Philosophy Compass 4/1 (2009): 255–270, 10.1111/j.1747-9991.2008.00192.

Philosophical Issues in Electromagnetism

Mathias Frisch*
The University of Maryland

Abstract
This paper provides a survey of several philosophical issues arising in classical
electrodynamics arguing that there is a philosophically rich set of problems in
theories of classical physics that have not yet received the attention by philosophers
that they deserve. One issue, which is connected to the philosophy of causation,
concerns the temporal asymmetry exhibited by radiation fields in the presence of
wave sources. Physicists and philosophers disagree on whether this asymmetry
reflects a fundamental causal asymmetry or is due to statistical or thermodynamic
considerations. I suggest that an explanation appealing to the asymmetry of
causation is more promising. Another issue concerns the conceptual structure of
the theory. Despite its empirical success, classical electrodynamics faces serious
foundational problems. Models of charged particles involve what by the theory’s
own lights are idealizations, I maintain, and this is a feature that is not readily
captured by traditional philosophical accounts of scientific theories. Other issues
I discuss concern (i) the relation between Lorentz’s theory of the electron and
Einstein’s Theory of Special Relativity; (ii) the notion of the domain of a theory,
the question of theory reduction, and the relation between classical and more
fundamental quantum theories; and (iii) the role of locality constraints, their
relation to the concept of causation; and the status of locality conditions in the
semi-classical theory of the Aharanov-Bohm effect.

1. Overview
Traditionally research in the philosophy of physics has focused predominantly
on interpretative puzzles in quantum mechanics and on the revolution in
our understanding of the structure of space and time ushered in by the
Special and General Theories of Relativity. This overly narrow focus may
give the impression that classical physics is metaphysically and methodo-
logically unproblematic or uninteresting, with the consequence that many
philosophically interesting issues in the foundations of physics have been
largely ignored by philosophers of physics. Fortunately, this situation has
begun to change.
In this essay I will survey several philosophical issues arising in the
context of classical electrodynamics. I will focus on the following topics,
some of which overlap with more traditional concerns. First, we can ask
© 2008 The Author
Journal Compilation © 2008 Blackwell Publishing Ltd
256 Philosophical Issues in Electromagnetism

what role theories of classical physics play in contemporary physics, given

that these theories have in a sense been superseded by quantum theories.
What, for example, is the status of the ontologies postulated by classical
theories? Can we accept quantum mechanics and nevertheless rationally
remain realists about classical fields and charges? Second, historically, the
development of microscopic classical electrodynamics and ‘theories of the
electron’ has been intimately linked with the development of the Special
Theory of Relativity. But how exactly did our understanding of space and
time change from Lorentz’s theory of the electron to Einstein’s theory of
relativity? According to one traditional way of understanding the change,
Lorentz’s theory of the electron proposed dynamical explanations for
phenomena that Einstein’s theory has shown to be purely geometric and
kinematic effects. Yet this orthodox view has recently been challenged.
Third, at the heart of classical electrodynamics are the Maxwell equations
and the Lorentz force law.1 The microscopic Maxwell equations allow us to
determine the electromagnetic fields, the electric field E and the magnetic
induction B, associated with a given microscopic charge and current
configuration. The equations are usually solved in terms of a modified
initial value problem: The initial fields at some time t0 together with the full
trajectory of the charges are posited as input into the Maxwell equations,
which then determine the evolution of the fields. According to the
Lorentz-force law, fields exert a force on charged particles. For continuous
distributions of charges (so-called ‘charged dusts’) the Lorentz force law
can be derived from the Maxwell equations together with the standard
formulation of the principle of energy-momentum conservation. The
situation is far more complicated in the case of discrete particles coupled
to an electromagnetic fields. While we can model both kinds of interactions
separately in the case of discrete charges; and while the resulting models
are empirically highly successful within the domain of classical physics, it
is unclear whether there is a conceptually fully satisfactory theory of the
interaction between localized microscopic charges and fields that takes
both modes of interaction into account simultaneously. Thus, despite its
tremendous empirical success, the theory is beset by serious, and perhaps
irresolvable, foundational problems.
Fourth, electromagnetic waves in the presence of wave sources exhibit
a striking temporal asymmetry. We observe waves that coherently diverge
from a source, but we generally do not observe the temporal inverse –
waves coherently converging onto a wave source – despite the fact that the
underlying dynamical equations of the theory are time-reversal invariant
and permit both kinds of behavior. What is the source of this asymmetry?
Existing answers fall into two classes: Some physicists and philosophers argue
that the radiation asymmetry can ultimately be reduced to thermodynamic
and statistical considerations, while others argue that the asymmetry is an
instance of a general causal asymmetry. This is an issue where the philosophy
of classical electrodynamics makes contact both with the philosophy of
© 2008 The Author Philosophy Compass 4/1 (2009): 255–270, 10.1111/j.1747-9991.2008.00192.x
Journal Compilation © 2008 Blackwell Publishing Ltd
 Philosophical Issues in Electromagnetism 257

thermodynamics and with a core issue in metaphysics – the place of

causation in our conception of the world.
Fifth, according to the standard interpretation, the ontology of classical
electrodynamics consists of discrete charged objects and continuous
electromagnetic fields transmitting the influence from one charge on
another. The fact that interactions between charged objects are mediated
by fields ensures that the theory satisfies various locality conditions –
conditions which are violated, for example, by Newtonian gravitational
theory. Indeed, arguably the very fact that eletromagnetic fields act locally
provides an important reason for interpreting fields realistically. However,
the traditional interpretation comes under pressure from the semi-classical
Bohm-Aharanov effect, which has been taken to suggest either that fields
can act non-locally or that the so-called electromagnetic potentials, which
in the traditional interpretation are viewed as mere calculational devices,
ought to be interpreted realistically.

