JARDINE,R. J., SYMES,M. J. & BURLAND,J. B. (1984). Giotechnique 34, No.

3, 323-340

The measurement
of soil stiffness
triaxial apparatus
R. J. JARDINE.*

M. J. SYMES*

This Paper describes a simple technique for accurately

measuring the mean local axial strains of triaxial samples over a central gauge length. The technique makes
use of an axial displacement gauge which is a development of one devised by Burland
& Symes (1982)
which makes use of electrolytic
levels. The device can
resolve to less than 1 cm over a range of 15 mm, is
simple to mount on the specimen and is not damaged
when the sample is taken to failure. The results of
undrained triaxial tests are presented for a wide spectrum of soil types ranging from sands through intact,
reconstituted
and remoulded
low plasticity till, undisturbed London clay to intact unweathered
chalk. The
test results show that conventional
external measurements of displacement
contain errors which are frequently so large that their use in the determination
of
soil stiffness at working levels of stress is invalid. The
errors mainly result from tilting of the sample, bedding
at the end platens and the effects of compliance
in the
apparatus.
Although much more experimental
work is
required
before general
conclusions
can be drawn
about the small strain behaviour
of soils the results
presented lead to some important
observations
on the
undrained
stiffness, linearity and yielding behaviour of
soils at small strains.
Cet article d&it une technique t&s simple pour mesurer de faGon prCcise les dkformations
locales moyennes dtchantillons
triaxiaux sur une jauge centrale. La
technique emploie une jauge de d&placement axial qui
reprksente une amelioration de celle invent&e par
Burland et Symes (1982) et qui utilise des niveaux
tlectrolytiques.
Lappareil est sensible 2 moins de
1 pm sur une longueur
de 15 mm. I1 est facile B
monter
sur ltchantillon
et reste intact
m&me si
ICchantillon est dCtruit. Les rCsultats des tests triaxiaux non-drain&
sont prCsentCs pour une large gamme
de types de sol, commenqant
par des sables, suivis de
moraines intactes de faible plasticitt
reconstitukes
et
remaniCes et de largile de Londres intacte jusqu8 la
craie intacte
non-altCrte
par les intempkries.
Les
rtsultats
des tests montrent
que les mesures conventionnelles du d&placement cOrnportent des erreurs qui
sont souvent si considerables
que les mesures sont ma1
adapt&es pour la d&termination
de la rigidit
du sol g
des niveaux operationnels
de la contrainte.
Les erreurs
proviennent
principalement
du
basculement
de
lCchantillon,
de la liaison imparfaite
au niveau des
plateaux terminaux
et des effects du d&placement
de
lappareil.
Discussion
on this Paper closes on 1 January
For further details see inside back cover.
* Imperial College of Science and Technology.

1985.

323

in the

and J. B. BURLAND*

Bien
que
beaucoup
de
travail
expCrimenta1
suppltmentaire
soit nCcessaire afin de pouvoir tirer
des conclusions
g&&ales
au sujet du comportement
des sols sous des d&formations
mineures, les r&hats
p&en&
fournillent
des observations
importantes
concernant
la rigidit
dans 1Ctat non-drain&,
la
IinearitC
et
ICcoulement
des
sols
sous
des
d&formations
mineures.
NOTATION
C

El
F
KO
L

Lo
LI
P
PO
RI3

T
6

compliance

of

loading

system =

(A,_+ 4&F
undrained
shear strength
undrained
stiffness
E,co.ol,-E,
at 0.01%
strain, etc.
deviator force on sample
u~/u, at rest
E,(,,.l,/E,(o.o,,
an index of linearity
length of sample
liquidity index
(a, + 2a,)/3 the mean effective stress
p at the start of the undrained
test
relative density
(0,, - &&(&
+ &) tilt ratio
sample rotation

A, A,, AT, A,,,

As, ARB, A,,,

components

of

measured
deflexions
(see Fig. 1)
corrected
overall axial strain
larger local axial strain
smaller local axial strain
mean local axial strain
larger incremental
rotation
of electrolevel
(see Fig. 2(c))
smaller incremental
rotation of electrolevel
(see Fig. 2(c))
vertical effective stresses
radial effective stress
INTRODUCTION
Accurate
determination
of soil
stiffness
is
difficult to achieve in routine laboratory
testing.
Conventionally,
the determination
of the axial
stiffness of a triaxial sample is based on external
measurements
of displacement
which include a
number of extraneous
movements.
For example,
the true soil strains developed
in triaxial
tests
can be masked by deflexions
which originate
in
the compliances
of the loading system and load
measuring
system. Such equipment
compliance

324

JARDINE,

SYMFS

errors add to a variety of sample bedding effects

to give a poor definition of the stress-strain
behaviour of the material under test, particularly
over the small strain range. Most triaxial tests
therefore tend to give apparent soil stiffnesses
far lower than those inferred from field behaviour.
The importance of such errors has long been
recognized and many diverse techniques have
been employed in attempts to improve strain
measurements. One solution has been to measure relative displacement between two reference
footings over a central length of a sample using
displacement transducers (e.g. Yuen, Lo, Palmer
& Leonards, 1978; Daramola, 1978; Brown,
Austin & Overy, 1980; Costa Filho, 1980).
Strictly these techniques are suitable only for
very small strain levels, since bulging of the
sample will cause the footings to rotate in later
stages of the test. Although important results
have been obtained with such techniques, they
are cumbersome and can suffer from jamming
and damage at large strains.
X-ray and optical methods have also been
used to follow reference points within the sample or on its membrane (Roscoe, Schofield &
Thurairajah,
1963; Arthur & Phillips, 1975).
However, the accuracy of these methods is
limited.
The resonant column apparatus offers a different approach for the determination
of the
dynamic stiffness of soils. The technique involves the application of periodic small strain
perturbations
to a sample as described by
Richart, Woods & Hall (1970). However, the
technique does not provide direct measurements
of the elemental behaviour of the soil under test,
since the states of stress and strain vary continuously both with time and in their distributions
within the sample.
In summary present methods of soil strain
measurement have a number of serious limitations. There is an urgent need for a simple but
precise method for the routine measurement of
the stress-strain behaviour of soil specimens
under controlled stress or strain paths, particularly where the soil exhibits high stiffness at
small strains.
In this Paper a simple technique for accurately
measuring the mean local axial strains during
triaxial testing is described. Local axial strains
are taken as those developing over a central
gauge length of the sample. The origins of some
of the more significant strain measurement errors which develop in standard testing are examined and their magnitudes assessed using the
new techniques. Results of experiments performed on a wide range of material are pre-

