The opinion of the Court was delivered by
HANDLER, J.
The State may seek guardianship of a child placed in foster care if it believes that the child has been abandoned by his or her natural parents or would suffer injury if returned to them. Transferring guardianship to the State terminates all the parental rights of the natural parents and is a prerequisite to having the child adopted by the foster parents or by another family. The Court in this case, as in the companion case, In re K.L.F., 129 N.J. 32, 608 A.2d 1327 (1992), is required to determine whether the parental rights of a natural mother should be terminated based on the need to protect children from potential harm that may result from being separated from foster parents with whom the children may have formed parental bonds.
The mother in this case has three children. Unable to cope with the difficulties of raising them, primarily because of homelessness, domestic abuse, and her own substance abuse, the mother voluntarily placed her children in foster care through the Division of Youth and Family Services. The children remained in foster care from 1986 until today, and although the mother regularly visited with them, the Division eventually determined that she lacked parental fitness and that the children required permanent homes. Acting on what it determined to be in the best interest of the children, the Division brought an action to terminate the mother’s parental rights in order that the children could be adopted.
The trial court held that the mother’s parental rights should be terminated, emphasizing the psychological harm that would result from breaking the bonds that each child had formed with foster caretakers. The Appellate Division affirmed in a per curiam opinion. The Court granted the mother’s petition for certification. 127 N.J. 549, 606 A.2d 362 (1991).
[6]*6I
A.C., who was bom in Colombia and came to this country as a teenager, is the natural mother of three children. Two girls, J.C. and J.M.C., were born in July 1983 and in January 1985, respectively, and J.C., a boy, was born in August 1986. A.C. voluntarily placed her two girls in foster care with the Division of Youth and Family Services (DYFS, Division, or agency) in August 1985. The children were returned to her after three months. Almost a year later, in October 1986, A.C. again placed the two girls, along with her new child, J.C., in foster care, where they have remained for the past five and a half years.
A.C. began unsupervised weekend visits with her children soon after their placement in foster care, seeing them regularly twice a month during the following year. Although DYFS had intended to reunite the family, in November 1987 the agency stopped unsupervised visits out of concern that the children were not being properly cared for. DYFS also came to believe that A.C. was addicted to drugs and was being abused by her husband (who, she claims, was not the father of any of the children). However, bi-monthly visits at the DYFS office continued. In April of the following year, A.C. entered drug treatment. By November 1988 the agency concluded that the children could not be returned successfully and that preparation should be initiated for their permanent placement and adoption. The agency transferred the case to its Adoption Resources Center (A.R.C.), which subsequently terminated visitation.
DYFS filed a petition for guardianship over the three children on July 7, 1989. It sought the termination of A.C.’s parental rights on the grounds that A.C. was unable and unwilling to stop causing the children harm and that to delay permanent placement would add to the harm facing the children.
At the time that DYFS moved for guardianship, the oldest child, J.C., had lived with at least two foster families. She was moved to her current pre-adoptive parents a week later, on July [7]*713, 1989. J.M.C. had been living with her current foster parents since October or November 1988. A.C. has since consented to the adoption of her youngest child, J.C., and his status is not an issue in the case.
The case was initially tried on November 9,1989, and December 15, 1989. Following a remand and additional hearings held in March 1991, the trial court concluded that termination of A.C.’s parental rights was necessary in the best interests of the children. It determined that A.C. had not, as a matter of law, abandoned her children even though she had placed them in foster care and had failed to achieve the requisite fitness to secure their return. However, it did find that the children would suffer serious psychological harm if they were removed from their foster or pre-adoptive homes and returned to A.C., and that the harm in part was attributable to A.C.’s own inability to plan for their future and her failure to rehabilitate herself. The Appellate Division affirmed.
II
Foster care is one of several social services that the Division of Youth and Family Services is empowered to provide to troubled families. A child may come into the custody of the Division and be placed in foster care pursuant to either a voluntary-placement agreement or a court order. In this case, the children were placed voluntarily without the involvement or review of a court. Under the voluntary-placement scheme, the decision to place a child in foster care rests solely with child’s parent or guardian. However, before providing foster care services DYFS must itself determine that a child’s welfare is endangered and that the child’s needs cannot be met either through financial assistance or placement with family or friends. N.J.S.A. 30:4C-11.
The law governing DYFS reflects a strong societal bent in favor of the integrity of the natural family. The law clearly favors keeping children with their natural parents and resolv[8]*8ing care and custody problems within the family. When children are placed in foster care, the law requires diligent efforts by the State to maintain a relationship between children and their parents and to return children home as quickly as possible. Nonetheless, when efforts to reunite families repeatedly fail, permanent plans must be made for children, justifying the termination of parental rights. The latter action requires court approval pursuant to a petition by the agency for guardianship. N.J.S.A. 80:40-15.