2. Classical Physics in a ‘Quantum World’

An immediate question that arises when we discuss theories of classical
physics is what the present status of these theories is in light of our
acceptance of quantum physics. Does our best current physics not tell us
that we ‘live in a quantum world’? Has classical electrodynamics, thus, not
been replaced by quantum theories and is today of merely historical
interest? An affirmative answer to these questions is suggested by many of
the major methodologists of science of the twentieth century, including
by Thomas Kuhn in his famous book Structure of Scientific Revolutions
(1996), but also by many of Kuhn’s critics. According to this view, when
scientists adopt a new theory, such as quantum mechanics, the new theory
replaces its predecessors, which are abandoned. While Kuhn and his critics
disagree on whether the process of theory acceptance and rejection is
rational and while there is widespread disagreement on the circumstances
that lead to a theory’s demise, there seems to be broad agreement that the
history of physics does indeed consist of theories that are accepted only
to be eventually discarded in favor of more successful rivals.
This view is seriously flawed, however. To be sure, there are examples
of theories that do fit the standard view and which were completely
abandoned in favor of their successors. Phlogiston theory is one such case.
Yet neither classical electrodynamics nor Newtonian physics fit this
traditional picture and both theories continue to play an important role
in contemporary physics. A more adequate way of thinking about theories
such as these has been proposed by Fritz Rohrlich, among others (Rohrlich
and Hardin). On Rohrlich’s view, quantum mechanics has not fully
replaced classical electrodynamics but rather has helped to establish the
limits of the latter theory’s domain of validity – that is the range of
phenomena that, using the theory, we can model to a desired level of
© 2008 The Author Philosophy Compass 4/1 (2009): 255–270, 10.1111/j.1747-9991.2008.00192.x
Journal Compilation © 2008 Blackwell Publishing Ltd
258 Philosophical Issues in Electromagnetism

accuracy. Despite the fact that classical electrodynamics has been superseded
by quantum mechanics in some sense – we take quantum theories to be
more fundamental and there are phenomena which cannot be modeled
classically – the classical theory remains the most appropriate theory for
modeling and explaining phenomena in its domain of validity. In Rohrlich’s
terminology, classical electrodynamics is an ‘established theory’: it is
well-confirmed within a certain domain of applications but with clear
limits of validity given by a more fundamental quantum theory.
Yet this view of classical theories also raises a number of questions that
deserve a fuller discussion. Does classical electrodynamics indeed provide us
with the best explanations of phenomena within its domain of application,
and if ‘yes’, what theory of scientific explanation can account for this
fact? How can we characterize in more detail the relation of classical
electrodynamics to the putatively more fundamental quantum theories?
What, for example, is the relation between the different ontologies
postulated by classical and quantum theories and what ought our ontological
commitment to the classical theory be? Can we coherently be committed
both to the existence of classical fields and to quantum field theory as a
more fundamental theory? Or is our commitment better described as a
commitment to the claim that the world is such that on a certain level it
is best represented as consisting of classical fields interacting with charged
material objects? (Rohrlich, ‘Realism Despite Cognitive Antireductionism’)
argues for a non-reductionism that allows us to be realists about the
classical ontology even while we acknowledge that the classical theory’s
mathematical formalism can be shown to arise from the more fundamental
quantum mechanical formalism in the limit. Whether there is room for such
a (weak) non-reductive position, which collapses neither into a full-blown
ontological reductivism nor into a pragmatism or instrumentalism about
the higher-level ontology is an issue that deserves to be explored further.

3. Classical Electrodynamics and Special Relativity

The development of microscopic electrodynamics at the turn of the twentieth
century was intimately bound up with that of the special theory of
relativity. From a natural assumption Lorentz showed that when the
Maxwell equations allow a certain configuration of charges and fields in
a system at rest, then they allow the same configuration of charges in an
inertial system moving through the ether. For Lorentz, however, only
measurements in frames that are at rest with respect to the ether reveal
the real fields; the fields satisfying the Maxwell equations in a moving
frame are ‘fictitious’. Lorentz was also able to derive equations equivalent
to the relativistic equation for length contraction (and eventually also for
time dilation) from dynamical considerations.2
One of Einstein’s main motivations in developing the special theory of
relativity appears to have been the fact that even though the Maxwell
© 2008 The Author Philosophy Compass 4/1 (2009): 255–270, 10.1111/j.1747-9991.2008.00192.x
Journal Compilation © 2008 Blackwell Publishing Ltd
 Philosophical Issues in Electromagnetism 259

