Memory of Early Maltreatment: Neonatal Behavioral
and Neural Correlates of Maternal Maltreatment
Within the Context of Classical Conditioning
Tania L. Roth and Regina M. Sullivan
Background: While children form an attachment to their abusive caregiver, they are susceptible to mental illness and brain
abnormalities. To understand this important clinical issue, we have developed a rat animal model of abusive attachment where odor
paired with shock paradoxically produces an odor preference. Here, we extend this model to a seminaturalistic paradigm using a
stressed, “abusive” mother during an odor presentation and assess the underlying learning neural circuit.
Methods: We used a classical conditioning paradigm pairing a novel odor with a stressed mother that predominantly abused pups
to assess olfactory learning in a seminaturalistic environment. Additionally, we used Fos protein immunohistochemistry to assess brain
areas involved in learning this pain-induced odor preference within a more controlled maltreatment environment (odor-shock
Results: Odor-maternal maltreatment pairings within a seminatural setting and odor-shock pairings both resulted in paradoxical
odor preferences. Learning-induced gene expression was altered in the olfactory bulb and anterior piriform cortex (part of olfactory
cortex) but not the amygdala.
Conclusions: Infants appear to use a unique brain circuit that optimizes learned odor preferences necessary for attachment. A fuller
understanding of infant brain function may provide insight into why early maltreatment affects psychiatric well-being.
Key Words: Attachment, learning, memory, neonate, maltreat-
Sullivan 2001; Sullivan 2003). Indeed, infant chicks, dogs, non-
human primates, and humans continue to show strong attach-
ments to abusive caregivers (Arling and Harlow 1967; Hess 1962;
Maestripieri 1998; Rajecki et al 1978). Clinical data show mal-
treated children exhibit strong, albeit disordered, attachment to
their caregiver, with abuse and neglect producing distinct clinical
outcomes (Bowlby 1965; Carlson et al 1990; Cicchetti and Toth
1995; Helfer et al 1997; Hesse and Main 2000; Morton and
Browne 1998; Schore 2002). Though most functional imaging
studies have focused on adults with a history of childhood
maltreatment, studies suggest that neglect or abuse compromises
brain development, most notably for the limbic system (hip-
pocampus, amygdala), stress axis (locus coeruleus, hypothala-
mus, amygdala), and cerebellum (Bremner 2003; Glaser 2000;
Schore 2002; Teicher et al 2003).
Research using animal models of maternal deprivation in
rodents and nonhuman primates parallel imaging studies, sug-
gesting that neglect also produces long-term compromises in the
limbic system, stress axis, and cerebellum (Dettling et al 2002;
Glaser 2000; Huot et al 2002; Liu et al 2000; Pryce et al 2004;
Rosenblum et al 1994; Sanchez et al 2001; van Oers et al 1998).
These findings are further supported by models of enrichment
correlating high levels of maternal care (licking/grooming, nurs-
ing) with the enhancement of neurobehavioral outcome (Caldji
n species requiring parental care, evolution has ensured
that infants quickly learn and express robust preferences to
the caregiver, regardless of the quality of care (Hofer and
et al 1998; Liu et al 2000; Meaney 2001; van Oers et al 1998).
Overall, studies demonstrate that maternal behaviors affect de-
velopment of these brain areas and thus offer potential sites to
understand the damaging effects of early maltreatment on sub-
sequent behavioral development (Bremner 2003; Dent et al 2001;
Francis et al 1999; Heim and Nemeroff 2001; Kaufman et al 2000;
Levine 2001; Nemeroff 2004; Sanchez et al 2001; Teicher et al
Trauma within the attachment system leaves the infant par-
ticularly vulnerable to childhood and adult psychiatric disorders,
behavioral changes in fear and anxiety, and alterations in neural
circuits, particularly those regulating stress and emotion (Connor
et al 2003; Gunnar 2003; Heim and Nemeroff 2001; Teicher et al
2003; Zeanah et al 2003). The overlap in the aforementioned
brain structures and those active in attachment suggests a possible
mechanism for the commanding effects of early adverse experi-
ences on the etiology of psychiatric disorders. Using an infant
animal model that incorporates maternal abuse within the con-
text of attachment may offer a more direct approach to under-
standing the damaging effects of maltreatment on brain devel-
opment, especially concerning the overlap in neural circuitry
supporting attachment and the etiology of psychiatric disorders.
Furthermore, use of such an animal model that approximates an
environment in which an infant receives both abusive and
positive caregiving may yield data that are more relevant to
Rat neonates form memories of odors associated with both
pleasant (milk or warmth) or aversive stimuli (tail-pinch or
shock) (Camp and Rudy 1988; McLean et al 1993; Sullivan et al
1986, 2000a), and odor learning is critical to the development of
mother-infant attachment (Hofer and Sullivan 2001). These mem-
ories are expressed as odor preferences, as indicated by pup
orientation toward the odor and even the climbing of an obstacle
to approach the odor (Camp and Rudy 1988; Roth and Sullivan
2001; Sullivan et al 2000a). The unique neural circuitry respon-
sible for this early olfactory learning is being elucidated and
includes the locus coeruleus, the olfactory bulb, and amygdala
(Sullivan 2003). A similar neural circuit has been characterized
for attachment related to reproduction in the adult (Brennan and
From the Department of Zoology, University of Oklahoma, Norman,
Address reprint requests to Tania Roth, Department of Zoology, University
of Oklahoma, 730 Van Vleet Oval, Norman, OK 73019; E-mail: Tania.
Received September 17, 2004; revised December 15, 2004; accepted Janu-
ary 13, 2005.
BIOL PSYCHIATRY 2005;57:823–831
© 2005 Society of Biological Psychiatry
Keverne 1997; Carter et al 1995; Fleming et al 1999; Insel and
Young 2001; Kendrick et al 1997).
To assess how abuse within the mother-infant dyad supports
attachment in the rat, we used two approaches to examine the
neurobiology contributing to the memory of neonate odor
preferences. First, we examined neonate learning and memory
with a new paradigm that uses maternal abuse and odor within
the context of attachment. Second, we used immunohistochem-
ical marking of the Fos protein to examine neonate brain areas
activated following a more controlled maltreatment environment
(odor-shock conditioning, a fear-conditioning paradigm used in
Methods and Materials
Subjects and Husbandry
We used male and female pups born of Long-Evans rats
(Harlan, Indiana) in the University of Oklahoma animal vivarium.
