Stud. Hist. Phil. Mod. Phys., Vol. 31, No. 2, pp.

201}226, 2000
 2000 Elsevier Science Ltd. All rights reserved.
Printed in Great Britain
1355-2198/00 $ - see front matter

The Coordinate-Independent 2-Component

Spinor Formalism and the Conventionality of
Simultaneity
Jonathan Bain*
In recent articles, Zangari (1994) and Karakostas (1997) observe that while
an e-extended version of the proper orthochronous Lorentz group Ot (1,3)
>
exists for values of e not equal to zero, no similar e-extended version of its
double covering group SL(2, C) exists (where e"1!2e , with e the
0 0
non-standard simultaneity parameter of Reichenbach). Thus, they maintain,
since SL(2, C) is essential in describing the rotational behaviour of half-
integer spin "elds, and since there is empirical evidence for such behaviour,
e-coordinate transformations for any value of eO0 are ruled out empiric-
ally. In this article, I make two observations:
(a) There is an isomorphism between even-indexed 2-spinor "elds and
Minkowski world-tensors which can be exploited to obtain generally
covariant expressions of such spinor "elds.
(b) There is a 2-1 isomorphism between odd-indexed 2-spinor "elds and
Minkowski world-tensors which can be exploited to obtain generally
covariant expressions for such spinor "elds up to a sign. Evidence that
the components of such "elds do take unique values is not decisive in
favour of the realist in the debate over the conventionality of simultaneity
in so far as such "elds do not play a role in clock synchrony experiments
in general, and determinations of the one-way speed of light in particular.
I claim that these observations are made clear when one considers the coordi-
nate-independent 2-spinor formalism. They are less evident if one restricts
oneself to earlier coordinate-dependent formalisms. I end by distinguishing
these conclusions from those drawn by the critique of Zangari given by Gunn
and Vetharaniam (1995).  2000 Elsevier Science Ltd. All rights reserved.
Keywords: Spacetime; Conventionalism; Simultaneity; Spinors; Formalism

* Department of Humanities and Social Sciences, Polytechnic University, Brooklyn, NY 11201,

U.S.A. (e-mail: jbain@duke.poly.edu).

PII: S 1 3 5 5 - 2 1 9 8 ( 9 9 ) 0 0 0 3 3 - 7
201
202 Studies in History and Philosophy of Modern Physics

1. Introduction

Recently a new twist has been given to a long-standing debate in philosophy of

spacetime, the debate over the conventionality of simultaneity (CS, hereafter) in
Special Relativity. In recent articles, Zangari (1994) and Karakostas (1997)
observe that while an e-extended version of the proper orthochronous Lorentz
group Ot (1, 3) exists for values of e not equal to zero, no similar e-extended
>
version of its double covering group SL(2, C) exists (where e"1!2e , with
0
e the non-standard simultaneity parameter of Reichenbach). Thus, they main-
0
tain, since SL(2, C) is essential in describing the rotational behaviour of half-
integer-spin "elds, and since there is empirical evidence for such behaviour,
e-coordinate transformations for any value of eO0 are ruled out empirically.
Gunn and Vetharaniam (1995) respond to Zangari by claiming that the exist-
ence of half-integer-spin "elds does not force restriction to standard coordinate
charts. They substantiate this claim by formulating the Dirac equation in
e-coordinates. Furthermore, they suggest that Zangari con#ates the internal
symmetries of spin space with the spacetime symmetries of the PoincareH
group. Finally, they claim that, regardless, the SL(2, C) &complex representation'
of spacetime points, while simpler than the Ot (1, 3) representation, does
>
not admit a parity operator. What the latter requires is the full Lorentz group
O(1, 3).
In this article, I contend "rst that Zangari and Karakostas base their critique
of the CS thesis on a coordinate-dependent description of 2-component spinors
(found in most of the early literature on the subject) and fail to do justice to
coordinate-independent techniques. This is signi"cant in so far as the coordi-
nate-independent 2-component spinor (alternatively &2-spinor') formalism indi-
cates three points. First, 2-spinor "elds, just like tensor "elds, can be considered
geometrical objects independent of particular coordinate representations of
them. Second, the 2-spinor formalism makes clear the relation between 2-spinor
"elds and general tensor "elds: even-indexed 2-spinor "elds can be put into a 1-1
correspondence with Minkowski world-tensors; whereas odd-indexed 2-spinor
"elds can be put into a 2-1 correspondence with Minkowski world-tensors.
Since Minkowski world-tensors can always be extended to general tensors, it is
evident that the information contained in even-indexed 2-spinor "elds can be
represented in arbitrary general linear coordinates (i.e. such "elds can be given
tensorial, generally covariant expressions), whereas the information contained
in odd-indexed 2-spinor "elds can only be given in arbitrary coordinates up to
a sign.
Third and "nally, the coordinate-independent description of 2-spinors makes
clear the basis of Zangari's and Karakostas's critique of the CS thesis; namely,

 In this article, the abstract index notation is in force with the following quali"cations: lower-case
Latin indices denote tensor "elds, and lower-case Greek indices denote their components.
Upper-case Latin indices denote spinor "elds, and upper-case Greek indices denote their
components.
¹he Coordinate-Independent 2-Component Spinor Formalism 203

that the existence of 2-spinor "elds requires spacetime to possess a particular

structure, viz that given by a global Minkowski tetrad "eld. This last point is
obscured in the coordinate-dependent description of 2-spinors. I will claim that
it is not decisive in the CS debate, in so far as it amounts to a variant of a typical
realist argument against the CS thesis that appeals to criteria of epistemic
warrant (unifying power, simplicity, etc.) that a traditional conventionalist will
not accept. I thus agree with Gunn and Vetharaniam's overall conclusion, but
for di!erent reasons. Moreover, I shall argue that some of the particulars of
Gunn and Vetharaniam's critique are a bit misleading.
In Section 2, I review the debate over the Conventionality of Simultaneity. In
Sections 3 and 4, I review the signi"cance of the group SL(2, C) in physical
theories and indicate the nature of the relation between 2-spinors and tensors. In
Sections 5 and 6, I criticise Zangari's and Karakostas's argument against CS and
indicate the options for the conventionalist. Finally, in Section 7, I criticise
Gunn and Vetharaniam's contribution to this debate.

2. The CS Debate

Recall that the CS debate centres on the value of Reichenbach's simultaneity

parameter e . Spacetime realists claim that the value e "1/2 is uniquely
0 0
speci"ed by (at the least) the conformal structure of Minkowski spacetime.
Hence two clocks A and B may be judged to be in synchrony just when their
readings are identical when both clocks lie on the same spacelike hypersurface.
For the realist, such synchrony may be established operationally in the follow-
ing manner: emit a light signal from clock A to clock B at time t (as judged by
?
A). Upon reception at clock B at reading t (as judged by B), re#ect it back to A.
@
Record the reception time t  at A. Then clocks A and B are judged to be in
?
synchrony just when
t "t #(t  !t ). (1)
@ ?  ? ?
Conventionalists, on the other hand, insofar as they reject a priori ontological
commitments to spacetime structure, claim that equation (1) presupposes that
the one-way speed of light is independent of direction; i.e. that the speed c of
>
light from clock A to clock B is identical to the speed of light c from B to A;
\
 The conventionality of simultaneity thesis should be made distinct from the relativity of simultane-
ity. The latter follows explicitly from the metrical structure of Minkowski spacetime (M, g ) and
?@
describes the fact that the relation of simultaneity between events in (M, g ) is de"ned uniquely only
?@
relative to a given inertial reference frame. In general, observers in di!erent inertial frames will not
agree on judgements of simultaneity. The conventionality thesis claims that the simultaneity relation
is not uniquely speci"ed within a given inertial frame and can only be given by imposing one of
a continuum of conventional choices (viz 0(e (1). For recent reviews of the debate see Janis
0
(1998), Anderson et al. (1998), and Norton (1992).
 Malament (1977) demonstrates that standard simultaneity is the only non-trivial equivalence
relation that can be implicitly de"ned in terms of the light cone structure of Minkowski spacetime
and the temporally-oriented worldline of an inertial observer. Sarkar and Stachel (1999) demon-
strate that additional &null-cone' simultaneity relations can be de"ned if one gives up time symmetry.
204 Studies in History and Philosophy of Modern Physics

and this cannot be ascertained without a prior determination of synchrony

between A and B. Without this presupposition, (1) must be written as
t "t #e (t  !t ), (2)
@ ? 0 ? ?
with the value of e being chosen by convention and having the value 1/2 just
0
when it is stipulated that c "c .
> \
Equation (2) de"nes an e-coordinate transformation given by
xG "xG, i"1,2,3,
C (3)
x"x!xG(1!2eG ), eG "(e , e , e ),
C A 0 0 0 0 0
where, in general, e depends on spatial direction. De"ning eG"(1!2eG ), (3)
0 0
can be written as (Zangari, 1994, p. 269)



1 e e e cx
  
0 1 0 0 x
xI"N(eG)I xJ" , (4)
C J 0 0 1 0 x
0 0 0 1 x
with " eo "(1. If the N(eG) are composed with a Lorentz transformation, one
obtains the &e-extended' homogeneous Lorentz group, call it e-O(1, 3), with
elements
K "NKN\. (5)
C
This amounts to an e-coordinate-induced similarity transformation on K. If the
invariants of a transformation group of a theory are taken as the observables of
the theory, the conventionalist can claim that the invariants of e-O(1, 3) are
observationally indistinguishable from the invariants of O(1, 3). For example,
under (5), all invariants of O(1, 3) map onto invariants of e-O (1, 3). In particular,
we have ds"c(dx )!(dxG)"ds. Hence, making use of (3),
 C
2eG (eG)!1
ds"(dx)# dxG dx# (dxG ), (6)
C C c C C c C
with associated metric
(g ) "1, (g ) "eG, (g ) "eGeHdGH!dGH. (7)
C  C G C GH
The realist in the CS debate will view the e-coordinate transformation (3)}(5)
as a passive coordinate transformation, and the &e-extended' Lorentz group
e-O(1, 3) obtained by the induced similarity transformation (5) as just O(1, 3)
in kooky coordinates.

