but it was not always the case. Studying the emergence of an explicit local-global articulation is tricky for a number of reasons. It concerns more or less all mathematicsFootnote 24; the meta level terms “local” and “global” have definitions which differ in every specific mathematical context—actually they can be used with no definitions at all. Moreover, the question of the explicit is crucial. When, at the turn of the 20th century, some mathematicians began to explicitly express such a distinction, was the general context one in which it was clear to everyone that this mattered (though it went without saying), or one in which no clear distinction was made between local and global statements, resulting in a wealth of faulty proofs and ambiguously-worded theorems? These questions were addressed in Chorlay (2011), who provided answers based on a combination of quantitative and qualitative methods.

This short list of examples illustrates how historians endeavour to design non-trivial questions, the means they use to answer these questions, and the kind of answers they tend to consider relevant and innovative. Although historians provide a great wealth of material that is of prime interest in MER, they do not usually provide this material in a form which directly meets the needs of the MER community.

8.3 Methodological Discussion of Some Classical Works

In the second part of this chapter, we would like to review several papers from the didactics of mathematics and the sciences which explicitly carried out epistemological investigations in the sense of Artigue. Our primary goal is methodological: we will present only a brief selection of papers in order to discuss concepts such as obstacles, aspects of a concept, stages and didactical reconstruction.

8.3.1 Epistemological Obstacles

In a paper published in 1990, Artigue discussed in detail two studies which aimed to investigate epistemological obstacles in the sense of BrousseauFootnote 25: Glaeser’s (1981) paper on negative numbers, and Sierpinska’s (1985) paper on limits. We will focus on the second paper, so as to summarise and further Artigue’s methodological analysis.

8.3.1.1 One Example in Analysis

Sierpinska’s paper is divided into two very different parts. The first part describes and analyses two classroom experiments, which included high school students who had received no prior teaching on limits or the derivative. In both situations, the students had to determine the tangent to a given curve at a given point, on the basis of a loose, intuitive and non-verbal description of the sought-after object. As Sierpinska points out, in contemporary mathematics, the notion of limit is the relevant tool for performing this task. Hence, by focusing on what students do and say (when trying to describe and justify what they do), she aims to capture what their conception of a limit is; to discover to what extent it differs from the current formal definition; and to reveal what obstacles stand in the students’ way when trying to pass from naïve and context-dependent conceptions to efficient general procedures and proto-definitions (définitions opératoires).

In the second part of the paper, Sierpinska turns to the history of mathematics to establish the epistemological natures of the obstacles that were identified:

If a given behaviour manifests itself in history just as it does with today’s students, then we are justified in regarding it as a specific feature of the development of the given concept, as opposed to an effect of teaching conditions. (Sierpinska 1985 8), (Trans. RC)

On the basis of two surveys on the history of mathematics, Sierpinska (1985, p. 38) lists and classifies “obstacles” in the history of mathematics, as the Fig. 8.3 shows.

Fig. 8.3
figure 3

Sierpinska’s classification of epistemological obstacles regarding limits

Artigue summarises Sierpinska’s list of obstacles (adapted from Artigue 1990, p. 253):

  • Horror infinity Footnote 26 brings together the obstacles which stem from the refusal to consider passing to the limit as an operation; those stemming from an automatic use of algebraic methods designed for the handling of finite quantities to the case of infinite quantities; those consisting in transferring all properties of the terms of a convergent sequence to its limitFootnote 27; and eventually those obstacles which consist of regarding the passage to the limit as a physical movement, or a way of coming ever closer.

  • Obstacles related to the notion of function include failure to identify the underlying functions; restriction to a sequence; monotonous reductionFootnote 28; failure to distinguish between limits and upper/lower bounds.

  • Geometric obstacles: geometric intuition generates serious obstacles which stand in the way of the formulation of a rigorous definition, both by preventing the determination of what is to be taken to be the difference/distance between two geometric magnitudes, and by conflating the notion of limit with that of point in the closure of a subset.

  • Logical obstacles reflect the failure to use quantification, or to take into account the order of quantifiers.

  • The obstacle of the symbol reflects the reluctance to coin a specific symbol (such as \( \mathop { \lim }\limits_{ \ldots } \ldots \)) for the operation “passing to the limit”.

Before we list what we take to be structural methodological shortcomings in this paper, it is only fair to underline several of its most positive features. The paper has a heuristic value, and indeed, several of the problems it points out have become central in the didactics of analysisFootnote 29: such as the inadequacy of everyday language to capture the fundamental concepts of analysis; complicated and deceiving connections between limit-positions of geometric objects, limits of associated magnitudes, and limits of associated numerical functions; sequential reduction and monotonous reduction; implicit transfer of calculation rules and of properties from the finite to the infinite, or from a converging sequence to its limits; and the need, at some point in the curriculum, to regard a numerical sequence or a numerical function as one object and not only as a collection of numbers. Also, as is often noted, the didactics (and the teaching) of analysis is particularly thorny because of the entanglement of all the basic concepts (limit, derivative, integral, function, the real continuum); hence, the difficulty in designing experimental situations which are rich and open-ended enough to yield interesting data, yet not so rich and open-ended that all possible problems crop up and interfere. Finally, we are not aware of any well-established theoretical framework (or even reasonably sea-worthy set of conceptual tools) adapted to the description and analysis of epistemological obstacles. In this respect, Sierpinska’s paper both points to a problem of general significance for MER, and offers a stimulating proposal.

8.3.1.2 Focusing on “Obstacles” in MER

As Artigue pointed out, the notion of “epistemological obstacle” may not be the best conceptual tool to describe the wealth of phenomena captured here, both in the classroom experiments and in the historical literature. Here, we first focus on the purely didactical aspects.