DYFS brought its legal action for guardianship over J.C. and J.M.C. under N.J.S.A. 30:40-15 (section 15), which provided for the termination of parental rights either when it is required in “the best interests of [the] child” [subsection c] or the parent “has failed substantially and continuously or repeatedly for a period of more than 1 year to maintain contact with and plan for the future of the child, although physically and financially able to do so.” [subsection d] Ibid. The Division relied on subsection (c) of section 15 as the basis for seeking termination of A.C.’s parental rights. N.J.S.A. 30:4C-20 (section 20) defines the court’s authority to terminate parental rights based on an application by DYFS under section 15. It may do so “[when it] is satisfied that the best interests of [the] child require that he be placed under proper guardianship.”
The Court in New Jersey Division of Youth and Family Services v. A.W., 103 N.J. 591, 512 A.2d 438 (1986), undertook a comprehensive examination of the foregoing statutory scheme. The Court, through Justice O’Hern, reasoned that although the parental interest in a relationship with children is fundamental and constitutionally protected, it is limited when the physical or mental health of children is jeopardized. Both the statute and the State Constitution, the Court explained, require that DYFS demonstrate clearly and convincingly that “the child’s best interests will be substantially prejudiced” if parental rights are not terminated. Id. at 603, 512 A.2d 438 (quoting In re Guardianship of Cope, 106 N.J.Super. 336, 340-41, 255 A.2d 798 (App.Div.1969)).
[9]*9The Court then set out four specific findings that a trial court should make before it terminates parental rights. The first finding is that the child’s health and development have been or will be seriously impaired by the parental relationship. Id. at 604-05, 512 A.2d 438. Secondly, the court must conclude that the parents are unable or unwilling to eliminate the harm and that a delay in permanent placement will add to the harm. Id. at 607, 512 A.2d 438. Third, the court should be convinced that alternatives to terminating parental rights have been thoroughly, explored and exhausted, including sufficient efforts made to help the parents cure the problems that led to the placement. Id. at 608-10, 512 A.2d 438. Fourth, all of those considerations must inform the determination that termination of parental rights will not do more harm than good. Id. at 610, 512 A.2d 438.
In 1991, the Legislature amended section 15, L. 1991, c. 275, § 7, to require that before the Division petitions for guardianship based on the child’s best interest, it be satisfied that the four standards articulated in A.W. have been met. Although the language amends section 15 governing agency discretion, rather than section 20 defining the judicial responsibility, it refers to the inquiry that a court must make in the termination proceedings under section 20. L. 1991, c. 275, § 7c. That the Legislature intended also to clarify the substantive standard for termination under section 20 and that the best interest standard is identical for both agency and court decisions is thus inferable.
The standards for terminating parental rights established by the Legislature and this Court are fully consistent with constitutional doctrine. State law allowing for the termination of parental rights must satisfy the protections that surround family autonomy under the United States Constitution. Parents have a constitutionally-protected, fundamental liberty interest in raising their biological children, even if those children have been placed in foster care. Santosky v. Kramer, [10]*10455 U.S. 745, 102 S.Ct. 1388, 71 L.Ed.2d 599 (1982). Balanced against the constitutional protection of family rights is the State’s parens patriae responsibility to protect the welfare of children. Because that power involves the State acting in the place of parents, it is limited to situations in which the state has demonstrated that the child’s parent or custodian is unfit, Developments in the Law, The Constitution and the Family, 93 Harv.L.Rev. 1156, 1219 (1980), or the child has been neglected or harmed, see State v. Perricone, 37 N.J. 463, 181 A.2d 751, cert. denied, 371 U.S. 890, 83 S.Ct. 189, 9 L.Ed.2d 124 (1962).
Termination of parental rights permanently cuts off the relationship between children and their biological parents. “Few forms of state action are both so severe and so irreversible.” Santosky, 455 U.S. at 759, 102 S.Ct. at 1398, 71 L.Ed.2d at 610; In re Adoption by J.J.P., 175 N.J.Super. 420, 426, 419 A.2d 1135 (App.Div.1980). When the child’s biological parents resist the termination of their parental rights, the court’s function will ordinarily be to decide whether the parents can raise their children without causing them further harm. In most cases proofs will focus on past abuse and neglect and on the likelihood of it continuing. See, e.g., In re Guardianship of J.E.D., 217 N.J.Super. 1, 524 A.2d 1255 (App.Div.1987). However, the cornerstone of the inquiry is not whether the biological parents are fit but whether they can cease causing their child harm. A.W., 103 N.J. at 607, 512 A.2d 438. The analysis of harm entails strict standards to protect the statutory and constitutional rights of the natural parents. The burden falls on the State to demonstrate by clear and convincing evidence that the natural parent has not cured the initial cause of harm and will continue to cause serious and lasting harm to the child. Santosky, supra, 455 U.S. at 768, 102 S.Ct. at 1402, 71 L.Ed.2d at 616-17.