 Suppose the distance between clocks A and B is d. Then the round-trip speed of light c is given by
c"2d/(t  !t ). Hence c "d/(t !t )"c/2e , and c "d/(t  !t )"c/2(1!e ). Explicitly,
? ? > @ ? 0 \ ? @ 0
then, knowledge of c requires prior knowledge of e .
! 0
 Suppressing indices, a homogeneous Lorentz transform K acts on 4-vectors x according to
x"Kx. Hence x "Nx"NKx,K x "K Nx, for any x. Thus K "NKN\.
C C C C C
¹he Coordinate-Independent 2-Component Spinor Formalism 205

In general, the conventionalist is a type of semantic anti-realist. Such a crea-

ture holds that, in the light of multiple intertranslatable descriptions of a phys-
ical system P that agree on all observational aspects of P, but do not agree on
in-principle unobservable aspects of P, any particular descriptive framework can
only be chosen by convention. In the CS context, the multiple descriptive
frameworks in question are coordinate charts, the physical system P consists of
the determination of clock synchrony, the in-principle unobservable aspect of
P is the one-way speed of light, and intertranslatability is secured by (3)}(5).

3. SL (2, C) and the Lorentz Group O(1, 3)

In this section, I review the relation between the group SL(2, C) (generated by
linear 2;2 complex matrices with determinant"1) and the Lorentz group
O(1, 3). Readers already familiar with this standard material may skip directly
to Section 4.
I "rst describe how the group SL(2, C) forms a 2-1 faithful representation of
the restricted Lorentz group Ot (1, 3). In other words, there exists a group
>
homomorphism o: SL(2, C)POt (1, 3) such that to every element K 3Ot (1, 3)
>  >
there corresponds exactly two elements $A3SL(2, C). I now brie#y indicate
how o is constructed.
Let M be Minkowski vector space and Herm(2) be the collection of all 2;2
complex Hermitian matrices. Any matrix H3Herm(2) can be expanded as


x#x x#ix
H"xp #xp #xp #xp "xIp " , (8)
    I x!ix x!x

where the xI are real and the p are the Pauli matrices. This de"nes a map
I
S: xI C xIp , from R to Herm(2). Under this correspondence, the determinant
I
of H is just the Lorentz length of xI:

det S(xI)"g xIxJ, (9)

IJ
where g are Minkowski metric components. The left-hand side of (9) is
IJ
preserved under transformations of the type ¸ : Herm(2)PHerm(2) given

 Recall that the Lorentz group O(1, 3) with elements K? has 4 components, O(1, 3)"Ot (1,3)
@ >
6Os (1, 3)6Ot (1, 3)6Os (1, 3), where $ indicates a determinant of $1 and !( ) indicates K '
> \ \ 
(() 0. Those components with det"#1 are referred to as proper (and alternatively denoted by
SO(1, 3)). The proper components are connected with the identify. Time- and space-re#ections are
given by elements of Os (1, 3) and Ot (1, 3), respectively.
> >
 Explicitly,





1 0 0 1 0 !i 1 0
p " , p " , p " , p " .
 0 1  1 0  i 0  0 !1
These form a basis for Herm(2). Hence (8) is basis-dependent description of the elements of Herm(2).
206 Studies in History and Philosophy of Modern Physics

explicitly by
¸ (H),H"AHAR, (10)

where A3SL(2, C) is a 2;2 complex matrix with unit determinant. The map
¸ thus induces a transformation on x? that preserves its length; i.e. it induces

a homogeneous Lorentz transformation K : MPM given by K "S\¸ S
  
and acting on elements of M according to H"xIp , where xI"KI xJ. The
I J
explicit form of K is given by

(K )I "tr(pIAp AR). (11)
 J J
This de"nes the map o: SL(2, C)POt (1, 3). It is to the proper orthochronous
>
component of O(1, 3) since SL(2, C) is connected to the identity and, upon
explicit calculation using (11), (K ) '0 for any A3SL(2, C). It is a group
 
homomorphism (i.e. it preserves group multiplication) since K K "

S\¸ ¸ S"S\¸ S"K . Moreover, it is onto (viz, faithful), since both
  
SL(2, C) and Ot (1, 3) are six-parameter groups. Finally, it is 2-1, since for any
>
A, B3SL(2, C), if K "K , then A"$B. (To see this, suppose A, B3SL(2, C)

and K "K . Then AB\3SL(2, C) and K \ "K K \ "K (K )\"
   
K (K )\"I (the identity element of Ot (1, 3)). By (11), it follows that
  >
AB\"$I (identity on SL(2, C)). Hence A"$B.) Thus elements of SL(2, C)
form a 2-valued faithful representation of the proper orthochronous Lorentz
group.
Furthermore, since SL(2, C) is simply connected and Ot (1, 3) is doubly
>
connected, the homomorphism o is a covering map, identifying SL(2, C) as the
universal covering group of Ot (1, 3). This topological property turns out to
>
have physical signi"cance. It entails that, under a rotation of 2p, carriers of
representations of SL(2, C) change sign, whereas carriers of representations of
Ot (1, 3) do not, as can be seen by the following considerations.
>
From (11), a path from positive identity I to !I in SL(2, C) corresponds to
a path from I to I in Ot (1, 3). In SL(2, C), such a path is not homotopic (i.e.
>
continuously deformable) to the trivial curve (since it does not close). If the path
is continued back to I in SL(2, C), then, since SL(2, C) is simply connected, the

 Suppose H3Herm(2). Then ¸ (H)R"(AHAR)R"(AR)RHRAR"AHAR"¸ (H), hence ¸ (H)3

  
Herm(2). Furthermore, ¸ preserves the determinant, det(AHAR)"det(A)det(H)det(AR)"det(H).

 To derive this, note that, from (8), H"xIp "(K )I xJp "AxIp AR"xIAp AR. Identifying
I  J I I I
coe$cients of xI yields (K )I p "Ap AR. Multiplying both sides of this latter with p and taking
 J I J M
the trace then produces tr(p (K )I p )"(K )I tr(p p )"(K )I (2d )"tr(p Ap AR). Equation
M  J I  J M I  J MI M J
(11) then follows.
 Let G and H be groups. Then a representation of G is a map D: GPH which is a group
homomorphism; i.e. D(g)D(g)"D(gg), ∀g, g3G, D(e)"identity on H, for e the identity on G.
Furthermore, for later reference, if G and H are matrix groups, and H is of order n and <L is an
n-dimensional vector space, then by selecting a basis for <L, elements of H can be considered linear
transformations (for changes of basis matrices) on <L. <L is then referred to as the representation
space of G and elements of <L are referred to as carriers of the representation of G.
 For lucid accounts of the topological concepts involved, see the discussions in Wald (1984, pp.
344}346), Naber (1992, Appendix B), and Penrose and Rindler (1984, pp. 41}46).
¹he Coordinate-Independent 2-Component Spinor Formalism 207

path is now homotopic to the trivial curve. The corresponding path in

Ot (1, 3) retraces itself. To see this concretely, note that a series of rotations
>
given by


ei h 0
$A" , (12)
0 e!ih

as h ranges from 0 to 2p, corresponds to the "rst path described above, from
I to !I and from I t> to I t> . (Acting on (8) via (10), the
1* ! 1* ! -   -  
matrix A above rotates x? by h in the (x , x )-plane.) Extending the range to
 
0)h)4n corresponds to the second path, homotopic to the trivial curve in
both SL(2, C) and Ot (1, 3). Hence carriers of representations of SL(2, C) change
>
sign under 2p rotations, whereas carriers of Ot (1, 3) do not.
>
Note that the carriers of the representations of SL(2, C) and Ot (1, 3) are the
>
main concern here (see footnote 10). They are the physical/geometrical objects
of concern to the physicist. It is the components of these objects with respect to
a given coordinate chart that are measured. Ontologically, they may be con-
strued in one of two ways. One may consider the carriers to be real, independ-
ently of coordinate chart representation or decomposition in a given frame "eld.
Alternatively, one may consider only the values of the components of the
carriers in a given coordinate chart to be real (i.e. what is real is what actually
gets measured). In this coordinate-dependent reading, the carrier itself is con-
sidered no more than an array of the values of its components relative to a given
coordinate chart that transforms between charts under a particular transforma-
tion rule. To see this more concretely, I will now indicate how an intrinsic,
coordinate-independent description of the carriers of SL(2, C), i.e. 2-spinors,
may be given. This will then be compared to a coordinate-dependent descrip-
tion.