AND BURLAND

sented, and it is shown that, as expected, routine

tests which employ external measurements of
strain lead to apparent soil stiffnesses which are
much too low. For the purposes of this Paper
only undrained behaviour is considered, since
the no-volume change condition obviates the
need to measure radial strains. However, in tests
designed to investigate more general effective
stress behaviour the local measurement of radial
and axial strains is equally important. Symes &
Burland (1984) describe the use of proximity
transducers for radial strain determination and
Maswoswe (1984) describes the use of a high
accuracy, submersible, linear variable differential transformer (LVDT) for the same purpose
on 38 mm dia. triaxial samples.
AXIAL STRAIN ERRORS IN TRIAXIAL TESTS
In a conventional triaxial test there are several
sources of movement that develop during shear
testing which may give rise to an overestimate of
the axial strain. One such source is the compliance of the loading system itself. For example, the construction of a Bishop & Wesley
(1975) cell is such that the lower reference point
for the vertical displacement transducer is attached to the ram while the upper reference
point is located on top of the cell, so that small
but nevertheless significant deflexions accumulate from the straining of the rolling Bellofram
diaphragms. For present purposes the sum of
such loading system deflexions will be termed
A,,. An internal load cell will also produce a
significant deflexion, which is termed A,.
The more important sources of error are illustrated in Fig. 1. Some of the deflexions shown
in this figure may be quantified by careful calibration, but large unaccountable errors remain
due to
(a) the difficulty of trimming a sample so that
the end faces are perpendicular to the vertical axis of symmetry
(b) play in the connection between the load cell
and the sample top cap, and
(c) the inevitable bedding down at the ends of
the sample, due to local surface irregularities or voids.
In the testing of rock samples the importance
of such errors has long been recognized, and the
careful grinding of sample ends combined with
the use of ground cylindrical seated platens is
commonly specified (e.g. Vogler & Kovari,
1978). The observation that many rocks fail in a
brittle fashion at axial strains of O-l% or less
has led to the specification of flatness limits of
+O.Ol mm and parallelism requirements
of

MEASUREMENT

around 3 minutes of arc for high quality sample

preparation. The preparation techniques employed for rock testing are unsuitable for most
soils, and it is probably not possible to approach
the same standards of sample regularity.
Recent work has demonstrated the rather
surprising finding that soils can be equally as
brittle as rocks and that an understanding of
their behaviour at levels of shear strain below
0.05%
is very important (see Gens, 1982;
Simpson, ORiordan & Croft, 1979). Indeed, it
is shown in this Paper that K. normally consolidated clays may reach peak strength in the
triaxial apparatus at axial strains as low as
0.1%. Moreover, even when the behaviour is
not brittle, the strains prior to yield are usually
very small.
Measures can be taken to reduce the errors
implicit in external strain measurement. The
results obtained from tests carried out on soil
which has been an&tropically consolidated to a
high level of mean effective stress suggest that
these procedures considerably reduce sample
bedding and tilting errors (see Gens, 1982). The
SHANSEP methods of testing soft clays can also
be expected to lead to significant improvements
in strain measurement. However, where swelling
stages are included in such tests tilting and other
1978).
errors
(Daramola,
may redevelop
Moreover, it is often desirable to obtain accurate strain measurements in tests which have not
involved anisotropic consolidation.
A more satisfactory approach is to make use
of local instrumentation which can be attached
to a central gauge length of a sample. Symes &
Burland (1984) have given a description of the
design of instruments which employ electrolytic
levels to measure combined horizontal shear
strain and axial strain in a hollow cylinder apparatus. The same principles have been adapted
to develop a vertical displacement measuring
system for use in a 100 mm dia. triaxial apparatus (see Burland & Symes, 1982). This
Paper describes a further development and improvement of the earier devices which enables
mean axial strains to be determined to within a
range of +0.002%
in triaxial stress path cells
designed for the testing of 38 mm dia. samples.
DESCRIBIION OF THE ELECIROLEVEL
GAUGES
Cooke & Price (1974) describe the use of
electrolytic liquid levels for the local measurements of ground strains around test piles. Their
reliability, simplicity and accuracy make these
transducers attractive in a wide variety of applications, and by mounting the capsules in simple
mechanisms it is possible to develop a range of

OF SOIL STIFFNESS

325

reorientation

Sample
compression

Fig. 1. Sources of emx in external strain measurements (+,

is the larger of the two shins;
FL= (ELI + E&3

devices to measure axial, radial and shear strains

in laboratory tests (Symes & Burland, 1984).
The liquid level transducers consist of an
electrolyte sealed in a glass capsule. In the simplest devices three coplanar electrodes protrude
into the capsule and are partially immersed in
the electrolyte. The impedance between the central electrode and the outer ones varies as the
capsule is tilted. A variety of levels with different sensitivities are commercially available.
The transducers employed in the triaxial
strain measurements were supplied by IF0 International Ltd and have a working range of
*lo. The system was excited using a 5 V a.c.
power supply of 4 kHz frequency. The gains
were adjusted to give a *3 V full scale output
which was monitored to *O-l mV with typical
scatters of *0.2 mV. The levels are sensitive to
temperature and vibration and should be operated in still conditions which are temperature
controlled to within *3C. Under such conditions the gauges can be stable over periods of
weeks.
The principles of the new axial strain measuring systems are essentially similar to those of the
earlier devices in that a hinged arrangement
converts displacements between two footings
mounted on the sample into a rotation of the
capsule, as shown in Fig. 2(a).

326

JARDINE,

Stanless

SYMES

steel tubing

Hinge C

Hinges A and B
(bl

IO

Fig. 2.