Ill
In this case both parties presented evidence relating generally to the harm that would befall the children if they were [11]*11removed from the custody and care of foster parents and returned to their natural parent. DYFS also presented evidence that addressed the conduct of A.C. as indicative of her own responsibility for having caused the harm to the children and her parental fitness as a factor in determining whether that harm coúld be remediated if the children were returned to her custody.
The evidence strongly indicated that A.C. showed an interest in her children while they were in foster care, visiting them regularly and frequently. DYFS recognized that interest but stopped short of returning the children to A.C. on several occasions due to concerns about her housing situation and drug and alcohol addiction. Toward the end of 1987, after the two girls had been in foster care for more than a year, DYFS made plans to return the children, one by one, to A.C. Those plans were cancelled after reports of domestic violence in A.C.’s home, as well as renewed concerns over her continuing drug and alcohol abuse. In February 1988, DYFS had a psychiatric examination of A.C. undertaken by the Urban League of Hudson County. That report, written by Dr. Willy Hoffmeister, a psychologist, reached equivocal conclusions. It noted A.C.’s troubled and unstable personality, but expressed an inability to make a recommendation concerning whether she could care for her children.
In April 1988, with the help of the Urban League, A.C. entered a twenty-eight-day in-patient drug- and alcohol-treatment program in Newark. Nydia Farias, the social worker in the case, testified that she had told A.C. that DYFS had made a decision that the case would be transferred to the A.R.C. for purposes of arranging for the adoption of the children. However, she had also told A.C. that if she wanted the children returned, she would have to continue receiving outpatient drug treatment and find a larger apartment. Ms. Farias testified that A.C.’s failure to have completed the in-patient portion of a drug-rehabilitation program that she began in April 1988 was a significant factor in DYFS’s decision to seek a permanent home [12]*12for the children. She produced letters from C.U.R.A., Inc., the sponsor of A.C.’s drug-rehabilitation program, stating that A.C. had successfully completed a twenty-eight-day residential program, but had declined to participate in an extended inpatient program and had begun, but failed to complete, a supplemental out-patient program. Whether DYFS reasonably relied on that correspondence from C.U.R.A., however, was a question raised by the testimony of Maria Baugh, A.C.’s drug counselor at C.U.R.A. Ms. Baugh, testifying in December 1989, expressed the view that A.C. had successfully completed C.U.R.A.’s thirty-two-day residential program and had remained free from drug and alcohol addiction since the conclusion of the program in May 1988. Ms. Baugh observed that after completing the program A.C. was “a totally different person. She had another view of her life, her values, her growth, her self-esteem.” She testified that A.C. had maintained regular contact with her, either monthly or semi-monthly, since May 1988.
Ms. Farias testified further that A.C.’s housing situation remained highly unstable between the time she finished the drug-treatment program in May and the time the case was officially transferred to the A.R.C. in November 1988. Tanya Rodriguez, who then became the case worker, testified that she kept in communication with A.C., discussing her plans for further drug counseling. She also was concerned with the inadequate size of A.C.’s apartment and the degree to which she had recovered from her drug addiction.
In addition to the evidence relating to A.C.’s conduct and fitness, the bulk of the evidence addressed the issue of harm to the children emanating from the prospect of their being removed from their respective foster parents and returned to their mother. Most of the evidence related to the older girl, J.C.
The DYFS social workers and the psychologist who evaluated J.C. concurred that she was a child with serious emotional problems and potentially-significant learning disabilities. None[13]*13theless, Ms. Rodriguez testified that J.C. “[was] doing very well [in a new home, i.e., that of Mr. and Mrs. D., her foster parents] since July of this year. She [was] well adjusted. She has done remarkably well, especially in school where she has been so behind.” Referring to the report of Dr. Paul Kennedy, a child psychiatrist, the witness stated, “[J.C.] is able to attach and trust someone again. And she is very bonded to this new caretaker.” Dr. Kennedy’s written psychiatric evaluation described the fragile and volatile nature of J.C.’s personality, observing that her “anger is that of a deprived child who feels easily threatened” and that “[h]er primary problem is emotional and is the result of her history.” Dr. Kennedy’s report also mentioned the positive relationship between J.C. and her foster parents, noting that “[t]here seemed to be a mutually affectionate relationship.”