 More abstractly, one need not restrict talk to matrix representations of SL(2, C) and carriers of
matrix representations; one can talk of general representations. Recall "rst that the homogeneous
Lorentz group SO(1, 3) is generated by 3 boosts K and 3 rotations J. De"ning new generators
A"(J#iK), B"(J!iK), it follows that [A , A ]"ie A , [B , B ]"ie B , [A , B ]"0
  G H GHI I G H GHI I G H
(i, j"1, 2, 3). Hence A, B each generate SU(2). Thus any representation of SO(1, 3) can be labelled by
two angular momenta (A, B), the "rst associated with A and the second with B. (Note, however, that
since SO(1, 3) is non-compact, its "nite-dimensional representations are not unitary; so while the
4-dimensional compact rotation group SO(4) decomposes as SO(4)"SU(2)SU(2), the homogene-
ous Lorentz group decomposes as SO(1, 3)"SL(2, C)SL(2, C).) In general, a "eld that transforms
according to the (A, B) representation of SO(1, 3) has components that rotate like objects with spins
j"A#B, A#B!1,2, "A!B". In particular, a scalar "eld transforms like the (0, 0) representa-
tion, a 4-vector "eld transforms according to (, ) (since it has a scalar (spin"0) time component
 
plus a 3-vector (spin"1) component), and a 2-component spinor "eld transforms either according
to the (, 0) or the (0, ) representation. When SO(1, 3) is extended by adding to it a parity
 
transformation, these two representations can no longer be considered separately (parity trans-
formations interchange them). One is thus led to the direct sum (, 0)(0, ), which is invariant under
 
parity and constitutes an irreducible representation of SO(1, 3) extended by parity (i.e. ISO(1, 3)).
This latter is the Weyl representation of a Dirac 4-spinor (see footnote 13 below).
208 Studies in History and Philosophy of Modern Physics

4. Coordinate-Independent Description of 2-Spinors

The representation space of the matrix representations of SL(2, C) considered

above is a 2-dimensional complex linear vector space S endowed with a bilinear
skew-symmetric 2-form 3. The structure (S, 3) is often referred to as spin space
and abbreviated simply by S. An element i of S (i.e. a carrier of the 2;2 matrix
representations of SL(2, C)) is referred to as a 2-component spinor. Its
Hermitian conjugate iY is an element of the 2-dimensional complex vector
space denoted S. An element kY of S is distinguished from an element of S by
the primed index convention. The spaces S and S are distinct: Hermitian
conjugation does not identify S with S insofar as it de"nes an anti-isomorphism.
For i, k3S and j3C, conjugation takes i#jk into i#jM k, rather than
i#jk. In addition to the vector spaces S, S, one can form the dual spaces
SH, SH with elements u , n , distinguished by lowered indices. They are anti-
 Y
linear maps from S and S into C.
The bilinear skew-symmetric 2-form 3 is a map from S;SPC and can be
identi"ed with an element 3 of SH;SH. It amounts to a skew-symmetric

&metric' on S, identifying S with its dual SH by means of the mappings
i C i "3 i and i C i!"3! i , where 3 3 !"d ! is the identity
 !  
element on S. Hence the 3-symbol raises and lowers spinor indices (the stan-
dard convention uses contraction over the "rst index of 3 to lower indices

and contraction over the second index to raise indices). The linear transforma-
tions A: SPS that preserve 3 are elements of SL(2, C). They act on

elements of S via i"A i . A general spinor of valence (p, q; r, s) is an
element of the tensor product of the four spaces S, SH, S, SH:

U2 Y2 Y 2 3(S).;(SH)O;(S)P;(SH)Q, (13)

!2" !Y "Y
where (S). is the tensor product of p copies of S, etc. The standard techniques of
tensor algebra can now be employed to generate relations between spinors of
various valences.

 Alternatively, a univalent spinor, or a Weyl spinor. Such 2-spinor elements of S should be

distinguished from the 2-spinors that appear in non-relativistic quantum mechanics and that encode
the two degrees of freedom of the non-relativistic wave function associated with its spin along
a particular axis (&up' or &down'). The 2-component spinors of S are carriers (of matrix representa-
tions) of SL(2, C); non-relativistic 2-component spinors are carriers of SU(2), the 2-fold covering
group of SO(3). The move from SU(2) spinors to SL(2, C) spinors may be associated heuristically
with the move from Galilean-invariant quantum mechanics to Lorentz-invariant quantum mechan-
ics. The &relativistic' spinors that usually appear in the latter are Dirac spinors t?, a"0, 1, 2, 3.
These are 4-component elements of the direct sum space SHS: t?"(iY, k ) (see equations (22)

below). Note, "nally, that there is nothing inherently quantum mechanical about spinors (of either
the SU(2) or the SL(2, C) variety). Indeed, SL(2, C) spinors have been used to great extent in classical
general relativity.
 In particular, if the determinant of A is de"ned by det(A)"3 3!"A A , the det(A)"1
 ! "
entails A A 3 "3 . This is analogous to the de"ning property of the homogeneous Lorentz
! "  !"
group O(1, 3) given by K? K@ g "g , where K3O(1, 3) and g is the Minkowski metric.
A B ?@ AB ?@
¹he Coordinate-Independent 2-Component Spinor Formalism 209

No mention of coordinates has been made up to this point. Equation (13)

de"nes a general valence spinor as a de"nite geometrical object in a coordinate-
independent manner. Just as with the standard tensor formalism, a coordi-
nate-dependent description of spinors can also be used. A spin basis for S con-
sists of two orthonormal (with respect to 3 ) univalent 2-spinors (o, n).

Taken together, they are sometimes referred to as a spin basis and can be
schematically represented by +3R ,, where R"0, 1 labels the basis and A is the
abstract index label for a (1, 0; 0, 0) 2-spinor. The components of a (p, q; r, s)
2-spinor then can be obtained by expanding the 2-spinor in the appropriate spin
basis:
UR 2RN C 2CO R 2RP C 2CQ "U2 Y2 Y 2 +3R , 2+3C ,"Y. (14)
!2" !Y "Y  Q

Note here that the components on the left are, in general, complex-valued
functions; i.e. they are scalar quantities that take values in the coordinate chart
adapted to the bases. The 2-spinor U2 2 Y2 2 and the bases, on the other
! !Y
hand, are de"nite, coordinate-independent geometrical objects. Since the bases
are invariant under SL(2, C), the components transform between spin bases
according to the rule
A(U)R 2R. R 2RP C 2CO C 2CQ "AR D 2AM P1 C1 UD 2DN D 2DP P 2PO P 2PQ , (14)
where the A are SL(2, C) transformations.
Similarly, a time-oriented orthonormal basis in Minkowski vector space is
called a tetrad basis and consists of a set of 4 mutually orthonormal (with
respect to g ) vectors. Such a tetrad is given by +e ,?, where k"0}3 labels the
?@ I
basis and the abstract index &a' indicates a vector. The components of a (p, q)
Minkowski world-tensor ¹ ?2 2 (to be distinguished from a general tensor)
U A
relative to a tetrad basis are given by expanding the world-tensor in the basis:
¹ J 2JN  2 O "¹ ?2@ 2 +eJ , 2+e O ,B. (15)
U I I U A B ? I
They transform between tetrad bases according to the rule
K(¹ )J 2JN  2 O "KJ  2(K\)NO O ¹ M2 2 O , (15)
U I I M I U N
where the K are homogeneous Lorentz transformations. Again, one should
clearly distinguish between the coordinate representation of the world-tensor,

 Alternatively, the general valence spinor U2 2 Y2 Y 2 may be viewed as a multilinear map
! " !Y "Y
U: (S).;(SH)O;(S)P;(SH)QPC (for spinor "elds, replace C with the space of complex-valued
functions). Note that, strictly speaking, spinor "elds cannot be considered geometrical object "elds
on the standard de"nition of the latter (see e.g. Trautman, 1965, p. 86). I am considering a widening
of this de"nition to allow in general cross sections of relevant "ber bundles to count as geometrical
object "elds (see for instance Torretti (1983, p. 98) and Stachel (1992, 1986).
 Hence 3 "o, 3 "n. The normalisation condition requires 3 on "1. Penrose and
  
Rindler (1984) use the term &spin basis' to refer to such a pair of 2-spinors, reserving the term &dyad
basis' for an unnormalised pair that satis"es 3 on "j, for arbitrary complex j.

 Here and below I distinguish a world-tensor with a &w' subscript. World-tensors are carriers of
matrix representations of Ot (1, 3). General (4-dimensional) tensors are carriers of matrix repre-
>
sentations of GL(4, R), the 4;4 general linear group over the reals.
210 Studies in History and Philosophy of Modern Physics

given by its components ¹ J 2JN  2 O relative to coordinates adapted to

U I I
a tetrad basis; and the coordinate-independent geometrical objects ¹ ?2 2 and
U A
+e ,?.
I
The early literature on spinors de"ned a 2-spinor simply as an array of
complex numbers that transforms according to (14). Such literature often
stresses that, in general, the As in (14) cannot be replaced by arbitrary non-
degenerate linear transformations, whereas the Ks in (15) can. Hence, whereas
a Minkowski world-tensor can be extended to a general tensor whose compo-
nents can take values in arbitrary, general linear coordinate charts, a 2-spinor
cannot be so-extended (e.g. Gel'fand et al. (1963, p. 252), Cartan (1966, pp.
150}151)). This is a consequence of the fact that, while Ot (1, 3) can be embedded
>
in the general linear group GL(4, R), SL(2, C) cannot, due to its topological
properties, as Karakostas (1997, p. 261) points out.
At this point, three things should be made clear concerning the fact that
SL(2, C) cannot be embedded in GL(4, R).
First, the fact that SL(2, C) cannot be embedded into GL(4, R) does not entail
that GL(4, R) has no spinorial (i.e. double-valued) representations. Ne'eman and
Sijacki (1987) have shown that GL(4, R) does have a double covering, denoted
by GL(4, R), which admits in"nite dimensional representations. Ne'eman and
Sijacki (1997) and Sijacki (1998) are recent reviews canvassing ways in which
physical "elds can be described using such in"nite dimensional spinorial repre-
sentations of GL(4, R). Their work indicates two options for the physicist:

 Note that the distinction between a coordinate chart and a basis (or frame) should be made here,
and in the spinor case above. In the general case of an n-dimensional manifold M, a coordinate chart
is a pair (;,
) where ;LM is a region of M and
: ;PRL is a map that labels each point in
; with an n-tuple of real numbers +xI,, k"12n. A frame "eld, on the other hand, is in general a set
of linearly independent vector "elds on M. (In the spinor case above, the frame "eld consists of
2-spinor "elds). If the "elds of the set mutually commute, then a coordinate chart can be adapted to
them by taking their integral curves as coordinate curves (if they do not mutually commute, they
constitute a non-coordinate, or non-holonomic, basis). Conversely, to every coordinate chart
characterised by the n-tuple +xI, one can adapt a frame "eld by the identi"cation e ?"(*/* I)?. The
I V
coordinates adapted to a Minkowski tetrad "eld are Minkowski inertial (standard) coordinates.
 I thank Michael Redhead for bringing this reference to my attention. That GL(4, R) admits
(in"nite dimensional) spinorial representation is not made all that clear in the physics literature. Two
representative statements are:
Like SO(n), the general linear group GL is not simply connected. However, its universal
L
covering group has no linear representation other than GL representations. This is why
L
physicists tried in vain for some time to de"ne spinors in curved space using Einstein's gauge
(GloK ckeler and SchuK cker, 1987, p. 191).

