Content uploaded by Robert Gerlai
Author content
All content in this area was uploaded by Robert Gerlai on Feb 12, 2019
Content may be subject to copyright.
Behavioural Brain Research 95 (1998) 191–203
Research report
Contextual learning and cue association in fear conditioning in mice: a
strain comparison and a lesion study
Robert Gerlai *
Genentech,Inc., Neuroscience Department,Mail stop c
72
,
460
Point San Bruno Boule6ard,South San Francisco,CA
94080
-
4990
,USA
Received 11 June 1997; received in revised form 21 October 1997; accepted 21 October 1997
Abstract
Fear conditioning with electric shock (unconditioned stimulus, US) paired with tone cue (conditioned stimulus, CS) has been
extensively applied in recent molecular neurobiological analysis of hippocampal dysfunction in mice because the context-depen-
dent test phase of this learning paradigm is claimed to detect hippocampal impairment in a specific manner, whereas the
cue-dependent test serves as a control situation independent of hippocampal function. These claims are based on hippocampal
lesion studies performed with rats and have not been conclusively confirmed with mice with specific hippocampal lesion.
Therefore, I investigated how hippocampal ibotenic acid lesion affects conditioned fear in mice. I confirm that extensive lesions
localized to the hippocampus impair context-dependent learning but also show that, unlike in the original rat studies, the
behavioral impairment is only partial. Furthermore, studying two inbred strains of mice (C57BL/6 and DBA/2) with highly
different hippocampal function, I show that the presence or absence of CS during training may influence the mouse’s ability to
learn complex multiple contextual stimuli in a genotype-dependent manner. I conclude that performance at the ‘context’ test may
be based on complex configural (hippocampal) learning but it can also be based on a more simple elemental (non-hippocampal)
learning thus leading to potentially false-negative findings in the analysis of hippocampal dysfunction. © 1997 Elsevier Science
B.V. All rights reserved.
Keywords
:
Contextual stimuli; Configural learning; Fear conditioning; Hippocampus; Ibotenic acid; Inbred strain; Mouse
1. Introduction
The hippocampus serves a crucial behavioral func-
tion. In humans it has been found to play roles in
declarative memory and forms of non-procedural
configural (relational) learning [35,36]. Animals with
hippocampal abnormalities have been shown to per-
form badly in tasks requiring exploration [8,16,27],
learning of novel stimuli (for examples see Ref. [36]),
spatial learning [12,26], contextual learning [23,31] and
in tasks involving the learning of multiple or complex
relationships of cues [6,32]. The accumulating data
from these studies suggest that the common feature of
hippocampal tasks is that they require learning multiple
environmental stimuli and the complex relationships
between such stimuli [10,11]. In addition to its interest-
ing behavioral function, the hippocampus has been in
the focus of attention because a neurophysiological
phenomenon, long-term potentiation (LTP), suggested
to underlie relational learning and memory [5], can be
elicited in a robust manner in this structure. Using
molecular genetic approaches such as transgenic and
gene targeting techniques, several investigators have
attempted to study hippocampal function, and the role
LTP plays in learning in vivo in mice (for examples see
Refs. [15,17]). Many of these studies used a behavioral
paradigm called fear conditioning to investigate
* Tel.: +1 415 2254120; fax: +1 415 2256240; e-mail: ger-
lai@gene.com
0166-4328/98/$ - see front matter © 1997 Elsevier Science B.V. All rights reserved.
PII
S0166-4328(97)00144-7
R.Gerlai
/
Beha6ioural Brain Research
95 (1997) 191–203
192
hippocampal dysfunction at the behavioral level [1–
3,21,34].
However, fear conditioning was originally worked
out for rats and not mice. The paradigm involved
training the experimental subjects in a conditioning
chamber where they received a mild electric shock
(unconditioned stimulus, US) that could be paired with
a conditioned stimulus (CS), usually a short tone signal.
A test trial to investigate the types of stimuli with which
the subject might have associated the shock was per-
formed later. Fear conditioning evokes a characteristic
natural fear response in rats that can be quantified as
the amount of immobility or freezing [4]. Rats have
been shown [23,31] to be able to associate two distinc-
tively different types of cues with the shock. They can
learn the simple association between US and CS (ele-
mental learning) but they also learn the so called con-
textual stimuli, i.e. the stimuli of the shock chamber
itself (configural learning). The latter stimuli represent a
complex set of cues that may not individually be associ-
ated with the shock itself, but together they define the
place where the shock occurred. According to the au-
thors of the rat fear-conditioning studies [23,31], learn-
ing the contextual stimuli requires a normally
functioning hippocampus. In support of this claim,
these authors showed that rats with electrolytic lesions
localized to the hippocampus were unable to remember
the contextual stimuli (retrograde amnesia), i.e. they
exhibited an almost complete abolition of freezing when
put in the chamber where they were shocked before.
This impairment was not due to performance factors
because in the cue test the lesioned rats were perfectly
capable of responding to a single associative cue (CS)
with which the shock was previously paired. Thus fear
conditioning was claimed to detect hippocampal dys-
function in a specific manner in rats. However, rats and
mice may behave differently in learning paradigms (for
examples see Refs. [15,18]). Furthermore, the timing of
hippocampal disruption, i.e. whether pre-training ‘le-
sion’ such as related to a mutation in the mouse genetic
studies, or post-training lesion such as in the quoted rat
studies were used, may also modulate results [25]. Since
the primary subject of transgenic and gene targeting
studies is the house mouse and since a potential muta-
tion-related hippocampal impairment is already present
before training, it is crucial to investigate whether mice
with specific hippocampal pre-training lesion show a
context-specific impairment in fear conditioning similar
to that of rats, an assumption that has not been confi-
rmed despite the frequent usage of the fear-condition-
ing paradigm in transgenic mice studies. To answer this
question I generated mice with lesions specific to the
hippocampus and analyzed their fear-conditioning
performance.