(a) Conversion

of axfal strain to rotation of

electrolevel capsule; (b) constructfon of electrolevel

gauges; (c) effects of tilting

AND BURLAND

The major difference between the instruments

described in this Paper and the earlier designs
lies in the geometrical configuration which permits their use on 38 mm dia. samples. Fig. 2(b)
shows the construction of the new devices. In
addition to geometrical changes, the hinge
mechanisms have been improved by replacing
the original brass pivots with polylluorotetraethylene (PFTE) and by simplifying the construction of the hinges themselves. The capsule
which protects the electrolytic level from the
action of pressure and water is constructed from
stainless steel, as are the tubular arms BC and
AC. The gauges are fully submersible and have
been tested at pressures of up to 1500 kPa. The
electrolevels are mounted in diametrically opposite pairs on a sample using a rapidly curing
contact adhesive which bonds the brass footings
to the membrane. The gauges rely on the radial
effective stresses to anchor the footings to the
sample under test.
It should be noted that if the sample tilts
when loaded the output from each gauge is
made up of a strain component and a tilt component, as shown in Fig. 2(c). Provided the
sample is homogeneous the mean axial strain is
given by half the sum of the outputs of a pair of
two diametrically opposed gauges and the tilt is
given by half the difference of the outputs. The
ability to detect sample tilt is a valuable feature
of the gauge.
Jardine & Brooks (1984) have carried out
simultaneous measurements of surface strains
for chalk specimens using foil strain gauges
bonded to the sample and electrolevel gauges
mounted on the membrane. The experiments
showed that, over the considered strain range of
0.15 %, any relative movement between the
membrane and the sample could be neglected.
Moreover, Gens (1982) used an optical technique to demonstrate that the membrane only
moves in relation to the sample when large
strains are developed.
The resolution and range of the gauge have
been determined by a two-part procedure.
Firstly, routine calibrations were performed over
a displacement range of 15 mm by mounting two
opposing gauges on a micrometer winding frame
graduated to 0.01 mm. A typical displacement
voltage characteristic is presented in Fig. 3. A
third order polynomial regression analysis can
then be used to model the characteristic (with a
typical correlation coefficient of 0.999 99) within
the limits shown. To determine the resolution a
second stage of calibration was carried out by
mounting a high resolution, small travel, LVDT
on the central axis of the winding frame so that
the changes in output could be determined for

MEASUREMENT

329

OF SOIL STIFFNESS

Table 1.
Name

Material

Sea
Sea
Sea
Sea
Sea
Sea

Sample
preparation

Rl
R1.4
R2
R4
R8
11

North
North
North
North
North
North

clay
clay
clav
cla;
clay
clay

Reconstituted
Reconstituted
Reconstituted
Reconstituted
Reconstituted
Intact

I2

North Sea clay

Intact

13

North Sea clay

Intact

RMl

North Sea clay

RM2

North Sea clay

HRSl

Ham river sand

HRS2

Ham river sand

LCl

London clay

Remoulded
LI=O.18
Remoulded
LI=O.O9
Pluviated
R, = 0.149
Pluviated
R, = 0.848
intact

Cl

LC2

London clay
Upper Chalk

intact
1intact

c2

Upper Chalk

Intact

types. Unbonded low plasticity clays are materials which may be expected to demonstrate many
of the features incorporated into critical state
descriptions of soil behaviour (Schofield &
Wroth, 1968) where stiffness would be principally conditioned by the initial stresses and preconsolidation stress level. The London clay samples were considered to be typical of weathered
lower London clay, which is a weakly bonded
material that can develop a reorientated fabric
on thin shear bands after failure, and thus, when
tested, often displays a number of characteristics
which diverge from the predictions of critical
state soil models (see Lupini, Skinner & Vaughan, 1981). The Ham river sand is a uniformly
graded, angular sand in which stiffness could be
expected to be mainly related to its mode of
deposition, initial stress and density. In contrast
the intact, unfissured, chalk used for tests Cl
and C2 was a strongly cemented material in
which bond type and strength might be expected
to dominate the stress-strain behaviour.
EXPERIMENTAL

RESULTS

Reconstituted samples of low plasticity clay

The effective stress paths followed by the
reconstituted samples Rl, R1.4, R2, R4 and R8

Consolidation
details

K,, (see Fig. 6)

K, (see Fig. 6)
K, (see Fig. 6)
K, (see Fig. 6)
K,, (see Fig. 6)
Lightly overconsolidated
in situ, then sampled
As above, reconsolidated
field stresses
Heavily overconsolidated
in field. Swelled back
after sampling
Not consolidated

OCR
before
shearing

(initial):
kPa

1.0
1.4
2.05
3.73
7.4
-1.1

267
206
158
106
65
474

El.1

508

>50

46

10
43

Not consolidated
Isotropically
consolidated
Isotropically
consolidated
Overconsolidated

in situ
then sampled
As above
Cut from quarry face
isotropically consolidated
As above

PO

132

404

226

199
345

363

during undrained shear are presented in Fig. 6,

together with the deduced contours of developed axial strain (the strain shown in this
figure is the average strain from diametrically
opposite pairs of electrolevels). From Fig. 6 two
important observations can be made.
First, the effective stress paths followed by the
tests were initially both nearly vertical and
straight. However, in each case there was a
certain stress level where the paths sharply deviated and then travelled on to failure. The
latter portions of the effective stress paths were
taken as representing the post-yield portion of
each test. In every test yield was approached
after the development of only very small strains.
Test Rl reached peak deviator stress at an axial
strain of O.l%, and tests R1.4, R2, R4 and R8
all demonstrated sharp changes in stress path
direction at axial strains of less than 0.2%. It is
important to appreciate that for many practical
problems the working stresses will lie on the
vertical portions of the stress paths where the
strains are very small.
Second, the stress path for the normally consolidated sample Rl shows brittle behaviour
with a marked reduction in strength beyond
0.1% strain. Tests carried out by Gens (1980)
on another low plasticity clay showed similar

330

JARDINE,

SYMES AND BURLAND

125(-

-contours

of axialStrain%I

loo-

150

200

250

300

(IT,'+03')/2,kPa

-251

Fig. 6.