Since a general, curved spacetime possesses no isometries or any other preferred classes of
di!eomorphisms and since even in Minkowski spacetime there is no natural action of the full
group of di!eomorphisms on spinor "elds, we cannot expect to de"ne a &transformation law'
of the type (2.2.10) [i.e. a general linear transformation] under di!eomorphisms for spinor
"elds in curved spacetimes (Wald, 1984, p. 360).
These statements are correct if by &spinor' is meant &"nite dimensional representation of SL(2, C)'.
Cartan's (1996, p. 151) &no-go' theorem is in fact restricted to the "nite representation case.
¹he Coordinate-Independent 2-Component Spinor Formalism 211

(a) Stick with "nite dimensional representations to describe physical "elds.

Thus we may use "nite dimensional representations of GL(4, R) (i.e. stan-
dard general tensor "elds) for integer-spin "elds, and "nite dimensional
representations of SL(2, C) (i.e. 2-spinor "elds) for half-integer-spin "elds.
The components of 2-spinor "elds cannot take values in general linear
coordinate charts because SL(2, C) cannot be embedded in GL(4, R).
(b) Use in"nite dimensional representations of GL(4, R) (what Ne'emen and
Sijacki call &mani"elds') to represent physical "elds of both integer and
half-integer spin. Since there is a natural action of Di!(4, R) (the group of
di!eomorphism of R) on these, we can use arbitrary general linear coordi-
nates to describe them.
Option (b) e!ectively refutes Zangari and Karakostas's argument against the CS
thesis. In brief (see Sections 5 and 6 below), they may be taken to argue that:
(1) The components of 2-spinor "elds (i.e. "nite dimensional representations of
SL(2, C)) do not take values in e-coordinates;
(2) Half-integer-spin "elds must be represented by 2-spinor "elds;
(3) Therefore, the existence of half-integer-spin "elds prohibits the use of e-
coordinates.
Option (b) indicates that (2) is false. In the rest of this paper, I shall argue that
even if one adopts option (a) and allows that 2-spinor "elds are the best way to
describe half-integer-spin "elds, the CS thesis is not thereby put in jeopardy.
The second point concerning the fact that SL(2, C) cannot be embedded into
GL(4, R) is that it does not follow from this that 2-spinor components only take
values in spin bases. The consequence of replacing the As in (14) with general
linear transformations is essentially to extend the group SL(2, C) to the general
linear group of 2;2 complex matrices GL(2, C). Mathematically, there is
nothing to stop one from doing this, thereby obtaining values for the compo-
nents of 2-spinors in general linear 2-dimensional complex coordinates. What is
prohibited is a physical interpretation of such coordinates; and this prohibition
rests on a mathematical result; namely, that carriers of GL(2, C), which might be
called GL(2, C)-&spinors', are not related to carriers of GL(4, R) (i.e. general
tensors) in the same way (detailed below) that SL(2, C) 2-spinors are related
to carriers of Ot (1, 3) (i.e. world-tensors). GL(2, C) cannot be embedded into
>
GL(4, R); hence general tensors cannot be decomposed into products of
GL(2, C)-&spinors'. (For this reason, calling such general linear 2-component
complex vectors &GL(2, C)-spinors' is an abuse of standard terminology, which
restricts the term &spinor' to a carrier of a double-valued representation. In this
sense, saying that 2-spinor components can only take values in spin bases is true
by de"nition (in the same way that saying world-tensor components can only
take values in tetrad bases is true by de"nition).)
Thus carriers of GL(2, C) do not admit physical interpretations; there
are no physical phenomena that they can be said to describe. But this does not
prevent one from viewing the extension to GL(2, C) as a passive coordinate
212 Studies in History and Philosophy of Modern Physics

re-description of a carrier of SL(2, C). Again, such an extension is mathemat-

ically possible; the obstacle it faces is interpretational; what is prohibited are
physical interpretations of general linear 2-dimensional complex coordinates.
(Of course this is just what underlies the CS debate; namely, the physical
interpretation of coordinate charts, and not just their mere existence. I would
claim that emphasis on coordinate-dependent techniques obscures this.)
The third and "nal point concerning the non-embeddability of SL(2, C) in
GL(4, R) is the following: it turns out that even-indexed 2-spinors are isomor-
phic with world-tensors. Hence, since there are generally covariant expressions
of world-tensors, there are generally covariant expressions of even-indexed
2-spinors. Thus the information content of even-indexed 2-spinors can be ex-
pressed in arbitrary general linear coordinates. (Recall that an expression is
generally covariant just when it is given purely in terms of general tensors; such
an expression can be expanded in arbitrary coordinate bases.) Moreover, the
information content of odd-indexed 2-spinors can be expressed in arbitrary
charts up to a sign, in so far as odd-indexed 2-spinors can be given world-tensor
expressions up to a sign. This will be made clear in discussion of the relation
between spinors and tensors in the coordinate-independent approach, to which
I now turn.

4.1. Isomorphism between M4 and Re(S;S)

To relate 2-spinors and tensors, one can construct an isomorphism between

the real subspace Re(S;S) of S;S and Minkowski vector space M in the
following manner. Let (oo Y, on Y, no Y, nn Y) be a basis for S;S, where (o, n)
and (o Y, n Y) are spin bases for S and S. Then Re(S;S) is a real 4-dimensional
vector space. Moreover, it can be shown that Re(S;S) is endowed with
a Lorentz metric of signature (1, 3). To see this, note that a suitable orthonormal
basis for Re(S;S) is given by

1 i
tY" (oo Y#nn Y), yY" (on Y!no Y),
(2 (2

1 1
xY" (on Y#no Y), zY" (oo Y!nn Y). (16)
(2 (2

This can be checked by forming the (0, 2; 0, 2) spinor g "3 3 with

Y Y  Y Y
respect to which the above basis has the desired properties:
(i) g tYx Y"0, and likewise for other combinations; and,
Y Y
 To see this, note that any spinor qY3S;S can be expanded as qY"aoo Y#bon Y#
cno Y#dnn Y, where a, b, c, d3C. qY is real just when qY"qY. This entails that a, b3R and
c"dM . Hence the real elements of S;S can be parameterised by four real constants each and thus
form a 4-dimensional real subspace.
 An alternative orthonormal basis for Re(S;S) is given by lY"oo Y, mY"on Y, nY"
nn Y, m Y"no Y. The corresponding basis +l?, m?, n?, m ?, in M is referred to as a null tetrad.
¹he Coordinate-Independent 2-Component Spinor Formalism 213

(ii) g tYt Y"1"!g xYx Y"!g yYy Y

Y Y Y Y Y Y
"!g zYz Y; (17)
Y Y
hence g is a Lorentz metric on Re(S;S).
Y Y
Now suppose (t?, x?, y?, z?) is a tetrad basis for M. De"ne the map p:
Re(S;S)PM by
p? "t?t !x?x !y?y !z?z . (18)
Y Y Y Y Y
Then p? is an isomorphism preserving the Lorentz metric g "
Y ?@
p Yp Yg . In components relative to a spin basis, pIRR "(1/(2)pI,
? @ Y Y Y
where pI are the Pauli matrices. In such a basis, the components of the image
vY of a world-tensor (4-vector) v ? 3 M, with components (v, v, v, v)
U
relative to the basis (t?, x?, y?, z?), are given by



1 vY vY 1 v#v v#iv
vRRY" " . (19)
(2 vY vY (2 v!iv v!v
This is the coordinate representation of the 2-spinor vY in the basis
(oo Y, on Y, no Y, nn Y). Again, it should be stressed that vY is a well-de"ned
mathematical object independent of this particular coordinate representation.
To summarise, the isomorphism p: Re(S;S)PM
(1) is de"ned in terms of a tetrad basis in M; and
(2) only relates even-indexed 2-spinors with world-tensors (the extension of p to
isomorphisms between tensor product spaces M;M;2 and
Re(S;S);Re(S;S);2 is trivial).
This last fact should be emphasised. It indicates that the isomorphism p just is
an isomorphisms between even-indexed 2-spinors (carriers of SL(2, C)) and
world-tensors (carriers of Ot (1, 3)). This is what the explicit appearance of
>
standard coordinates in its construction indicates. To extend the isomorphism
to general tensors, one needs to explicitly include a tetrad basis. Hence under p,
the (1, 0; 1, 0) 2-spinor vY corresponds uniquely to the world-tensor v ? and to
U
the general tensor pair (v?,+e ,?), where v? is a general tensor and +e ,? is a tetrad
I I
"eld. This general tensor pair is the generally covariant expression of the
even-indexed 2-spinor.