Mothers were housed in polypropylene cages with wood shav-
ings and kept in a temperature- (20°C) and light-controlled
(12h:12d) environment with food and water continually avail-
able. Day of parturition was termed 0 days of age, and litters
were culled to 5 male pups and 5 female pups on postnatal (PN)
1 or 2. No more than one male pup or one female pup from a
given litter was used for any training/test condition. The Univer-
sity of Oklahoma Institutional Animal Care and Use Committee
approved all procedures.
Experiment 1 Odor-Maternal Maltreatment Conditioning.
Forty-eight pups from seven litters were used in the maltreatment
experiment. On PN 7 and 8 (13.9–18.8 g) pups were assigned to
one of four training conditions: 1) Paired (n ? 13): pups received
simultaneous maltreatment by a mother and a novel odor
(peppermint); 2) Unpaired (n ? 8): pups received the odor 30
minutes before receiving the maltreatment; 3) Odor Only (n ?
9): pups received only the odor; and 4) Maltreatment Only (n ?
8): pups received only maltreatment by a mother. The training
chambers consisted of a 45.5 ? 30.5 ? 45 cm opaque Plexiglas
box. To serve as an effective maternal stressor, we provided
limited clean aspen shavings (100 mL) on the chamber floors
(Gilles et al 1996). Chambers were lit with a red light and covered
with Privacy Mirror Film to ensure behavior was not disturbed
by experimenter observations (Gila, CPFilms Inc., Martinsville,
Mothers (nonbiological to experimental pups) were placed in
the chambers 5 minutes prior to receiving pups (1 male pup and
1 female pup trained simultaneously within each condition). This
short adaptation time served as a stressor to disrupt normal
maternal behavior and potentiate behaviors that seemed painful
to pups. Odor was presented with a Kimwipe (Kimberly-Clark
Corporation, Roswell, Georgia) (25 ?L of pure McCormick [Hunt
Valley, Maryland] peppermint extract; 5 minutes in a fume hood)
placed on the chamber lid. A training session lasted 30 minutes,
during which maternal and pup behaviors were recorded in
5-minute intervals. Abusive maternal behaviors were: 1) step-
ping: the mother steps or jumps on the pup; 2) throwing: the
mother throws the pup some distance; 3) dropping: the mother
drops a pup during retrieval or transport; 4) dragging: the mother
drags a pup across the chamber; 5) pushing away/actively
avoiding: the mother runs from a pup’s approaches or pushes a
pup away from her, crushing the infant onto the floor; and 6)
rough handling: the mother aggressively grooms a pup or
transports a pup by a limb. These behaviors typically elicited pup
vocalization, indicative of neonate distress and pain (White et al
1992). Several of these behaviors are categorized as abusive
in primates (Brent et al 2002; Maestripieri 1998), qualitatively
different from the normal maternal behavior repertoire (Alberts
and Cramer, 1988; Denenberg et al 1969; Fleming and Rosenblatt
1974), and may harm the infant or interfere with development
(Maestripieri and Carroll 1998; Righthand et al 2003; Zigler and
Hall 1990). Less frequently, positive maternal behaviors were
observed and included pup grooming, anogenital licking, and
nursing. These behaviors did not elicit pup vocalization. Follow-
ing training, pups were placed in a 30°C incubator for 15 minutes
and then returned to the biological mother until testing in a
For comparative purposes, an additional 25 pups (PN 7–8;
13.7–20.2 g) were trained using a similar training paradigm as
described above; however, pups received pairings with mothers
who were nonabusive, thus representing a more normal and
positive learning environment. In this natural paradigm, mothers
were placed inside the training chambers and given two treat-
ments that prevented the stress-induced maltreatment: copious
shavings on the floor (2 cm layer) and a 1-hour habituation time
before receiving pups. The four training conditions were: 1)
Paired (n ? 7): pups received simultaneous maternal care and
odor; 2) Unpaired (n ? 6): pups received the odor 30 minutes
before receiving the care; 3) Odor Only (n ? 6): pups received
only odor; and 4) Maternal Care Only (n ? 6): pups received
only care from a mother.
Conditioning. To assess the neonate brain under exclusively
painful learning conditions, pups were trained using odor-shock
conditioning. Pups feel pain and vocalize and try to escape in
response to shock (Barr 1995; Emerich et al 1985; Stehouwer and
Campbell 1978; Sullivan et al 2000a). Eleven pups from four
additional litters were used for IHC neural analysis. On PN 7 and
8 (15.0–19.4 g) pups were randomly assigned to a condition: 1)
Paired odor-shock (n ? 4); 2) Unpaired odor-shock (n ? 4); and
3) Odor Only (n ? 3). Once placed inside the training apparatus
(individual 600 mL beakers), pups were given a 10-minute period
to recover from handling. During a 1-hour training session, pups
received 14 presentations of a 30-second peppermint odor and a
1-second .5 mA tail-shock, with a 4-minute intertrial interval.
Paired odor-shock subjects received 14 pairings of the odor with
shock during the last second of the odor, while Unpaired
odor-shock subjects received a 1-second shock 2 minutes after
an odor presentation. Odor Only subjects received only the
peppermint odor. Peppermint odor was delivered with a flow-
dilution olfactometer at 2 L/min and at a concentration of 1
peppermint vapor:10 clean air. Following training, pups were
placed in a 30°C incubator, and after 90 minutes, pups were
quickly decapitated and their brains were removed, frozen in
2-methylbutane (?45°C), and placed in ?70°C until cutting and
postfixation for IHC.
To verify associative learning, we recorded behavioral re-
sponses to the odor during training to construct acquisition
curves. Due to motor immaturity, generalized movements were
recorded (0 ? “no movement of the four limbs or head” ? 5 ?
“movement of all four limbs and head”) (Hall 1979). Nineteen
additional pups (PN 6–7, 13.3–18.1 g; Paired n ? 6, Unpaired
n ? 6, and Odor Only n ? 7) were trained in the same manner
to test for an odor preference in a Y-maze.
ment 1, pups were tested in a two-odor-choice test. The testing
One day following training in Experi-
824 BIOL PSYCHIATRY 2005;57:823–831
T.L. Roth and R.M. Sullivan
apparatus was a Plexiglas arena (24 ? 14 cm) with a wire mesh
floor. The floor was divided into two areas by a 2-cm midline:
one area contained the conditioned odor (Kimwipe scented with
25 ?L of peppermint extract placed in a ventilation hood for 15
minutes), and the other contained 100 mL of clean aspen
shavings. The time spent on each side was recorded (Video-
mex-V, Columbus Instruments, Columbus, Ohio). Each pup
received three 60-second trials, with a counterbalanced orienta-
tion for each trial. The floor was cleaned between trials.