John Frederickson, the guardian of the children, visited each child at her foster home, interviewing the children and their pre-adoptive parents. As to J.C., he submitted a report in which he “found that [J.C.] is very attached to Mr. & Mrs. [D.] as well as Nicole [Mr. D.’s daughter by an earlier marriage].” He noted further, “J.C. indicated that she is very happy here and wished Mr. & Mrs. [D.] to be ‘her mommy and daddy’ and wants very much to live with them and have Nicole as her sister.”
After receiving Mr. Frederickson’s report, the court requested a “bonding evaluation between the children and their respective foster parents” from a court psychologist, Regina Johnson. With respect to J.C., Ms. Johnson concluded:
[J.C.] is most definitely bonded with the [D] family. She is expressive and stated, “I love them a lot, I want to stay with them — they are nice to me.” ... [Separation from the [D’s] would be detrimental to this child’s emotional and physical well-being____ Further separation will leave permanent scars.
The evidence relating to the younger child, J.M.C., was not as extensive. Ms. Rodriguez testified that J.M.C. has been in her current foster home since the fall of 1988 and “is very well adjusted. She is a very happy child. She relates to the foster [14]*14mother as ... mommy” and is very close to another child in the family.
Frederickson also submitted a report about J.M.C.. He visited her foster home on January 9, 1990. J.M.C. had lived in the foster home for two years, and Frederickson found that she had “bond[ed] with her potential adopted sister, Josephine[,] as well as her potential adoptive mother [Ms. P.].” Based on his finding of close relationships between J.M.C. and her foster mother and sister, the guardian recommended, subject to the foster mother meeting certain conditions concerning her living situation, that J.M.C. be adopted by Ms. P. and that the child’s parental relationship with her natural mother be terminated. He also wrote that even if the placement did not work out, “it [was] in the best interest of the child to terminate the parental relationship with [J.M.C.’s] natural mother.”
Ms. Johnson testified with respect to J.M.C.: “There is no doubt that she is well bonded to Mrs. [P.] or her ‘step-sister’ ____ She was affectionate to Mrs. [P] and referred to her as ‘Mother.’ ... [H]er relationship is ‘firm’ and separation from Mrs. [P] and the other children, at this time, would be traumatic and lead to poor affect for this pre-school child.” She recommended expressly that there be a “termination of parental rights.”
Based on the trial testimony and the subsequent report the trial court on May 22, 1990, ordered that parental rights be terminated and guardianship transferred to DYFS. The trial court found that A.C. had failed for well over a year to plan for the return of the children, she failed to follow through on drug and alcohol counseling, and the children had bonded to their foster parents. The Appellate Division remanded because Ms. Johnson’s bonding evaluation had not been disclosed until the trial court’s opinion had been filed and A.C. should have had an opportunity to cross examine her, as well as to present additional expert evidence.
[15]*15On remand, beginning in March 1991, the court heard six days of additional testimony from Ms. Johnson as well as from other experts. First, the parties stipulated to several facts about A.C.’s rehabilitation. With respect to her housing situation, they agreed that she was living in the same apartment in which she had been living since December 1989. Concerning her work, they agreed that she had a steady job, that she had been working there for the previous two years, and that there was on-site after school child care at her workplace. Finally the parties “stipulated” that A.C. asserted that she was drug and alcohol free and that DYFS had no evidence to suggest otherwise.
In addition to the testimony of Ms. Johnson, the court-appointed counselor, DYFS produced its own expert, Dr. Martha Page. Dr. Page testified only with respect to J.C., with whom she had a counseling relationship. Dr. Page stated that when she met with J.C. in November 1990, she “was having a great deal of difficulty in school” and “[h]er behavior was very uneven” and “she was extremely hard to handle.” She described J.C. as an “emotionally disturbed child” with special needs and a low resiliency to change, one who “needs stability ... [and] a sense of identity.” According to Dr. Page, separating J.C. from her foster parents would “reinforce her notion that she’s somehow failed again.” She mentioned J.C.’s fear of being rejected and emphasized her need for consistent care and permanency in planning for placement. Dr. Page also believed that if the child were returned to A.C. and the placement did not work out, the result would be particularly damaging to the emotional health of the child. She also concluded that any visitation with A.C. and delay in J.C.’s permanent placement would be “disastrous” and result in “transplant shock.”