For Brousseau, an obstacle (be it didactic or epistemologicalFootnote 30) is an element of knowledge (either explicit or “in act”) which proves valid and efficient in some contexts, but becomes systematically error-generating in other contexts. A standard example is that of the comparison of decimal numbers: some rules which are valid and efficient for natural numbers (for instance, the more digits it takes to write a number, the larger the number) become invalid and error-generating when applied with decimals (it takes three digits to write 1,23 and only two to write 1,3; but 1,23 < 1,3). An obstacle is epistemological if it depends on mathematical facts only, regardless of teaching paths. When confronted with an error-generating in-act-theorem, it is not always easy to tell whether it is epistemological or didactic; Brousseau suggested that a distinctive feature of epistemological obstacles was their presence in mathematics of the past.

In the classroom experiments, students face several challenges: changing registers or frames (from the graphical to the numerical); experiencing the inadequacy of the expressive resources available to them (everyday language, body motion, basic symbolic algebra); hitting upon a paradoxical core (division by zero yielding a finite result, straight line defined by two points which overlap); and regarding a process/procedure as an object/concept (the notion of procept would later be designed to describe this phenomenon). The fact that, when confronted with these challenges, students do not come anywhere close to an ε-δ definition of the derivative as limit of the differential quotient probably does not qualify as an error.

A difference between a statement and a reference statement (say, a given formal definition) may be telling without pointing to any difficulty, misconception or error. Some differences may indicate difficulties, but not all difficulties are errors. All errors do not derive from epistemological obstacles. The epistemological character of some obstacles is not always easy to ascertain; and, more often than not, the history of mathematics may not be useful or even necessary to ascertain it.

To substantiate the last statement, let us mention the case of the “natural numbers/decimals” obstacle, which is clearly (partly) epistemological but with no significant historical basis. The same holds for a well-documented error-generating belief, namely, all the magnitudes associated with a given class of objects vary in the same way: for instance, a polygon has a length and an area; but it is not true that when the length increases, the area always increases as well. This error-generating belief is probably epistemological insofar as reveals some aspect of the acts of knowing and forecasting, independently of teaching paths and curricula. As Artigue (1990, p. 261) stressed, it is reminiscent of Bachelard’s original notion of epistemological obstacle in physics, and points to a larger class of obstacle-generating features of the understanding process. As such, instances of errors ascribable to this infelicitous thought-process may very well be found in historical texts. If some were, they would probably not be indicative of any global dynamic of the development of mathematical thought; and if none were found in history, it would not make this error-generating belief any less important in mathematics education.

8.3.1.3 Do “Obstacles” Help Us to Make Sense of History?

The notion of “obstacle” is probably not the best conceptual tool to make sense of the historical data collected either. Here we spell out several arguments, and suggest alternative investigative paths.

First of all, this notion depends on that of error. Writing, for instance, that Archimedes failed to consider “passing to the limit” as an operation, and avoided the use of the actual infinite which (Sierpinska claims) is a core element of the Weierstrassian ε-δ definition of a limit, seems slightly disrespectful and highly questionable (respectively). Archimedes’ proofs by exhaustion may be written in a style differing from the current standards in the first year of tertiary education; it does not make them mathematically incorrect, and does not testify to any cognitive shortcoming. Whether or not passing to the limit is an “operation” (and has to be seen this way) seems debatable; so is the assertion to the effect that the ε-δ definition of limits depends on the actual infinite. As to the last point, one could argue that the ε-δ definition is actually very close to the Archimedian proof-scheme. For instance, to establish that two areas are equal, Archimedes proves that this difference is less than any given area. To establish this, he relies on a fact which is made explicit in Euclid’s Elements, and warrants convergence to zero for a class of sequences of magnitudes.Footnote 31 As to Weierstrass, many of his contemporaries regarded his construction of the real numbers and his definition of the limit as a major step toward the reduction of analysis to arithmetic, and the subsequent elimination of infinites. When comparing proofs by exhaustion with the current Weierstrassian definition, their fundamental similarity raises a series of potentially fruitful questions: what is the difference between a uniform, formalFootnote 32 proof-scheme, and a definition? What is the role of symbolic notations (the question is all the more tricky since Weierstrass worked in a partly rhetorical context when discussing limits)? Why was the Archimedian proof-scheme criticized in the 17th century, and how are these criticisms related to the emergence of calculus?

As we mentioned earlier, all conceptual difficulties are not errors. The case of negative numbers could also illustrate this point: from the middle of the 17th century to the middle of the 19th century, some mathematicians, mathematics educators and philosophers debated the meaning and legitimacy of isolated negative numbers, and of the multiplication rule (Schubring 2005). This phenomenon was quite independent from the fact that negative coefficients and the multiplication of negatives had been used for centuries.Footnote 33 The same holds for imaginary numbers or calculus: at some points, their meaning and legitimacy were discussed in spite of the consensus on how to use them. Two fruitful questions would be: what were the arguments? Why were these issues controversial in some contexts and not in others?Footnote 34

Gathering historical data to document differences between various statements and a reference definition can also bring to light the multiplicity of aspectsFootnote 35 of a given concept. For instance, in her paper, Artigue listed a number of aspects of the notion of a tangent to a curve: such as straight-line such that no other straight-line can be inserted between it and the curve (local convexity); straight-line defined by two infinitely close points; and straight-line defined by the direction of the velocity vector of any point gliding along the curve. These aspects are not mathematically equivalent and from an epistemological viewpoint their ecology is different; they target different classes of curves and of problems and the associated signifiers (be they graphical, symbolic or rhetoric) are different. Similarly, historical investigation (based on texts) and didactical investigations (based on live empirical data) could go hand in hand; which does not mean they would be carried out along the same line or with the same expected outcome. Firstly, because the didactical study attempts to capture the total and personal cognitive structure of students associated with a mathematical concept (such as Tall and Vinner’s concept image); while historians study context-embedded rational actors, not cognitive subjects. Secondly, historical texts usually display the mature and genre-dependent productions of mathematical experts. It does not mean that all they say is correct, but it means they usually make meta-level choices which reflect both a large overview of mathematics, and the intended middle-scale structure of the text (such as letter, research paper, treatise, textbook). For instance, it is true that reading the first lesson of Lagrange’s Théorie des fonctions analytiques (first published in 1797) shows a specific definition of the derivative: if f(x) denotes of function of variable x, then substituting x + i for x (i being an indeterminate quantity) gives rise to a development of the form