Y-Maze. For the 19 additional pups trained in Experiment 2,
on the day following training, they were removed from the
mother and tested using a Y-maze. The Y-maze consisted of a
habituation chamber (7 cm long and 9 cm wide) and two alleys
(22 cm long and 9 cm wide) extending at 45° angles. The
habituation chamber was separated from the alleys via two
removable doors. One arm of the maze contained aspen wood
odor (20 mL of clean, aspen shavings in a petri dish), while the
other arm contained the peppermint odor (25 ?L of peppermint
extract on a Kimwipe placed in a ventilation hood for 5 minutes).
Each pup was placed in the starting chamber and given 5
seconds for habituation before the doors to the alleys were
removed. Each subject had 60 seconds to make a choice, which
required the pup to enter one of the alleys. Each subject was
given five sequential trials, and the floor was cleaned between
each trial. Pup orientation was counterbalanced between trials.
Observations of each pup were made blind to the training
Fos Protein IHC
Immediate early genes serve as markers of changes in neu-
ronal activity and thus are indicative of changes in neuronal
plasticity reflective of learning and memory (Dragunow and
Bilkey 2002; Herrera and Robertson 1996; Kaczmarek 2002;
Tischmeyer and Grimm 1999). Brains (from pups trained for IHC
in experiment 2) were coronally sectioned (20 ?m); every sixth
section was collected on pretreated slides (Fisherbrand Plus,
Fisher, Pittsburgh, Pennsylvania) for Fos processing, and every
seventh section was collected for cresyl violet staining. Fos
sections were postfixed for 1 hour in 4% paraformaldehyde/.1
mol/L phosphate buffer (PB) (pH 7.2) and then rinsed in .1 mol/L
PB (pH 7.2) and dried in a cool airstream. Slides were stored in
boxes with Drycap capsules (Ted Pella Inc., Redding, Californ-
nia) in ?20°C until Fos processing. To eliminate peroxidase
activity, sections were incubated in .1 mol/L phosphate buffer
saline (PBS) (pH 7.2) containing 3% hydrogen peroxide (H2O2)
and 10% methanol. Following PBS rinses and incubation in .2%
Triton X-100 (Sigma, St. Louis, Missouri), slides were incubated in
3% Bovine Serum Albumin (Sigma) for 1 hour. After additional
PBS rinses, slides were treated overnight at 4°C with the primary
antibody (c-fos, sc-52, Santa Cruz Biotechnology, Santa Cruz,
California) diluted 1:500 in PBS. Afterward, they were rinsed in
PBS, incubated in the secondary biotinylated antibody (goat
anti-rabbit, Vector Laboratories, Burlingame, California) for 2
hours at room temperature, and then incubated for 90 minutes in
avidin-biotin-peroxidase (ABC) complex solution. Following,
slides were treated with PB containing .1% 3,3’-diaminobenzi-
dine and H2O2. Slides were then dehydrated in alcohol and
Histoclear (National Diagnostics, Atlanta, Georgia) and cover-
slipped for microscope examination.
Fos-positive cells were counted using a microscope (Olympus
Optical Co., Tokyo, Japan; 10? objective) equipped with a
drawing tube. Brain areas were outlined using the corresponding
cresyl violet sections and a stereotaxic atlas (Paxinos and Watson
1986), and all Fos-positive cells were counted bilaterally without
knowledge of the training condition. A Fos-positive cell was
distinguished from the background by the density of staining, the
shape, and the size of the cell. The mean number of Fos cells per
brain area for an animal was determined by averaging the counts
from all sections (two sections counted per brain area). Brain
areas examined were the granule cell layer of the olfactory bulb,
the anterior and posterior piriform cortex, and the basolateral/
lateral and central amygdaloid nuclei.
We used analysis of variance (ANOVA) and post hoc Fisher
tests to analyze differences between training conditions and drug
treatment groups for both behavioral and Fos experiments. In
addition, we used paired and unpaired t tests to compare the
frequency of abusive and normal maternal behaviors in experi-
Neonates Learn a Preference from Odor-Maternal
To our knowledge, this is the first study to use physical abuse
from rat mothers within a neonate classical conditioning para-
digm. As shown in Figure 1, pups that received contiguous
presentations of maltreatment and peppermint odor learned an
odor preference, as demonstrated by the significant amount of
time spent over the odor during the test relative to control
subjects [F(3,34) ? 4.745, p ? .01]. Post hoc tests indicate that the
paired pups spent significantly more time over the peppermint
odor (p ? .05). Likewise, pup training within the natural para-
digm resulted in a learned odor preference [F(3,21) ? 4.395, p ?
.02; Figure 2]. Post hoc tests indicate that pups receiving simul-
taneous odor and maternal care spent significantly more time
over the odor (p ? .05).
Maternal behaviors observed and classified as abusive in-
cluded stepping/jumping, throwing, dropping or dragging, push-
ing away or actively avoiding, and rough handling. Since behav-
iors were only a few seconds in length and mothers exhibited
several within each observational period, we analyzed the
frequency of behaviors over the course of the training session. To
Figure 1. Training within the maltreatment paradigm (odor-“abusive”
stressed dam) resulted in a conditioned odor preference, as expressed dur-
peppermint odor during testing is shown for pups that received Paired
presentations of odor-maltreatment, Unpaired presentations of odor-mal-
treatment, Odor Only, or Maltreatment Only. *p ? .05 between these
T.L. Roth and R.M. Sullivan
BIOL PSYCHIATRY 2005;57:823–831 825
show there were differences in maternal behavior within and
between the two training paradigms, t tests were used to
compare the frequency of behaviors displayed for all the mothers
used. As illustrated in Figure 3, within the maltreatment para-
digm, there were significantly more abusive behaviors observed
than normal maternal behaviors [paired t test, t(14) ? ?3.932,
p ? .01], producing an adverse learning environment. In contrast,
in the natural paradigm, mothers were rarely abusive [paired
t test, t(9) ? 9.390, p ? .01], creating a positive learning
environment. Between the paradigms, mothers were significantly
more abusive toward infants in the maltreatment paradigm than
the natural paradigm [unpaired t test, t(23) ? 9.253, p ? .01]. In
addition, these mothers displayed significantly less normal and
nonabusive behaviors [unpaired t test, t(23) ? ?5.933, p ? .01].