Dr. Matthew Johnson testified on behalf of A.C. He found a strong and enduring bonded relationship between J.C. and A.C. He believed “erasing” the biological mother from the child’s life would cause the child serious emotional harm, particularly regarding the child’s identity and development in adolescence. [16]*16Thus, although he recognized potential danger in moving J.C. out of her current foster home, he also felt that terminating her relationship with her natural mother could also cause serious long-term harm. Dr. Johnson expressed concerns about A.C.’s ability to care for the children, mentioning among other things that she might have “residual emotional difficulty” in caring for the children and that she appeared to lack social networks necessary to support her in times of crisis. On the other hand, he noted that she had a number of strengths, that she was intelligent, had largely cleaned up her life, and had maintained contact with the children. Nevertheless, when asked on direct examination whether he believed A.C. could “assume custodial care of [J.C.],” Dr. Johnson again said that “to be frank there are concerns,” and emphasized that “this child [J.C.] had emotional disturbance.”
There was much less focus during the second trial on the younger child, J.M.C. Dr. Johnson found that she had significant relationships with both her foster mother and her natural mother. However, he was not able to say conclusively that there was bonding in either case.
The experts testified to some extent about the foster parents as prospective adoptive parents. Dr. Page addressed a number of concerns raised about the foster parents seeking to adopt J.C. She believed that the problems could be worked out and that the child was making progress with the D.’s. Dr. Johnson also expressed certain concerns about the D.'s and their relationship with J.C. He recommended that the foster parents receive professional counseling. Nevertheless, Dr. Johnson made no recommendation on who should have final custody over the two children, but suggested continued visitation and, at least at that time, the preservation of parental rights in A.C.
IV
This case presents no basis for terminating the parental rights of the mother based on abandonment under section 15(d). [17]*17Abandonment requires a finding that a parent, although physically and financially able to fulfill her parental responsibilities, willfully forsook them. The concept of abandonment entails volitional and purposeful conduct that equates with a willful giving up of parental rights and duties. See In re K.L.F., supra, 129 N.J. at 38-40, 608 A.2d 1327. A.C. was not guilty of that kind of conduct.
The critical question is whether termination of parental rights is justified under the broad statutory standard of section 15(c) and section 20 predicated on the best interests of the child.
The trial court found that the children had bonded to their foster parents, and that this bonding had been caused by or exacerbated by A.C.’s conduct. With respect to J.C., the court found the child to be emotionally disturbed as the result of her mother’s actions. Given her special needs and her earlier traumatic experiences with A.C., the court concluded that J.C. would suffer greatly from being moved. The court also expressed concern about the consequences of a potential relapse by A.C. into drug and alcohol abuse. In the case of J.M.C., the trial court did not find her to be suffering from emotional harm caused by A.C. and suggested that A.C. was “capable of caring for a normal child.” It based its decision to terminate parental rights on the substantial psychological harm to J.M.C. that would result from severing her relationship with her pre-adoptive mother.
The Appellate Division affirmed the trial court’s conclusions with respect to both children, finding that the lower court’s determination was supported by clear and convincing evidence in the record. The Appellate Division accepted the conclusion that although A.C. might have rehabilitated herself, the children enjoyed stability in the care of their foster parents and faced serious harm on being removed from their foster homes, which would be “the proximate result of A.C.’s failures over an [18]*18extended period.” (Citing Sorentino v. Family & Children’s Soc., 74 N.J. 313, 324, 378 A.2d 18 (1977)).
Without disparaging the reasoning and determinations made by the courts below, we are compelled by the record as it currently stands to conclude that there is not clear and convincing evidence to support the findings necessary to terminate parental rights. Although a significant amount of testimony has been taken in this case, much of the evidence was either flawed or insufficient to answer the central question of serious psychological harm. The problem with the sufficiency and relevance of the evidence as reflected in the record no doubt derives from the unsettled state that has existed about the legal significance of bonding and psychological parenting theories in termination disputes. In this case and the companion case, In Re KLF, supra we have sought to clarify how these theories comport with the standards for termination and to provide a framework within which essential evidence implicating those theories can be identified, produced and evaluated.