4.2. Odd-indexed spinors

Again, the isomorphism (18) e!ects a 1-1 translation between even-indexed

2-spinors and Minkowski world-tensors. For odd-indexed 2-spinors, no such
1-1 translation exists owing to the topological properties of SL(2, C).

 Recall that univalent 2-spinors, as carriers of representations of SL(2, C), change sign under 2p
rotations, whereas 4-vectors, as carriers of representations of Ot (1, 3), do not. Any odd-indexed
>
2-spinor is, in general, the sum of outer products of an odd number of univalent 2-spinors, hence will
also change sign under 2p rotations.
214 Studies in History and Philosophy of Modern Physics

However, a 2-1 translation can be obtained, which I now review (cf. Penrose and
Rindler, 1984, pp. 125}129).
Every univalent 2-spinor i de"nes a null 4-vector l ? by l ?"iiY. But
U U
eGFi de"nes the same null vector l ? for real h. This phase is encoded up to sign
U
in the bivector de"ned by
(F ) "i i 3 #i i 3 . (20)
U ?@  Y Y Y Y 
To see this, one "rst completes i to a spin basis (i, q), where qq is unique up
to an additive multiple of i. The bivector (F ) can then be written as
U ?@
(F ) "l m !l m #l m !l m "l X , (21)
U ?@ ? @ @ ? ? @ @ ? ? @

where (l?, m?, n?, m ?) is null tetrad (footnote 21) and X "(1/(2)(m #m ) is
@ @ @
real, spacelike and orthogonal to l?. Hence (F ) contains l ? and lies in the
U ?@ U
spacelike 2-plane spanned by X and > "(1/(2)(m !m ). The pair
? ? @ @
(l ?, (F ) ) is called a &null #ag', consisting of a #agpole l ? with #ag (F ) .
U U ?@ U U ?@
Under the phase change i C eGFi, we have q C e\GFq and m? C eGFm?.
Hence X Ccos 2hX #sin 2h> and > C!sin 2hX #cos 2h> . Thus
? ? ? ? ? ?
under a phase change of h, the #agpole l ? remains invariant but the #ag (F )
U U ?@
rotates about l ? by 2h. In particular, under a phase change of n, the spinor
U
i changes sign, while the null-#ag remains invariant. In this sense, a null-#ag
only encodes a univalent 2-spinor up to sign.
Note that such null-#ags are world-tensors; they assume the existence of
a tetrad "eld. A generally covariant expression up to sign of a univalent 2-spinor
i is obtained by replacing the world-tensor pair (l ?, (F ) ) with a general
U U ?@
tensor triple (l?, F , +e ,?) satisfying the appropriate conditions.
?@ I
4.3. Fields with half-integer spin as carriers of SL(2, C)

The above correspondences between 2-spinors and world-tensors may be

extended to Minkowski spacetime (M, g ) (for M a di!erential manifold and
?@
g a Minkowski metric) by replacing M with the tangent spaces ¹ (M) at
?@ N
points p of M, and de"ning a 2-spinor "eld of valence (p, q; r, s) in (M, g ) as
?@
a map from points in M to the space (S)N;(SH)O;(S)P;(SH)Q. Spinor "eld
equations can now be written by introducing a spinor derivative, and transla-
tions for various "eld equations can be given (see e.g. Wald (1984), Penrose and
Rindler (1984)). One "nds, in general, that "elds with spin s may be represented
by 2-spinors with 2s indices. Hence, in the context of Minkowski spacetime, the
2-spinor formalism is expressively equivalent to the tensor formalism (insofar as,

 To see this, note that, under the isomorphism p, such an l? is indeed null: g l?l@"
?@
3 3 iiYi iY"0.
 Y Y
 To derive this, use is made of the relation i q !q i "3 which holds in general for any spin
  
basis (i , q ). X is unique up to an additive multiple of l determined by the choice of q. Under
  ? ?
qCq#jo, j3C, we have m Cm #jm , m Cm #jM m , hence X CX #(j#jM )l .
? ? ? ? ? ? ? ? ?
¹he Coordinate-Independent 2-Component Spinor Formalism 215

for any theory ¹ restricted to Minkowski spacetime, dynamically possible

models of ¹ expressed in the tensor formalism can be expressed in the 2-spinor
formalism).
However, as we have seen with odd-indexed 2-spinors, the converse does not
in general hold. For half-integer-spin "elds given by odd-indexed 2-spinor "elds,
no fully equivalent tensor expressions can be constructed; rather, one must be
content with tensor expressions up-to-a-sign. For example, in the coordinate-
independent 2-spinor formalism, the Dirac equation is given by
* tY"ku , *Yu "ktY, (22)
Y  
where k"!i2\m
\ and the 2-spinors tY, u form the 4-component

Dirac spinor W"(tY, u ), appearing in the standard expression

(i
c *?!m)W"0. Equation (22) can be translated into the tensor formalism
?
by (Penrose and Rindler, 1984, pp. 221}222):
F
F B#F
F B"!2kF C ,
?@ B A ?B A @ ?@ A
G
G B#G
G B"!2kG C , (22)
?@ B A ?B A @ ?@ A
where F "u u 3 , G "t t 3 and C "u t . These transla-
?@  Y Y ?@ Y Y  Y ?  Y
tions require F (resp. G ) to be null, skew, and anti-self-dual (resp. self-dual):
?@ ?@
(a) F F?@"G G?@"0,
?@ ?@
(b) F "!F ; G "!G ,
?@ @? ?@ @?
(c) HF "!iF ; HG "iG . (23)
?@ ?@ ?@ ?@
Note that the null and (anti-) self-dual conditions can be satis"ed by introducing
a tetrad +e ,? (in the case of condition (c), such a tetrad provides the orientation
I
4-form e ). Hence a Dirac 4-spinor W corresponds up to a sign to the general
?@AB
tensor 4-tuple (F , G , C , +e ,?) satisfying conditions (a)}(c) and the "eld
?@ ?@ ?Y I
equations (22). (The translation is essentially based on the one between
a univalent 2-spinor and a null-#ag, taking into account that the Dirac 4-spinor
is composed of two univalent 2-spinors. Penrose and Rindler (1984, p. 222) give
the extension of (22) for nth-order di!erential equations involving odd-indexed
2-spinors).
At this point, it should be noted that there is empirical evidence for the
rotational behaviour of carriers of SL(2, C). The mere existence of half-integer-
spin "elds does not in and of itself provide reason to use carriers of SL(2, C)

 Equations (22) are obtained from the standard expression of the Dirac equation by identifying
c with the map c =Y: MCHom(SSH) given by c =Yv?(tY, u )"(2(vYu , v tY),
? ? ?  Y Y
where v?3M and Hom(SSH) is the set of linear transformations from SSH to itself. The e!ect of
the linear transformation c =Yv? is to "rst exchange the 2-spinor components of a Dirac 4-spinor
?
and then contract them with the p-image of the 4-vector.
 The *-operator in condition (c) is the hodge-dual operator, de"ned for bivectors by
HF "e FAB, where e is a volume element 4-form.
?@  ?@AB ?@AB
216 Studies in History and Philosophy of Modern Physics

to describe them. It might be argued that the phase change associated with
SL(2, C)-induced rotations is unobservable (insofar as such phases are, from the
standpoint of quantum mechanics, unobservable). Hence one might conclude
that representations of half-integer-spin "elds that are unique up to a sign are
su$cient. However, macroscopic spinorial behaviour has been observed.
This indicates that the additional degree of freedom associated with carriers of
SL(2, C) is physically signi"cant, hence carriers of SL(2, C) are better suited to
describe half-integer-spin "elds than are carriers of Ot (1, 3).
>
Finally, it should be mentioned that 2-spinor "elds may be de"ned on curved
spacetimes as sections of an appropriate spinor bundle. This is the vector bundle
associated with a principal SL(2, C)-bundle (or dyad bundle). The latter may be
heuristically considered as the &double covering' of the bundle of time-oriented
and oriented orthonormal frames (more precisely, the bundle space of a dyad
bundle is the double covering of the bundle space of a tetrad bundle). One might
expect then that 2-spinor "elds, as cross sections of spinor bundles, can only be
de"ned on manifolds that support a global tetrad "eld. Indeed, Geroch (1968)
has shown that, for non-compact spacetime manifolds, a necessary and su$cient
condition for the existence of spinor "elds is the existence of a global tetrad "eld
(see Wald (1984, pp. 365}369) for a lucid discussion of this and other conditions
for the existence of spinor bundles).
One point to be kept in mind here is that the existence of a global tetrad "eld
does not, in and of itself, force us to use a standard Minkowski inertial
coordinate chart adapted to it. Indeed, the arena of the traditional CS debate is
Minkowski spacetime which trivially admits a global tetrad "eld, and this has
not prevented conventionalists from renouncing standard coordinate charts. It
is a realist interpretation of the global tetrad "eld that can be de"ned
in Minkowski spacetime to which the conventionalist objects. Again, more
speci"cally, her objection is to a realist interpretation of the isotropic temporal
structure provided by such a tetrad. She claims that the temporal structure of
spacetime is an in-principle unobservable object; and, as I have described it
above, her conventionalism is motivated by skepticism towards such objects.
However, as described above, her conventionalism is only possible in so far as
there are multiple intertranslatable descriptions of the phenomena that agree on
all observable aspects. Zangari and Karakostas in e!ect claim that this is not the
 Aharonov and Susskind (1967, p. 1237) proposed the following thought experiment to detect the
rotational behaviour of half-integer-spin "elds:
Imagine two systems having free electrons and exhibiting tunnelling current when they are
close together. We then separate the systems spatially and rotate one of them n times relative
to the other and bring them together so that tunnelling current can #ow again. It turns out
that the direction of current #ow depends on n modulo 2.
In Werner et al. (1975), a similar e!ect was observed for neutrons. A neutron beam is split and one
component in directed through a magnetic "eld that can be adjusted in order to force the magnetic
moments of neutrons passing through it to precess by a given amount (2p, say). The beams are
recombined and relative phase shifts can then be observed in the form of interference patterns. See
also Weingard and Smith (1982) for a discussion of the interpretation of such experiments.
¹he Coordinate-Independent 2-Component Spinor Formalism 217

case; that the wealth of inter-translatable descriptions reduces to just a single

description when one attempts to account for the empirically-grounded rota-
tional behaviour of half-integer-spin "elds. I turn now to their critique.

