Bruce Krank appeals from an amended district court judgment granting joint legal custody and sole physical custody of his son, Andrew Krank, to Andrew’s mother, Annette Jones. Jones cross-appeals, claiming the district court’s child support determination was clearly erroneous. We affirm the trial court’s child custody decision and dismiss the cross-appeal.
The facts underlying this appeal are reported in Krank v. Krank, 529 N.W.2d 844 (N.D.1995), and will be repeated only as necessary to resolve the issues raised in this appeal.
In March 1994, the trial court awarded joint legal custody of Andrew to Krank and Jones, with Krank having physical custody. Jones appealed. Both parties alleged domestic violence, and we reversed and remanded because the trial court had failed to “make detailed findings on the domestic violence issue.” Krank, 529 N.W.2d at 850. We stated that, in eases of reciprocal domestic violence, section 14-09-06.2, NDCC, requires the trial court to “measure the amount and extent of domestic violence inflicted by both parents.” Id. We instructed the trial court that:
“If the amount and extent of domestic violence inflicted by one parent is significantly greater than that inflicted by the other, the statutory presumption against awarding custody to the perpetrator will apply only to the parent who has inflicted the greater domestic violence, and will not apply to the parent who has inflicted the lesser. However, if the trial court finds that the amount and extent of the violence inflicted by one parent is roughly proportional to the violence inflicted by the other parent, and both parents are otherwise found to be fit parents, the presumption against awarding custody to either perpetrating parent ceases to exist.” Id.
*716On remand, the trial court issued a memorandum decision in which it stated it could not “in good faith find that the domestic violence was roughly equal.” The trial court found Krank had perpetrated “disproportional violence” and found the statutory presumption against awarding custody to a perpetrator of domestic violence to operate against Krank. The trial court ordered entry of an amended judgment awarding physical custody of Andrew to Jones. Krank appeals from the amended judgment.
Krank claims the trial court erred in granting physical custody of Andrew to Jones. Child custody is a finding of fact that we will not disturb on appeal unless clearly erroneous. Rule 52(a), NDRCivP; Helbling v. Helbling, 532 N.W.2d 650 (N.D.1995). A finding of fact is clearly erroneous if no evidence exists to support it, if an erroneous view of the law induced it, or if we are left with a definite and firm conviction that the trial court has made a mistake. Ryan v. Flemming, 533 N.W.2d 920 (N.D.1995).
Krank first argues that the trial court’s findings were insufficient to support its award of custody to Jones. Krank faults the trial court for failing to make separate findings on each domestic violence allegation raised by the parties and for failing to decide whether a 1988 shooting incident involving Krank and Jones was “intentional or accidental, unjustified or justified.”
In Krank, 529 N.W.2d at 850, we did not instruct the trial court to make findings on each allegation, but on the “domestic violence issue” as a whole. When domestic violence is a factor in a custody decision, the court “shall cite specific findings of fact.” NDCC § 14-09-06.2(1)©. Rule 52(a), NDRCivP, also requires trial courts to “find the facts specially.” Findings are especially important in cases involving domestic violence, because the presence of findings on the issue shows that the trial court did not ignore the issue in reaching its conclusion. Helbling, 532 N.W.2d at 653. If we can understand the factual basis of the trial court’s decision from its findings, the findings are adequately specific. Severson v. Hansen, 529 N.W.2d 167 (N.D.1995); Gross v. Sta-Rite Industries, Inc., 322 N.W.2d 679 (N.D. 1982). Here, the factual basis of the trial court’s decision is clear. Its findings show that it weighed the evidence before it and found that the domestic violence inflicted by both parties was not equal. The trial court did not err in failing to make individual findings on each allegation of violence.
We did express concern about the trial court’s original findings on the shooting incident. Krank, 529 N.W.2d at 850. We stated:
“We do not intend to suggest that a slap in the face is roughly proportional to rapes and physical beatings, if, in fact, the trial court finds that those incidents did occur. But an unjustified shooting, if that is what the trial court finds transpired here, might reasonably be considered to be roughly proportional to those acts of domestic violence.” Id.