North Sea day stress paths for tests Rl, R2, R4. R8

behaviour. Small loops are apparent in the stress

paths for samples R1.4 and R2 close to failure.
If, instead of measuring pore pressures at the
base, a central piezometer probe had been used
(Hight, 1983) it is probable that these loops
would not have been observed.
Figure 7(a) shows the stress-strain characteristics of the reconstituted samples of low plasticity clay. Again it can be seen that the strains
over the initial range of stresses are exceedingly
small. In order to allow a meaningful analysis of
the initial stiff zone the strains have been replotted to a logarithmic scale in Fig. 7(b). The latter
figure shows a remarkably consistent trend, with
the strain required to achieve peak strength
steadily increasing with OCR. The scatter in the
early stages of test R2 was caused by vibrations
from a nearby motor and demonstrates that the
new gauges perform best in a still environment.
In Fig. 7(c) the stiffness characteristics of the
samples are examined by plotting the normalized secant modulus EJc, up to and including peak deviator using the same strain axes.
The use of the secant modulus E, is not meant
to imply that the soil behaviour is strictly elastic,
and has merely been taken as a convenient
measure of soil stiffness.
It is apparent that the initial stiffnesses obtained using local instrumentation are very much
higher than the values commonly measured in
routine soil triaxial testing. The stress-strain be-

haviour is non-linear and at strain levels above

1.0% the ratio of E,/c, can be seen to fall to
more familiar levels. While the existence of high
initial stiffnesses has been postulated to explain
anomalies between observed and predicted field
behaviour (see Simpson et al., 1979) the results
given in Figs 6 and 7 demonstrate that laboratory tests are capable of revealing both the high
stiffness and the detailed nature of pre-yield
behaviour. The characteristic variation of stiffness with strain is similar in all tests, but the
results from tests R1.4 and R2 demonstrate that
lightly overconsolidated clay shows a particularly
high normalized stiffness at low strain. The data
from tests Rl, R1.4, R2, R4 and R8 are summarized in Table 2. The column giving the times
to reach El = 0.1% strain gives a measure of
the differences between local and external strain
rates. In each case the local rate slowly increased until, at large strains, it equalled 4.5%
per day. As discussed later, normal anisotropic
consolidation reduced many of the potential errors in external strain measurement.
Intact samples of tow plasticity clay
The stress paths for tests 11, 12 and I3 are
given in Fig. 8(a) where the initial, postsampling, effective mean stress for sample I1 is
represented by point A and the reconsolidation
effective stress path for sample 12 is given by the
broken line BC. The initial applied effective

MEASUREMENT

stress for sample 13 is represented by point D.

axial strains which developed during shear
are indicated in the same figure, and details
of the samples initial conditions are given in
Table 1.
The observed errors in conventional overall
measurements of strain are discussed in a later
section. However, it is of interest to compare the
externally and locally measured strains for test
11 since this test is typical of routine high quality
testing of intact samples. The comparison is
shown in Fig. 8(b). It is apparent that the strains
deduced from external measurements of deflexion, even though corrected for load cell and
apparatus compliance, give much larger strains
than the values measured locally on the sample.
Indeed the conventional measurements completely mask the initial stiff behaviour of the
intact material. These errors are discussed more
fully later.
Referring again to Fig. 8(a), as was the case
for the reconstituted tests, all three intact samples demonstrate yield with a sharp deviation in
the effective stress path. The post-yield effective
stress path for the anisotropically consolidated
sample 12 differs markedly from that for the
comparable reconstituted sample Rl (see Fig. 6)
whereas the path followed by the heavily overconsolidated sample 13 is similar to those followed by R4 and R8. Ageing, bonding, sampling, or macrofabric features could all be responsible for such differences. With regard to the
strains, samples I2 and 13, like the reconstituted
samples, yielded at axial strains of 0.1% to
0.2%. In contrast, sample 11, which was tested
unconsolidated undrained, showed a less stiff
behaviour between the attainment of 0.1%
axial strain and the peak deviator condition.
The detailed stress-strain characteristics for
tests 11,12 and 13 are shown in Fig. 9 as plots of
(or-u3)/2
and EJc, against strain on semilogarithmic axes. A comparison between the
stress-strain response of samples 11 and 12
shows that reconsolidation of 12 produced only a
slight change in stiffness. The values of EJc, for
11 and 12 fall within the limits of the stiffnesses
found from the reconstituted tests (see Fig.
7(b)). The EJc, curve for sample I3 can be seen
to lie below the band of stiffness values determined for 8 2 OCR 3 1.0 with reconstituted material, but within a range that might be extrapolated for highly overconsolidated
samples.
Parameters from tests 11, 12 and 13 are given in
Table 2.
The stress paths for the two experiments on
the remoulded samples RMl and RM2 are
shown in Fig. 8(a), as are the strain levels at
various stages of the tests. Although the samples

331

OF SOIL STIFFNESS

The

a$
R4

asa
0

(a)

(4

3200

Tests Rl, R1.4, R2, R4 and R8: (a)

Fig. 7.
stress-sti
data; (b) stiessstraia data; (c) StsIwss
characteristics

had not been preloaded, the shapes of the stress

paths and the pattern of strains are similar to
those given by the overconsolidated samples of
intact and reconstituted
clay. The detailed
stress-strain and stiffness plots are given in Fig.
9 and may be seen to fall in the range extrapolated for overconsolidated intact or reconstituted samples. Summary parameters for tests
RMl and RM2 are given in Table 2.
Tests on London clay, Ham river sand and
chalk
The intial conditions for the tests LCl and
LC2 (London clay), HRSl and HRS2 (Ham
river sand) and the two chalk tests Cl and C2
are given in Table 1.
The stress paths followed during tests HRSl,
HRS2 and Cl and C2 are given in Fig. 10(a),
which also shows the strain levels at appropriate

332

JARLXNE,
Table 2.

Test

Rl
R1.4
R2
R4
R8
11
I2
13
RMl
RM2
HRSl
HRS2
LCl
LC2
Cl
c2

summaryof

test results

c,:

E, @ 0.01 %:

kPa

kPa

122
122
108
94
67
255
275
173
39.5
85.0
1085
1142
123
100
1350
1600

SYMES AND BURLAND

2.22 x
4.50 x
359 x
2.26 x
1.13x
5.10x
7.43 x
9.4 x
2.6 x
9.3 x
2.9 x
4.9 x
1.24x
1.20 x
5.7 x
4.0 x

5
c,

@Jo.01 %

1820
3690
3320
2400
1690
2000
2700
540
660
1090
270
430
1010
1200
4220
2500

lo5
lo5
lo5
10
lo5
lo5
lo5
lo4
lo4
lo4
10
lo5
lo5
lo5
lo6
lo6

@O.Ol

I:(.I)t:

830
2 180
2 270
2 130
1740
1080
1460
2 030
2 430
2 180
2 200
1210
550
600
15 500
11000

min
0.185
0.270
0.353
0.386
0.407
0.333
0.187
0.340
0.331
0.278
0.518
0.503
0.371
0.387
0.723*
0.854*

38
49
52
65
105
100
59
126
156
72
90
59
55
65
510
587

* Since both samples failed at +<O.Ol

L was taken here as E,~,.,,,IE,~O.OO,~.
7 r(c.r) corresponds to the time taken to develop E,=O.l
% in each test. For a rate of strain of
4.5 % per day tcO.i)would be 32 min.

water, and neither sample achieved an undrained critical state condition.