5. Zangari and Karakostas on SL(2, C) and e-Coordinates

Zangari (1994, p. 273) and Karakostas (1997, p. 260) observe that, while
e-extended Lorentz coordinate transformations can be de"ned via (5), there are
no corresponding e-extended SL(2, C) coordinate transformations. Using (8)
and (10), such a transformation M would satisfy




a b ct#x x!ix a c
H "MHMR" , (24)
C c d x#ix ct!x bM dM
where


(ct#e xG)#x x!ix
H "x Ip " G , (25)
C C I x#ix (ct!e xG)!x
G
and a solution M exists only for eG"0. Zangari and Karakostas conclude that
the possibility of de"ning e-coordinates vanishes in the 2-spinor formalism:

Thus, contrary to what the CS thesis (and even its opponents) assert, it is not
always possible to de"ne coordinates with [e O1/2] on which representations of
0
the Lorentz group can act. Spinor representations do not have Lorentz trans-
formations in non-standard coordinates systems, and one cannot then formulate
STR [special theory of relativity] (Zangari, 1994, p. 273).

Hence the possibility of recoordinatising the temporal component of a local

coordinate system according to t -synchrony relations no longer exists, if the
C
well-de"ned transformation properties of the spinor representation of the Lorentz
group are to be preserved. Thus the standard synchronisation relation, contrary to
the CS thesis, seems to be singled out by the very nature of things when the spinor
(universal covering) group of Lorentz transformations is taken into consideration
(Karakostas, 1997, p. 260).

What the non-existence of an e-version of M indicates is that an e-transforma-

tion K cannot be decomposed into the product of two Herm(2) matrices, and, in
C

 Zangari (1994, pp. 270}271) interprets (8) as a complex representation of a spacetime point (Gunn
and Vetharaniam (1995, p. 604) follow his lead). This is a bit misleading. Such an interpretation
con#ates the notion of a vector space (M) with that of an a$ne space (Minkowski spacetime).
Equation (8) establishes a correspondence between Minkowski 4-vectors and 2;2 complex Her-
mitian matrices. At the most, this can be interpreted as a correspondence between tangent vectors at
a point in Minkowski spacetime and elements of Herm(2); and not as a correspondence between
Minkowski spacetime points and elements of Herm(2). The signi"cance of the group SL(2, C) is not
that it provides &complex representations' of Minkowski spacetime points, but rather it is the
dynamical symmetry group for half-integer spin "elds (see footnote 12).
218 Studies in History and Philosophy of Modern Physics

particular, into the product of two SL(2, C) matrices. This means that SL(2, C)
cannot be embedded in e-O(1, 3), and this, again, is a special consequence of the
fact that SL(2, C) cannot be embedded in GL(4, R). Hence the components of
2-spinor "elds cannot take values in e-coordinates. From this Zangari and
Karakostas conclude that, once we require the use of 2-spinor "elds in descrip-
tions of physical phenomena, we cannot employ e-coordinates.
While this claim is straightforwardly correct from a mathematical point of
view, its signi"cance to the CS debate is questionable. Before canvassing pos-
sible conventionalist options, a few initial quali"cations should be made explicit.
(a) First, arguably, 2-spinors (even and odd-indexed) can be expanded in
arbitrary GL(2, C) coordinates (see Section 4 above). So while e-coordinates are
ruled out by 2-spinors, standard coordinates, qua coordinates, are not thereby
given priority.
(b) Again as indicated in Section 4, one may adopt a coordinate-dependent
de"nition of 2-spinors that restricts their components, by de"nition, to spin
bases. However, given this choice, not only are e-coordinates and 4-dimensional
real general linear coordinates prohibited; so also are standard Minkowski
inertial coordinates. The only coordinates permitted are those adapted to
spin-bases.
(c) While even-indexed 2-spinor components cannot take values in e-coordi-
nates, the information content they contain can be expressed in e-coordinates;
and the information content contained in odd-indexed 2-spinors can be ex-
pressed in e-coordinates up to a sign. These facts follow from the ability to
construct generally covariant expressions for even-indexed 2-spinors, and gener-
ally covariant expressions up to a sign for odd-indexed 2-spinors.
(d) Finally, note that what the non-existence of an e-version of M establishes is
that the isomorphism between M and Re(S;S) is by de"nition one between
world tensors and even-indexed 2-spinors. The appearance of standard charts in
the construction of the isomorphism entails that an even-indexed 2-spinor can
only be translated into a general tensor (as opposed to a world-tensor) in the
presence of a tetrad basis. In other words, if we want a generally covariant
description of even-indexed 2-spinor "elds, then we cannot do without a tetrad
basis. Hence the charge against the conventionalist can be phrased in terms of
the following (Spinor Realism) claim.

(SR) The existence of half-integer-spin "elds requires using odd-indexed

2-spinor "elds to describe them. The existence of odd-indexed 2-spinor

 Of course these are just &spin-entangled' Minkowski inertial coordinates (i.e. inertial coordinates
with two extra degrees of freedom). Such spin-coordinates do provide the temporal structure needed
to de"ne the standard simultaneity relation. Still, the point remains that if the concern is solely with
what coordinate charts are permitted by the existence of 2-spinor "elds, then we see that such "elds
prevent the use of both non-standard e-coordinates and standard coordinates, as the latter are
normally de"ned. I agree that the moral at this point is that coordinates are not what are at stake,
but rather a realist interpretation of temporal structure. This further urges the adoption of (SR)
below as the point of debate for a spinor realist.
¹he Coordinate-Independent 2-Component Spinor Formalism 219

"elds as geometrical objects requires the existence of a global Min-

kowski tetrad "eld. In particular, it entails the existence of an isotropic
temporal structure on spacetime that cannot be done away with, if we
wish to account for all the observable phenomena.

This, I would submit, is the substantive charge that can be levelled against the
conventionalist by a spinor realist. It indicates how the conventionalist's seman-
tic anti-realism in undercut by the privileging of one descriptive framework over
competing alternatives. Note that (SR) can be justi"ed simply by reference to
Geroch's (1968) theorem that the existence of 2-spinor "elds on a spactime
manifold M requires the existence of a global Minkowski tetrad "eld on M. One
need not bring coordinates into the discussion at all. I now consider ways the
conventionalist can respond to (SR).

6. Options for the Conventionalist

If the conventionalist is read as claiming that there is no fact of the matter as

to the local metrical structure of spacetime (in particular, the local temporal
structure), i.e. there is no fact of the matter that determines whether, locally, the
metric of spacetime is the Minkowski metric g or an e-metric (g ) , then
?@ C ?@
a Spinor Realist will counter with the claim that the existence of half-integer-
spin "elds provides the fact of the matter that privileges a preferred temporal
structure. And this indicates that there is a preferred simultaneity relation;
namely, the standard one.
In response, the conventionalist can "rst and foremost claim that the existence
of half-integer-spin "elds does not require using odd-indexed 2-spinor "elds to
describe them. As indicated in Section 4, we can use spinorial representations of
GL(4, R) in descriptions of half-integer-spin "elds; and such representations do
admit e-coordinates.
Furthermore, even if the realist is granted the use of 2-spinors in descriptions
of half-integer-spin "elds, the conventionalist can simply claim that anything
you can do with even-indexed 2-spinor "elds, you can do just as well with
general tensors. And, while what you can do with odd-indexed 2-spinor "elds,
you can only do up to a sign with general tensor "elds, this is "ne, since
odd-indexed 2-spinor "elds have nothing to do with determinations of clock
synchrony. Simply put, clocks and measuring rods are not spinorial (i.e. half-
integer-spin) objects. Hence, for the purposes of describing procedures involving
the determination of clock synchrony, the tensor formalism is adequate. To
argue otherwise is to engage the conventionalist in an argument the premisses of
which she will reject; namely, that unifying power, or simplicity, count as criteria
when it comes to justifying belief in in-principle unobservable objects.
Recall that the conventionalist's position is a particular type of anti-realist's
position; speci"cally, the conventionalist is an anti-realist with respect to the
in-principle unobservable quantity given by the speed of light in a given
220 Studies in History and Philosophy of Modern Physics