$$ {f\left( {x + i} \right) = f\left( x \right) + ip + i^{2} q + \cdots } $$

where p is a function of x only; function p will be called the derivative of f, denoted by \( {f^{\prime} \left( x \right)} \) (we are paraphrasing) (Lagrange 1867, p. 21). However, the complete title of the book shows that this is a choice: Théorie des fonctions analytiques, contenant les principes du calcul différentiel, dégagés de toutes considérations d’infiniment petits, d’évanouissants, de limites Footnote 36 et de fluxions, et réduits à l’analyse algébrique des quantités finies. In the introduction, Lagrange (1867, trans. RC, p. 16) even discussed the definition of the derivative as the limit of the differential quotient: “One has to acknowledge that this idea, in spite of being right in itself, is not clear enough to serve as principle for a science whose certainty must be founded on evidence, and most of all, not clear enough to be presented to beginners” . To make sense of this statement would require that the meaning of “limit”, “algebraic analysis”, “principle” and “evidence” be investigated. Since his definition depends on a theorem on the existence of power-series expansions for functions, Lagrange established in a later chapter that this expansion holds “generally”; to make sense of the latter statement would require that the meaning of the terms “function” and “general” be investigated. Finally, Lagrange also engages in proofs in which the limit-definition of the derivative is used in a ε-δ fashion (Chorlay 2013), demonstrating his ability to switch between aspects when necessary. The fact that expert mathematical thinkers (not necessarily mathematicians and not necessarily of the past) exert meta-level control on a variety of representations is a well-identified key to advanced mathematical thinking, as Artigue stressed (Artigue et al. 2007).

Finally, gathering historical documents is only one of the ways to document the multiplicity of aspects making up a mathematical concept. In particular, studying the avatars of the concept in various fields of contemporary mathematics can, to some extent, serve a similar purpose. For instance, as far as the derivative is concerned, differential geometry and algebraic geometry offer new vistas.Footnote 37 The standard differential geometric approach provides an extension of calculus in which the primitive notion ramifies into several distinct notions (differentials, Lie derivative, covariant derivative with respect to a connection), with different invariance properties, each of them involving new spaces associated with the original domain. From a dynamic and epistemological viewpoint, this testifies to a process of conceptual differentiation rather than to a process of reduction to a unique correct definition on the basis of formerly loose and partly faulty proto-concepts. The algebraic geometric approach is further from the standard calculus approach, since it relies neither on real numbers nor on notions of limit, velocity or approximation. Extending the original multiplicity-of-intersection approach to the derivative, the scheme version of algebraic geometry provides rigorous notions of double point or infinitesimal thickening of a point (or even a subvariety).

On the one hand, this multiplicity of viewpoints/aspects within mathematics (be it contemporary or historical) for a given concept provides food for thought; on the other hand, it raises methodological questions. As to the food-for-thought part, let us mention two fields of investigation: first, the epistemological analysis of the aspects (mathematical properties, ergonomic and semiotic properties) and of their connections to other aspects of the same notion. Second, the study of the connections between scholarly knowledge and school knowledge (as documented in syllabi); the history of education, and the theory of didactic transposition provide tools for the latter. As to the methodological issues, we will mention only one. For a working mathematician endeavouring to give mathematical answers to mathematical questions, a given theoretical context provides a stable and unquestioned framework including definitions, fundamental theorems and standard proof techniques. For the researcher in mathematics education as well as for the historian of mathematics, the very notion of a background reference mathematical theory is problematic in a number of ways: which reference mathematical theory should be chosen for a specific investigation? What is the role of the reference theory in the didactical or historical investigation? Is a reference theory needed at all?

8.3.2 Large Scale Narratives: Episodes or Stages

We now proceed with our methodological discussion by presenting several papers written by researchers in mathematics education and mathematics educators with a sustained interest in the history of mathematics. This will enable us to discuss two issues: the relevance of large scale narratives dealing with “stages” or “genesis”; and the heuristic use of history in didactics.

In 2007, two papers published in the same issue of Educational Studies in Mathematics suggested different ways of using a historical perspective to question the teaching of algebra: Katz’s (2007) Stages in the history of algebra with implications for teaching; and Syntax and meaning as sensuous, visual, historical forms of algebraic thinking, by Puig and Radford (2007).

The abstract of the Katz paper reads:

In this article, we take a rapid journey through the history of algebra, noting the important developments and reflecting on the importance of this history in the teaching of algebra in secondary school or university. Frequently, algebra is considered to have three stages in its historical development: the rhetorical stage, the syncopated stage, and the symbolic stage. But besides these three stages of expressing algebraic ideas, there are four more conceptual stages which have happened along side of these changes in expressions. These stages are the geometric stage, where most of the concepts of algebra are geometric ones; the static equation-solving stage, where the goal is to find numbers satisfying certain relationships; the dynamic function stage, where motion seems to be an underlying idea, and finally, the abstract stage, where mathematical structure plays the central role. The stages of algebra are, of course not entirely disjoint from one another; there is always some overlap. We discuss here high points of the development of these stages and reflect on the use of these historical stages in the teaching of algebra. (Katz 2007, p. 185)