The nature, e.g. the modality and relative salience, of
cues assumed to be part of the context in fear condi-
tioning is not understood. The implicit assumption that
all contextual cues are equal under all circumstances
and for all genotypes of mice may be incorrect. Conse-
quently, the complexity, and thus the hippocampal-de-
pendent nature, of the context test may be questionable
under certain circumstances. For example, CS presenta-
tion, or the lack of it, may influence whether mice use
a configural learning strategy and learn complex multi-
ple contextual stimuli or apply a simple elemental learn-
ing strategy and memorize only a subset of cues or
maybe a single salient cue from the context. Using two
inbred strains of mice, DBA/2 and C57BL/6, I investi-
gated this question. These mouse strains have been
found to exhibit marked differences in hippocampal
learning tasks [26,28,37,39– 41], as well as in hippocam-
pal neuroanatomy and biochemistry [8,14,22], but per-
form well in non-hippocampal tasks. Comparison of
these strains does not require an invasive surgical ma-
nipulation and the differences between them are due to
genetic factors. Therefore their analysis may be more
relevant to studies applying gene targeting and trans-
genic approaches. Using these strains of mice, I investi-
gate whether CS presentation during training interferes
with acquisition of contextual stimuli and with the
ability of mice to respond to such stimuli at a later test
(for other aspects of differences between these strains in
fear conditioning see Ref. [24]).
2. Methods
2.1.
Animals and housing
For hippocampal ibotenic acid lesioning, 4-month-
old C57BL/6 mice (males) were used. Sixteen animals
received the lesion (for details see below) and 14 were
given vehicle injection only. For the strain comparison,
3-month-old C57BL/6(n=33) and DBA/2(n=32)
males were tested. In addition to these animals, another
group of naive mice (C57BL/6, n=6; DBA/2, n=9)
were also tested for their ambulatory activity in the
open field. All mice were tested once after training, i.e.
they were monitored either in the context or the cue
test. Mice were kept in groups of 10 in standard plastic
cages (45×24 ×15 cm) on sawdust bedding in the
same room in which temperature (2091°C) and light
cycle (12/12 h light/dark, lights turned on at 06:00 a.m.)
were controlled. Food and water were available ad lib.
For fear conditioning (training phase), mice from each
genotype (i.e. both DBA/2 and C57BL/6) were ran-
domly divided into two groups: tone cue present (cue)
or tone cue absent (no cue). For the test phase, these
mice were further divided and assigned to either the
context test or the cue test so that, after training, a
mouse was tested only once (for sample sizes see
figures). In the open field, previously untested, naive
mice were measured.
R.Gerlai
/
Beha6ioural Brain Research
95 (1997) 191–203
193
2.2.
Lesioning and staining procedures
Mice (20– 30 g body weight) were anaesthetized with
a ketamine (9 ml/kg body weight), xylazine (4.5 ml/kg),
acepromazine (0.9 ml/kg) i.p. injection, and were placed
in a stereotaxic apparatus. They received four injections
of either ibotenic acid (0.1 ml volume/injection, 10
mg/ml of freshly made ibotenic acid in physiological
saline) or vehicle (physiological saline) in their
hippocampi [16] bilaterally (site 1, AP −1.2, ML 1.4,
DV 1.8; site 2, AP −2.2, ML 2.5, DV 1.8; site 3, AP
−3.0, ML 3.2, DV 3.0; site 4, AP −3.0, ML 3.2, DV
4.0; as measured from Bregma in mm). All mice recov-
ered from surgery within 24 h and showed no signs of
health problems or abnormal motor behavior as long as
observed (6 months post-operation). Mice were tested
in the fear-conditioning paradigm for 3 weeks after the
operation. All mice tested behaviorally were later sac-
rificed to study the extent of hippocampal damage.
They were anaesthetized with nembutal, and perfused
with 4% paraformaldehyde in phosphate-buffered sa-
line (PBS, pH 7.2) transcardially. The brains were
removed and first placed in 4% paraformaldehyde PBS
solution at 4°C for 12 h and then in a 20% sucrose
solution at 4°C for 12 h. Subsequently the brains were
cryosectioned using a sliding microtome with a freezing
stage. Forty-mm thick frontal sections were collected
throughout the entire hippocampus and later stained
using cresyl violet (NISSL) stain. The stained sections
were mounted on slides, dried, baked at 50°C, dehy-
drated with a series of ethyl alcohol solutions, and then
covered. Processing of brains, sectioning, and staining
procedures were carried out at the same time for the
lesioned and control mice.
2.3.
Fear-conditioning apparatus and procedure
Fear conditioning was carried out using the Gemini
active/passive avoidance apparatus (San Diego Instru-
ments). The shock chamber (25×28 ×16.5 cm
(width×depth ×height)) had three black acrylic walls
and one transparent wall. Its ceiling was made of white
acrylic, through which a 10-W light bulb could illumi-
nate the chamber. The floor of the chamber was made
of metal bars that delivered the scrambled electric foot
shock (1 s, 0.7 mA). A tone generator mounted on the
ceiling could provide a constant high pitch (80 dB, 2900
Hz) tone cue (20 s). The apparatus was placed in a
sound-attenuating hood with a glass window, through
which mice could be observed. The recording sequence
of mice was randomized across genotypes and cue
groups. The training and test sessions lasted for 6 min
for each mouse. After each individual training session,
the apparatus was cleaned with water and thoroughly
dried. For context and cue test sessions, the test ap-
paratuses were cleaned with clydox and then thor-
oughly dried, so that a potential similarity in olfactory
cues between training and test was reduced.
The stimulus presentation was as described in Ref.
[2]. It was determined by a stimulus script file written
on an IBM PC that controlled the apparatus as follows:
0 s recording start; 160 s tone on; 179 s shock on; 180
s tone and shock off; 220 s tone on; 239 s shock on; 240
s tone and shock off; 280 s tone on; 299 s shock on; 300
s tone and shock off; 360 s recording end. Thus the
mice were allowed to explore the apparatus undisturbed
for 160 s then they were given a 20-s tone signal and, at
the 179th second, a 1-s long electric shock. The shock
and the tone cue were terminated at the same time
(180th s). This procedure was repeated for a total of
three tone– shock pairings during the 6-min training
session. The tone was administered only to the ‘cue’
group. The ‘no cue’ group received the shocks alone.
After training was finished, mice were returned to their
community cage and left undisturbed. The next day
they were exposed to either the context test or to the
cue test. The context test was carried out in the same
shock chamber where the mice were trained the day
before. However, no shock and no tone cue was deliv-
ered. The other group of mice was observed in the cue
test. For this test, mice were placed individually in a
chamber similar in dimensions to those of the shock
chamber. This new chamber, however, was made highly
different from the shock chamber. All of its walls were
transparent. Its bottom had no metal bars and it was
placed in a part of the room with different extra maze
visual cues compared to what could be seen from the
original location of the training shock chamber. The
new chamber was equipped with a tone generator iden-
tical to the one used during training. Both the ‘cue’ and
the ‘no cue’ group received the tone cue during the test.