direction. As indicated in Section 2, any value (constrained by 0(e(1) is

consistent with the observational consequences of special relativity. Hence any
simultaneity relation de"ned by a given value for e is observationally adequate.
In the past, realists have argued that the standard simultaneity relation is
privileged because it is more simple or more unifying than non-standard rela-
tions in so far as it is adapted to the conformal structure of Minkowski
spacetime; or because we should be realists with respect to Robertson}Walker
spacetimes (or, in general, globally hyperbolic spacetimes) with privileged time
coordinates (cf. Friedman (1983), p. 320, Torretti (1983), p. 230). But the die-hard
conventionalist will claim these responses beg the question over the in-principle
unobservable one-way speed of light. In this context, the claim that half-integer-
spin "elds privilege a given time coordinate is just another argument that begs
this same question. Half-integer-spin "elds have nothing to do with determining
clock synchrony, and, in particular, the one-way speed of light. Hence their
existence should do nothing to convince an anti-realist who claims the one-way
speed of light has no determinate value.
Furthermore, it is not all that apparent that arguments for using the 2-spinor
formalism based on simplicity and/or unifying power should be convincing even
to a spacetime realist. The existence of physical phenomena that can be de-
scribed by the 2-spinor formalism, and that cannot be described by the general
tensor formalism, should not automatically force us to use the former formalism.
Note that, while the 2-spinor formalism is expressively equivalent to the world-
tensor formalism, both are not expressively equivalent to the general tensor
formalism. There are models of GR that can only be expressed using general
tensors; that cannot be expressed, in particular, using world-tensors or 2-
spinors. To restrict all descriptions of physical phenomena to the 2-spinor
formalism would require viewing models of GR that do not admit global tetrad
"elds as unphysical. This seems a bit hasty. Part of the central problem facing
contemporary theoretical physics is to reconcile phenomena best described by
the 2-spinor formalism (viz, half-integer-spin "elds) with phenomena best de-
scribed by the general tensor formalism (viz, phenomena displaying di!eomor-
phism invariance). It seems the lesson to be learned, at least until a quantum
theory of gravity is established, is that which formalism one should use will
depend on what type of phenomena is being described. Hence the conventional-
ist would be warranted in refusing to use (even-indexed) 2-spinors in descrip-
tions of clock synchrony procedures.
I submit, therefore, that the adoption of the 2-spinor formalism in descrip-
tions of half-integer spin "elds does not require adopting coordinates adapted to
spin bases, let alone standard (Minkowski inertial) coordinates, in determina-
tions of clock synchrony.

7. Gunn and Vetharaniam (1995)

In this section, I assess three points of criticism that Gunn and Vetharaniam
(1995) level against Zangari (1994).
¹he Coordinate-Independent 2-Component Spinor Formalism 221

7.1.

First, Gunn and Vetharaniam claim that the Dirac equation governing the
dynamical behaviour of a massive spin-1/2 "eld can be given an e-coordinate
covariant expression. In particular, they write,
(i
(c )I* !m)t?"0, (26)
C I
where t? are the components of a Dirac 4-spinor and the e-versions of the
c matrices are given by



1 n pG 1 pG
(c )" G , (c )G" . (27)
C !n pG 1 C pG 1
G
They thus conclude that, pace Zangari, the SL(2, C) &complex representation' of
spacetime points (see footnote 28) is not necessitated, in particular, by the
existence of spin-1/2 "elds.
Karakostas (1997, p. 271) observes that the c do not form a Cli!ord algebra
C
since (c ) (c ) #(c ) (c ) O2g . However, the c are closed under Cli!ord
C? C@ C@ C? ?@ C
multiplication de"ned by (c ) (c ) #(c ) (c ) "2(g ) , with (g ) de"ned by
C? C@ C@ C? C ?@ C ?@
(7). This is perfectly consistent with the de"nition of a Cli!ord algebra. What is
problematic with (26) in the context of the CS debate is the fact that the Dirac
4-spinor components in (26) are unde"ned in general linear coordinates. In
particular, the components of a Dirac 4-spinor cannot take values in e-coordi-
nates. Recall that in the 2-spinor formalism, a Dirac 4-spinor is simply the direct
sum of two univalent 2-spinors, and the components of the latter cannot take
values in e-coordinates, as discussed above.
To be fair to Gunn and Vetharaniam, they treat the t in (26) as a scalar with
respect to di!eomorphisms on M. (This follows the standard practice of con-
structing a curved-space version of the Dirac equation by gauging the Lorentz
group. Karakostas (1997, pp. 273}274) provides a good exposition. See also
Kaku (1993, p. 641).) Doing so, I would argue, does not directly address the
objection to the CS thesis given by Zangari and Karakostas, which is that

 In general, if < is a vector space over a commutative "eld K with unit element 1 and equipped with
a quadratic form q: <CK, then the Cli!ord algebra C(<, q) associated with < is de"ned as the
quotient C(<, q),¹(<)/J, where ¹(<) is the tensor algebra over <, and J is the two-sided ideal in
¹(<) generated by elements of the form xx#2q(x)1, for x3<. The Cli!ord product in C(<, q) is
then the product induced by the tensor product in ¹(<). In terms of the associated metric g (the
bilinear form de"ned by g(x, y)"q(x#y)!q(x)!q(y)), the Cli!ord product is given by
xy#yx"2g(x, y)1, for x, y3<. Only in an orthonormal basis is this product anticommutative. In
general, a Dirac spinor is an irreducible representation of a Cli!ord algebra.
 This essentially means that the Cli!ord algebra C(R, g ) generated by the c 's does not have
C C
irreducible representations that are, in addition, representations of the direct sum SL(2, C)SL(2,
C). Arguably, e-Dirac spinors can simply be de"ned as the irreducible representations of C(R, g ),
C
but such mathematical objects cannot be physically interpreted (in the same sense that GL(2, C)
&2-spinors' cannot be physically interpreted): such e-Dirac &spinors' cannot encode the rotational
behaviour of half-integer-spin "elds.
222 Studies in History and Philosophy of Modern Physics

2-spinor components cannot take values in e-coordinates. This is trivially the

case for t when it is treated as a scalar with respect to coordinate transforma-
tions on M; as such, it has no components with respect to coordinate charts on
M. It is only expandable in spin bases on the tangent spaces. Gunn and
Vetharaniam are motivated by this to declare that spin space is an internal
symmetry space that has nothing to do with spacetime. I think this stance is
problematic, as I shall now attempt to demonstrate.

7.2.

In thinking that the SL(2, C) &complex representation' of spacetime points is

necessitated by the existence of odd-indexed spinor "elds, Gunn and
Vetharaniam accuse Zangari of con#ating spin space as an internal symmetry
parameter space with spacetime as the arena governing the kinematics of special
relativity:
one regards quantum mechanical spin as an internal property of particles. The
symmetries corresponding to the transformation properties of spinors are symmet-
ries of an internal spinor space. [2] Internal spinor space must, therefore, not be
confused with the external space of Minkowski space time (1995, p. 605).

Apparently, Gunn and Vetharaniam claim that the two degrees of additional
freedom provided by spinors should be associated with an internal state of the
"eld being described, in analogy with isotopic spin, electroweak charge, colour,
and other internal gauge symmetries that appear in standard Yang}Mills
theories. These additional two degrees of freedom should not, they argue, be
associated with degrees of freedom in spacetime. In particular, spacetime coordi-
nates need not transform in the same way as spinor "eld components.
Gunn and Vetharaniam are correct in noting that spacetime coordinates need
not transform in the same way as spinor components. Simply put, as indicated
above, clocks and measuring rods are not spinorial objects. However, it is
misleading to view spin as an internal symmetry on a par with gauge symmet-
ries. Spin is one of two fundamental properties of a relativistic spacetime "eld,
the other being mass. Wigner's famous analysis (1939) has shown that a relativis-
tic "eld is determined by its mass and spin, insofar as mass and spin are the only
two Casimir invariants of the PoincareH group, the symmetry group of Min-
kowski spacetime.
Another way of making the distinction between gauge symmetries and spin
can be seen in the "ber bundle formulation of gauge theories. In this formalism,
for a given gauge theory with gauge symmetry group G, the internal symmetry
spaces are "bers in a principle G-bundle over a spacetime manifold M.
The associated vector bundle is a tangent bundle with sections identi"ed as

 The curved-space treatment of the Dirac equation requires the existence of a global tetrad "eld;
hence, it does not directly address the Spinor Realist's claim given above that such a global tetrad
"eld provides the fact of the matter as to the temporal structure of spacetime.
¹he Coordinate-Independent 2-Component Spinor Formalism 223

spacetime "elds that interact with the gauge "eld. If these "elds are half-integer-
spin "elds, then the associated vector bundle must be a spinor bundle, and
a third bundle over M arises, namely, an SL(2, C) dyad-bundle. This latter di!ers
from the principle gauge bundle insofar as it is soldered to the base space
M (technically, this means that a soldering form exists on the dyad bundle). The
presence of soldering amounts to the following distinction: whereas the connec-
tion and curvature on the gauge bundle do not de"ne structures on M, the
connection and curvature on the dyad bundle do project down to structures on
M. In this sense, spin is intimately connected with the spacetime manifold in
a way that typical Yang}Mills gauge symmetries are not.

7.3.

Third, Zangari claims that the SL(2, C) &complex representation' of spacetime

points is to be preferred for its topological simplicity. Gunn and Vetharaniam
contend, however, that this simplicity is o!set by the inability within the SL(2, C)
representation to account for parity violations. They claim that the O(1, 3) &real
representation' of spacetime points is needed in order to be able to represent parity
operations, which have physical signi"cance, in the light of con"rmed parity
violation in electroweak theory. Gunn and Vetharaniam thus argue not only that
Zangari's empirical argument for the SL(2, C) representation is wrong, but, more-
over, that there is an empirical argument against the SL(2, C) representation.