In cases in which DYFS seeks termination of parental rights, not on grounds of current unfitness but because of potential harm to the child based on separation from a foster parent with whom the child has bonded, the quality of the proof adduced must be consistent with the interests at stake. To the extent that the quality of the child’s relationship with foster parents may be relevant to termination of the natural parents’ status, that relationship must be viewed not in isolation but in a broader context that includes as well the quality of the child’s relationship with his or her natural parents. As suggested by In re Guardianship of J.R., 174 N.J.Super. 211, 223, 416 A.2d 62 (App.Div.1980), prolonged inattention by natural parents that permits the development of disproportionately stronger ties between a child and foster parents may lead to a bonding relationship the severing of which would cause profound harm — a harm attributable to the natural parents and cognizable under the standards set forth in A.W., supra, 103 N.J. at [19]*19604-11, 512 A 2d 438. To show that the child has a strong relationship with the foster parents or might be better off if left in their custody is not enough. See In re Baby M, supra, 109 N.J. 396, 445, 537 A.2d 1227 (1988). DYFS must prove by clear and convincing evidence that separating the child from his or her foster parents would cause serious and enduring emotional or psychological harm. Santosky, supra, 455 US. at 768, 102 S.Ct. at 1402, 71 L.Ed.2d at 616-17; see also Division of Youth & Family Servs. v. T. C., 251 N.J.Super. 419, 440, 598 A.2d 899 (App.Div.1991) (“termination (justified] in order to spare the child grievous and irreparable psychological harm”). Such proof should include the testimony of a well qualified expert who has had full opportunity to make a comprehensive, objective, and informed evaluation of the child’s relationship with the foster parent. In hearing a petition for termination in which the fitness of natural parents is neither relied on nor disputed by the agency, the trial court must also consider parallel proof relating to the child’s relationship with his or her natural parents in assessing the existence, nature, and extent of the harm facing the child.
As the contrasting opinions of the experts in this case illustrate, there are competing psychological theories of the effects of parental bonding. In large measure, the variances in their recommendations derive from different assumptions concerning the fragility versus resiliency of the child psyche. Compare Joseph Goldstein, Ana Freud, & Albert Solnit, Beyond the Best Interests of the Child (1973) with Everett Waters & Donna Noyes, Psychological Parenting vs. Attachment Theory: The Child's Best Interests and the Risks of Doing the Right Things for the Wrong Reasons, 12 N.Y.U.Rev.L. & Soc. Change 505, 513 (1983-84). Those who, like Dr. Page, urge the wider use of psychological parenting theory see children as highly vulnerable and fragile. Their psyches are easily injured by traumatic events and those injuries can adversely shape their subsequent development. See Goldstein, Freud & Solnit, supra, at 33. In contrast, others, presumably like Dr. [20]*20Johnson, posit more flexibility in children, arguing that attachments “support[] the development of independence.” Waters & Noyes, supra, 12 N.Y.U.Rev.L. & Soc. Change at 509-10. Change under the right circumstances can play a positive role in children’s development. Ibid,.; Carol Stack, Cultural Perspective on Child Welfare, 12 N.Y.U.Rev.L. & Soc. Change, 539 (1983-84). Indeed, a good deal of recent literature on the subject argues that psychological parenting theory overestimates the importance of continuity in care in relation to other factors that affect child development, such as the quality of care children receive. See Marsha Garrison, Why Terminate Parental Rights?, 35 Stan.L.Rev. 423, 458-59 (1983). In addition, experts and commentators differ over the importance of ongoing relationships between children and their natural parents. Although natural parents can be a disruptive influence for children who have been adopted, some commentators and psychologists believe that trying to eliminate the natural parents from the children’s lives and memory is impossible, and therefore wrong. See Peggy Davis, Use and Abuse of the Power to Sever Family Bonds, 12 N.Y.U.Rev.L. & Soc. Change, 557, 567-70 (1983-84); Fernando Colon, Family Ties and Child Placement, 17 Family Process 289 (1978).
Moreover, there are the grave pitfalls that may be encountered in the application of otherwise sound psychological parenting and bonding theories. Scholars and some courts suggest that theories of parental bonding may be relied on too often to keep children in foster care rather than return them to their parents. E.g., In re Interest of L.J., 220 Neb. 102, 368 N.W.2d 474, 483 (1985); Malcolm Bush & Harold Goldman, The Psychological Parenting and Permanency Principles in Child Welfare: A Reappraisal and Critique, 52 Amer. J. Orthopsychiatry, 223, 226 (1982). Parents with few resources rely on foster care to protect their children during difficult periods, including but not limited to experiences of homelessness and domestic violence. A single-minded focus on continuity in care can result in parents who rely temporarily on foster [21]*21care for needed assistance, finding it impossible to regain custody over their children. Parents, particularly those with limited incomes and unstable housing and work experiences, should be able to turn to the foster-care system without fear of losing their children.
In addition, the uncritical use of bonding theory can increase the risk of institutional bias. Use and Abuse of the Power to Sever Family Bonds, supra, 12 N.Y.U.Rev.L. & Soc. Change, at 567. That risk may be reflected in attitudes that tilt the process in favor of the agency and its social workers and foster parents. See Martin Guggenhein, The Political Implications of the Psychological Parenting Theory, 12 N.Y.U.Rev.L. & Soc. Change, 549. The theories of bonding also may be misused to determine only which set of parents is optimum or even “better” in some vague social sense, rather than capable of rearing the child without serious harm. See In re Baby M, supra, 109 N.J. at 445, 537 A.2d 1227 (1988) (“mere fact that a child would be better off with one set of parents than with another is an insufficient basis for terminating the natural parent’s rights.”).