The timing of tone presentation was as explained in
training above. However, no shock was given.
During recording, the experimenter remained ob-
scured to the experimental subjects whose behavior was
recorded by a video camera. The recordings were later
replayed and quantified. Freezing has been found to be
a natural behavioral response elicited by pain or fear
[4,13,23], and it has been used as a measure of learning
and retention of memory associated with fearful or
painful experience. Although freezing has been a useful
indicator of fear, the techniques with which it has been
measured may not be optimal. Automated recording of
immobility using photocell detector systems or observa-
tion-based time-sampling methods [2,24,28] may not be
precise. The former technique, for example, is unable to
differentiate between non-locomotory behaviors and
freezing, and the latter method has low resolution if a
large sampling interval is employed or makes scoring
difficult if a small sampling interval is used. Therefore,
I employed an alternative method frequently applied by
ethologists. With an event recorder computer software
R.Gerlai
/
Beha6ioural Brain Research
95 (1997) 191–203
194
(Noldus Observer) written for the Macintosh, I
recorded the sequence of motor/posture patterns with
high precision in a continuous manner (for detailed
description of the technique and behavioral parameters
see Ref. [20]). In addition to freezing, a number of
behavioral elements were also recorded as described in
Ref. [20]. Briefly, either the relative duration of, i.e.
percent of time spent with a particular element, or its
frequency, i.e. the number of times it occurred during
the recording, session was measured and later analyzed.
The following behavioral elements were recorded: mo6-
ing (relative duration), locomotion and active behaviors
including rearing and leaning against the wall; grooming
(relative duration), stereotypical fur licking and face
cleaning movements; jumping (frequency), quick leaping
movements whereby the body is lifted off from the
ground; 6ocalization (frequency), squeaking; escaping
(frequency), a single bout of quick stereotypical run-
ning most often to a corner of the cage.
In addition to fear conditioning, previously untested
naive mice were monitored in the open field, a 45×
45×37-cm transparent acrylic cage (San Diego Instru-
ments) that was placed in a gray acrylic shelf unit
illuminated by dim light. The experimental animal was
placed in the center of the open field and its movement
was monitored by an 8×8 array of infrared photocell
detectors for a period of 20 min. The activity counts
(number of photobeam breaks) were recorded by an
IBM-compatible microcomputer and expressed in
counts per 1-min bins. After each session, the open field
was thoroughly cleaned with clydox, water, and then
dried.
2.4.
Statistical analysis
Relative duration of freezing was calculated for each
1-min interval of the training, context test, and cue test
sessions. In the case of the other behavioral elements,
only the training session data were used and the data
were analyzed separately for the first half (0– 179 s, i.e.
no shock given) and second half (179– 360 s, i.e. shock
given) of the session. For statistical calculation and
data handling, SYSTAT 5.2, Macintosh version, statis-
tical software package was used. Two- and three-way
repeated measure variance analyses (ANOVAs) were
carried out to test the effect of lesion (ibotenic acid-le-
sioned, vehicle-injected control), genotype (C57BL/6or
DBA/2), cue (tone-cue present or absent during train-
ing) and time (six 1-min intervals, the repeated measure
factor). In the case of significant main effects or interac-
tion terms, post hoc multiple-comparison tests, such as
Tukey Honestly Significant Difference (HSD) tests were
applied. The variance homogeneity criterion was
checked by Bartlett’s test. In case of inhomogeneous
variances, non-parametric tests such as Mann-Whitney
U-tests were used to confirm statistical significance.
3. Results
Ibotenic acid injection led to severe hippocampal
lesions (Fig. 1B,D,F) but affected no other brain areas.
Proximal to the four injection sites, the hippocampus
was completely ablated, whereas further away the stain-
ing showed pyknotic, most probably dead, neurons.
The lesion could be observed throughout the entire
hippocampus in both sides of the brain. The affected
brain area and the severity of lesion was consistent
across all ibotenic acid-injected mice. In comparison, all
vehicle-injected animals (Fig. 1A,C,E) showed cell body
layers in their hippocampus with normal cellularity and
cytostructural features.
Ibotenic acid-lesioned animals showed no increased
mortality, were healthy, moved actively, and exhibited
no obvious signs of any defects after recovering from
surgery. When exposed to the novel fear-conditioning
chamber, they exhibited no increased immobility or
lack of grooming behavior (Fig. 2A) for the first 3 min
of the session, during which no shocks were delivered
(moving t=0.045, df=16, p\0.95; grooming t=
1.343, df=16, p\0.19) compared to vehicle-injected
control mice. When delivered a shock (Fig. 3A), they
exhibited a significant freezing response which was not
different from that of the control mice. ANOVA
showed no significant lesion effect (F
1,32
=0.126, p\
0.725) a significant time effect (F
5,160
=116.12, pB
0.0001) and a non-significant lesion×time interaction
term (F
5,160
=1.17, p\0.30). The relative duration of
moving and grooming was dramatically decreased (Fig.
2B) after shock delivery in both the lesioned and con-
trol mice, which were not different from each other
(moving t=1.371, df=32, p\0.18; grooming t=
1.626, df=32, p\0.11). It is also notable that the
frequency of jumping, vocalization, and escaping (Fig.
2C) was also not significantly different between the
lesioned and control mice (jumping t=0.683, df=32,
p\0.49; vocalization t=1.608, df=32, p\0.11; es-
caping, Mann-Whitney U=103, p\0.06), suggesting
that hippocampal ibotenic acid lesion did not lead to
nociceptive changes and had no significant effect on
motor responses.
In the context test, ibotenic acid-lesioned animals
showed a significant impairment (Fig. 3B) compared to
control (ANOVA, lesion F
1,20
=7.71, pB0.012).