equations allow the same configurations of charges and fields at rest and in
motion, there is no frame-independent answer as to what the electric and
magnetic fields are. Rather electric fields in one frame can appear as magnetic
fields in another inertial frame and vice versa. Thus, according to relativity
theory, the object that exists frame-independently is the electromagnetic
field, which can be represented in terms of the electromagnetic field tensor
Fµν which has electric and magnetic field strengths as coordinate-dependent
components.
It is generally held today that Einstein’s special theory of relativity has
changed our understanding of space and time and of the motion of
objects in two important ways. First, we no longer think that there is a
privileged class of inertial frames, the ‘ether rest frames’; and, second, we
think that relativistic phenomena, such as length contraction, do not require
an explanation in terms of electromagnetic forces or quantum mechanical
interactions, a dynamical explanation, but are purely kinematic effects –
that is, artifacts of the relative state of motion of the frames used for
representation – and a consequence of the geometry of spacetime, of the
fact that spacetime is Minkowskian. This orthodox view has recently been
challenged by Harvey Brown (Brown; Brown and Pooley), who agrees
with the first part of the orthodox view but not with the second. According
to what Brown takes to be the orthodox view, the structure of spacetime
– that is, the fact that spacetime is Minkowskian – explains the fact that
the laws of our theories are Lorentz invariant and accounts for the universal
behavior of rods and clocks. But Brown argues that this view has the
arrow of explanation backward. On his view, relativistic phenomena
ultimately require a dynamical explanation; it is a brute fact, which itself
permits of no further explanation, that the laws are Lorentz invariant and
it is this fact that explains that Minkowski spacetime is the proper arena
in which to represent non-gravitational physical phenomena.
One of Brown’s targets is the view that spacetime substantivalism plays
an important role in explaining relativistic effects. The substantivalist takes
spacetime to be an entity in its own rights, the affine structure of which
determines the geodesics – that is, the paths along which unaccelerated
particles move. Once we know that objects ‘live’ in Minkowski spacetime,
and satisfy the constraints of Minkowski geometry, there is a simple well-
known geometric construction that allows us to derive length contraction
and time dilation. Thus, length contraction and time dilation appear to
be purely geometric effects, which are a straightforward consequence of
the structure of Minkowski spacetime. But Brown argues that merely
appealing to the structure of spacetime does not answer the question as
to why objects obey the constraints of Minkowski geometry. In positing
that force-free objects follow the geodesics of Minkowski spacetime the
substantivalist needs to assume that objects are somehow able to sense ‘the
ruts and grooves’ of spacetime and Brown finds this assumption utterly
mysterious. What is needed instead, Brown argues, is a ‘constructive’
© 2008 The Author Philosophy Compass 4/1 (2009): 255–270, 10.1111/j.1747-9991.2008.00192.x
Journal Compilation © 2008 Blackwell Publishing Ltd
260 Philosophical Issues in Electromagnetism

explanation that appeals to ‘the details of the bodies’ microphysical

constitution’ (Brown and Pooley), where he invokes a distinction famously
drawn by Einstein between ‘constructive theories’ that explain a phenomenon
in terms of underlying mechanisms and ‘theories of principle’ that explain
a phenomenon by showing how it can be derived from general principles,
generalized from experience.3
Brown’s account has been forcefully challenged by Michel Janssen, who
takes himself to defend the orthodox view that the special theory of relativity
has shown length contraction and time dilation to be purely kinematic
phenomena (Janssen; see also Balashov and Janssen). Janssen agrees with
Brown in his rejection of spacetime substantivalism but argues that even
within a relationalist framework, which takes spatiotemporal relations to have
no existence independent of physical objects instantiating them, the
explanatory arrow is from the structure of Minkowski space to the Lorentz
invariance of the dynamical laws, rather than the other way around.
On first sight there seems to be a rather stark disagreement between
Brown’s and Janssen’s views. Brown, it seems, believes that only an account
of the particular forces pushing and pulling the microscopic constituents
of a rod can explain length contraction, while Janssen maintains that the
structure of Minkowski spacetime explains the phenomenon. But on closer
inspection the differences might strike one as rather more subtle. While
Brown stresses that the behavior of rods and clocks requires a dynamical
and constructive explanation he also appears to concede that the only
feature of the dynamics that is required to explain length contraction or
time dilation is its Lorentz invariance. The full dynamics may be needed
to account for the existence of stable solid objects and of time-keeping
devices, but given that there are stable macroscopic objects, no more than
the Lorentz-invariance of the laws is required to account for their behavior
in different inertial frames. And while Janssen seems to insist that the
proper explanation of length contraction is geometrical, he also suggests
that the explanation proceeds by appealing to Lorentz invariance as a
‘meta-law’, and that the role of Minkowski spacetime is to ‘encode’ this
nomological fact. Thus, on one reading of the two competing views,
the disagreement reduces to the question whether Lorentz invariance is
postulated as brute constraint, as Brown maintains, or as meta-law or
overarching nomic constraint, as Janssen claims. But then the disagreement
would perhaps be best captured neither in terms of the kinematics-dynamics
distinction nor in terms of the distinction between principle- and con-
structive theories, but rather as a disagreement about the role of laws in
science and about the explanatory force of appeals to nomic constraints.

4. The Coherence of the Theory

The fundamental equations at the heart of microscopic classical electro-
dynamics are the microscopic Maxwell equations, which determine the
© 2008 The Author Philosophy Compass 4/1 (2009): 255–270, 10.1111/j.1747-9991.2008.00192.x
Journal Compilation © 2008 Blackwell Publishing Ltd
 Philosophical Issues in Electromagnetism 261