The distinction between a study of the semiotic aspects and the “conceptual” aspects provides depth to the analysis by spanning a two-dimensional grid, and suggests further investigations into the connections between both aspects. However, whether or not Katz identifies “stages” in the history of “algebra” is questionable; actually, the core of the paper is less schematic, displaying Katz’s historical culture and sense of nuance. For instance, in regard to the “stage” aspect, Katz mentions Al-Tusi’s (d. 1274) study of the number of solutions for a class of third-degree equations in terms of the maximal value reached by what we would call the left-hand side of the equation; this exemplifies a typically “functional” way of thinking in an algebraic context, long before the 17th century. As to “algebra”: whether or not the algorithmic solving of riddles bearing on width and area in Babylonian clay tablets, or establishing a list of basic geometric identities as in book II of the Elements, exemplify “algebraic” practices is highly questionable, and has been (and sometimes still is) hotly discussed in the historical community. Katz is well aware of this fact and points out in the introduction of the paper that the term “algebra” is not well-defined. Babylonian methods and Greek “geometric algebra” may not be stages in the history of “algebra”; however it is necessary to study them since they provide the necessary context to understanding something which clearly belongs to the history of algebra, namely Al-Khwarizmi’s Kitab al-jabr. Comparison is the key methodological tool, and what is specific in Al-Khwarizmi can only be spelled out by comparing it to other mathematical practices; deciding whether or not these practices are algebraic is a somewhat byzantine question.

More generally, setting out to identify “the stages” in the “development /historical genesis” of a theory or concept is a methodologically risky endeavour, even for careful thinkers who warn their readers against dubious analogies between ontogeny and phylogeny, and abstain from drawing direct implications for teaching. We’ve already mentioned two reasons: the use of a-historical categories (such as “algebra”) to select and organise historical content on a very large scale raises intrinsic methodological difficulties, and leads to simplifications which leave out all that does not fit within the frame; but then, what are we to learn from a history which is not the actual history? This delineation of homogeneous blocks also wipes out the role of agents. For instance, in La Géométrie, Descartes presents a method to determine the tangent to algebraic curves; however, his correspondence shows that he also relies on ingenious kinematic arguments when studying non-algebraic curves such as the cycloid.

There are two further difficulties. First, thinking in terms of “stages of development” has a definite positivistic flavour: in this setting, the next stage replaces the previous one; the fact that the next one comes after the former one is implicitly taken to mean that it is better in some way, and this improvement is taken to be an adequate explanation or cause for the historical change. Instead of using loaded terms such as “stage”, lighter terms could probably be used, such as “aspects”, “viewpoints” or “conceptual polarities”. What Katz convincingly points out is that algebraic equations can be (and have been) considered from a static-numerical viewpoint, or from a more dynamic-functional viewpoint. Viewpoints can coexist, and their relative virtues may depend on circumstances. Moreover, if one becomes obsolete at some point in time, its obsolescence can be studied as a historical phenomenon: for whom does it become obsolete? Was the drive for change conceptual, semiotic, instrumental or institutional? Was the change actively promoted by some, or did things just fade away and die out? We do not mean to say that loosely defined terms such as “aspect” or “polarities” are the only descriptive tools which didactical analysis should use when attempting to derive relevant information from the history of mathematics. Rather, we argue that starting the investigation with a priori conceptual tools such as “stages” (and “obstacles”) not only leads to many difficulties, it also hides a wealth of interesting phenomena. However, we do not discuss here the conceptual tools relevant to later analysis of these phenomena in a way which makes full use of what history can offer and provides relevance from a mathematics education perspective.

Second, more often than not, large scale narratives are implicitly expected to cover all the stages. For instance, in his paper, Katz mentions a final structural stage, in spite of the fact that it is not central to his argument and relies on a rather dated view of the history of modern mathematics. Indeed, recent historical work has greatly improved our understanding both of the history of algebra in the 19th and early 20th centuries,Footnote 38 and of the history of structuralism, both in algebra (Corry 2004) and in other branches of mathematics (Chorlay 2010).

The transition from rhetorical algebra to symbolic algebra is also the focal point of Puig and Radford (2007). Their semiotic analysis focuses on the way meaning was/is attached either to words or to symbols; and they point out the fact that the connection between signs and what they represent is different when writing an isolated equation, and when the equation is algebraically transformed to yield solutions. On the basis of their knowledge of rhetorical and geometrically flavoured algebraic practices (such as that of Al-Khwarizmi), Katz, Puig and Radford question what they take to be the commonly held view that algebra grew from arithmetic, and that it is only natural that school curricula should follow the same path.

The latest point illustrates what Artigue mentioned as one of the main roles of historical knowledge for the researcher in mathematics education, namely to distance oneself from one’s own training-induced mathematical knowledge and image of mathematics. The role of historical knowledge is used heuristically so as to suggest new vistas, or to identify elusive articulations. This heuristic use bears on epistemological moments rather than actual historical episodes, even if these moments are suggested by historical texts and can be illustrated with thought-provoking excerpts. As such, their heuristic value is not affected by the fact that, for instance, Al-Khwarizmi’s practice is more multi-faceted when it comes to the relationship between algebra, geometry and arithmetic. Indeed, in the Kitab al-jabr, Al-Khwarizmi usually resorts to geometry to justify numerical algorithms. However, on several occasions, for lack of a justification “by the (geometric) cause” (al-‘illa), he resorts to justification “by the expression” (al-lafz) (Rashed 2007, pp. 49–56). On several occasions, algebraic rewriting rules (in a rhetorical context) are derived from the numerical algorithms in the Indian numerical system (base 10 positional system) which Al-Khwarizmi had introduced in his book, Hindu Art of Reckoning. Footnote 39 Just as well, when analysing the emergence of symbolic algebra in the 16th and early 17th centuries, a historian would probably feel the need to bring into the picture elements which were central for mathematicians at that time, in particular, the distinction between analysis and synthesis, as well as the method of false position (regula falsi).