Changes across time proved to be non-significant
(ANOVA, time F
5,100
=2.025, p\0.08; time×lesion
interaction F
5,100
=0.290, p\0.91), suggesting that
both lesioned and unlesioned mice exhibited a stable
freezing response throughout the test session. It is
notable, however, that although impaired, ibotenic
acid-lesioned mice exhibited a significant amount of
freezing (on average 55% compared to the 0% pre-
shock values), a response that appears to be specific to
the context (compare the context test freezing values to
R.Gerlai
/
Beha6ioural Brain Research
95 (1997) 191–203
195
Fig. 1. Representative transverse brain sections showing the hippocampal formation in mice. Panels A, C, E (from anterior to posterior):
vehicle-injected mice; panels B, D, F: ibotenic acid-injected mice. Scale bar is indicated. Note the significant ibotenic acid-induced lesion
throughout the hippocampus of the lesioned mice.
those obtained in the cue test before cue presentation,
i.e. Fig. 3B,C). Therefore, one may conclude that mice
with hippocampal ibotenic acid lesion, although im-
paired, are able to exhibit a significant fear response
when exposed to contextual stimuli. Unlike in rats
[23,31], the fear response is not abolished by the
hippocampal lesion in mice.
In the cue test, both lesioned and unlesioned mice
were capable of responding to the tone cue (CS), how-
ever, before tone presentation both groups of mice
showed only low levels of freezing. ANOVA revealed
no significant lesion effect (F
1,10
=4.682, p\0.07), a
significant time effect (F
5,50
=15.357, pB0.0001) and a
non-significant lesion×time interaction (F
5,50
=0.417,
p\0.80), suggesting that both lesioned and unlesioned
mice responded equally well to the tone cue. These
findings suggest that lesioned mice, similarly to control,
were able to freeze and were motivated to respond to
CS but exhibited no generalized immobility. These re-
sults also show that the impairment seen in the context
test is specific to that task.
Analysis of C57BL/6 and DBA/2 inbred strains
yielded results somewhat analogous to those of the
control and hippocampal-lesioned mice. Both geno-
types of mice were active in the shock chamber before
any shock was administered (Fig. 4A), and although
C57BL/6 appeared to be somewhat more active the
difference between the strains was not significant (mov-
ing t=1.712, df=63, p\0.09). Furthermore, the
strains performed an almost identical amount of
grooming (t=0.959, df=63, p\0.34). After shock
presentation, however, both genotypes of mice de-
R.Gerlai
/
Beha6ioural Brain Research
95 (1997) 191–203
196
Fig. 2. Behavioral elements of vehicle- (black) and ibotenic acid-lesioned (striped) C57BL/6 mice recorded during training. Bars represent the mean
of relative duration or frequency of the element. Error bars represent standard error. Sample sizes (n) are also indicated. (A) Moving and
grooming during the first 3 min of training (no shocks were given during this period). (B) Moving and grooming during the second 3 min of
training (three shocks were given during this period). (C) Frequency of jumping, vocalization and escaping during the second 3 min of training
(three shocks were given during this period). Note that the latter three behaviors did not occur during the first 3-min period. Also note that no
significant differences (p\0.05) were found between vehicle- and ibotenic acid-injected mice (escaping bordered the significance level).
creased their moving and grooming duration dramati-
cally. Nevertheless, moving duration was significantly
smaller in DBA/2 compared to C57BL/6 (Mann-Whit-
ney U=842, pB0.001) but grooming did not differ
(t=0.168, df=63, p\0.86). Analysis of freezing re-
sponses during training (Fig. 5A) showed that neither
C57BL/6 nor DBA/2 mice exhibited any freezing before
the first shock was applied. This suggests that they were
not afraid of the chamber or the handling procedure
itself. It also shows that neither strain of mice re-
sponded with freezing to the (first) tone cue alone. In
response to the shock, however, both strains exhibited a
dramatically increased freezing reaction. Two-way re-
peated measure ANOVA revealed a significant geno-
type (F
1,63
=16.61, pB0.0001) and time
(F
5,315
=416.41, pB0.0001) effect as well as a signifi-
cant time×genotype interaction (F
5,315
=8.19, pB
0.0001). Tukey HSD test indicated that DBA/2
exhibited more freezing at intervals 4 and 5 (pB0.05)
compared to C57BL/6. Interestingly, the transient dif-
ference in freezing responses between the two strains
appears to correlate well with the number of jumping
the mice performed: DBA/2 exhibited a significantly
higher number of jumping, a direct response to shock
(t=3.402, df=63, pB0.001) compared to C57BL/6,
and also showed an elevated frequency of escaping
(Mann-Whitney U=411.5, pB0.02) suggesting that
DBA/2 were more sensitive to the shock itself. The
frequency of vocalization showed no significant differ-
ences between the strains (t=0.210, df=63, p\0.83).
In addition to shock sensitivity, the differences seen
between the strains in their freezing response may also
R.Gerlai
/
Beha6ioural Brain Research
95 (1997) 191–203
197
Fig. 3. Behavior (relative duration of freezing) of hippocampal ibotenic acid- and vehicle-injected mice in the fear-conditioning paradigm. (A)
Training; (B) context test; (C) cue test. Error bars represent standard error. Sample sizes (n) are indicated. The three tone signals are shown as
solid lines under the graphs (training and cue test). During training the tone signals were paired with shock (arrows). No significant differences
could be observed during training between lesioned and unlesioned mice. During the context test (B), although ibotenic acid-lesioned mice
performed significantly below the level of vehicle-injected control mice, they still exhibited a significant freezing specific to the context. In the cue
test (C), all mice responded with significant increase of freezing to the tone cues.
be due to their general activity level, C57BL/6 being
more active than DBA/2. This is supported by our data
(Fig. 5D) obtained in another apparatus, an open-field
activity-monitoring cage in which C57BL/6 mice exhib-
ited a somewhat elevated activity upon exposure to
novel stimuli compared to DBA/2: ANOVA showed
that the strain effect bordered significance level (F
1,13
=
3.52, p=0.08), the time effect was significant (F
19,247
=
4.22, pB0.0001) and strain×time interaction
non-significant (F
19,247
=0.62, p\0.89).
The results of the context test are shown in Fig. 5B.
Three-way ANOVA revealed a significant genotype
(F
1,33
=56.19, pB0.0001), cue (F
1,33
=5.84, pB0.02)
and time (F
5,165
=14.68, pB0.0001) effect. Among the
interaction terms, only time×genotype interaction bor-
dered the level of significance (F
5,165
=2.07, p=0.07),
the other interaction terms were non-significant. These
results suggest that: first, C57BL/6 mice responded to
the contextual stimuli with higher freezing values com-
pared to those of DBA/2. This is a notable finding since
C57BL/6 were found generally more active before
shock presentation and they were also less responsive to
shock. The second important conclusion from the re-
sults of the overall ANOVA is that mice that received
no tone cue during training the previous day froze
longer in response to the context compared to those
that were trained with the tone cue. And, finally, the
results also show that extinction of freezing occurred in
both strains.