We therefore instructed the trial court to make “factual determinations” regarding the “amount and extent” of domestic violence inflicted by each party. Id. In its April 11, 1995, Memorandum Decision, Findings and Order for Amended Judgment, the trial court specifically found that Krank inflicted “disproportional” violence on Jones. In its May 3, 1995, supplemental memorandum decision, the trial court stated:
“The guardian ad litem made specific findings that it was virtually impossible to ascertain how the shooting took place. The Court cannot state a specific finding on this either. The Court finds in effect that [Krank] initiated the violence that evening by forcibly stopping [Jones], when there was a protection order in effect, and confronting her. She stated that he made comments about ‘fucking her one more time.’ [Krank] during the relationship showed a domineering and bullying personality. [Krank] would find [Jones] and would threaten and intimidate her whether it was when she was cleaning up the bar after work or when she was with another gentleman.”
The trial court’s findings show that it evaluated the “amount and extent” of domestic violence inflicted by the parties, and that it could not conclude the shooting was “unjusti*717fied.” The trial court’s findings may not have been extensive, but the trial court clearly took note of our concerns and addressed them.
Given the evidence in the record, it is doubtful whether any court could satisfy Krank’s demand for findings on whether the shooting was “intentional or accidental, unjustified or justified.” The state brought no charges, and the case never went to trial, even though the police conducted a thorough investigation at the time of the shooting. The truth about the shooting therefore remains a mystery, and we cannot fault the trial court for failing to solve it seven years after the event took place.
Krank next argues that the trial court’s findings of fact on the quantum of domestic violence were clearly erroneous. He argues, in effect, that the evidence of his acts of domestic violence was not credible because it was uncorroborated.
Evidence need not be corroborated to be credible. See, e.g., State v. Schill, 406 N.W.2d 660 (N.D.1987) [A child’s uncorroborated testimony can support a conviction for child sexual abuse if it is found to be credible]; State v. McLain, 312 N.W.2d 343, 347 (N.D.1981) [“[T]he uncorroborated testimony of a victim is sufficient to establish all the elements of the crime of rape.”]. Under the clearly erroneous standard, we defer to the trial court on matters of credibility because it has had the better opportunity to assess the credibility and demeanor of witnesses. Severson, 529 N.W.2d at 169. The trial court “is in a much better position to ascertain the true facts than an appellate court, which must rely upon a cold record.” Schmidkunz v. Schmidkunz, 529 N.W.2d 857, 859 (N.D.1995).
Here, there is evidence of domestic violence on both sides. Jones admitted that she slapped Krank and threw a glass of tea at him. She claims she shot him accidentally. There is evidence that Krank harassed Jones, shoved her, hit her, threw her, pulled her hair, and threw glasses at her. There is evidence that Krank continued to harass Jones after she obtained an adult abuse protection order against him. The shooting occurred while this order was in force, and there is evidence that the shooting came at the end of an evening of abuse and harassment by Krank. Given the evidence in the record, we cannot say that the trial court erred by crediting Jones’s assertions and concluding “there was disproportional violence on [Krank’s] part.”
Krank’s last argument is that the trial court should not have applied the statutory presumption against awarding custody to a perpetrator of domestic violence against him because of the time that has elapsed since the alleged abuse took place. The parties agree that the last incidents of abuse involving them occurred in 1988 on the night of the shooting. Krank argues that there is substantial evidence that he “rehabilitated his life” in the five years between the night of the shooting and the custody hearing.
We stated in Krank, 529 N.W.2d at 849, that lapse of time alone is not enough to overcome the statutory presumption against awarding custody to a perpetrator of domestic abuse. This is because domestic violence is a “learned behavioral pattern.” Id.; see also Mary Ann Dutton & Catherine L. Waltz, Domestic Violence: Understanding Why It Happens and How to Recognize It, Family Advocate, Winter 1995, at 14 (hereinafter Dutton) [reporting a 40 percent recidivism rate among batterers five years after treatment]. We suggested, based on our analysis in Heck v. Reed, 529 N.W.2d 155 (N.D.1995), that Krank may have been able to show his rehabilitation by evidence of participation in a treatment program or counseling for domestic violence. Krank, 529 N.W.2d at 849. We observed in Krank that Krank introduced no evidence showing he had sought treatment. Id. Nor did Krank introduce any such evidence on remand. Krank has not yet reached the threshold level we have established for showing rehabilitation over time; there is no evidence that he has taken any positive steps to become a non-abuser.