Table 1 provides a comparison of the frequencies of maternal
and pup behaviors observed within the paradigms. Infant step-
ping/jumping on and rough handling were the most frequently
observed abusive behaviors, followed by pushing away/avoid-
ing, infant dragging, dropping, and throwing. Within the mal-
treatment paradigm, 45% of the time pups emitted audible
vocalizations in response to mother-infant interactions. Mothers
also displayed normal behaviors toward infants within the mal-
treatment paradigm. Thus, odor preference learning may also be
attributable to these behaviors, though the occurrence of abusive
behaviors far exceeds those of nonabusive behaviors. Within the
natural paradigm, there was the rare observation of a mother
stepping on or roughly handling a pup, and pups vocalized less
than 3% of the conditioning time. In sharp contrast to the
maltreatment paradigm, mothers spent significant time display-
ing normal behaviors toward neonates, such as frequent licking
Neural Circuitry Supporting Odor-Pain Conditioning
To examine changes in the neonate brain only attributable to
pain-induced learning, we used odor-shock conditioning in
which pups learn odor preferences (Sullivan et al 2000a). Figure 4
indicates that tail-shock is an effective stimulus in producing a
conditioned odor preference, as pups that had received Paired-
odor-shock presentations made significantly more choices to-
ward the conditioned odor during the Y-maze behavioral test
[F(2,16) ? 10.708, p ? .01]. Analysis of acquisition curves of pups
used for IHC neural analysis indicated that all subjects had similar
preconditioning behavior, [F (12,48) ? .450, p ? .934; Figure 5A],
but behavior in response to the odor indicated a significant effect
of training condition, [F (12,48) ? 2.981, p ? .01; Figure 5B]. Post
hoc tests showed that subjects receiving Paired presentations of
odor and shock had significant acquisition (learning) in compari-
son with control subjects (p ? .05).
Figures 6 and 7 illustrate learning-induced changes in Fos
protein expression following odor-shock conditioning. Cellular
staining was not observed without the primary antibody. Ninety
minutes following the conditioning, there were learning-induced
changes in the number of Fos-positive cells in the olfactory bulb
granule cell layer [F (2,7) ? 8.151, p ? .02; Figure 7]. Analysis
with post hoc Fisher tests revealed that the Paired presentations
of the odor-shock, which again generate a behavioral odor
preference, induced significantly less Fos in the granule cell layer
in comparison with control presentations (p ? .05).
normal maternal behavior) resulted in a conditioned odor preference, as
spent over the peppermint odor during testing is shown for pups that
received Paired presentations of odor-maternal care, Unpaired presenta-
tions of odor-maternal care, Odor Only, or Maternal Care Only. *p ? .05
between these groups.
Mothers within the maltreatment paradigm predominately displayed abu-
sive behaviors toward neonates, thus providing an adverse conditioning
environment. Within the natural training paradigm, mothers were rarely
Table 1. Frequency of Abusive or Normal Maternal Behaviors Observed
During Mother-Infant Interactions Within Experiment 1
Percent of Observation Periods in Which Behaviors Occurred
Maltreatment Paradigm Natural Paradigm
Step or jump on
826 BIOL PSYCHIATRY 2005;57:823–831
T.L. Roth and R.M. Sullivan
As changes in mitral cell activity (output neurons of the
olfactory bulb) have been suggested to reflect learned associa-
tions in pups (Sullivan and Wilson 2003; Wilson et al 1987; Yuan
et al 2003), we were interested in the activity of cortical areas
known to process olfactory information transmitted from the
bulb. Analysis of variance revealed a training effect on Fos
expression in the anterior piriform [F(2,7) ? 8.779, p ? .02;
Figure 7] but not the posterior piriform cortex [F (2,7) ? .018, p ?
.982; Figure 7]. Post hoc Fisher tests showed that Paired odor-
shock presentations evoked significantly more Fos expression in
the anterior piriform than Unpaired odor-shock or Odor Only
presentations (p ? .05). In agreement with autoradiography data
(Sullivan et al 2000a), learned odor-shock associations did not
(0–5 rating of limb and head movement) and indicates that pups used for
that received either Paired or Unpaired odor-shock presentations or Odor
Only presentations. Each data point represents the summation of behavior
from two consecutive trials; vertical lines indicate SEM. Conditioning treat-
ing) in response to the odor in comparison with the control subjects. IHC,
Figure 4. Odor-shock conditioning effectively induces a paradoxical odor
(? SEM) number of approaches (out of five test trials) toward the pepper-
mint odor during Y-maze testing is shown for the 19 additional pups that
shock presentations, or Odor Only presentations. *p ? .05 between these
Figure 6. Representative images of Fos protein expression 90 minutes fol-
lowing odor-shock conditioning in neonates (10x, scale bar ? 100 ?m). (A)
through (C) illustrates Fos expression in the granule cell layer of the olfac-
tory bulb in pups with (A) Paired odor-shock, (B) Unpaired odor-shock, or
(C) Odor Only presentations. Paired subjects showed significantly less Fos
expression than Unpaired or Odor Only subjects. Likewise, (D) through (F)
ing. Paired subjects showed significantly more Fos expression than the
control subjects. Arrows represent examples of Fos-positive cells in each
cortex; I, Layer I; II, Layer II; III, Layer III.
Figure 7. Odor-shock conditioning induces Fos protein expression in neo-
nate olfactory circuitry. Paired presentations of odor and shock induce
of the olfactory bulb and the anterior piriform cortex (ant PIR) of the olfac-
tory cortex. Conditioning did not produce significant changes within the
posterior piriform cortex (post PIR), the basolateral/lateral amygdala (BLA/
Fos-positive cells counted bilaterally in each brain area. Paired n ? 3–4,
Unpaired n ? 3–4, Odor Only n ? 3. *p ? .05 between these groups.
T.L. Roth and R.M. Sullivan
BIOL PSYCHIATRY 2005;57:823–831 827
induce significant Fos expression in the basolateral/lateral [F(2,8) ?
.648, p ? .545; Figure 7] or central amygdaloid nuclei [F (2,7) ?
.065, p ? .937; Figure 7].