Further, to keep termination proceedings based on bonding theory focused on whether the children have a reasonable opportunity for stable and continuous development may be difficult. Termination of parental rights does not always result in permanent placement of the child. Use and Abuse of the Power to Sever Family Bonds, supra, 12 N.Y.U.Rev.L. & Soc. Change, at 559, 567 (quoting Borgman, Antecedents and Consequences of Parental Rights Termination for Abused and Neglected Children, 60 Child Welfare 391, 392, 402 (1981)). Much of the literature on foster children makes clear that too many children “freed up” for adoption do not in the end find permanent homes. See Borgman, supra, 60 Child Welfare at 402.
The evolving nature of this complex field of knowledge and the limitless circumstances surrounding its application prevent [22]*22courts from easily selecting one principal theory or approach over the other or from placing greater reliance on the proponents of adoption over the advocates of natural parents. A particular theory suffices if it has substantial acceptance within the community of experts. The scientific theory is not unsound simply because it is novel or controversial. See State v. Kelly, 97 N.J. 178, 478 A.2d 364 (1984). The credibility and reliability of witnesses, including institutional witnesses, invites critical awareness, not suspicion. The court has a responsibility, albeit difficult, to make sense of the competing views presented by the experts and to sift the attitudes of all witnesses. Because the welfare of a child is the central concern, it is important that courts thoroughly inform themselves of the subject matter. See Peggy Davis, “There is a Book Out____,” An Analysis of Judicial Absorption of Legislative Fact, 100 Harv.L.Rev. 1539 (1987) (emphasizing need for trial courts to be aware of “competing expert perspectives” on psychological parenting theory). A court must assure a complete and balanced presentation of all relevant and material evidence sufficient to enable it to make a sound determination consistent with the child’s best interests.
Because psychologists and psychiatrists play a critical role in reaching an ultimate decision in termination cases both sides should be able to present expert witnesses. See Mathews v. Eldridge, 424 U.S. 319, 334, 96 S.Ct. 893, 902, 47 L.Ed.2d 18, 33 (1976) (process due participants depends on the nature of interest at stake and decision making process at issue); see also Alan B. Handler, The Judicial Pursuit of Knowledge: Truth and/or Justice, 41 Rutgers L.Rev. 1, 17 (1988) (“parties who do not have the resources to obtain expert testimony are at a distinct disadvantage when social science theories assume a central role in the judicial process.”). Moreover, the court should not hesitate to call on independent experts, as was done in this case by the trial court. Id. at 21 (cases in which courts may have “an enhanced role in the production and use of scientific evidence” include those “where critically relevant [23]*23scientific knowledge is controversial and experts are polarized”). The Appellate Division was correct to remand after the first trial so that A.C.’s attorney would have an opportunity to cross examine Ms. Johnson and to enable the trial court to supplement the record with the testimony of additional witnesses.
The main point in weighing expert evidence is the fit between the expert opinion based on scientific theory and professional experience and the facts of the case. The psychologist with actual experience treating or counseling children in foster care can provide invaluable aid to the court in helping to predict the reaction of a child to a transfer. The other major source of guidance for the trial court is the testimony of agency social workers and case managers. Such persons have a direct and ongoing knowledge of the parties involved. While their testimony may not be sufficient alone to justify the conclusion that removing a child from a foster home would cause the child serious harm, it nonetheless can provide needed subtext for the opinions of the experts.
The tangles and snares that surround bonding theory are evident in this case. Important testimony was presented by DYFS employees and valuable information by the guardian. They furnished probative evidence establishing empirically the affection and strength of the relationship of the children with their respective foster parents. However, they were not qualified to express opinions concerning psychological bonding and the harmful consequences to the children from its disruption. The written reports of the psychologists in the case were also useful, but were conclusory and expressed determinations without supporting explanations. Further, they were not incorporated into the opinions of any of the experts so they could be assessed, albeit indirectly, in the context of evaluating the opinions of the testifying experts.
In many respects the expert testimony was not sufficiently directed to the central issue of bonding itself. Dr. Page, a [24]*24qualified expert who actually counseled J.C., contributed material testimony with respect to J.C.’s personality and psychological and emotional state. She also presented the theoretical basis for bonding theory. However, she acknowledged that she did not conduct a bonding evaluation. Nor did she indicate that she evaluated, accepted or used the testimony of other witnesses as a basis for determining that bonding occurred.