fields in the presence of a charge and current configuration, and the

Lorentz force law, which specifies the force acting on a charge in an
electromagnetic field and is used as input into Newton’s second law to
give the momentum change of a charge. Together they constitute a coupled
set of quasi-linear equations that, one would hope, determine the temporal
evolution of charges and fields specified at some initial time t. In the case
of continuous charge distributions or ‘charged dusts’ there are indeed
existence and uniqueness proofs for the coupled Maxwell-Lorentz
equations. The only problem arising in this case is that these proofs only
guarantee the existence of local solutions – that is solutions in some small
neighborhood of the initial data in question. Moreover, there are intuitively
plausible initial charge distributions for which local solutions cannot be
arbitrarily extended and for which the equations have no global solutions
(see Frisch, ‘Laws and Initial Conditions’).
But many of the applications of microscopic classical electrodynamics
concern the interaction of discrete charged particles, such as electrons,
with electromagnetic fields and it is much less clear in this case whether
there is a coherent and fully satisfactory way of integrating the two ways
in which charges and fields interact. At the heart of the problem lies the
fact that a completely satisfactory treatment should include so-called
self-interaction effects – the effects that the field of a charged particle has
on the motion of the particle itself – and arguably there is no conceptually
completely unproblematic way of incorporating self-interactions into
the theory.4
The standard way of modeling interactions between electromagnetic
fields and charges, familiar to physics students from both undergraduate and
graduate courses in electrodynamics, is simply to ignore self-interactions
– usually without making explicit that this is being done. The resulting
set of equations – the Maxwell equations and a Newtonian equation of
motion for charged particles that takes into account only the Lorentz
force due to fields external to the charge is inconsistent with the principle
of energy-momentum conservation, even though the theory is taken to
satisfy this principle. Frisch (‘Inconsistency in Classical Electrodynamics’;
Inconsistency, Asymmetry, and Non-Locality) argues that this inconsistency has
consequences for how we think about scientific theories and that classical
electrodynamics does not fit standard philosophical accounts of scientific
theorizing that take theories to delineate conceptually coherent possible
worlds. These arguments are criticized in Belot, Muller, and Vickers. Fred
Muller – incorrectly, I believe – criticizes the inconsistency argument,
while Gordon Belot and Peter Vickers agree that the assumptions in
question are indeed inconsistent but question what conclusions can be
drawn from this fact. Belot, in particular, argues that the inconsistency is
merely an instance of the common phenomenon that approximations that
we make in modeling physical systems are strictly speaking inconsistent
with the underlying fundamental equations.
© 2008 The Author Philosophy Compass 4/1 (2009): 255–270, 10.1111/j.1747-9991.2008.00192.x
Journal Compilation © 2008 Blackwell Publishing Ltd
262 Philosophical Issues in Electromagnetism

The overarching philosophical issue at the heart of this debate is

whether standard formal philosophical accounts of scientific theories – be
it as deductively closed sets of sentences or as classes of models representing
the possible worlds allowed by the theory – allow us adequately to capture
the conceptual structure of classical electrodynamics. The problem is that
any theory with self-interactions has to posit a model for charged particles
and arguably there is no physically ‘well-behaved’ model of a discrete
finitely charged particle that does not involve what by the theory’s own lights
are idealizations.
Models of charges can be grouped into two classes, depending on
whether charges are treated as point particles or as extended particles.
Point-particle models fit better with relativistic constraints, yet an immediate
problem faced by these models is that, according to the Maxwell equations,
the field of a finitely charged point particle is infinite at the location
of the charge and the field-energy diverges for any volume that includes
the charge. Extended-particle models avoid the infinities, but run afoul of
relativistic constraints such as the prohibition against superluminal
propagation. (Note, also, that there can be no purely electromagnetic
extended charged particles, since such a particle would explode due to
the repulsive forces between its different parts.) There is a rich history of
attempts to overcome these problems or to arrive at physically reasonable
particle equations of motion despite these problems, many of which are
reviewed in (Rohrlich, Classical Charged Particles; Parrott; Spohn).
Even though this question is far from being completely settled, it
appears that none of these solution attempts result in a treatment of the
self-energy problem that do not involve what by the theory’s own lights
are idealizations of some form or other. But if that is indeed the case,
then classical electrodynamics does not merely involve ‘benign’ kinds of
approximations to a fully coherent and in-principle axiomatizable formalism,
and hence may only ill fit with traditional philosophical accounts of
scientific theorizing. Rather, idealizations and approximations are present
at the very first step and in the very construction of a classical model of
the electron. Classical electrodynamics, thus appears to be an example of
what Mark Wilson calls a ‘theory façade’: a patchwork of laws that are
‘stitched together’ in ways that do not easily submit to formal axiomatization
and that contain weak spots – in our case, the self-interactions – which
are not readily covered by any of the patches, even though the resources
of the theory can in often ingenious ways be extended beyond the patches
and into the weak spots (see Wilson).

5. The Asymmetry of Radiation

A further issue concerns possible explanations of an apparent temporal
asymmetry that characterizes electromagnetic radiation phenomena. It is
a consequence of the Maxwell equations that accelerated charges radiate.
© 2008 The Author Philosophy Compass 4/1 (2009): 255–270, 10.1111/j.1747-9991.2008.00192.x
Journal Compilation © 2008 Blackwell Publishing Ltd
 Philosophical Issues in Electromagnetism 263

Radiation fields exhibit a temporal asymmetry, which intuitively can be

characterized as follows: We observe electromagnetic fields coherently
diverging from sources but we do not, or only rarely seem to observe the
time-reverse of such phenomena – that is, fields coherently converging
into a source. This asymmetry has struck many as puzzling, since the
Maxwell equations are time-reversal invariant. One should note, however,
that there is nothing puzzling about the fact that individual models of
fields in the presence of charged particles exhibit a temporal asymmetry.
What is puzzling, it seems, is that all or most of the situations we observe
exhibit the same temporal asymmetry.
The first surprisingly controversial issue is the question as to what
exactly the asymmetry consists in. In order to get a sense of the issue, a
little bit of background is needed. The Maxwell equations imply an
inhomogenous wave equation for the fields in the presence of charges or
wave sources. This equation defines an initial value problem and the
general solution can be written as a sum of diverging fields associated with
the sources – the so-called retarded fields Fret – and source-free incoming
fields Fin, propagating from the initial value surface. But the fields in a
given region can equally as well be represented in terms of a final value
problem and the total field then consists of source-free outgoing fields Fout,
propagated backward in time from the final value surface in accord with
the source-free, homogeneous wave equation, together with fields converging
on the source, the so-called advanced fields Fadv. That is, the same total field
Ftotal can be represented as a combination of source-free incoming and
retarded fields or of source-free outgoing and advanced fields:
Ftotal = Fret + Fin = Fadv + Fout.
Thus, it seems that the asymmetry cannot consist in the putative fact
that the fields associated with sources are retarded, since both retarded and
advanced representations are possible (as well as linear combinations of
the two).
However, if the incoming field is approximately equal to zero, then the
total field exhibits an asymmetry: in this case the total field is fully
retarded, but generally will not be fully advanced. Unless the retarded and
advanced fields are equal, the advanced field representation will involve a
non-zero source-free outgoing field Fout = Fret − Fadv, which will destructively
interfere with the advanced field before the source is turned on and will
be equal to the retarded field after the source turns on. This fact has led
Zeh to characterize the radiation asymmetry as follows: ‘Why does the
condition Fin = 0 (in contrast to Fout = 0) approximately apply in most
situations?’ (Zeh 21).
But it is not clear that there is an asymmetry of the kind postulated by
Zeh. First, if we assume that all forces are electromagnetic, then a charge
cannot accelerate, and hence cannot radiate, unless it experiences a non-
zero external field. Second, the incoming field is not zero in ‘most’
© 2008 The Author Philosophy Compass 4/1 (2009): 255–270, 10.1111/j.1747-9991.2008.00192.x
Journal Compilation © 2008 Blackwell Publishing Ltd
264 Philosophical Issues in Electromagnetism