Krank’s rehabilitation argument rings especially hollow given the facts of this case. It is true that there have been no incidents of abuse between Krank and Jones since 1988. A major reason for this may be the fact that *718Jones and Krank have not lived with each other or in the same town since shortly after the 1988 shooting incident. They currently live separated by approximately 350 miles, and their only contact has been during brief visits.
From what we know of domestic violence, the distance separating Krank and Jones may be a more important factor in the cessation of violence in their relationship than the “rehabilitation” of one of the parties. It is common for domestic abuse to occur as a stage in a cycle of violence. See State v. Leidholm, 334 N.W.2d 811 (N.D.1983); Laura Crites and Donna Coker, What Therapists See That Judges May Miss, The Judges Journal, Spring 1988, at 9 (hereinafter Crites). Typically, the cycle includes a tension building stage, followed by an explosion, followed by a period of loving and caring. Crites, supra at 10. Once the cycle is complete, it begins again. Id. Separation can interrupt the cycle of abuse, and has long been society’s recommended solution to domestic violence.1 See Martha R. Mahoney, Legal Images of Battered Women: Redefining the Issue of Separation, 90 Mich.L.Rev. 1 (1991) (hereinafter Mahoney); Dutton, supra at 17. Separation, however, does not change the psychological characteristics of the parties. See Mahoney, supra at 57-60; Dutton, supra at 17; Crites, supra at 10. The cycle of abuse can begin again, unless “a strong intervening force” is applied. Crites, supra at 10. Action by the battered person alone is unlikely to stop the cycle: “once the violence starts, only the batterer can stop it.” Dut-ton, supra at 17.
The trial court found that Krank displayed a domineering personality throughout his relationship with Jones. The record shows that Krank engaged in a pattern of bullying Jones. Viewed in the context of their relationship as a whole, the fact that domestic abuse has not occurred lately between Krank and Jones may show no more than an interruption in the cycle of abuse caused by the parties’ long separation. Given the evidence of a pattern of abusive behavior on Krank’s part throughout his relationship with Jones, and given the fact that Krank has not sought treatment for his abusive tendencies, the trial court did not err in finding that the statutory presumption against awarding custody operates against Krank despite the passage of time since the last incident of abuse between the parties.
In her cross-appeal, Jones argues that the trial court’s determination of the child support amount was clearly erroneous. In its amended judgment, the trial court awarded support of $266 per month based on a monthly net income of $1,100. Jones claims that both the child support amount and the net income determination are incorrect.
In its Order Setting Child Support and Related Issues, the trial court stated that Krank needed to provide a complete 1994 tax form and information about his “medical insurance situation.” The trial court did not wait to receive this information before setting the support amount, noting that Krank was anxious to have the amended judgment entered so that he could appeal. The trial court acknowledged that it set the support amount based on “incomplete information” and it indicated that “the Court may modify this amount back to June 1st if more complete financial information so dictates.” The trial court restated this condition in the Amended Judgment. This court “will not entertain appeals from orders from which we cannot affirmatively establish were meant to be, in all aspects, final.” Sime v. Tvenge Assoc. Architects, 488 N.W.2d 606, 608 n. 1 (N.D.1992). Here, the trial court asked Krank to produce additional financial information and indicated that it would modify its order, if necessary, on receipt of the requested information. Because the trial court anticipates further proceedings on the child support question, it is clear that its child support order is not final. Therefore, it is not appealable at this point.
*719Accordingly, we affirm the trial court’s child custody decision and dismiss the cross-appeal.
VANDE WALLE, C.J., and MESCHKE, JJ., concur.. Separation, however, is not a cure-all for domestic violence. "In the most extreme cases, battered women are more likely to be killed after the separation.” Dutton, supra at 17. The personality of the batterer dictates whether separation will end or intensify abuse. Mahoney, supra at 58.