The primary support for the trial court’s decision came from Ms. Regina Johnson, whose report was initially requested by the trial court. Aside from the expert offered by A.C., Johnson was the only expert who testified to the children’s relationships with their foster parents. However, her testimony revealed that she had little or no formal training in conducting bonding evaluations or comprehensive knowledge of the relevant scientific literature. Nor did she have an opportunity to evaluate A.C. or her relationship with the children. Her conclusion that bonding had occurred and that harm would result if those bonds were severed lacked the support and cogency that should surround an expert’s opinion in this kind of case.
A.C.’s expert, Dr. Johnson was qualified. His testimony in part supported the conclusion that a strong bond existed between the children and their respective adoptive parents. That evidence was available to the court to support a termination decision, even though it was proffered on behalf of the natural mother. The court, we emphasize, need not be limited to choosing the opinion of one expert over another. Nevertheless, we are unable to say here that Dr. Johnson’s conclusion that there was a strong relationship between the children and their adoptive parents demonstrated that serious harm would ensue if the children were returned to their mother. That is particularly so in light of the expert’s own inability to reach a firm conclusion with respect to custody.
The issue of custody is intertwined with that of termination. That issue explains another factor that weakens the evidence in support of the likelihood of serious harm attendant to a change [25]*25of custody. Although Dr. Page and Ms. Johnson testified broadly about the value of stability and permanency for children in foster care, neither considered the specific relationships that A.C. has with her two daughters. Dr. Johnson, on the other hand, stressing the flexibility and resilience of children, did consider their specific relationships. In that regard, the case of J.C. presents perhaps the most difficult problem. The experts of both parties indicated that J.C. had serious emotional problems and that her special needs for stability suggest that she is particularly at risk from being removed from her foster home. Dr. Johnson acknowledged that A.C. might not be able to successfully care for J.C. given her special needs. On the other hand there was also substantial evidence suggesting that J.C. has a significant relationship with her natural mother. Weighing the potential harm that terminating J.C.’s relationship with her mother against that which might come from removing her from her foster home is painfully difficult, but it is a decision that necessarily requires expert inquiry specifically directed to the strength of each relationship.
We thus conclude that there is not clear and convincing evidence to support the determination to terminate A.C.’s parental rights. That does not mean, however, that termination may not be an appropriate resolution. DYFS has presented substantial evidence of the harm that may come to these children if separated from their foster parents. Hence, that evidence may not be disregarded, even though, as the record now stands, it does not meet the strict statutory and constitutional standards that govern the termination of parental rights.
Consequently, we remand to the trial court in order that additional evidence may be adduced directly addressing whether the two children have bonded with their foster parents and if so whether breaking such bonds would cause the children serious psychological or emotional harm. To determine these questions the trial court should conduct hearings and direct any of the witnesses to undertake any necessary investigations and examinations. Further, the parties and the court should give [26]*26due regard to the relationships between A.C. and the two children. A.C. should be allowed visitation with the children sufficient to enable the experts to consider those relationships. Because the children have an essential and overriding interest in stability and permanency, it is inimical to their welfare that their legal status remain unresolved. The hearing on remand should be undertaken and concluded as expeditiously as possible.
Finally, we recognize that A.C.’s parental rights may ultimately be terminated even though her contact with the children, in contrast to her custody over them, exposes them to no harm. The risk to children stemming from the deprivation of the custody of their natural parent is one that inheres in the termination of parental rights and is based on the paramount need the children have for permanent and defined parent-child relationships. See N.J.S.A. 9:3-50(b) (“the entry of judgment of adoption shall establish the same relationships, rights, duties and obligations between the child and the adopting parent as if such child were born of such adopting parent in lawful wedlock”). In this case there is expert testimony indicating a significant relationship between A.C. and J.C. and a potential for harm to J.C. if contact does not continue. This evidence raises the question of whether the welfare of children under certain circumstances reasonably requires continued contact with natural parents subsequent to guardianship being granted to DYFS or an adoption. See In re Guardianship of R.O.M.C., 243 N.J.Super. 631, 633-34, 581 A.2d 113 (App.Div.1990); Kattermann v. DiPiazza, 151 N.J.Super. 209, 376 A.2d 955 (App.Div.1977); Michaud v. Wawruck, 209 Conn. 407, 551 A.2d 738 (1988). We note that this question may arise at some future time but do not address it here.
V
We reverse the judgment of the court below terminating the parental rights of A.C. and remand the matter for further [27]*27proceedings consistent with this opinion. We do not retain jurisdiction.