It is notable that ANOVA has been shown to be
insensitive to detect interactions [42], therefore a Tukey
HSD test was carried out to investigate whether the
effect of tone cue presentation during training (i.e. its
presence or absence) affected the performance of the
two strains of mice the next day at the context test
differently. Tukey HSD confirmed what can be seen on
Fig. 5B. C57BL/6 mice were unaffected by cue presen-
tation: the apparent difference between the ‘no cue’ and
‘cue’ groups is not significant (p\0.05). However, in
DBA/2 mice the ‘no cue’ group exhibited a significantly
increased (pB0.05) freezing value at intervals 1, 2 and
4 compared to the ‘cue’ group. Although a potential
ceiling effect cannot be ruled out in case of C57BL/6,
these results show that the absence of tone cue during
training significantly increased freezing in DBA/2 but
not in C57BL/6 mice during the context test the next
day.
The results of the cue test are summarized in Fig. 5C.
ANOVA showed a significant genotype (F
1,24
=15.66,
pB0.001), cue (F
1,24
=21.30, pB0.0001) and time
(F
5,120
=9.87, pB0.0001) effect. The interaction terms
time×genotype (F
5,120
=2.36, pB0.05) and time×cue
(F
5,120
=14.35, pB0.0001) were also significant. The
other interaction terms were non-significant. These re-
sults, as well as Fig. 5C, clearly demonstrate that
freezing response in the cue test was genotype and cue
treatment dependent. Tukey HSD test showed that
before any tone presentation (intervals 1 and 2) C57BL/
6 mice performed significantly more (pB0.03) freezing
R.Gerlai
/
Beha6ioural Brain Research
95 (1997) 191–203
198
Fig. 4. Behavioral elements of C57BL/6 (black) and DBA/2 (hatched) mice recorded during training. Bars represent the mean of relative duration
or frequency of the element. Error bars represent standard error. Sample sizes (n) are also indicated. (A) Moving and grooming during the first
3 min of training (no shocks were given during this period). (B) Moving and grooming during the second 3 min of training (three shocks were
given during this period). (C) Frequency of jumping, vocalization and escaping during the second 3 min of training (three shocks were given
during this period). Note that the latter three behaviors did not occur during the first 3-min period. Also note that moving relative duration was
significantly smaller and jumping and escaping frequency was significantly higher in DBA/2 during the shock period compared to C57BL/6.
than DBA/2, irrespective of whether they received a
tone cue during training or not. It appears that C57BL/
6 mice exhibited a transfer effect between the old and
the new context and were fearful in the new context,
whereas DBA/2 mice were not. Upon tone presenta-
tion, those DBA/2 and C57BL/6 mice that were trained
with the tone cue responded with a significant (pB
0.05) increase of freezing (interval 4), i.e. both made the
association between shock and tone and were able to
respond to tone alone. Interestingly, however, the ele-
vated freezing response extinguished significantly (inter-
vals 5 and 6, pB0.05) in DBA/2, whereas it remained
unchanged in C57BL/6, a phenomenon also observed in
aCaMKII mutant vs. control mice [34]. It is also nota-
ble that, as expected, those mice that received no tone
cue during training (‘no cue’ DBA/2 and C57BL/6
groups) did not change their freezing response upon
tone presentation confirming that the tone itself was
most probably a neutral stimulus to these animals.
4. Discussion
The results presented here show that hippocampal
ibotenic acid lesion in mice, unlike in rats [23,31], does
not lead to a complete abolition of freezing in response
to contextual stimuli. Interestingly, this is highly
analogous to what has been observed in null mutant
R.Gerlai
/
Beha6ioural Brain Research
95 (1997) 191–203
199
Fig. 5. Behavior of two inbred strains of mice (C57BL/6 and DBA/2) in the fear-conditioning paradigm (freezing relative duration: (A) training;
(B) context test; (C) cue test) and in the open field (ambulatory activity count, D). Error bars represent standard error. Sample sizes (n) are also
indicated. The three tone signals are shown as solid lines under the graphs (training and cue test of fear conditioning). Note, however, that the
‘no cue’ groups received no tone signal during training. The shocks are represented by the arrows. Note the significantly increased freezing of
DBA/2 at training (A) intervals 4 and 5. Also note that DBA/2 mice are generally impaired at the context test (B), however, the impairment is
more pronounced in the cue-trained group. In the cue test (C), note that C57BL/6 mice performed significantly more freezing before tone
presentation compared to DBA/2. In the open field, C57BL/6 mice show a somewhat increased activity compared to DBA/2. For statistical
calculations and detailed results see the Section 3.
(KO) mice generated by gene-targeting techniques. KO
mice exhibited a freezing response to contextual stimuli
considerably above preshock baseline (0%) level. For
example, aCaMKII heterozygous mutant mice showed
a 62% freezing response compared to 73% of the con-
trol mice for the first minute of the context test after
being trained with five CS– US pairing the day before
[34]. mGluR1 null mutant mice with impaired
hippocampal LTP showed approximately 50% freezing
response compared to the 75% value of control mice
[2]. PKCgnull mutant mice, also with impaired LTP as
well as impaired spatial learning, exhibited approxi-
mately 50% freezing compared to 70% of the control
mice [1]. Transgenic mice carrying a mutant form of
CaMKII gene (with a point mutation at Asp-286)
exhibited significantly altered hippocampal synaptic
plasticity and spatial learning, yet they were found to
show ‘normal’ contextual learning performance (unal-
tered freezing levels) in the fear-conditioning paradigm
[3]. Two strains of mice with disrupted genes of iso-
forms of PKA showed impaired mossy fiber LTP but
no context-dependent impairment in fear conditioning
[21]. The only study [7] in which a remarkable reduc-
tion of freezing to context was found in mice applied
aspiration lesion techniques. However, these mice suf-
fered from an extensive brain damage, a lesion that
extended well beyond the hippocampal formation, and
no data were presented (e.g. a control cue test) to
investigate general learning or performance deficits. A
recent work [24] also appears to contradict the findings
of the aspiration lesion study: two inbred strains of
mice and their F1 hybrids showed a robust freezing
R.Gerlai
/
Beha6ioural Brain Research
95 (1997) 191–203
200
after lesioning. Unfortunately, however, investigators of
this study applied an unusual neurotoxic lesioning pro-
cedure (using a kainic acid and colchicine cocktail)
which led to brain damage affecting not only the
hippocampus but also neocortical, entorhinal and tha-
lamic areas, as well as the amygdala. In summary, it
appears that specific hippocampal dysfunction, at least
when induced by gene targeting, does not abolish freez-
ing to context in mice.