A third way of making heuristic use of the history of mathematics over a long time-scale is illustrated by Artigue and Deledicq’s text of 1992 on Quatre étapes dans l’histoire des nombres complexes: quelques commentaires épistémologiques et didactiques. The authors study four groups of texts related to four stages in the history of complex numbers: the use of new and uninterpreted operators in the symbolic algorithms for solving 3rd degree equations in the Italian Renaissance; the dispute over the logarithm of complex numbers, and the multivaluedness of the measurement of angles at the beginning of the 18th century; the first geometric representations of complex numbers at the turn of the 19th century; and the formal algebraic constructions of complex numbers by Cauchy (as residue classes of real polynomials modulo \( {X^{ 2} + 1} \)) and Hamilton (\( C \) as \( {R \times R} \)). The method of identifying “stages”, “episodes” or “key moments” is clearly epistemological. Artigue and Deledicq select four meaningful epistemological challenges (such as: to extend the symbolic system in order to provide more uniform algorithms to solve some equations), without aiming for a (questionable) comprehensive narrative in terms of successive homogeneous stages. It so happens that, in the case of complex numbers, these four epistemological challenges do correspond to well-delineated historical contexts.Footnote 40 Again, the fact that they include an episode such as the dispute over complex logarithms—which played a significant role in history but does not clearly correspond to a contemporary teaching issue—illustrates the “distancing ourselves from what we think complex numbers are about” function of history; a function which, as we mentioned earlier, knowledge of advanced contemporary mathematics can serve just as well.

For each group of texts, Artigue and Deledicq explicitly distinguish between three types of commentary: historical, epistemological, and didactical. They call “epistemological” the comments which relate to the nature of mathematics or the nature of the active engagement with mathematics, regardless of teaching contexts and learning issues. In this case, the “didactical” comments do not relate directly to teaching and learning issues, since the authors do not wish to draw quick conclusions from the study of the action of mathematical “experts” such as Cardano or Cauchy. Rather, the didactical comments relate to two different aspects: the comparison between the historical situations/challenges and teaching situations/challenges; and the conceptual tools which researchers in mathematics education are familiar with when describing and analysing engagement with mathematics in teaching contexts. For instance, after remarking that the formal constructions of Cauchy and Hamilton used real numbers but took place before the formal constructions of the set of real numbers, Artigue and Deledicq stress the similarity with teaching contexts in which the order of the introduction of notions differs from the purely deductive order. More often than not—and this is typical of a heuristic approach—the final output of the reflection is a series of questions rather than answers, or the delineation of research projects.Footnote 41

In this context, the distinction between an epistemological comment and a didactical cannot be absolute. However, even when similar terms can be used in both cases, a slight but definite difference in meaning remains. For instance, discussing (on an epistemological level) the tool/object polarity leads to questions such as: when did mathematicians consider that symbolic tools such as \( {\sqrt { - 1} } \) challenged the notions of number or magnitude? What were the relative roles of the geometric semantic and the formal constructions in the passage from complex-numbers-as-tools to a new and autonomous class of objects? In the didactical context, the dialectique outil/objet of Régine Douady provides tool to analyse the function of a given concept in a given teaching context, and to design teaching paths.

8.3.3 Epistemological Analysis of an Advanced Field: Linear Algebra

In the 1990s, Jean-Luc Dorier’s dissertation and subsequent publications exemplify the rare case in which both historical and didactical issues are studied at research level, simultaneously. Dorier combined a clear methodological distinction between the two fields of inquiry, and a long-run practice of co-problematisation.Footnote 42

The topic of study was linear algebra and its teaching in the transition from secondary to tertiary education. The starting point was a diagnosis of the didactical system around 1990 in France, and its structural shortcomings. In higher-secondary education, students dealt with free-moving vectors (defined either as translations, or as equivalence classes of couples of points) in the context of Euclidean geometry. These vectors differed from those used in physics (which are usually fixed), and with their 3D-geometrical interpretation and their products (dot and cross), they also differed strikingly from the elements of an abstract, axiomatically-defined, vector space. As a result, the teaching of abstract linear algebra in the first year of higher education usually combined a poorly-motivated abstract approach as far as lessons were concerned (and those who tried to abstract the general structure from the geometrical case faced foreseeable difficulties) and repetitive algorithmic tasks as far as exercises were concerned; tasks for which the general theory was usually not necessary, since a reasonable command of linear systems could usually solve the problem.

Dorier endeavoured to both analyse the reasons for this state of affairs, and design alternative teaching paths. In both cases, the analysis and the proposals were based on his knowledge of history; however, the analysis and the proposals do not derive directly from history, for at least two reasons. First, there is no such thing as the history of linear algebra, and the problem of delineation of the object of study is a purely historical problem. Second, because of the heterogeneity of the two fields of research, Dorier selected what he considered to be key episodes in the history of mathematics, analysed them from both a historical and epistemological viewpoint, and used these studies as raw material for his didactical reflection. This didactical reflection combined other elements, both theoretical and empirical.

A key episode took place in the inter-war period, with the formulation and (partial) adoption in the mathematical community of axiomatic algebraic structures:

This final step has its roots in the late nineteenth century, but only really started after 1920. It corresponds to the axiomatization of linear algebra, that is to say the reconstruction, with the concepts and tools of a new axiomatic central theory, of what used to be operative (but not explicitly theorized or unified) methods for solving linear problems. It is important to realize that axiomatization did not, in itself, allow mathematicians to solve new problems, but it gave them a more universal approach and language to be used in many varied contexts (functional analysis, quadratic forms, arithmetic, geometry …). In fact, the theory of determinants, which was very prosperous in the first half of the nineteenth century, is sufficient to solve all linear problems in finite dimension. (Dorier 1995, p. 176)

This epistemological reading of a historical episode paves the way for didactical analysis. Contrary to many concepts taught at primary and secondary levels, the concepts of abstract linear algebra do not primarily fulfil a problem solving function; rather, they serve formalising, unifying and generalising purposes (FUG conceptsFootnote 43). This specific connection between new concepts and more familiar elements of knowledge create specific teaching challenges: painstakingly designed problem-solving situations will probably not be adequate; and abstracting, unifying and formalising on the basis of analogies between many different specific fields is probably not a feasible teaching path. For FUG concepts, alternative teaching strategies can be based on using reflective analysis with students, carried out at a meta level:

My hypothesis, based on the epistemological analysis presented above, is:

students have to anticipate the power of generalization due to the use of vector spaces.