Functional imaging studies suggest that child maltreatment
alters brain development in areas that regulate stress, cognition,
and emotion. Behaviorally, short-term and long-term effects of
child abuse and neglect include problems in self-regulation,
excessive anxiety and fearfulness, aggressive behaviors (abuse,
violence), substance abuse, mood disorders, and posttraumatic
stress disorder (Brown 2003; Connor et al 2003; Pollak and
Tolley-Schell 2003; Righthand et al 2003; Teicher et al 2002;
Zeanah et al 2003). How maltreatment produces the brain and
behavioral changes remains unclear, hindered by our limited
understanding of the age at which brain function emerges within
specific behavioral systems. What is evident within the attach-
ment system is that the neural circuitry of the infant brain is
probably specialized to optimize the learning necessary for
mother-infant attachment. Thus, use of an animal model with
abuse in the context of attachment offers an approach to
assessing how the unique circuitry of the infant brain responds to
early maltreatment, how an infant can still form an attachment to
an abusive caretaker, and ultimately how changes in the infant’s
brain contribute to childhood and adult psychiatric disorders.
Attachment Within an Adverse Environment
In rat neonates, a painful stimulus paired with an odor
induces a conditioned odor preference (Camp and Rudy 1988;
Roth and Sullivan 2001; Sullivan et al 1986, 2000a), while a
similar treatment in older pups and adults readily causes odor
aversions and fear (Camp and Rudy 1988; Otto et al 1997;
Paschall and Davis 2002; Richardson et al 2000; Sullivan et al
2000a; Ressler et al 2002). This study has provided additional
validity for our odor-shock model by demonstrating that actual
physical abuse from the mother in the presence of an odor
supports a learned odor preference. We induced abusive behav-
iors from mothers through stress, which models the effects of a
stressful environment as a risk factor for potentiating infant
maltreatment, including in humans (Cicchetti 1990; Field 1983;
Maestripieri and Caroll 1998; Righthand et al 2003; Schapiro and
Mitchell 1983). In our maltreatment paradigm, pups received
multiple counts of abusive behaviors; yet, they pursued contact
with the abusive mothers and later approached the odor that had
been paired with the abuse. A similar increase in caretaker
attachment following abuse has been documented in nonhuman
primates (Arling and Harlow 1967; Maestripieri 1998), chicks
(Hess 1962), infant dogs (Rajecki et al 1978), and humans
(Cicchetti 1998). A common factor in these diverse species is that
the caregiver provided both abusive and positive maternal
behaviors, suggesting that our abusive mother model may more
closely approximate the complexity of the abusive attachment
The Unique Neurobiology Responsible for Attachment
During the first 10 days of a rat neonate’s life (sensitive
period), the brain is optimized to ensure rapid and robust odor
preference learning, regardless of the hedonic value of stimuli.
This learning is expressed by neonates’ enhanced ability to
acquire learned odor preferences and a decreased ability to
acquire learned odor aversions. At least two brain structures
underlie this unique odor learning: the hyperfunctioning norad-
renergic locus coeruleus (LC) underlies heightened preference
learning and the hypofunctioning amygdala appears to underlie
attenuated aversion learning (Moriceau and Sullivan 2004a,
2004b; Okutani et al 1998; Rangel and Leon 1995; Sullivan and
Wilson 1993; Sullivan et al 2000a, 2000b; Yuan et al 2003). As the
sensitive period ends, maturation of LC autoinhibition (reducing
the release of norepinephrine [NE]) greatly attenuates the rapid
odor preference learning, and amygdala participation permits
fear conditioning (Moriceau and Sullivan 2004a, 2004b; Naka-
mura and Sakaguchi 1990; Rangel and Leon 1995; Sullivan et al
2000a; Sullivan 2003).
To understand how painful experiences influence attach-
ment, we assessed the neonate’s olfactory neural circuit using
Fos protein immunohistochemistry and odor-shock condition-
ing. Odor-shock conditioning offers a more controlled maltreat-
ment environment, as pups only receive shock. Thus, brain
changes reflect a learned association based exclusively on aver-
sive stimulation. This the first study to examine changes in gene
expression in the neonate brain following painful learning
conditions, showing that a shock-induced odor preference is
attributable to a decrease of Fos protein expression in the
granule cell layer of the olfactory bulb, an increase of Fos
expression in the anterior piriform cortex, and the lack of
significant changes in the posterior piriform cortex and basolat-
eral/lateral and central amygdaloid nuclei.
Olfactory Bulb. We found learning-induced changes in the
olfactory bulb, further confirming that the olfactory bulb encodes
odor learning during early life (Johnson et al 1995; McLean et al
1999; Sullivan and Leon 1987; Sullivan and Wilson 1995; Wilson
and Sullivan 1992; Wilson et al 1987; Woo and Leon 1987; Woo
et al 1996; Yuan et al 2003, 2004). The olfactory bulb learning-
associated changes are dependent on the reward causing the LC
to release high levels of NE, which prevent mitral cell habituation
to repeated odor presentations (Okutani et al 1998; Sullivan and
Wilson 1994, 2003; Yuan et al 2003) and are only acquired during
the early attachment period (Moriceau and Sullivan 2004b;
Sullivan and Wilson 1995; Woo and Leon 1987). Granule cells
(inhibitory interneurons of the olfactory bulb) modulate mitral
cell activity (Trombley and Shepherd 1992; Wilson and Leon
1988), and our learning-induced decrease in Fos expression in
granule cells suggests that learning also reflects less inhibition
within the bulb (Sullivan and Wilson 1994, 2003).
The axons of olfactory bulb mitral cells
project directly to the piriform cortex and other areas of the
olfactory cortex (Haberly 2001; Schwob and Price 1984). The
present results suggest that a section of the olfactory cortex, the
anterior piriform cortex, also encodes for neonatal odor learning,
as learning resulted in a significant increase in neural activity and
therefore adds another brain area to the neonate’s neural cir-
cuitry for odor learning and memory. The piriform cortex has
also been implicated in adult odor perception and learning
(Barkai and Saar 2001; Datiche et al 2001; Linster and Hasselmo
2001; Litaudon et al 1997; Mouly et al 2001; Ressler et al 2002;
Schoenbaum and Eichenbaum 1995; Tronel and Sara 2002;
Wilson and Stevenson 2003; Wilson et al 2004; Zinyuk et al 2001).
Amygdala. Consistent with previous results, we found that
the neonate amygdala does not appear to significantly participate
in odor learning during the sensitive period (Moriceau and
Sullivan 2004a; Moriceau et al 2004; Sullivan et al 2000a; Sullivan
and Wilson 1993), presumably contributing to readily learned
odor preferences (despite aversive conditioning) until PN 10.