 The existence of a soldering form on frame bundles is the main distinction between "ber bundle
formulations of Yang}Mills gauge theories and a "ber bundle formulation of general relativity as
a theory of a metric and connection over a principle frame bundle (see e.g. Trautman, 1980).
Before leaving this subject, I should note that Karakostas (1997, pp. 261}263) describes spinor
"elds (conceived as sections of a spinor bundle) as &Lorentz covariant and coordinate invariant'. This
seems to be at odds with the standard notions of covariance and invariance (see e.g. Earman (1974),
Friedman (1983), Norton (1993)). Under these notions, covariance refers to the form an equation
takes with respect to a coordinate chart. A generally covariant equation retains its form in arbitrary
general linear coordinates. A Lorentz covariant equation retains its form only in Minkowski inertial
coordinates. A test for general covariance is whether the equation can be given an expression purely
in terms of tensors. (Similarly, a test for Lorentz covariance is whether the equation can be given an
expression purely in terms of world-tensors.) Invariance, according to the standard notion, refers to
the symmetries of a given mathematical object; in particular, how the object behaves under active
transformations. Tensors are generally (or, one might say, &coordinate') invariant: they are invariant
under di!eomorphisms (viz, active general linear transformations). World-tensors and even-indexed
2-spinors are Lorentz invariant. Note that odd-indexed 2-spinors are not: they can change sign
under an active Lorentz transformation. Odd-indexed spinors are invariant only under actions of
SL(2, C).
Hence an equation involving only even-indexed 2-spinors is Lorentz invariant. It can be put in
a Lorentz covariant form as well as a generally covariant form; in the same way that an equation
involving only world-tensors can be put in a generally covariant form. A equation involving only
odd-indexed 2-spinors, on the other hand, is not even Lorentz invariant; it is SL(2, C)-invariant.
Furthermore, it can be put into a generally covariant form only up to a sign.
 From footnote 6, parity operations may be represented by elements of the Ot (1, 3) component
\
(determinant"!1) of O(1, 3). SL(2, C) is connected to the identity, hence none of its elements can
represent a re#ection about a spatial axis.
224 Studies in History and Philosophy of Modern Physics

As noted above, the issue should not be parsed in terms of an SL(2, C)

representation of points versus an O(1, 3) representation; rather, the issue
concerns the choice between a 2-component spinor formalism and a tensor
formalism. In this context, one might thus claim: the 2-component spinor
formalism does not admit parity transformation, whereas the tensor formalism
does. (,I do not claim that this is what Gunn and Vetharaniam hold; I grant that
they are simply responding to Zangari's distinction between O(1, 3) representa-
tions of points vs SL(2, C) representations of points, which, I do claim, is
technically incorrect.)
However, this claim is patently false. A homogeneous Lorentz transformation
K @3O(1, 3) can be decomposed in the following manner in the 2-component
?
spinor formalism (Penrose and Rindler, 1984, pp. 167}175):

K Y"$h hM Y or K Y"$

M Y, (28)
Y Y  Y  Y
where det(h )"det(
)"1, with h and
determined uniquely up to
Y  Y 
a sign. Speci"cally, it can be shown that

#h hM Y3Ot (1, 3), #

M Y3Ot (1, 3),
Y  \  Y >
#h hM Y3Os (1, 3), !

M Y3Os (1, 3). (29)
Y  \  Y >
The improper (parity) transformations are of the form $h hM Y. The point
Y 
then is that, within the 2-component spinor formalism, such operators do
occur. It is thus incorrect to say that the tensor formalism is superior to the
2-component formalism with regard to representing parity operations.

8. Conclusion

The existence of half-integer-spin "elds, in and of itself, does not commit us to

the standard simultaneity relation. For, according to current theory, such "elds
have nothing to do with determinations of the in-principle unobservable one-
way speed of light that generates the skepticism that underlies the conventional-
ist's anti-realism. Simply put, as indicated above, clocks and measuring rods are
not spinorial objects. Hence, while half-integer-spin "elds may best be described
by odd-indexed 2-spinor "elds, and the components of the latter cannot take
values in non-standard coordinate charts, this does not prevent the conven-
tionalist from using non-standard coordinate charts in descriptions of clock
synchrony experiments. In this article, I have argued that these observations
become clear when one considers how half-integer-spin "elds are described in
the coordinate-independent 2-spinor formalism.

Acknowledgements*Thanks to Jeremy Butter"eld and an anonymous referee for insightful com-

ments and suggestions.
¹he Coordinate-Independent 2-Component Spinor Formalism 225

References

Aharonov, Y. and Susskind, L. (1967) &Observability of the Sign Change of Spinors under
2p Rotations', Physical Review 158, 1237}1238.
Anderson, R., Vetharaniam, I. and Stedman, G. E. (1998) &Conventionality of Syn-
chronisation, Gauge Dependence and Test Theories of Gravity', Physics Reports 295,
93}180.
Cartan, E. (1966) The Theory of Spinors (Dover Edition, New York: Dover, 1981).
Earman, J. (1974) &Covariance, Invariance, and the Equivalence of Frames', Foundations
of Physics 4, 267}289.
Friedman, M. (1983) Foundations of Spacetime Theories (Princeton: Princeton University
Press).
Gel'fand, I., Minlos, R. and Shapiro, Z. (1963) Representations of the Rotation and Lorentz
Groups and their Applications, translated by G. Cummins and T. Boddington (New
York: Macmillan).
Geroch, R. (1968) &Spinor Structure of Spacetimes in General Relativity. I', Journal of
Mathematical Physics 9, 1739}1744.
GloK ckeler, M. and SchuK cker, T. (1987) Diwerential Geometry, Gauge Theories, and Gravity
(Cambridge: Cambridge University Press).
Gunn, D. and Vetharaniam, I. (1995) &Relativistic Quantum Mechanics and the Conven-
tionality of Simultaneity', Philosophy of Science 62, 599}608.
Janis, A. (1998) &Conventionality of Simultaneity', in E. Zalta (ed.), Stanford Encyclopedia
of Philosophy, plato.stanford.edu/entries/spacetime-convensimul/.
Kaku, M. (1993) Quantum Field Theory (Oxford: Oxford University Press).
Karakostas, V. (1997) &The Conventionality of Simultaneity in the Light of the Spinor
Representation of the Lorentz Group', Studies in History and Philosophy of Modern
Physics 28, 249}276.
Malament, D. (1977) &Causal Theories of Time and the Conventionality of Simultaneity',
NouL s 11, 293}300.
Naber, G. (1992) The Geometry of Minkowski Spacetime (New York: Springer-Verlag).
Ne'eman, Y. and Sijacki, D. (1987) &GL(4, R) Group-Topology, Covariance and Curved-
Space Spinors', International Journal of Modern Physics A2, 1655}1668.
Ne'eman, Y. and Sijacki, D. (1997) &World Spinors*Construction and Some Applica-
tions', Foundations of Physics 27, 1105}1122; Los Alamos e-print archive gr-
qc/9804037.
Norton, J. (1992) &Philosophy of Space and Time', in M. H. Salmon et al., Introduction to
the Philosophy of Science (New Jersey: Prentice Hall), pp. 179}231.
Norton, J. (1993) &General Covariance and the Foundations of General Relativity: Eight
Decades of Dispute', Reports on Progress in Physics 56, 791}858.
Penrose, R. and Rindler, W. (1984) Spinors and Spacetime, Vol. 1: Two-Spinor Calculus and
Relativistic Fields (Cambridge: Cambridge University Press).
Sarkar, S. and Stachel, J. (1999) &Did Malament Prove the Non-Conventionality
of Simultaneity in the Special Theory of Relativity?', Philosophy of Science 66,
208}220.
Sijacki, D. (1998) &World Spinors Revisited', Acta Physica Polonica B29, 1089}1097; Los
Alamos e-print archive gr-qc/9804038.
Stachel, J. (1986). What a Physicist Can Learn from the Discovery of General Relativity',
in R. Ru$ni (ed.) Proceedings from the Fourth Marcel Grossmann Meeting on General
Relativity (Amsterdam: Elsevier Science Publishers), pp. 1857}1862.
Stachel, J. (1992) &How Einstein Discovered General Relativity: A Historical Tale with
some Contemporary Morals', in M. MacCallum (ed.), General Relativity and Gravi-
tation (Cambridge: Cambridge University Press), pp. 200}208.
Torretti, R. (1983) Relativity and Geometry (New York: Pergamon Press).
226 Studies in History and Philosophy of Modern Physics

Trautman, A. (1965) &Foundations and Current Problems in General Relativity', in

A. Trautman, F. A. E. Pirani and H. Bondi (eds.), Lectures on General Relativity.
Brandeis Summer Institute in Theoretical Physics, 1964 (New York: Englewood Cli!s),
pp. 1}248.
Trautman A. (1980). &Fiber Bundles, Gauge Fields, and Gravitation', in A. Held (ed.),
General Relativity and Gravitation, One Hundred Years after the Birth of Albert Einstein
(New York: Plenum), pp. 287}308.
Wald, R. (1984) General Relativity (Chicago: University of Chicago Press).
Weingard, R. and Smith, G. (1982) &Spin and Space', Synthese 50, 213}231.
Werner, S., Colella, R., Overhauser, A. and Eagen, C. (1975) &Observation of the Phase
Shift of a Neutron Due to Precession in a Magnetic Field', Physical Review Letters 35,
1053}1055.
Wigner, E. (1939) &On Unitary Representations of the Inhomogeneous Lorentz Group',
Annals of Mathematics 40, 149}204.
Zangari, M. (1994) &A New Twist in the Conventionality of Simultaneity Debate',
Philosophy of Science 61, 267}275.