In this sense, I tried to build a teaching sequence which introduces the learner to a condensed form of reflective analysis, which has been proved to be one of the fundamental stages in the genesis of unifying and generalizing concepts. (…) The sequence, built on the basis of an epistemological analysis, creates an artificial context, which motivates the explicitation of the vector space axioms by the students themselves. (Dorier 1995, p. 186)

The adjective “artificial” testifies to the fact that this is not a rediscovery approach. Of course, both the challenge (teaching a FUG concept) and the design (enabling students to reflect on the properties of the mathematics they are dealing with) are pretty specific to late-secondary and tertiary education. However, they are not specific to abstract algebra, as Robert’s work on the notion of limit shows.

A different historical investigation was motivated by didactical and epistemological reasons. Starting from the hypothesis that a key concept for learners was that of linear dependence (and its avatars: dimension, rank of a family of vectors, rank of a linear system of equations), Dorier investigated historical episodes of explicit formulations of similar concepts; as mentioned above, mathematicians were familiar with properties of linear dependence long before they were integrated into the abstract-algebraic theory of vector spaces, and had efficient tools to deal with linear problems. Without attempting to cover everything even barely related to linear-thinking, Dorier focused on the work of several authors, in particular Euler (1707–1783), Grassmann (1809–1877) and Frobenius (1849–1917). His in-context then comparative studies enabled him to make out several (mathematically equivalent) viewpoints on linear dependence, thus providing epistemological depth to the mathematical concept, and to identify conditions for the formulation of a more abstract and context-independent notion of linear dependence.

In his paper on Cramer’s paradox in the theory of algebraic curves, Euler introduced a notion which Dorier christened “inclusive dependence”: in a linear system,Footnote 44 the equations are not independent if at least one of them is “included” in the others, a phenomenon which manifests itself through an obstruction to the usual elimination procedure. In his paper of 1875 on the Pfaff problem, Frobenius also discussed linear systems, and the context is also elimination theory—although with the central tool of determinants, which was not the case for Euler. In spite of these similarities, the Frobenius paper explicates many fundamental notions which were only implicit in Euler: the fact that the set of solutions is stable under linear combinations; the fact that other systems of equations can be considered equivalent to the first, in two different (but equivalent) ways: they have the same linear space of solutions; and they consist of linear combinations of the first equations, satisfying certain conditions (we would now regard this as a change of basis for a subspace of the dual space). The duality between coefficients for the equationsFootnote 45 and solution n-uples is brought to light, as is the relationship between the ranks m and n-m of the two systems of n-uples. Leaving technicalities aside, we can say that Frobenius formulated fairly general concepts by considering sets of solutions and sets of equations, the properties and equivalent representations of such sets, and a numerical characterisation of their “size” (the rank).

On the basis of the analysis we just outlined, Dorier designed a teaching module in which the notions of linear dependence and independence were gradually formulated in an ever more abstract way, starting from the context of linear equations. Instead of determinant theory (which was the historical context), Dorier relied on Gaussian elimination. At several points, meta level questions were discussed such as generality of the procedure, invariance of the rank, and the relative virtues of various proto-definitions of the concept of linear independence.

Of course, this summary does not do justice to a decade of investigations. In particular, Dorier also studied the transposition process through which the abstract algebraic structure of the 1930s was gradually divided and transformed into teaching objects. In particular, he showed how the geometric vectors—whose history is quite different from the one we just recapitulated—were brought into the picture in the hope of paving the way for the abstract version. For a comprehensive view of this work, see Dorier (2000a).

8.3.4 History of Physics and Physics Education Research

The following detour through physics may be surprising. Nevertheless it enables us to discuss a new set of examples based on recent research which documents a wealth of fruitful opportunities for interaction between the history of science and science education research. This section of the chapter will also reflect the fact that the two authors come from history of mathematics and physics education research respectively.

8.3.4.1 “Obstacle” in Early Physics Education Research

This detour through physics education research echoes Artigue’s paper. She positioned her thinking on the concept of “epistemological obstacle” in the wake of results developed by early physics education researchers, particularly those of Viennot (1979, 2001). In doing so, Artigue connects the concept of obstacle to more generic reasoning that could explain recurrent mistakes or confusions produced by both novices and experts facing several physical situations or problems to be solved. As an example, Artigue refers to the linear causal reasoning where several variables changing simultaneously are considered one after another, interacting in a chronological way (i.e., multiple variable relationships interpreted as temporal relationships). This general trend of reasoning (generally incompatible with rationality in physics) has been identified as a powerful obstacle “generator” in several domains of physics (for example, electrokinectics, thermodynamics, waves). Focusing on trends of reasoning, physics education research reactivated to a certain extent the Bachelardian concept of “epistemological obstacle”. In the late 70s and early 80s, several physics education researchers sought to identify similarities between historical ideas and students’ trends of reasoning.Footnote 46 This orientation very likely opened the way for a long-standing interest from physics education researchers (and more generally from science education researchers) in the history of science. Nevertheless, this interest has changed substantially. Indeed, the large volume of research which has been carried out worldwide tends to promote the history of science as a powerful science teaching (or training) tool, especially in physics education.