situations. For example, the cosmic background radiation is always present

(see North), and unless we are conducting an experiment in a darkened
room there will also be incoming fields in the visual spectrum. Finally,
every hyperplane, on which F = 0 and which we can use as initial value
surface to determine the fields in its future, can equally be used as a final
value surface to determine the fields in its past. Thus, there seem to be
exactly as many situations in which Fin = 0 as there are situations with
Fout = 0.
One might reply that in many situations in which we are interested the
cosmic background radiation or background radiation from the sun can
be ignored and hence the models physicists use to represent these
situations posit zero incoming fields. Thus, Zeh’s formulation of the
asymmetry might apply to the models that are appropriate in representing
systems of radiating charges rather than to the complete physical situations
modeled. More importantly, perhaps, paradigmatic situations exhibiting
the asymmetry usually only involve only a small number of charges. Thus,
Frisch (‘Tale of Two Arrows’) proposes a way of capturing an asymmetry
that characterizes our representations of systems involving small numbers
of charges.
Another way of characterizing the asymmetry might be this. Consider
once more a single radiating charge in a zero incoming field. Not only is
the free outgoing field generally not zero in such cases, but it also has a
rather special shape in that it consists of a linear combination of the
retarded and advanced fields associated with the charge in question. Now,
in most realistic situations the incoming field will not be zero, but there
nevertheless will be a difference between the free incoming and outgoing
fields in retarded and advanced representations, respectively: the kind of
incoming fields we observe do not have collapsing and rediverging waves
as components that are centered on the trajectory of the source, while the
outgoing field does have diverging waves as its components. (Thus, given
what we know about realistic fields, we could in principle use knowledge
of the ‘free’ outgoing field in an advanced field representation to retrodict
where and when a radiating source was present, but we could not similarly
use our knowledge of the free incoming fields alone to predict where a
source will radiate.) Again, this asymmetry is restricted to situations
involving a small number of charges, for in the case of a very large
number of charges, the retarded and advanced fields of these charges can
combine to result, for example, in approximately constant fields that are
not centered on any of the charges. (One can see this intuitively by
thinking about the fields in the presence of an absorber. In the future
of a perfectly absorbing region the field is zero. Hence, the incoming
field Fin, whatever it may be, is equal to the sum over all absorber particles
of Fadv − Fret.)
How can we explain the asymmetry of radiation fields? Broadly, the
solutions that have been proposed fall into two classes: those that take the
© 2008 The Author Philosophy Compass 4/1 (2009): 255–270, 10.1111/j.1747-9991.2008.00192.x
Journal Compilation © 2008 Blackwell Publishing Ltd
 Philosophical Issues in Electromagnetism 265

radiation asymmetry to be ultimately reducible to considerations from

thermodynamics and statistical physics and those that take the asymmetry
to be an expression of a causal asymmetry.
Frequently writers cite a debate between Ritz and Einstein as an early
instance of the controversy, claiming that Einstein took the asymmetry to
have a thermodynamical origin while Ritz thought that the radiation
asymmetry was more fundamental than that of thermodynamics. But, as
Frisch (Inconsistency, Asymmetry, and Non-Locality) shows, Einstein’s position
is considerably more complex than the standard view suggests. While it is
true that in his joint paper with Ritz Einstein wrote that he ‘believes that
the irreversibility [of radiation processes] is exclusively based on reasons of
probability’ (Ritz and Einstein 324), he also in the very same year claimed
that according to classical wave theory, ‘an oscillating ion produces a diverging
spherical wave’ and that ‘the reverse process does not exist as elementary
process. . . . The elementary process of the emission of light is, thus, not
reversible’ (Einstein 819). The latter quote suggests that Einstein, too,
believed that there was an asymmetry characterizing elementary radiation
processes that was not of thermodynamic origin.
Many writers who try to explain the radiation asymmetry by appealing
to thermodynamic considerations invoke Wheeler and Feynman’s time-
symmetric absorber theory of radiation (Wheeler and Feynman). Wheeler
and Feynman propose that the universe is surrounded by a perfectly
absorbing region and that field associated with a charge is half-retarded
and half-advanced. The fully retarded, diverging fields we observe are
the result of the half-retarded field’s stimulating an advanced response
field in the absorber, which, as Wheeler and Feynman argue, combines
with the half-retarded field of the source to give a fully retarded field.
Fields do not similarly look fully advanced and the symmetry between
retarded and advanced representations is broken, according to them, since
a coherent absorber action in the past is extremely improbably. Huw
Price (Time’s Arrow & Archimedes’ Point) has criticized this argument
for committing what he calls ‘the temporal double-standard fallacy’:
since Wheeler and Feynman postulate a time-symmetric wave associated
with a charge, a coherent absorber response in the past seems to be no
more or less improbable than the future response wave their theory
requires.
There have been attempts to address this problem within Wheeler and
Feynman’s framework (Hogarth; Hoyle and Narlikar), which are criticized
in Frisch (Inconsistency, Asymmetry, and Non-Locality), and there are other
suggestions for how thermodynamic considerations might be marshaled
without an appeal to an infinite absorbers and time-symmetric particle
fields (North; Price, ‘Recent Work on the Arrow of Radiation’). But all
these attempts, it seems to me face the following dilemma. Either they
are in danger of committing the double standard fallacy: any account that
argues that incoming coherent radiation is extremely improbable has to
© 2008 The Author Philosophy Compass 4/1 (2009): 255–270, 10.1111/j.1747-9991.2008.00192.x
Journal Compilation © 2008 Blackwell Publishing Ltd
266 Philosophical Issues in Electromagnetism