Olfactory information projects directly to the amygdala, as well
as via the piriform cortex (Schwob and Price 1984). In adult and
older pup learning, the amygdala is intimately associated with
828 BIOL PSYCHIATRY 2005;57:823–831
T.L. Roth and R.M. Sullivan
learned fear (Ressler et al 2002; Rosenkranz and Grace 2002;
Sullivan et al 2000a; Tronel and Sara 2002; Fanselow and Gale
2003; Maren 2003; McIntyre et al 2003; Packard and Cahill 2001;
Schafe et al 2001; Walker and Davis 2002). The lack of amygdala
participation in neonate learning may not reflect immaturity,
since olfactory bulb afferent fibers are present in the amygdala at
birth, as are piriform afferents to the amygdala (Schwob and
Price 1984). Moreover, in other studies from our laboratory, we
have been able to prematurely incorporate the amygdala into the
neonate’s odor learning circuitry (before PN 10) and produce
fear conditioning by increasing pups corticosterone levels (Mor-
iceau and Sullivan 2004a; Moriceau et al 2004) or decreasing
pups’ opioid activity (Roth and Sullivan, unpublished data).
Thus, the results from this study suggest that the lack of
significant changes in gene expression within the neonate amyg-
dala appear to prevent fear learning at a time during develop-
ment that would hinder the attachment process.
Shared Circuitry of Attachment and Maltreatment
The neurobiological consequences of child maltreatment and
the neurobiology mediating attachment exhibit shared circuitry.
For example, childhood maltreatment produces long-term
changes in several brain areas, including the amygdala and locus
coeruleus, suggestive that altered development of these areas
contributes to the emergence of psychiatric disorders. Indeed,
studies continue to associate changes in amygdala function with
numerous psychiatric disorders (Liberzon and Phan 2003; Machado
and Bachevalier 2003; McEwen 2003a, 2003b; Pujol et al 2004;
Rothbaum and Davis 2003). We have shown the importance of
the amygdala and LC in supporting infant attachment, implying
that compromised development may jeopardize attachment. The
lack of amygdala participation during neonatal odor-pain learn-
ing and its susceptibility to changes induced by childhood
maltreatment suggests a common mechanism for attachment
despite abuse and the etiology of psychiatric disorders. An
understanding of this shared circuitry should provide insight into
the relationship between altered mother-infant attachment and
subsequent emotional health in maltreated children.
In summary, infants form attachments to their caregivers
regardless of the quality of the parental care (Bowlby 1965; Hofer
and Sullivan 2001). This is true for a wide range of species,
suggesting our model is useful in assessing principles with wide
phylogenetic importance (Rajecki et al 1978; Harlow and Harlow
1965; Helfer et al 1997; Hess 1962; Salzen 1970). Our assessment
of the neural basis of neonatal learning in the context of
maltreatment showed cellular changes within the olfactory bulb
and adds the anterior piriform cortex to the learning circuit,
indicating learning-related cortical processing. Our results further
confirm that a brain area important for emotional memory and
processing, the amygdala, does not appear to be significantly
participating in the neonate learning circuit. Moreover, we
suggest that our new maltreatment paradigm may provide an
understanding of how abusive treatment of offspring affects the
neural circuitry responsible for attachment, how an infant can
still form an attachment to an abusive caretaker, and the effect of
long-term experiences with maternal maltreatment on brain
development and psychiatric well-being.
This research was supported by Grant HHS-PHS NRSA F31
DA06082 (to TLR) and Grants NICHD-HD33402 and NSF-
IBN0117234 (to RMS).
For experimental assistance, we thank Dr. Donald Wilson,
Jennifer Cunningham, Ashley Hollingsworth, and Stephanie
Moriceau. We also thank Dr. Gordon Barr, Dr. Joseph Bastian,
Dr. Douglas Gaffin, and Dr. Donald Wilson for their comments
on an earlier draft of the manuscript.
Alberts JA, Cramer CP (1988): Ecology and experience. In: Blass EM, editor.
Barkai E, Saar D (2001): Cellular correlates of olfactory learning in the rat
piriform cortex. Rev Neurosci 12:111–120.
Barr GA (1995): Ontogeny of nociception and antinociception. NIDA Res
Bowlby J (1965): Attachment. New York: Basic Books.
Bremner JD (2003): Long-term effects of childhood abuse on brain and
learning. Prog Neurobiol 51:457–481.
Brent L, Koban T, Ramirez S (2002): Abnormal, abusive, and stress-related
behaviors in baboon mothers. Biol Psychiatry 52:1047–1056.
Maternal care during infant regulates the development of neural sys-
Camp LL, Rudy JW (1988): Changes in the categorization of appetitive and
Carlson V, Cicchetti D, Barnett D, Braunwald KG (1990): Finding order in
disorganization: Lessons from research on maltreated infants’ attach-
ments to their caregivers. In: Cicchetti D, Carlson V, editors. Child Mal-
treatment: The Theory and Research on the Causes and Consequences of
Child Abuse and Neglect. New York: Cambridge University Press, 494–
Carter CS, DeVries AC, Getz LL (1995): Physiological substrates of mamma-
lian monogamy: The prairie vole model. Neurosci Biobehav Rev 19:303–
Cicchetti D (1990): How research on child maltreatment has informed the
York: Cambridge University Press, 377–431.
Cicchetti D (1998): Child abuse and neglect–usefulness of the animal data:
Comment on Maestripieri and Carroll (1998). Psychol Bull 123:224–230.
Cicchetti D, Toth SL (1995): A developmental psychopathology perspective
on child abuse and neglect. J Am Acad Child Adolesc Psychiatry 34:541–
Connor DF, Doerfler LA, Volungis AM, Steingard RJ, Melloni RH Jr (2003):
Datiche F, Roullet F, Cattarelli M (2001): Expression of Fos in the piriform
cortex after acquisition of olfactory learning: An immunohistochemical
study in the rat. Brain Res Bull 55:95–99.
investigation and quantification of nest-building. Behaviour 34:1–16.
Dent GW, Smith MA, Levine S (2001): Stress-induced alterations in locus
coeruleus gene expression during ontogeny. Brain Res Dev Brain Res
infant common marmoset (Callithrix Jacchus, primates) and analysis of
its effects on early development. Biol Psychiatry 52:1037–1046.