Is not the hippocampus important for learning com-
plex contextual cues in mice? If it is, how could mice
with hippocampal lesion or dysfunction solve the con-
text test and freeze? One potential explanation is the
following. Under certain circumstances the experimen-
tal subjects may be able to focus their attention to a
single stimulus or a limited number of stimuli of the
context, and thus reduce the complex configural feature
of the context learning task to a simple elemental
learning task. Such an elemental learning strategy may
facilitate fear responses in the context test the next day
since the experimental subjects may recognize the single
stimulus learned. In support of this argument consider
that not all cues of the context may be equally salient
and some may be paid special attention by the mice due
to species-specific genetic predisposition. Ethologists
have shown the existence of ‘key’ stimuli (for review
and examples see Refs. [15,18,19,38]) and they also
confirmed that such stimuli may not have to be pre-
sented in temporal contingency or contiguity with the
unconditioned stimulus (the shock in fear condition-
ing), yet these stimuli may elicit vigorous learning under
certain circumstances. Such a salient cue from the
context thus may effectively reduce the task to a simple
elemental learning situation.
In addition to the characteristics of the contextual
stimuli, the timing or the type of lesioning may also
significantly affect whether elemental or configural
learning is used by rodents [25]. Experimental subjects
trained with a previously disrupted hippocampus may
not be able to perform configural learning and thus
may be forced to use an elemental learning strategy,
e.g. learn a single stimulus from the context, a strategy
that bypasses the hippocampus. Note that in the origi-
nal rat studies [23,31], experimental subjects were le-
sioned after training, which led to a severe impairment
at the context test. However, in the present study, as
well as in the analyses of null mutant mice [1– 3,21,34],
the hippocampal dysfunction was induced before train-
ing and the ‘lesioned’ mice were exposed to the training
procedure with an already dysfunctional hippocampus.
These mice apparently performed relatively well at the
‘context’ task implying that the task was solved by
them using a non-hippocampal strategy.
This suggestion is supported by the findings of a very
recent study [25] in which the effect of different types of
hippocampal lesions as well as the effect of pre- vs.
post-training lesioning was analyzed in rats. The extent
of anterograde impairments in contextual fear condi-
tioning of rats was found to be dependent upon
whether the hippocampal lesions were induced by neu-
rotoxic (NMDA) or electrolytic lesioning procedures.
The former (similar to our present ibotenic acid lesion-
ing) destroys the neurons themselves, the latter only the
axonal pathways. Interestingly, neurotoxic lesions per-
formed before training had very little effects on contex-
tual learning performance (a finding similar to our
results with mice) but electrolytic lesions led to impair-
ments. Moreover, both neurotoxic and electrolytic le-
sions performed after training impaired contextual
learning performance (a finding previously shown for
rats in the literature). The argument made by the
authors on the basis of these and other findings is very
similar to what has been put forth in the present study.
They argue that neurotoxic lesioning may not lead to
detectable anterograde deficits because the pre-training-
lesioned rats may be able to use alternative, elemental,
learning strategies that later allow the rats to perform
well at the context test. On the other hand, rats that
were lesioned after training could utilize their func-
tional hippocampus during training. The performance
of these rats at a later context test is impaired because
they are unable to use the configural strategy due to
their hippocampal lesion, and also are unable to use an
elemental strategy since they were not conditioned to
single elemental cues during training. Finally, the gener-
alized deleterious effect of pre- or post-training elec-
trolytic lesions is assumed to be due to disrupting
exploratory behavior [25].
The relative salience of CS vs. contextual stimuli may
also be important in determining whether an experi-
mental subject will use an elemental or configural learn-
ing strategy. For example, attending to a single cue
from the context may be more difficult if the experi-
mental subject is given a strong, salient associative cue
(CS) that is contiguous with the shock. Thus, CS
presentation may prevent (or lead to a reduction of)
using the elemental learning strategy to acquire contex-
tual cues, a suggestion which is in accordance with the
present findings obtained for mouse strains, DBA/2 and
C57BL/6. DBA/2 mice that were trained with shock
alone and no tone cue (CS) were found to exhibit a
fairly robust freezing response to ‘context’ when tested
1 day after the training. However, DBA/2 mice receiv-
ing shock in association with the tone cue during
training exhibited a significantly impaired freezing re-
sponse to the context the next day. This strain of mice
was previously found to exhibit impaired hippocampal
function at the behavioral [24,28,29,39– 41], neu-
roanatomical [8,9,33] and biochemical [14,22] levels.
They may be similar to hippocampally lesioned mice.
Thus their inability to learn contextual cues when a
clear CS– US association is given and their ability to
R.Gerlai
/
Beha6ioural Brain Research
95 (1997) 191–203
201
learn the ‘context’ when no such association is offered
may all be mediated by non-hippocampal processes. In
summary, these results support the notion that under
certain circumstances mice may not necessarily have to
apply multi-stimulus processing (configural learning re-
quiring the hippocampus) when presented with a con-
textual task. Application of elemental learning may
depend upon the magnitude (and kind) of hippocampal
impairment, upon how homogeneous the contextual
cues are, i.e. whether there are some salient ‘key’ stimuli
among the contextual cues, and upon the relative
salience and predictive nature of the CS compared to
the contextual cues.
Furthermore, our results with the two inbred mouse
strains also imply that the application of elemental vs.
configural learning strategies may depend upon
hippocampal function modulated by the genotype of
the mouse. Unlike DBA/2 mice, C57BL/6 animals re-
sponded with significantly high levels of freezing to the
context both when trained with or without the tone cue
the previous day. Thus the performance of this strain at
the context test could not be significantly disrupted by
the presentation of a CS during training, implying that
C57BL/6 mice, whose hippocampal function is unim-
paired, could employ configural learning despite the
clear US– CS association offered. It is also interesting
to note that, although required in contextual learning,
multi-stimulus processing may be disadvantageous in
simple learning paradigms. It may delay learning, e.g.
impair acquisition rate compared to what could be
achieved via a simpler strategy. Interestingly, DBA/2
mice showed an improved extinction, and decreased
their freezing duration faster, both during the context
test (Fig. 3B,C) and the cue test compared to C57BL/6
mice, again arguing that DBA/2 may not be using
complex multi-stimulus processing that requires a nor-
mally functioning hippocampus.