be able to explain why outgoing coherent radiation is not similarly

improbable. Or the account ends up presupposing the very asymmetry
they are meant to explain. Thus, North appears to argue that the asymmetry
is due to the fact that the universe began in a state in which there are
large hot sources in a background of fully thermalized, approximately zero
fields. Thermodynamic considerations are then meant to show that such
sources are overwhelmingly likely to be associated with retarded radiation.
But the assumption that initial fields are approximately equal to zero is
already sufficient for the fact that total fields are approximately fully
retarded and there remains no work to be done for thermodynamic
considerations (see Frisch, ‘Tale of Two Arrows’).
It is also not clear that a thermodynamic explanation of the asymmetry
could be general enough. For one, the asymmetry characterizes not only
macroscopic systems of fields and sources, but microscopic systems as well;
and thermodynamic explanations proceed by arguing that a certain
macroscopic evolution is overwhelmingly likely, given that the system in
question is a microstate that is in some sense typical for the system’s
macro-state. Moreover, the radiation asymmetry appears to be analogous
to the following asymmetry that does not have a thermodynamic origin.
Consider a box with a small hole containing a dense gas in an otherwise
empty space. What we expect to observe is a stream of gas particles
exiting through the hole, rather than the time-reverse: a stream of particles
coming in from infinity ‘carefully’ aimed at the hole, resulting in a net
influx of particles into the box. Yet the latter phenomenon is not
excluded on thermodynamic grounds, since we are not dealing with a
closed system.5
Instead of invoking thermodynamic considerations, physics textbooks
usually explain the preference for a retarded representation of the fields by
appealing to considerations of causality (see, e.g., Jackson; Griffiths).
According to a causal interpretation of the interaction between fields and
their sources, retarded and advanced representations of the total field are
mathematically equivalent, but only the retarded representation represents
the physical situation correctly: sources physically contribute retarded
fields to the total field, and thus, only retarded fields are physically real,
even though the linearity of the wave equation ensures that the total field
can always be mathematically represented in terms of advanced fields as
well (Rohrlich, ‘Time in Classical Electrodynamics’; Frisch, Inconsistency,
Asymmetry, and Non-Locality). How does a causal account fare with respect
to the dilemma I presented above? The account has an obvious way of
avoiding the temporal double standard fallacy, since coherently diverging
disturbances of the field in the future of a radiating charge are explained
by the charge’s action as common cause of the disturbances. By contrast,
coherently converging radiation is extremely implausible, unless it, too,
has a common cause explanation in its past. (For example, a coherently
converging wave might be produced by a radiating source in the center
© 2008 The Author Philosophy Compass 4/1 (2009): 255–270, 10.1111/j.1747-9991.2008.00192.x
Journal Compilation © 2008 Blackwell Publishing Ltd
 Philosophical Issues in Electromagnetism 267

of a spherical mirror, which reflects the diverging wave due to the source
back towards the center.) Moreover, the account is not simply question
begging. It explains the asymmetry between different mathematical
representations of the total field involving free incoming and free outgoing
fields, respectively, in terms of the physical asymmetry that field sources
produce diverging field disturbances.
The main criticism of causal accounts is that the notions of a source
‘causing’, ‘producing’, or ‘physically contributing’ retarded fields are ill-
defined and are an instance of scientifically illegitimate philosophy-speak.
In response, defenders of causal accounts can try to show how the notion
can be related to other concepts, such as an asymmetric counterfactual,
and how these notions can account for asymmetries characterizing our
experimental interactions with field sources. What advocates of an explanation
of the radiation asymmetry that appeals to an irreducible notion of
causation cannot do, however, is provide an account of the notion of
cause that will only invoke notions that a ‘Humean’ will find acceptable.
Rather, a causal explanation can only be defended against the criticism of
being ill-defined, by showing that the causal notions are related to a
cluster of other concepts, including those of counterfactuals, interventions,
and experimentation in rigorous and illuminating ways.
There is a second prima facie asymmetry in the theory: particle-equations
of motion that include the effects of radiation on the motion of a charged
particle seem, on first sight, not to be time-reversal invariant. Whether
appearances are correct in this case is a question that is controversially
debated in the literature (see Rohrlich, ‘Causality and the Arrow of
Classical Time’; ‘Time in Classical Electrodynamics’; Rovelli).