The stress paths of the chalk tests Cl and C2
are compared in the same figure. The samples
showed stiff behaviour up to brittle failure at
a, - 03 equal to 1331 kPa and 1620 kPa respectively. The failure strains for tests Cl and C2
were both around 0.075%.
The post-failure
behaviour can be seen to be characterized by a

The sand experiments

showed a stiff
response to loading over the initial portions of
each test but the samples rapidly lost stiffness as
the stress paths approached the dilatant part of
their state boundary surface. After yield the
stress paths curved to the right and climbed the
state boundary surface until, at large strains,
peak strengths were developed. In both tests
failure was initiated by cavitation of the pore-

intervals.

@ = 30

300-

(Anal

strams

mdlcaled

in %)

600
(9

+ n3)/2.

kPa

RM2
(31

Fig. 8(a).

Intact and remoulded stress paths for tests 11, I2, W, RMl

and RM2

MEASUREMENT

OF SOIL STIFFNESS

333

300

Ultlmatec

Local

measurements

Overall

corrected

measurements

100

(~2

of tuIcu
FL,
%

Apparent
linear
Eu = 4.8 X 10

Fig.

elasrlc
modulus
kPa, Eufcu = 188

Only test 12 athned

peak devtator at an
axial strain below 5%

\
\

0 005
0.01
0.1
1.0

calculaled

~~ and

E/C

E /c
f&l

from

FL

2353
2000
667
147

Ir&

E
'.

172
140

8(b). Stress-strain data for test I1

(al

255 kPa
__-

IELI

Comparison

2400

__u

progressive weakening with the effective stresses

roughly following unloading paths.
The stress paths for tests LCl and LC2 are
shown in Fig. 10(b). Both samples showed an
initially stiff response to loading which persisted
up to axial strains of around 0.1%. The stress
paths both deviated to the right after the attainment of 1.0% axial strain until peak strengths
were mobilized
at strains of 4.5% and 3.5%
respectively.
Both tests showed a steep postpeak loss of strength,
and examination
of the
samples after testing showed that polished shear
surfaces had formed within the specimens.
The
stress-strain
and stiffness characteristics
for the
tests described in this section are summarized
in
Fig. 11 together with Fig. 12, which summarizes
the results of all the tests reported in this Paper.
The plots
demonstrate
the following
main
points.

(a) The chalk samples showed brittle behaviour

(b)
Awal

strain

EC %

(b)

9. (a) Stress-stain data for tests 11,12,13, RMl

and RM2, (a) stiffness characteristics for tests 11, I2,
13, RMl, RM2 and R2
Fig.

with failure occurring

at ~~2 0.075%.
In
contrast, the London clay and sand samples
failed only after developing
large strains.
The chalk tests Cl and C2 gave the highest
normalized
stiffnesses, which equalled those
of the low plasticity clay at low strains but
exceeded them at strains above 0.0 1% . The
chalk samples also showed the most linear
behaviour.

334

JARDINE,

SYMES

AND

BURLAND

1600

1200

1
$

8OC

a"
I
-g

4oc

C
(0, + 03)/2

kPa

(4

20

/fA

50

K
LC2

m
%

80
1OU

120

. O

N,
c
6

o_eLCl

10

50

i:

0.7

&y

04

o-4

0.2

02

01
007

0.1

0 05

0.02z

0~0040.01
0.002

175

150

002
n /,1

200

250

(cl + 0,)/2:

300

kPa

@I

Fig. 10. (a) Tests on chalk and Ham river sand: stress paths for HRSl, HRs2, Cl and C2; (b) stress pati for
tests LCl and LC2 (axfaf strains: %)

MEASUREMENT

335

OF SOIL STIFFNESS
500 O-

400 O-

300 O"=
'1
UJ
200 O-!

300.

100 O-

OCI.01

Fig. 12. Summary of normafhd

s_eS

The
London clay tests showed stiffness
characteristics which were similar to that of
heavily overconsolidated or remoulded, low
plasticity clay.
(d) The normalized stiffness characteristics for
the Ham river sand, experiments HRSl and
HRS2, form a lower bound to all the results,
continuing the trend demonstrated by the
dilatant samples of low plasticity clay in tests
RMl, RM2, 13 and R8.
(c)

0 01

01

10

10

(b)

The test results from all the experiments are

further summarized in Table 2.
INTERPRETATION
Axial strain q:

0.001

0.01
0.1
Ax,alStrainEL %

WI

11. (a) Stressstraio data for tests Cl and C2;

(b) stres+stmfn data for tests HRSl and HRS2; (c)
stress-strafn data for tests LCl and LC2; (d) stitbws
characteristicsfor tests Cl, C2, HRSl, HRS2, LCl
andLC2
Fig.

In the past most laboratory studies of the

stress-strain characteristics of soils have been
hampered by the errors that are inherent in
conventional triaxial testing, particularly for
overconsolidated
soils,
and comprehensive
studies of soil stiffness at low strains are rare.
The test programme on the low plasticity clay
provides a body of data which can be used for
evaluating the small strain undrained stressstrain properties of that material. These properties may then be compared with the limited
number of results from the tests on the London
clay, the chalk and the Ham river sand in order
to highlight some of the factors influencing soil
stiffness. More detailed discussions of the small
strain behaviour of London clay and Ham river
sand are given by Costa Filho (1980) and
Daramola (1978).
It is recognized that much more experimental
work is required using the new techniques before
general
conclusions
can be drawn.

336

JARDINE,

SYMES

AND RURLAND

cell and ram). Four main conclusions

drawn as follows.

12.

can be

(a) For normally, anisotropically, consolidated

10.
8. unconsohdated

(b)

(cl

(4
Mean

axialstrain
EL:%

Fig. 13.