Dragunow M, Bilkey D (2002): Neuroanatomical and functional mapping
using activation of transcription factors. In: Kaczmarek L, Robertson HA,
editors. Handbook of Chemical Neuroanatomy: Immediate Early Genes
Emerich DF, Scalzo FM, Enters EK, Spear N, Spear L (1985): Effects of 6-hy-
droxydopamine-induced catecholamine depletion on shock-precipi-
tated wall climbing of infant rat pups. Dev Psychobiol 18:215–227.
Fanselow MS, Gale GD (2003): The amygdala, fear, and memory. Ann N Y
T.L. Roth and R.M. Sullivan
BIOL PSYCHIATRY 2005;57:823–831 829
tive. In: Reite M, Caine NG, editors. Child Abuse: The Nonhuman Primate
Data. New York: Alan R Liss, Inc, 151–174.
Fleming AS, O’Say DH, Kraemer GW (1999): Neurobiology of mother-infant
interactions: Experience and central nervous system plasticity across
development and generations. Neurosci Biobehav Rev 23:673–685.
Francis DD, Caldji C, Champagne F, Plotsky PM, Meaney MJ (1999): The role
of corticotropin-releasing factor-norepinephrine systems in mediating
the effects of early experience on the development of behavioral and
endocrine responses to stress. Biol Psychiatry 46:1153–1166.
an immature rat model of continuous chronic stress. Pediatr Neurol 15:
Glaser D (2000): Child abuse and neglect and the brain–a review. J Child
Psychol Psychiatry 41: 97–116.
in the study of early experiences. Ann N Y Acad Sci 1008:238–247.
Haberly LB (2001): Parallel-distributed processing in olfactory cortex: New
insights from morphological and physiological analysis of neuronal cir-
cuitry. Chem Senses 26:551–576.
Hall WG (1979): Feeding and behavioral activation in infant rats. Science
Harlow HF, Harlow MK (1965): The affectional systems. In: Schrier A, Harlow
ogy of mood and anxiety disorders: Preclinical and clinical studies. Biol
Helfer ME, Kempe RS, Krugman RD (1997): The Battered Child. Chicago: Uni-
versity of Chicago Press.
Herrera DG, Robertson HA (1996): Activation of c-fos in the brain. Prog
In: Brown R, Galanter E, Hess EH, Mendler G, editors. New Directions in
Psychology. New York: Rinehart and Winston, 159–199.
Hesse E, Main M (2000): Disorganized infant, child, and adult attachment:
Hofer MA, Sullivan RM (2001): Toward a neurobiology of attachment. In:
Nelson CA, Luciana M, editors. Handbook of Developmental Cognitive
Neuroscience. Cambridge, MA: MIT Press, 599–616.
Huot RL, Plotsky PM, Lenox RH, McNamara RK (2002): Neonatal maternal
separation reduces hippocampal mossy fiber density in adult Long
Evans rats. Brain Res 950:52–63.
Johnson BA, Woo CC, Duong H, Nguyen V, Leon M (1995): A learned odor
evokes an enhanced fos-like glomerular response in the olfactory bulb
of young rats. Brain Res 699:192–200.
Kaczmarek L (2002): c-Fos in learning: Beyond the mapping of neuronal
activity. In: Kaczmarek L, Robertson HA, editors. Handbook of Chemical
New York: Elsevier, 189–215.
Kaufman J, Plotsky PM, Nemeroff CB, Charney DS (2000): Effects of early
adverse experiences on brain structure and function: Clinical implica-
tions. Biol Psychiatry 48:778–790.
Kendrick KM, Da Costa AP, Broad KD, Ohkura S, Guevara R, Levy F, et al
(1997): Neural control of maternal behaviour and olfactory recognition
Levine S (2001): Primary social relationships influence the development of
Liberzon I, Phan KL (2003): Brain-imaging studies of posttraumatic stress
disorder. CNS Spectr 8:641–650.
Linster C, Hasselmo ME (2001): Neromodulation and the functional dynam-
ics of piriform cortex. Chem Senses 26:585–594.
Litaudon P, Mouly A, Sullivan R, Gervais R, Cattarelli M (1997): Learning-
induced changes in rat piriform cortex activity mapped using multisite
recording with voltage sensitive dye. Eur J Neurosci 9:1593–1602.
Liu D, Diorio J, Day JC, Francis DD, Meaney MJ (2000): Maternal care, hip-
pocampal synaptogenesis and cognitive development in rats. Nat Neu-
Maestripieri D (1998): Parenting styles of abusive mothers in group-living
rhesus macaques. Anim Behav 55:1–11.
animal data. Psychol Bull 123:211–223.
McEwen BS (2003a): Early life influences on life-long patterns of behavior
McEwen BS (2003b): Mood disorders and allostatic load. Biol Psychiatry 54:
lateral amygdala in memory consolidation. Ann N Y Acad Sci 985:273–
McLean JH, Darby-King A, Sullivan RM, King SR (1993): Serotonergic influ-
ence on olfactory learning in the neonate rat. Behav Neural Biol 60:152–
McLean JH, Harley CW, Darby-King A, Yuan Q (1999): pCREB in the neonate
ence-conditioned training. Learn Mem 6:608–618.
Meaney MJ (2001): Maternal care, gene expression, and the transmission of
individual differences in stress reactivity across generations. Annu Rev
Moriceau S, Roth TL, Okotoghaide T, Sullivan RM (2004): Corticosterone
controls the developmental emergence of fear and amygdala function
to predator odors in infant rat pups. Int J Dev Neurosci 22:415–422.
Moriceau S, Sullivan RM (2004a): Corticosterone influences on mammalian
neonatal sensitive-period learning. Behav Neurosci 118:274–281.
Moriceau S, Sullivan RM (2004b): Unique neural circuitry for neonatal olfac-
tory learning. J Neurosci 24:1182–1189.
Mouly AM, Fort A, Ben-Boutayab N, Gervais R (2001): Olfactory learning
induces differential long-lasting changes in rat central olfactory path-
ways. Neuroscience 102:11–21.
Nakamura S, Sakaguchi T (1990): Development and plasticity of the locus
imentation. Prog Neurobiol 34:505–526.
Nemeroff CB (2004): Neurobiological consequences of childhood trauma.
Okutani F, Kaba H, Takahashi S, Seto K (1998): The biphasic effects of locus
rat olfactory bulb. Brain Res 783:272–279.