6. Locality
There is one final philosophical issue that I want to mention briefly. It is
a widely held view that it is a desideratum that our best scientific theories
postulate only local actions. Most intuitively, perhaps, this demand can be
expressed and motivated in causal terms. There seems to be something
intuitively objectionable about the idea that causes can act where they are
not, and hence, it seems that a cause and a distant effect ought to be
connected by a spatiotemporally continuous sequence of causes. Newton’s
theory of gravity violates this demand, even though Newton himself said
that action-at-a-distance is a ‘great absurdity’ (qtd. in Lange 94). Classical
electrodynamics, by contrast, seems to satisfy the demand, if we interpret
electromagnetic fields realistically.
Indeed, interpretations of classical electrodynamics provide an ideal case
study of how general methodological and metaphysical demands can
influence the interpretation of our theories. Considerations of locality
seem to suggest that fields ought to be interpreted realistically. Yet there is
also a second entity that plays a role in describing fields, the electromagnetic
© 2008 The Author Philosophy Compass 4/1 (2009): 255–270, 10.1111/j.1747-9991.2008.00192.x
Journal Compilation © 2008 Blackwell Publishing Ltd
268 Philosophical Issues in Electromagnetism

potential Aµ. And consideration of observability and determinacy suggest

that the potential not be interpreted realistically and ought to be understood
as a mere calculational device, since the potentials are only determined up
to a so-called ‘gauge transformation’ and the same observable fields can
be represented by many different potentials.
This short gloss ignores a number of philosophically interesting questions
and must be qualified in various important ways. First, one can ask what
more precisely we mean when we demand that a theory be local. Lange
distinguishes several different senses in which a theory can be causally
local. But philosophers and physicists have also proposed locality conditions
that are not couched in causal terms and it is an interesting question to
ask how various causal and non-causal senses of locality might be related
(see Frisch, ‘Non-Locality in Classical Electrodynamics’). Second, if the
assumption that our best theories ought to be local, in some sense, is used
to derive metaphysical implications, we should ask what reasons we can
cite in support of this assumption. Lange argues that it is difficult to find
a fully convincing argument in favor of a causal locality condition, and it
seems to me that the case is even more difficult to make for non-causal
locality conditions, which cannot, for example, appeal to the intuition
that something cannot act where it is not.
Third, it is not entirely clear, whether introducing electromagnetic
fields is enough to ensure that the theory is local. For one, a point-particle
equation of motion that includes radiation reaction effects, the Lorentz-Dirac
equation, on its standard interpretation is causally non-local in that it
predicts that a charge accelerates in response to future fields. Moreover, the
standard picture of locally acting fields comes under pressure from the
semi-classical Bohm-Aharonov effect. This is the effect that quantum
mechanical interference patterns of a beam of electrons passing a very long
solenoid vary with the magnetic field inside of the solenoid, even though
the magnetic field is zero in the regions through which the electron beam
passes. Many philosophers have suggested that this effect forces us to give
up either locality or determinism. One might take the effect to be due to
the fields acting where they are not, and hence give up locality; or one
might argue that the effect shows that potentials ought to be interpreted
realistically, even though the ‘true’ potential is only determined up to a
gauge transformation; and finally one could take the Bohm-Aharanov
effect to point to the reality not of localized values of the potential,
but of non-local loop integrals of the potential, which turn out to
be gauge-invariant (see Healey for a detailed discussion of various
interpretive options).

Acknowledgment
The work for this paper was supported by the National Science Founda-
tion under Grant No. 0646677.
© 2008 The Author Philosophy Compass 4/1 (2009): 255–270, 10.1111/j.1747-9991.2008.00192.x
Journal Compilation © 2008 Blackwell Publishing Ltd
 Philosophical Issues in Electromagnetism 269

Short Biography
Mathias Frisch’s research focuses on issues in the philosophy of science
and the philosophy of physics, including the role of causal reasoning in
science, the nature of scientific explanation, and the role of models in
scientific theorizing. He has published articles in Philosophy of Science, The
British Journal for the Philosophy of Science, and Studies in the History and
Philosophy of Modern Physics, as well as a book, Inconsistency, Asymmetry, and
Non-Locality: A Philosophical Investigation of Classical Electrodynamics (Oxford
University Press), which examines conceptual problems arising in the
interpretation of classical particle-field theories.

Notes
* Correspondence address: Skinner Bldg, College Park, Maryland, United States. Email:
[email protected].
1
The standard graduate- and research-level textbook on classical electrodynamics is Jackson.
Books focusing on the problem of arriving at a classical particle equation of motion are
Rohrlich, Classical Charged Particles; Parrott; Spohn.
2
A summary of Lorentz’s theory of the electron can be found in Lorentz.
3
While the distinction between principle and constructive theories is usually traced to Einstein,
a closely related distinction had earlier been drawn by H. A. Lorentz (see Frisch, ‘Mechanisms,
Principles, and Lorentz’s Cautious Realism’).
4
Since self interactions are infinitesimal for a charged dust, the problem does not arise in this case.
5
Recent Boltzmannian accounts of the thermodynamic asymmetry begin with the assumption
that the world began in a state that is macroscopically special – a state of extremely low entropy
– but microscopically normal. Yet such an assumption does not disallow the incoming stream
of particles, since the system’s initial state with an incoming stream of particles is a macroscopically
distinguishable from one without incoming particle stream – that is, the state can be excluded
on the grounds that it is not microscopically ‘normal’.

Works Cited
Balashov, Yuri and Michel Janssen. ‘Presentism and Relativity’. Br J Philos Sci 54.2 (1 June
2003): 327– 46. doi:10.1093/bjps/54.2.327.
Belot, Gordon. ‘Is Classical Electrodynamics an Inconsistent Theory?’ Canadian Journal of
Philosophy 37.2 (June 2007): 263–282. doi:Article.
Brown, Harvey and Oliver Pooley. ‘Minkowski Space-Time: A Glorious Non-Entity’. The
Ontology of Spacetime. Amsterdam: Elsevier, 2006.
Brown, Harvey R. Physical Relativity: Space-time Structure from a Dynamical Perspective. Oxford:
Oxford UP, 2005.
Einstein, Albert. ‘Über die Entwicklung unserer Anschauung über das Wesen und die Konstitution
der Strahlung’. Physikalische Zeitschrift 10 (1909): 817–25.
Frisch, Mathias. Inconsistency, Asymmetry, and Non-Locality: A Philosophical Investigation of Classical
Electrodynamics. Oxford: Oxford UP, 2005.
——. ‘Inconsistency in Classical Electrodynamics’. Philosophy of Science 71.4 (2004): 525– 49.
doi:10.1086/423627.
——. ‘Laws and Initial Conditions’. Philosophy of Science 71.5 (2004): 696–706. doi:10.1086/
425056.
——. ‘Mechanisms, Principles, and Lorentz’s Cautious Realism’. Studies in History and Philosophy
of Science Part B: Studies in History and Philosophy of Modern Physics 36.4 (December 2005):
659 –79. doi:10.1016/j.shpsb.2005.07.002.