Tests on low pldkfty

clay: comparison of
internal and external corrected strain measurements

Nevertheiess, the results obtained are sufficiently encouraging to warrant a preliminary

discussion since a number of important observations can be made from the data presented in
the previous sections. To develop these points
the discussion is divided into three main parts
(a) an analysis of the strain errors implicit in
conventional triaxial testing, which will be
based on comparisons between observed
differences between the external and internal measurements of strain
(b) general features of the observed soil behaviour at small strains
(c) a discussion of the choice of parameters for
the comparison and normalization of the
experimental data.
Errors in conventional

stiffness measurements

Figure 1 shows that the overall measured

deflexion in a triaxial test is given by
A=AL+AT+ABT+AS+AeB+A,,,

(1)

Calibration of the load cell and ram characteristics for the apparatus used in this testing programme showed that their combined compliance
c could be taken, approximately,
as c =
5.4 x 10e4, where c = (A,+ A,,,)/F mm/N
and
F is the deviator force in newtons. Clearly such
deflexions are most important for strong materials, so that in tests Cl and C2, for example,
c was around SO times larger than the compliance of the samples themselves.
The significance of the remaining terms in
equation (1) may be assessed from Fig. 13 in
which the local measurements of axial strain, Ed,
are plotted against the ratio E,/E~ for all the
tests on the low plasticity clay (E, is the external
strain corrected for the compliance of the load

samples the corrected strain, E,, is close in

magnitude to the mean local strain.
For overconsolidated samples (e.g. R2, R4,
R8) the agreement between local and overall corrected measurements is far less satisfactory.
The difficulties in obtaining accurate load
cell stiffness calibrations can lead to overestimates of the stiffness and thus produce
values of FJQ_ less than unity.
For unconsolidated tests on intact or remoulded samples the disagreement between
local and external corrected measurements
is most severe, and E, can be an order of
magnitude greater than Ed..

The last observation is emphasized in Fig. 8(b)

(referred to previously) in which the locally and
externally measured strains are plotted against
shear stress for test Il. The bedding and other
errors implicit in the corrected strain E, give the
illusion of nearly linear straining up to about
0.6% axial strain, while the central portion of
the sample was behaving in a much stiffer and
less linear way. The initial slope of the apparent
stress-strain line corresponds to E,/c,190,
which is more than 12 times smaller than the
maximum secant EJc, deduced from the local
strain measurements at 0.05% strain.
In general the strains measured by each pair
of electrolevels during a test were dissimilar
until the average local strain exceeded 0.1%.
This can be explained by non-parallelism of the
sample ends, differential bedding and top cap
movements. It should be noted that if the nonparallelism of the ends were to cause the sample
to tilt when loaded, then the apparent strains
measured by the electrolevel gauges would
equally overestimate the larger local strain .sr.,
and underestimate the smaller strain E~.~.Although a number of dual axis gauges would be
required to describe fully the tilt experienced by
a sample the mean axial strain can be computed
from the data given by a pair of gauges as
cc = (E,., + ~&/2. A measure of tilt in relation to
axial strain is given by the ratio (or., - &,)/(e,, +
f&) = T the tilt ratio. The maximum values observed for this ratio at various mean strains are
summarized in Table 3. The results show that
the tilting action can be considerable and that
the use of paired local displacement gauges is
essential if the stress-strain behaviour below
e,_ = 0.1% is to be observed. For the remoulded
samples a ball seating was used and this accentuated tilting, particularly at large strains.

MEASUREhJENT

Table 3.

OF

SOIL

337

STIFFNESS

Maximum tilt ratios observed for tests on low plasticity day

Mean axial strain eL: %

Tilt ratio T for intact

and reconstituted samples*

0.005
0.01
0.05
0.1
1.0

2.0
1.5
0.9
0.4
0.1

Tilt ratio
T for remoulded
samples*
2.1
1.7
2.6
2.8
0.9

T = (on - H&I(HL, + 0L2).

* For parallel straining f3r1= or and T = 0, and if 19r_~
is negative T can exceed unity.
Summarizing, it is found that even the most
careful calibration of the load cell and ram
deflexions is not sufficient to allow external
measurements to be used to define the stressstrain characteristics of a soil accurately. Features such as bedding of the end platens and
tilting of the sample can lead to serious underestimation of soil stiffness. The electrolevel devices described earlier in this Paper offer a simple means of circumventing the errors which
invalidate the measurement of soil stiffness in
conventional triaxial tests.
Small srrain behaviour
It has been shown that the region of stress
space within which the tested soils exhibit very
stiff undrained behaviour is generally bounded
by the 0.1% axial strain contour. Such a low
strain region is shown in Fig. 14 for a number of
samples which have all been consolidated, under
K. conditions, to the same maximum stresses
before unloading to various overconsolidation
ratios prior to undrained compression (see also
Fig. 6). The 0.1% contour coincides with the
yield point for OCR= 1 but lies below it for
OCR>l.
The 0.1% strain contour shown in
Fig. 14 is not strictly a yield locus since drained
effective stress paths parallel to, but beneath, it
(e.g. along the swelling line) could involve yield
and large strains. Specimens undergoing differ-

ent stress history and/or modes of deposition

prior to undrained testing will usually have
different low strain regions. For example, for
specimen 11, which was sampled and tested unconsolidated undrained, the small strain region
lies well below the region observed by shearing
from the K0 swelling line (see Figs 6 and 8).
However, it is evident from Figs 6 and 14 that
the small strain region for undrained compression can be extensive and the stress paths for
many engineering problems will be within this
region.
For ease of comparison and presentation, the
initial undrained stress-strain
characteristics
may be represented by the following two indexes relating to stiffness and linearity.

(a) Stiffness is given by the undrained secant

(b)

modulus at 0.01% axial strain, E,o.o,,. It

may be expressed non-dimensionally
as
(E,/c,)~.,,,,
(E,lp~)o.ol,
etc., as discussed in
the next section.
An index of linearity is defined as L =
E,(,,.,JE,(,,.,,,,
where EUo.,, is the undrained
secant modulus at 0.1% strain. Straight line
behaviour then gives L = 1.0, and if the
modulus decreases with strain L < 1.0.

Values of L are given in Table 2 and it can be

seen that every test departed from straight line
behaviour over its small strain range. In general,
Undrained
stress path
for OCR = 1.0

Undrained

Fig. 14. Schematic drawiog of upper

reconstftuted low plasticity day

bound

to small strain range for

338

JARDINE,

SYMFS

Rl-4
R2
3000

Reconstituted

MLC2
-

HRS2

LCl

I3-.