Our intention is not to provide an exhaustive review of the aims and findings produced by physics education researchers working on or with historical materials. Instead, we address the issue of the choices that underlie the exploitation of the history of science when used for educational aims. Indeed, specific educational purposes may reflect specific and heuristic terms for using historical materials. In this regard, the history of science is mainly used in order to: (1) address the learning of a given concept (or law); and (2) improve students’ and/or teachers’ views on the nature of physics (from both epistemological and social viewpoints). Even if these two approaches are rarely exclusive from each other, researchers in physics education often fail to provide an explicit justification for the choices which govern the way they use, extract and organise the historical material in their work and studies. In this last part of our chapter, our aim is to provide some guidelines or benchmarks for a more explicit justification of the choices taken by physics education researchers when using historical materials. These guidelines can also benefit mathematics education researchers.

8.3.4.2 History of Science and the Learning of New Concepts

The above-mentioned approach promoted by Dorier can form a fruitful answer to questions which inevitably underlie the elaboration of teaching-learning sequences in physics involving the history of science. In our perspective, associating the history of science and physics education consists in creating an “epistemological dialectic” (Dorier 2000b, p. 10) between two inquiries: the first focuses on students’ reasoning or conceptions concerning a given physical phenomenon; the second concentrates on constructing knowledge in the historical context. This dialectic allows: (1) specifying the didactic constraints shaping the teaching process; (2) extracting the historical elements to be reorganised according to these constraints; and (3) ensuring that these elements take place in the didactic system with the aim of favouring the acquisition of a given knowledge by students. This last stage requires assuming a specific reorganisation of the extracted historical elements. Such reorganisation takes the form of a teaching-learning sequence designed on the basis of historical elements and named “didactical reconstruction”. In a didactical reconstruction, historical and an-historical elements are included and mixed up in order to address specific didactic constraints (such as targeted knowledge, students’ current conceptions or reasoning, visibility of the historical material, and the usual functioning of the class).Footnote 47 The idea is not to provide teachers or students directly with a history of science but to identify learning levers from a specific historical inquiry involving first-hand written sources.Footnote 48 These levers are articulated and completed with an-historical elements chosen and organised according to specific educational and conceptual purposes. Consequently, a didactical reconstruction is neither objective nor exhaustive but appears constrained, as with any reconstruction project. Indeed, the historical elements retained by a physics education researcher, as well as the way he or she chooses to organise them can lead to reconstructions that differ from those of historians of science. Because the motivations are specific on both sides, they produce particular readings. The legitimacy of these readings is guaranteed, not through a possible closeness with an ideal historical route but through the “fertility” of the program which underlies them:

There is no neutral reading, no reading that does not engage a previous decision for defining the detained events or for defining relevant materials, entities, mechanisms. Every time, a selection principle is applied that depends on the adopted program. Each story, each reconstruction, each model corresponds to a determined principle of reading. This principle remains from a program, that is, from a generic manner of explaining or giving sense to an object. (Berthelot 2002, p. 242, trans. CdH)

In this perspective, we admit that a didactic “program” exists that supports the search for elements that could favour a better appropriation of physic concepts and laws. When conducting research, using the history of science for physics learning in a conceptual perspective should take into account the type of reasoning a student can use concerning a phenomena to be studied. Searching, within the history of science, for situations that could be transformed into problems to be investigated by students is a way of challenging the didactic use of the history of science. From this standpoint, a search for certain proximity between students’ reasoning and ideas from the past can form a fruitful fulcrum for the historical literature.

The model of didactical reconstruction as we define it has two functions. First, it allows us to make more explicit and also to better understand the (implicit) choices governing some teaching-learning sequences based on historical grounds. It is in this way that we understand the principles governing Merle’s (2002) teaching sequences concerning the horizon line (skyline) in elementary astronomy. This sequence is based on an argument developed by Aristotle to address the spherical shape of Earth: the modification of the aspect of the night-sky for observers travelling south. It also takes into account the way students explain (with a drawing) why observers situated either in the north or in the south of the same meridian do not see the same stars in the sky. Generally, the drawing they provide presents a meridian sometimes flat and sometimes curved, whereas the field of vision of each of the observers is represented by a cone. If this way of geometrising the visible space of an observer fits perfectly with observations described by Aristotle, it does not allow discrimination between ideas of a flat Earth and those of a round Earth. In fact, the geometrical tool used by Aristotle to solve the puzzle of the visible stars is not a cone but a tangent line to the meridian, today known by the term “horizon”, and which forms the knowledge developed as Merle’s sequence. Here we face an asymmetry (not explicitly specified by the author) between historical hypotheses and stakes on the one hand and didactic ones on the other hand. From an historical point of view, the notion of horizon serves as a model for explaining the changes in the night-sky in order to justify the idea of a spherical Earth; from a didactic point of view, the changes in the night-sky connected to the spherical shape of the Earth allow the construction of the notion of horizon. Students face an incoherence which leads them to admit that their tool of “spontaneous” reasoning does not allow them to conclude that the Earth is spherical. This incompatibility between Aristotle’s reasoning process and their own conclusions leads them to build a new geometrising tool for the field of vision. Here, the history of science is not actively involved in resolving the problem posed to the students but the approach proposed by Merle can be interpreted in terms of a didactic reconstruction: the situation she grasped from history of science has been chosen in order to form a fruitful problem-to-be-solved by students. Her choice was governed by what she knew about students’ conception of “horizon”.

The second function of our model is to provide science education researchers with a framework for the design of a teaching-learning sequence. We have created and implemented several sequences according to the guidelines presented above (de Hosson and Kaminski 2007; de Hosson and Décamp 2014). In these sequences, a dialectic was settled in order to create a problem directly inspired by an historical episode that could meet students’ interest and thus, be accepted by them.