If the differences between the behavior of DBA/2 and
C57BL/6 mice are indeed due to differences in
hippocampal function and thus differences in multi-
stimulus processing, as the above indeed imply, the
results of the context test of this study will have to be
viewed as a cautionary example. The context test may
not always be appropriate to quantify hippocampal-de-
pendent learning performance because it may be solved
by an elemental learning strategy, a non-hippocampal
function (also see background vs. foreground contex-
tual cues, a concept put forth in a previous rat lesion
study [30]). This could explain why robust freezing
responses were obtained when mutant transgenic mice
with hippocampal dysfunction were exposed to the
context test. Our present study suggests that the appli-
cation of elemental vs. configural learning may depend
upon the genotype of the mice. Thus, conceivably,
introduced mutations may have similar differential ef-
fects. Depending upon the nature of hippocampal dys-
function induced by such mutations and upon
methodological details of conditioning such as dis-
cussed above (also see Refs. [16,25]), transgenic or null
mutant mice may be able to apply alternative learning
strategies and perform comparably to mice with normal
hippocampal function, a problem that may lead to
false-negative findings.
The above suggest that the context test may not
always test hippocampal function. The results of the
cue test also show a similar complexity: the cue test
may not only test hippocampal-independent behavior,
but may also have a hippocampal component. The
amount of freezing at the cue test differed between
DBA/2 and C57BL/6. The latter froze much more than
the former, but only after training. Furthermore, com-
parison of the pre-shock freezing values (obtained be-
fore any electric shock was applied during training)
with the performance obtained at the cue test after
training, shows that freezing in C57BL/6, but not in
DBA/2, was elevated in the new test chamber (new
context) even before any tone cue was given. This is a
remarkable finding suggesting that C57BL/6, but not
DBA/2, found the new context slightly aversive. It is
also notable that mice trained without a tone cue did
not respond to tone with freezing during the cue test.
The tone is a neutral stimulus. Moreover, increased
freezing in C57BL/6 is observable before tone cue pre-
sentation. Thus, increased freezing in C57BL/6 cannot
be due to a simple CS– US association. Hypoactivity
was not observed in the home cage of C57BL/6, nor in
the test apparatus before shock. Therefore, the increase
of freezing response of C57BL/6 in the cue test is likely
due to a transfer effect. C57BL/6 mice might have
interpreted the new situation as one, at least partly,
similar to the original shock chamber. Also, note that
the difference between C57BL/6 and DBA/2 may not
be due to C57BL/6 responding more strongly to pain or
fear, since in fact the opposite could be seen at the
training phase. Moreover, the initial absence of freezing
in the cue test in DBA/2 mice could not be due to some
putative motor impairment since DBA/2 could respond
to a tone cue with a very high level of freezing, not
significantly different from the response of C57BL/6
mice. Since all cues including visual, tactile, and olfac-
tory were different between the test chambers of the
two contexts, one may conclude that the only similarity
between the two contexts was the procedure: in both
cases the mice were removed from their home cage,
handled, and put in a novel situation. In conclusion,
the above observations again suggest that DBA/2 mice
are capable of making an association between a pair of
stimuli (CS and US), but are impaired at responding to
complex contextual cues and thus may not show a
transfer effect between partly different contexts. It is
also notable that, in a recent study, response to the tone
cue was also found to be hippocampal dependent [25].
R.Gerlai
/
Beha6ioural Brain Research
95 (1997) 191–203
202
The authors of this work argue that, under certain
circumstances, the tone itself may not be a unitary
stimulus since it may have complex characteristics, such
as a range of frequencies, amplitude changes, duration
and location, and it may activate a population of
neurons with different but overlapping tuning curves.
Configural coding of the tone itself thus may require
the hippocampus, a function that is more apparent if
some of the characteristics of the previously presented
tone cue are complex or modified at the cue test.
In summary, our present results as well as recent
findings in the literature suggest that, despite the con-
ceptual power of the clear cut hippocampal-dependent
(configural context learning) vs. hippocampal-indepen-
dent (elemental, simple cue learning) distinction preva-
lent in the literature, the fear-conditioning paradigm is
a complex situation in which the configural or elemen-
tal nature of the task may depend upon experimental
procedure, type of hippocampal dysfunction, timing of
induction of the dysfunction, and the genotype of ex-
perimental subjects.
References
[1] Abeliovich A, Paylor R, Chen C, Krim J, Wehner J, Tonegawa
S. PKCgmutant mice exhibit mild deficits in spatial and contex-
tual learning. Cell 1993;75:1263–71.
[2] Aiba A, Chen C, Herrup K, Rosenmund C, Stevens C, Tone-
gawa S. Reduced hippocampal long-term potentiation and con-
text-specific deficit in associative learning in mGluR1 mutant
mice. Cell 1994;79:365–75.
[3] Bach ME, Hawkins RD, Osman M, Kandel ER, Mayford M.
Impairment of spatial but not contextual memory in CaMKII
mutant mice with a selective loss of hippocampal LTP in the
range of the ufrequency. Cell 1995;81:905–15.
[4] Blanchard RJ, Blanchard DC. Crouching as an index of fear. J.
Comp. Physiol. Psychol. 1969;81:281–90.
[5] Bliss TVP, Collingridge GL. A synaptic model of memory:
long-term potentiation in the hippocampus. Nature
1993;361:34–9.
[6] Bunsey M, Eichenbaum H. Conservation of hippocampal mem-
ory function in rats and humans. Nature 1996;379:255–7.
[7] Chen C, Kim JJ, Thompson RF, Tonegawa S. Hippocampal
lesions impair contextual fear conditioning in two strains of
mice. Behav. Neurosci. 1996;110:1177–80.
[8] Crusio WE, Schwegler H, van Abeelen JHF. Behavioral re-
sponses to novelty and structural variation of the hippocampus
in mice. I. Quantitative-genetic analysis of behavior in the open-
field. Behav. Brain Res. 1989;32:75–80.
[9] Crusio WE, Schwegler H, Lipp H-P. Radial-maze performance
and structural variations of the hippocampus in mice: A correla-
tion with mossy fibre distribution. Brain Res. 1987;425:182–5.
[10] Eichenbaum H. Declarative memory: insights from cognitive
neurobiology. Annu. Rev. Psychol. 1997;48:547–72.