HRSl

RM2

RMl

OCR

Fig. 15.

summary
of au tests

the values of L increased with overconsolidation

ratios, and test Rl (OCR= 1.0) showed the
smallest L value of 0.185. The two chalk specimens showed almost linear behaviour.
Choice of parameter for normalizing E,
It is not, at present, common to carry out
triaxial tests to determine undrained stiffness for
purposes of practical design and analysis. Most
engineers rely on correlations between stiffness
and a related, but more readily obtained,
parameter. For example, Ladd, Foot, Ishihara,
Schlosser & Poulos (1977) presented stiffness
data from SHANSEP tests which are normalized
by c, plotted against OCR. In their plots E, was
determined over given proportions of shear
stress increment rather than the fixed strain
increments used in this work.
The stiffness data given here in Figs 7(b), 9(b),
11(c) and 12 have also been normalized with
respect to the peak undrained shear strength.
Fig. 15 shows the curve of EJc, at 0.01% axial
strain against OCR for the reconstituted low
plasticity clay. Results from the other tests reported here are shown as single points. The
stiffness of the normally consolidated sample Rl
is perhaps misleading as the initial behaviour is
probably controlled by the amount of time permitted for secondary consolidation. However,
the data show that (EU/c,)o.Ol quickly increased
from the value at OCR = 1 to a maximum at an
OCR of about 1.4, and then steadily reduced
with increasing overconsolidation. Although the
stiffness given by the intact samples of the same

AND BUFUAND

clay can be seen to follow approximately the

same relationship, the remoulded tests RMl and
RM2 give values which only correspond to the
most heavily overconsolidated
reconstituted
samples. The ratio (EJc,),.,,,
ranged between
540 and 3700 for the low plasticity clay and the
normally consolidated tests Rl, I1 and I2 gave
values between 1800 and 2700.
The data for the comparative soils are also
given in Fig. 15 and show that the (E,/c,)~.~,
values for the chalk were similar to the maximum given by the North Sea clay. The London
clay results fell roughly in the mid-range but the
Ham river sand tests gave the smallest
values of all.
(EJcJ,.,,
The data from the triaxial tests show that even
with fixed rate of displacement compression
tests there is a wide range in the ratio EJc, for
a single clay. The undrained stiffness clearly
depends on strain level, stress history, method of
formation and, probably, strain rate. With other
soils mineralogy,
grading, macrofabric
and
cementation could produce different characteristics. It has, for example, been suggested by Ladd
et al. (1977) that, in comparable tests, the ratio
EJc,
is higher in lean clays than in more
plastic soils.
Although it is convenient to use the ratio
EJc, to compare different soil types and initial
conditions, the parameter cannot be considered
to be fundamental since the undrained shear
strength also depends on rate, total stress path,
sample disturbance and soil macrofabric. In particular, the use of c, can be confusing in soils
which develop orientated, residual, structures in
thin shear zones.
The initial mean effective stress po acting in a
sample has been used as an alternative parameter with which to normalize stiffness measurements (see Atkinson,
1973; Wroth, 1971).
While the undrained shear strength depends on
the conditions of testing, pOcan be measured in
the laboratory without ambiguity. In the field,
however, po will depend on KO and cannot be
calculated with such certainty.
Figure 16 shows the same data as Fig. 15, but
with the stiffness EUo.,,, normalized by pO in
place of c,. The pattern demonstrated by the
reconstituted low plasticity clay is familiar. The
normally consolidated tests again showed relatively low normalized stiffnesses but the other
reconstituted, intact and remoulded results fall
within a far narrower scatter than the c, normalized result, with (EJp,),,.,,
lying between
1700 and 2400.
The results for the Ham river sand, perhaps
fortuitously, plot close to the curve for the reconstituted North Sea clay, but the London clay

MEASUREMENT

results fall distinctly below the full line. The

chalk tests produced very high (E,/p,),.,,, values
but since their stiffnesses are probably controlled by bonding, rather than effective stress,
these values may be arbitrary.
From the experiments described in this Paper
normalization by p,, would appear to be preferable for uncemented soils. The ratio Eu/po is
seen to be less dependent on method of formation and stress history than E,/c,, and an effective stress approach is likely to be more useful
when the small strain laboratory techniques are
applied to drained behaviour.
The work described here deals only with undrained stiffness in triaxial compression. It is
evident that investigations are required into the
more general behaviour of soils in the small
strain range, where accurate radial strain measurements would be required, and that the influence of many parameters (including rate and
ageing effects) must also be assessed.
SUMMARY AND CONCLUSIONS
A new technique is described for the accurate
measurement of local axial strains on soil specimens in the triaxial apparatus. The strains are
measured using an electrolytic level device
which is simple to use, resolves relative displacements to less than 1 urn over a range of 15 mm
and is not damaged when the sample is taken to
failure.
The new technique was employed in a programme of tests principally on a low plasticity
clay from the North Sea with additional comparative tests being conducted on Ham river
sand, London clay and intact chalk, thus covering a wide spectrum of soil types.
The test results show that conventional external measurements of displacement contain errors which frequently mask the initial stressstrain characteristics of the soil and invalidate
their use in the determination of soil stiffness.
The errors in the external measurement of displacement mainly result from tilting of the sample, bedding on the end platens and the effects
of compliance in the apparatus.
In every test the low plasticity clay showed
highly non-linear, but very stiff, initial behaviour. The attainment of 0.1% axial strain
generally coincided with a marked loss of stiffness and could be taken as the limit to the small
strain range. Correlation with the undrained
effective stress paths shows that such a range
extends over the main part of stress space in
which soil would usually be considered elastic.
Such stiff initial behaviour is therefore likely to
be important in the analysis of many practical
problems.

339

OF SOIL STIFFNESS
3000r

13

01

6 810

20

40

RMl

RMZ

100

OCR

Fig. 16.