A common conception hold by students concerning the relationship between force and motion was addressed (de Hosson 2011). It is difficult for students to admit that an object dropped from the top of the mast of a ship moving at a constant velocity lands at the bottom of the mast because it retains the horizontal movement of the ship. This difficulty echoes the problem staged by Galileo in his Dialogue concerning the two chief world systems. This proximity led us to elaborate a learning pathway in which some elements of Galileo’s dialogue were selected and reorganised according to specific educational constraints. The relevancy of the sequences have been asserted in the “didactic engineering” framework (Artigue 1994) and rely on the identification of students with the characters staged by Galileo. Here, the concept of “epistemologic obstacle” (Bachelard 2002) is considered as heuristic since it allows the search for anchoring problems, i.e., problems inspired by the history of science that could be appropriated easily by students, particularly as difficulties or ideas or reasoning on both (historical and cognitive) sides could be considered, to a certain extent, to be closed.Footnote 49

Beyond the restricted and specific perspective detailed above, the selection process conducted by the science education researcher working with historical materials can also aim to challenge students’ ideas of the nature of science.

8.3.4.3 Nature of Science (NoS)

History of science seems to play a significant role in helping teachers and/or students to develop more appropriate conceptions of the scientific enterprise. Nevertheless, the research carried out by Abd-el-Khalick and Lederman shows that the use of the HoS to enhance teachers’ NoS views operates under certain conditions (Abd-el-Khalick and Lederman 2000). In particular, they claim that only an explicit instructional approach that targets certain NoS aspects can enhance teachers’ NoS views:

Science educators cannot simply assume that coursework in HoS by itself is sufficient to help prospective science teachers develop desired understandings of NoS (Abd-el-Khalick and Lederman 2000, p. 1088).

Considering NoS as an expression that refers to ‘‘the epistemology of science, science as a way of knowing, or the values and beliefs inherent to the development of scientific knowledge’’ (Lederman 1992), some authors have used the history of science to promote a renewed image of the nature of science. This also engages different choices and foci. Indeed, if we wish to use the history of science to influence students’ understanding of science, we must treat historical material in ways which illuminate particular characteristics of science. In the research conducted by Maurines and Beaufils (2012), history of science has not been considered per se but as a means to introduce students to 20th century philosophical ideas of science in order to help them to acquire scientific literacy. This was considered to be a richer understanding of how science works mainly today, rather than in the past. Consequently, the authors sought to identify, through analysing the history of physics (the creation of the laws of refraction), which aspects of physics could be considered as temporal invariables. According to the authors, the intersection of the various studies on science is where the most authentic view of science is revealed. Thus, they based their analysis on the philosophy, history, psychology and sociology of science. From this perspective, scientific knowledge was considered to be the result of activities—intellectual and practical—performed by individuals, working collectively, in the socio-cultural context of a given historical period. Maurines and Beaufils elaborated a set of documents (a dossier) in order to address and reveal some consensual views on NoS (e.g., “relationships between scientists”). These documents were made up of two types: some comprised historical scientific information (e.g., facts, hypothesis, knowledge, experiment); while others were chosen in order to be analysed on the basis of the different dimensions (spatiotemporal range and degree of externality) which can be associated with the different characteristics of NoS. As an example, the texts of the dossier related to the objective ‘relationships between scientists’ provide not only some scientific information, such as the type of explanation advanced by a scientist about the law of refraction, but also some information on the interaction between this scientist and the others.

8.4 Conclusion

We conclude this methodological tour by commenting on a deceptively simple motto: history does not teach, yet there is a lot to learn from it. Historians do not provide direct (even if partial) answers to MER questions, for a number of structural reasons which we attempted to describe: the two main reasons being the deep heterogeneity of the objects of study (mathematics written in contexts different from ours, usually by experts/teaching and learning of mathematics); and epistemological differences between the two fields of study (which reflect this heterogeneity). Yet heterogeneity and autonomy do not imply incommensurability. We presented several cases of fruitful interactions, from the purely heuristic—albeit of a well-controlled nature—to instances of the dialectic of co-problematisation.

While discussing these examples, we touched upon theoretical frameworks and concepts of MER, but we did not focus on them. These concepts were designed to study teaching and learning situations, hence are probably not suited for historical investigation. However, when the research question is a MER question, didactical concepts do not only help organise data into an intelligible structure, they also play a key role in the first phases of an investigation, when shaping the question and delineating the object of study. How, then, can a research question framed by didactical concepts make fruitful use of the history of science without mistaking these concepts as tools for historical investigation (Barbin 1997)? We feel this tension is structural and call for methodological vigilance. In this chapter, we adopted an approach which is very close to Dorier’s, by using few and rather loose concepts—such as aspect—when discussing direct contact between historical and didactical investigations. This does not imply that, in another phase of research, theoretical frameworks cannot play their part. Indeed, we feel a natural continuation of the methodological discussion presented in this chapter would be the study of the extent to which different theoretical frameworks assign a role to epistemological investigation—and the role it might play.

As Artigue pointed out, discussing facts, knowledge and concepts cannot be all there is when discussing the role of epistemological awareness for the researcher in science education. Various forms of experience are also central, such as distancing oneself from one’s own mathematical culture, or enabling one to see new dimensions in otherwise rather flat and unproblematic elements of mathematics. We would like to conclude with a quotation from Michel Foucault, who, in a different context, strikingly described a similar experience:

As for what motivated me, it is quite simple; I would hope that in the eyes of some people it might be sufficient in itself. It was curiosity – the only kind of curiosity, in any case, that is worth acting upon with a degree of obstinacy: not the curiosity that seeks to assimilate what is proper for one to know, but that which enables one to get free of oneself [se déprendre de soi-même]. After all, what would be the value of the passion for knowledge if it resulted only in a certain amount of knowledgeableness and not, in one way or another and to the extent possible, in the knower’s straying afield [égarement] of himself? There are times in life when the question of knowing if one can think differently than one thinks, and perceive differently than one sees, is absolutely necessary if one is to go on looking and reflecting at all. (Foucault 1990, p. 8)