© 2008 The Author Philosophy Compass 4/1 (2009): 255–270, 10.1111/j.1747-9991.2008.00192.x

Journal Compilation © 2008 Blackwell Publishing Ltd
270 Philosophical Issues in Electromagnetism

——. ‘Non-Locality in Classical Electrodynamics’. Br J Philos Sci 53.1 (March 1 2002): 1–19.
doi:10.1093/bjps/53.1.1.
——. ‘A Tale of Two Arrows’. Studies in History and Philosophy of Science Part B: Studies in
History and Philosophy of Modern Physics 37.3 (September 2006): 542–58. doi:10.1016/
j.shpsb.2005.03.004.
Griffiths, David J. Introduction to Electrodynamics. 3rd ed. Upper Saddle River, NJ: Prentice Hall,
1999.
Healey, Richard. Gauging What’s Real: The Conceptual Foundations of Gauge Theories. New York,
NY: Oxford University Press, 2007.
Hogarth, J. E. ‘Cosmological Considerations of the Absorber Theory of Radiation’. Proceedings
of the Royal Society of London A267 (1962): 365 – 83.
Hoyle, F. and J. V. Narlikar. ‘Cosmology and Action-at-a-Distance Electrodynamics’. Reviews
of Modern Physics 67 (1995): 113 – 55.
Jackson, John David. Classical Electrodynamics. 3rd ed. New York, NY: Wiley, 1999.
Janssen, Michel. ‘Drawing the Line between Kinematics and Dynamics in Special Relativity’.
2008. <http://philsci-archive.pitt.edu/archive/00003895/>.
Kuhn, Thomas S. The Structure of Scientific Revolutions. 3rd ed. Chicago, IL: U Of Chicago P,
1996.
Lange, Marc. An Introduction to the Philosophy of Physics: Locality, Fields, Energy and Mass. Oxford:
Blackwell, 2002.
Lorentz, H. A. The Theory of Electrons and Its Applications to the Phenomena of Light and Radiant
Heat. 2nd ed. Leipzig: B. G. Teubner, 1916.
Muller, F. A. ‘Inconsistency in Classical Electrodynamics?’ Philosophy of Science 74.2 (April
2007): 253–77. doi:10.1086/520942.
North, Jill. ‘Understanding the Time-Asymmetry of Radiation’. Philosophy of Science 70.5
(2003): 1086 – 97. doi:10.1086/377391.
Parrott, Stephen. Relativistic Electrodynamics and Differential Geometry. New York, NY:
Springer-Verlag, 1987.
Price, Huw. ‘Recent Work on the Arrow of Radiation’. Studies in History and Philosophy of
Science Part B: Studies in History and Philosophy of Modern Physics 37.3 (September 2006): 498–
527. doi:10.1016/j.shpsb.2006.03.004.
——. Time’s Arrow & Archimedes’ Point: New Directions for the Physics of Time. New York, NY:
Oxford UP, 1996.
Ritz, Walter and Albert Einstein. ‘Zum gegenwärtigen Stand des Strahlenproblems’. Physikalische
Zeitschrift 10 (1909): 232 – 324.
Rohrlich, Fritz. ‘Causality and the Arrow of Classical Time’. Studies in History and Philosophy
of Science Part B: Studies in History and Philosophy of Modern Physics 31.1 (March 2000): 1–13.
doi:10.1016/S1355-2198(99)00030-1.
——. Classical Charged Particles. 3rd ed. Singapore: World Scientific, 2007.
——. ‘Realism Despite Cognitive Antireductionism’. International Studies in the Philosophy of
Science 18.1 (2004): 73. doi:10.1080/02698590412331289260.
——. ‘Time in Classical Electrodynamics’. American Journal of Physics 74.4 (2006): 313 –15.
—— and Larry Hardin. ‘Established Theories’. Philosophy of Science 50 (December 1983): 603–
17.
Rovelli, Carlo. ‘Comment on: “Causality and the Arrow of Classical Time”, by Fritz Rohrlich’.
Studies in History and Philosophy of Science Part B: Studies in History and Philosophy of Modern
Physics 35.3 (September 2004): 397–405. doi:10.1016/j.shpsb.2004.02.005.
Spohn, Herbert. Dynamics of Charged Particles and Their Radiation Field. Cambridge: Cambridge
UP, 2004.
Vickers, Peter. ‘Frisch, Muller, and Belot on an Inconsistency in Classical Electrodynamics’. The
British Journal for the Philosophy of Science 59.4 (2008): 767–92. doi: 10.1093/bjps/axn039.
Wheeler, John A. and Richard P. Feynman. ‘Interaction with the Absorber as the Mechanism
of Radiation’. Reviews of Modern Physics 17 (1945): 157– 81.
Wilson, Mark. Wandering Significance: An Essay on Conceptual Behaviour. New York, NY: Oxford
UP, 2006.
Zeh, H. D. The Physical Basis of the Direction of Time. 4th ed. Berlin: Springer, 2001.

© 2008 The Author Philosophy Compass 4/1 (2009): 255–270, 10.1111/j.1747-9991.2008.00192.x

Journal Compilation © 2008 Blackwell Publishing Ltd