Summary of all tests

The stiffness ratio (E,/c,)~.~~~ was shown to be

strongly dependent on OCR for the intact and
reconstituted samples. Lightly overconsolidated
conditions produced the highest values of the
ratio, and heavily overconsolidated
and remoulded samples showed the lowest. The alternative non-dimensional ratio (E,/p,,),.,, was less
sensitive to OCR and method of formation.
Normally consolidated samples of intact and
reconstituted samples of the low plasticity clay
showed the least linear initial behaviour, and
gave values of L = (= Euo.,,/E,~,.,,,,) which were
lower than 0.2. Although linearity steadily increased with overconsolidation ratio the largest
value of L recorded for the clay was 0.407.
The limited number of comparative tests on
other materials shows that the small strain
characteristics of the low plasticity clay are not
unique. The values of (E,/c,)~.~,~ and (Eu/p,,)o.cll
for specimens of chalk, sand and London clay
exceeded the results obtained in conventional
tests, and in each case the small strain behaviour
was non linear. The cemented chalk samples
showed both the highest normalized stiffness
and the nearest approximation to linear stressstrain behaviour.
In summary, the techniques described in this
Paper make it possible to detect, simply and
reliably, mean local axial strains in triaxial tests
with a resolution of approximately 0.001%. In
the first programme of tests using the new
equipment observations have been made of the
undrained stress-strain characteristics of soils
which, without local strain measurements, could
only be inferred from field measurements. Although more research is required into the factors controlling the stiffness of soils, finite element analyses have been carried out using constitutive models based on the experimental data,
as described by Jardine, Potts, Fourie & Burland

340

JARDINE,

SYMES

(1984), which demonstrate that the initial small

strain characteristics of a soil are of great importance in the analysis of engineering
problems
and the interpretation

of in situ tests.

ACKNOWLEDGEMENTS
The samples
of low plasticity
clay from the
North Sea were provided
by BP International
Ltd and the Authors
are grateful to h4r W. J.
Rigden for his interest in the work and his per-

mission to publish the results. Thanks are also due

to Dr J. H. Atkinson for his helpful comments.
The Authors wish to acknowledge the importance of the contribution to this topic of Dr P.
R. Vaughan, who supervised the first small
strain studies conducted by Dr L. C. Costa Filho
and Dr 0. B. Daramola at Imperial College, and
also to thank their colleagues who have generously donated time and practical help to the
work described. Mr P. Smith and Mr N. Brooks
both provided particularly valuable contributions to the work, which was funded by the
Marine Technology Directorate of the Science
and Engineering Research Council.
REFERENCES

Arthur, J. R. F. & Phillips, A. B. (1975). Homogene-

ous and layered sand in triaxial compression.

Giotechnique 25, No. 4, 799-1815.
Atkinson, J. H. (1973). The deformation of undisturbed
London clay. PhD thesis, University of London.
Bishop, A. W. & Wesley, L. D. (1975). A hydraulic
triaxial apparatus for controlled stress path testing.
Gt?oorech&ue 25, No. 4, 657-670.
Brown. S. F.. Austin. G. & Overv. R. (19801. An
instrumented triaxial cell for cyc& loading of clay.
ASK%4 Geotech. Test. J. 3, No. 4, 145-152.
Brooks, N. J. (1983). The settlement of foundations on
chalk. MSc thesis, University of London.
Burland, J. B. & Symes, M. (1982). A simple axial
displacement gauge for use in the triaxial apparatus. Giotechnique 32, 1, 62-65.
Cooke, R. W. & Price G. (1974). Horizontal inclinometers for the measurement of vertical displacement in the soil around experimental foundations. Field instrumentation in georechnical
engineering, pp 112-125. Butterworths: London.
Costa Filho, L. M. (1980). A laboratory inuestigation
of the small strain behaviour of London clay. PhD
thesis, University of London.
Daramola, 0. (1978). The influence of stress history on
the deformation of sand. PhD thesis, University of
London.

AND

BUFUAND

Gens, A. (1980). Discussion: Design parameters for

soft clays. Froc. 7th Eur. Conf. Soil Mech., Brighton, 1979 4, 25-26.
Gens, A. (1982). Stress-strain and strength characteristics of a low plasticity clay. PhD thesis, University
of London.
Hight, D. W. (1983). Simple piezometer probe for the
routine measurement of pore water pressure in
triaxial tests on saturated soils. Gkotechnique 32,4,
315-322.
Jardine, R. J. & Brooks, N. J. (1984). The use of a
new axial displacement gauge for the determination of rock stiffness. In preparation.
Jardine, R. J., Potts, D. M., Fourie, A. & Burland, J.
B. (1984). The importance of small strain behaviour in the analysis of soil structure interaction.
In preparation.
Ladd, C. C., Foot, R., Ishihara, K., Schlosser, F. &
Poulos, H. G. (1977). Stress deformation and
strength characteristics.
Proc. 9th Int. Conf. Soil
Mech., Tokyo 3, 293-305.
Lupini, J. F., Skinner, A. E. & Vaughan, P. R. (1981).
The drained residual strength of cohesive soils.
GPotechnique 31, No. 2, 181-213.
Maswoswe, J. (1984). PhD thesis, University of London. In preparation.
Richart, F. E., Woods, J. D. & Hall, J. R. 1970.
Vibrations of soils and foundations.
New Jersey:
Prentice-Hall.
Roscoe, K. H., Schofield, A. N. & Thurairajah, A.
(1963). An evaluation of test data for selecting a
yield criterion for soils. Prcc. Symp. Laboratory
Shear Testing, ASTM STP 361, 111-128.
Schofield, A. N. & Wroth, C. P. (1968). Critical state
soil mechanics. London: McGraw Hill.
Simpson, B., ORiordan, N. J. & Croft, 0. D. (1979).
A computer model for the analysis of ground
movements in London clay. Gkotechnique 29, No.
2, 149-175.
Symes, M. J. & Burland, J. B. (1984). The determination of local displacements on soil samples. ASTM
Geotech. Test. J. In press.
Vogler, U. W. & Kovari, K. (1978).
Suggested
methods for determining the strength of rock materials in triaxial compression. Int. J. Rock Me&
Min. Sci. Geomech. Abstr. 15,No. 2, 47-51.
Wroth, C. P. (1971). Some aspects of the elastic
behaviour of over-consolidated clay. Stress-strain
behaviour of soils. Proceedings
of the Roscoe
Memorial
Symposium,
347-361.
Cambridge:
Foulis.
Yuen, C. M. K., Lo, K. Y., Palmer, J. H. L. &
Leonards, G. A. (1978). A new apparatus for
measuring the principal strains in anisotropic clays.
ASTM Geotech. Test. J. 1,No. 1.24-34.