[11] Eichenbaum H. Is the rodent hippocampus just for ‘place’? Curr.
Opin. Neurobiol. 1996;6:187–95.
[12] Eichenbaum H, Stewart C, Morris RGM. Hippocampal repre-
sentation in place learning. J. Neurosci. 1990;10:3531–42.
[13] Fanselow MS. Associative vs. topographical accounts of the
immediate shock freezing deficit in rats: Implications for the
response selection rules governing species-specific defensive reac-
tions. Learning Motivation 1986;17:16–39.
[14] Fordyce DE, Wehner JM. Determination of hippocampal
protein kinase C using frozen tissue method: Comparison of
synaptosomal and total activity in C57BL/6 and DBA/2 mice.
Biochem. Biophys. Res. Commun. 1992;188:690–4.
[15] R. Gerlai, Ethological approaches in the analysis of behavior of
transgenic and knock-out mice, in: W.E. Crusio, R. Gerlai
(Eds.), Recombinant DNA Techniques in Behavioral Neuro-
science, Elsevier, Amsterdam (1998) (in press).
[16] R. Gerlai, A new continuous alternation task in T-maze detects
hippocampal dysfunction in mice: A strain comparison and
lesion study. Behav. Brain Res. (1997) (in press).
[17] Gerlai R. Gene targeting studies of mammalian behavior: Is it
the mutation or the background genotype? Trends Neurosci.
1996;19:188–9.
[18] Gerlai R. Molecular genetic analysis of mammalian behavior
and brain processes: Caveats and perspectives. Semin. Neurosci.
1996;8:153–61.
[19] Gerlai R. Can paradise fish (Macropodus opercularis) recognize
its natural predator? An ethological analysis. Ethology
1993;94:127–36.
[20] Gerlai R, Friend W, Becker L, O’Hanlon R, Marks A, Roder J.
Female transgenic mice carrying the human gene for S100bare
hyperactive. Behav. Brain Res. 1993;55:51–9.
[21] Huang Y-Y, Kandel ER, Varshavsky L, Brandon EP, Qi M,
Idzerda RL, McKnight GS, Bourtchouladze R. A genetic test of
the effects of mutations in PKA on mossy fiber LTP and its
relation to spatial learning and contextual learning. Cell
1995;83:1211–22.
[22] Ingram DK, Cofman TP. An overview of neurobiological com-
parisons in mouse strains. Neurosci. Biobehavi. Rev.
1980;4:421–35.
[23] Kim JJ, Fanselow MS. Modality-specific retrograde amnesia of
fear. Science 1992;256:675–7.
[24] Logue SF, Paylor R, Wehner JM. Hippocampal lesions cause
learning deficits in inbred mice in the morris water maze and
conditioned-fear task. Behav. Neurosci. 1997;111:104–13.
[25] S. Maren, G. Aharonov, M.S. Fanselow, Neurotoxic lesions of
the dorsal hippocampus and Pavlovian fear conditioning in rats.
Behav. Brain Res. (1997) (in press).
[26] Morris RGM, Garrud P, Rawlins JNP, O’Keefe J. Place naviga-
tion impaired in rats with hippocampal lesions. Nature
1982;297:681–3.
[27] J. O’Keefe, L. Nadel, The Hippocampus as a Cognitive Map,
Clarendon, Oxford, 1978.
[28] Paylor R, Tracy R, Wehner J, Rudy JW. DBA/2 and C57BL/6
mice differ in contextual fear but not auditory fear conditioning.
Behav. Neurosci. 1994;108:1–8.
[29] Paylor R, Morrison SK, Rudy JW, Waltrip LT, Wehner JM.
Brief exposure to an enriched environment improves perfor-
mance in the Morris water task and increases hippocampal
cytosolic protein kinase C activity in young rats. Behav. Brain
Res. 1992;52:49–59.
[30] Phillips RG, LeDoux JE. Lesions of the dorsal hippocampal
formation interfere with background but not foreground contex-
tual fear conditioning. Learning Memory 1994;1:34–44.
[31] Phillips RG, LeDoux JE. Differential contribution of amygdala
and hippocampus to cued and contextual fear conditioning.
Behav. Neurosci. 1992;106:274–85.
[32] Rapp PR, Kansky MT, Eichenbaum H. Learning and memory
for hierarchical relationships in the monkey: Effects of aging.
Behav. Neurosci. 1996;110:887–97.
[33] Schwegler H, Lipp H-P, Van der Loos H, Buselmaier W. Indi-
vidual hippocampal mossy fiber distribution in mice correlates
with two-way avoidance performance. Science 1981;214:817–9.
[34] Silva AJ, Rosahl TW, Chapman PF, Marowitz Z, Friedman E,
Frankland PW, Cestari V, Cioffi D, Su¨ dhof TC, Bourtchuladze
R. Impaired learning in mice with abnormal short-lived plastic-
ity. Curr. Biol. 1996;6:1509–18.
R.Gerlai
/
Beha6ioural Brain Research
95 (1997) 191–203
203
[35] Squire LR, Zola-Morgan S. The medial temporal lobe memory
system. Science 1991;253:1380–6.
[36] Squire LR, Knowlton B, Musen G. The structure and organiza-
tion of memory. Annu. Rev. Psychol. 1993;44:453–93.
[37] Sutherland RJ, Kolb B, Whishaw IQ. Spatial mapping: Defini-
tive disruption by hippocampal or medial frontal cortical dam-
age in the rat. Neurosci. Lett. 1982;31:271–6.
[38] Topa´ l J, Csa´nyi V. The effect of eye-like schema on shuttling
activity of wild house mice (Mus musculus domesticus). Context
dependent threatening aspects of eyespot patterns. Anim. Learn-
ing Behav. 1994;22:96–103.
[39] Upchurch M, Wehner JM. DBA/2Ibg mice are incapable of
cholinergically-based learning in the Morris water task. Pharma-
col. Biochem. Behav. 1988;29:325–30.
[40] Upchurch M, Wehner JM. Differences between inbred strains of
mice in Morris water maze performance. Behav. Genet.
1988;18:55–68.
[41] Upchurch M, Wehner JM. Inheritance of spatial learning ability
in inbred mice: A classical genetic analysis. Behav. Neurosci.
1989;103:1251–8.
[42] Wahlsten D. Insensitivity of the analysis of variance to heredity-
environment interaction. Behav. Brain Sci. 1990;13:109–61.
.