Packard MG, Cahill L (2001): Affective modulation of multiple memory sys-
Paschall GY, Davis M (2002): Olfactory-mediated fear-potentiated startle.
Paxinos G, Watson C (1986): The Rat Brain in Stereotaxic Coordinates. San
Diego: Academic Press.
Pollak SD, Tolley-Schell SA (2003): Selective attention to facial emotion in
physically abused children. J Abnorm Psychol 112:323–338.
Pryce CR, Dettling AC, Spengler M, Schnell CR, Feldon J (2004): Deprivation
of parenting disrupts development of homeostatic and reward systems
in marmoset monkey offspring. Biol Psychiatry 56:72–79.
Pujol J, Soriano-Mas C, Alonos P, Cardoner N, Menchon JM, Deus J, et al
(2004): Mapping structural brain alterations in obsessive-compulsive
Rajecki DW, Lamb ME, Obmascher P (1978): Toward a general theory of
Rangel S, Leon M (1995): Early odor preference training increases olfactory
Ressler KJ, Paschall G, Zhou X, Davis M (2002): Regulation of synaptic plas-
ticity genes during consolidation of fear conditioning. J Neurosci 22:
830 BIOL PSYCHIATRY 2005;57:823–831
T.L. Roth and R.M. Sullivan
Richardson R, Paxinos G, Lee J (2000): The ontogeny of conditioned odor Download full-text
potentiation of startle. Behav Neurosci 114:1167–1173.
Evaluation Guide. New York: The Haworth Press, Inc.
Rosenblum LA, Coplan JD, Friedman S, Bassoff T, Gorman JM, Andrews MW
functioning in adult primates. Biol Psychiatry 35:221–227.
Roth TL, Sullivan RM (2001): Endogenous opioids and their role in odor
preference acquisition and consolidation following odor-shock condi-
tioning in infant rats. Dev Psychobiol 39:188–198.
Salzen EA (1970): Imprinting and environmental learning. In: Aronson LR,
of Behavior. San Francisco: W.H. Freeman.
Sanchez MM, Ladd CO, Plotsky PM (2001): Early adverse experience as a
developmental risk factor for later psychopathology: Evidence from ro-
dent and primate models. Dev Psychopathol 13:419–449.
Schafe GE, Nader K, Blair HT, LeDoux JE (2001): Memory consolidation of
Schapiro SJ, Mitchell G (1983): Infant-directed abuse in a seminatural envi-
The Nonhuman Primate Data. New York: Alan R Liss, Inc, 29–48.
Schoenbaum G, Eichenbaum H (1995): Information coding in the rodent
prefrontal cortex. I. Single-neuron activity in orbitofrontal cortex com-
pared with that in pyriform cortex. J Neurophysiol 74:733–750.
Schore AN (2002): Dysregulation of the right brain: A fundamental mecha-
Schwob JE, Price JL (1984): The development of axonal connections in the
central olfactory system of rats. J Comp Neurol 223:177–202.
Sullivan RM (2003): Developing a sense of safety: The neurobiology of neo-
Sullivan RM, Hofer MA, Brake SC (1986): Olfactory-guided orientation in
neonatal rats is enhanced by a conditioned change in behavioral state.
Sullivan RM, Landers M, Yeaman B, Wilson DA (2000a): Good memories of
bad events in infancy. Nature 407:38–39.
Sullivan RM, Leon M (1987): One-trial olfactory learning enhances olfactory
locus coeruleus stimulation is sufficient to produce learned approach
responses to that odor in neonatal rats. Behav Neurosci 114:957–962.
tory associative learning. Behav Neurosci 107:254–263.
Sullivan RM, Wilson DA (1994): The locus coeruleus, norepinephrine, and
memory in newborns. Brain Res Bull 35:467–472.
Sullivan RM, Wilson DA (1995): Dissociation of behavioral and neural corre-
lates of early associative learning. Dev Psychobiol 28:213–219.
Teicher MN, Andersen SL, Polcari A, Anderson CM, Navalta CP, Kim DM
Tischmeyer W, Grimm R (1999): Activation of immediate early genes and
Trombley PQ, Shepherd GM (1992): Noradrenergic inhibition of synaptic
transmission between mitral and granule cells in mammalian olfactory
bulb cultures. J Neurosci 12:3985–3991.
Tronel S, Sara SJ (2002): Mapping of olfactory memory circuits: Region spe-
cific c-fos activation after odor-reward associative learning or after its
retrieval. Learn Mem 9:105–111.
van Oers HJJ, de Kloet ER, Whelan T, Levine S (1998): Maternal deprivation
tion and feeding but not by suppressing corticosterone. J Neurosci 18:
learning, fear-potentiated startle, and extinction. Pharmacol Biochem
White NR, Adox R, Reddy A, Barfield RJ (1992): Regulation of rat maternal
Wilson DA, Fletcher ML, Sullivan RM (2004): Acetylcholine and olfactory
perceptual learning. Learn Mem 11:28–34.
Wilson DA, Leon M (1988): Noradrenergic modulation of olfactory bulb
Wilson DA, Stevenson RJ (2003): The fundamental role of memory in olfac-
tory perception. Trend Neurosci 26:243–247.
Wilson DA, Sullivan RM (1992): Blockade of mitral/tufted cell habituation to
Wilson DA, Sullivan RM, Leon M (1987): Single-unit analysis of postnatal
olfactory learning: Modified olfactory bulb output response patterns to
learned attractive odors. J Neurosci 7:3154–3162.
development to learned odors. Brain Res 433:309–313.
Woo CC, Oshita MH, Leon M (1996): A learned odor response decreases the
number of fos-immunopositive granule cells in the olfactory bulb of
young rats. Brain Res 716:149–156.
Yuan Q, Harley CW, McLean JH (2003): Mitral cell ?1 and 5-HT2Areceptor
induced learning in the olfactory bulb. Learn Mem 10:5–15.
cell dendrites of olfactory bulbs of neonatal rats and mice during olfac-
tory nerve stimulation and ?-adrenoceptor activation. Learn Mem 11:
Zeanah CH, Keyes A, Settles L (2003): Attachment relationship experiences
Zigler E, Hall NW (1990): Physical child abuse in America: Past, present, and
future. In: Cicchetti D, Carlson V, editors. Child Maltreatment: Theory and
York: Cambridge University Press, 38–75.
T.L. Roth and R.M. Sullivan
BIOL PSYCHIATRY 2005;57:823–831 831