dissenting.
I respectfully dissent. In my opinion, the trial court misunderstood the two-stage deci-sional process for changing a child’s custody, mistakenly applied the law of behavioral conditions for custody, and failed to make the essential finding of necessity to change custody. On the entire evidence, I am firmly convinced that the trial court was mistaken in changing custody.
Although the opinion by Justice Levine recognizes the two-stage analysis for modifying an existing custody placement, which differs from an original custody decision, the trial court did not. The trial court expressly found “that there has been a significant change of circumstances since the 1991 Amended Judgment,” but the court did not say what the change was. The trial court simply “finds that the significant change of circumstances is such that a change in physical custody will serve the best interests of the child.” That is the wrong standard for changing custody.
The trial court failed to identify the crucial concern in changing custody. Justice Levine stressed this aspect in her concurrence in Ludwig I:
[T]o change custody, there must be a significant change of circumstances that requires the change.... In order to require or necessitate a custody transfer, the significant change of circumstances must weigh against the best interests of the child....
[T]he language used in some of our cases and cited by the majority, that we analyze whether a change in custody will “serve” or “foster” the best interests of the child, is misleading and inaccurate.... Whether the change of circumstances requires that there be a change of custody must be answered in the context of the recognized overriding benefit in maintaining the stability and continuity of the custodial parent-child relationship and not by looking at the particular change and whether it will “foster” the best interests of the child.
Ludwig v. Burchill (Ludwig I), 481 N.W.2d 464, 469-70 (N.D.1992) (Levine, Justice, concurring) (last three emphases added). Since only this writer signed that additional opinion by Justice Levine, I believe that the trial court misunderstood it, and was misled into using only the “best interests” analysis inaccurately recited by the majority opinion in Ludwig I.
The two-stage analysis is not a recent refinement in child custody doctrine. See Miller v. Miller, 305 N.W.2d 666 (N.D.1981); Orke v. Olson, 411 N.W.2d 97 (N.D.1987). In addition, the current court has unanimously endorsed the view that “the second decisional stage [for change of custody] is whether the change in circumstances so adversely affects the child that it compels or requires a change in custody in order to foster the child’s best interests.” Hagel v. Hagel, 512 N.W.2d 465, 467 (N.D.1994), citing Johnson v. Schlotman, 502 N.W.2d 831, 834 (N.D.1993) and other precedents. In my opinion, this change of custody should be reversed to reinforce the doctrinal teaching on changing custody, as well as to correct the mistaken application of law here.
Justice Levine’s opinion acknowledges that “[t]he trial court did not make an express finding that the change of circumstances in this case adversely affected the child.” Justice Levine would excuse this omission by implying that the trial court determined that the child was so adversely affected that change of custody was necessary. However, to give a large measure of finality to an initial custody placement, and thus to preserve a child’s stable relationship with a fit custodial parent, we require an express finding of the necessity for the change. Anderson v. Anderson, 448 N.W.2d 181, 182 (N.D.1989). In Anderson, the trial court removed an eight year old child from her mother, who had “a longstanding problem of alcohol abuse” and who left the child with the non-custodial father during one period of *679treatment at Heartview Foundation. We reversed the transfer of custody because the trial court “focused solely upon a determination of whether it would be in [the child’s] best interest to reside with her father or mother, without first considering whether there had been a significant change of circumstances since the original decree to [necessitate] a change of custody.” Id.
In her opinion, Justice Levine does not describe how Justin was “adversely affected,” nor did the trial court. Rather, Justice Levine predicts, from the generalized idea that “alcohol dependence ... has a negative effect on parenting skills,” that Justin might be affected by “the further-reaching effect ... of growing up with an alcohol-dependent parent.” However, this record does not reflect any existing adverse effect on Justin during his four years in Burchill’s custody. Indeed, the trial court found:
Though still having some behavior problems in school, Justin seems to have adjusted fairly well to being the child of a broken home.
And:
Ms. Morris [Justin’s teacher] believed that Justin now fits into the “normal” category of first graders.
There was no finding of harm to Justin.
Rather than focusing on how Justin fared with his father for four years, the trial court faulted Burchill’s lifestyle itself. The court concluded Burchill “has more negatives this time around,” although adding that it “still wants the parties to know that it thinks they are still basically good people.” Those conclusions came after findings faulting Burchill for his second DUI and following license suspension, while recognizing that transportation had been adequately provided for Justin by Burchill’s parents (who lived with them temporarily while budding a new home nearby), and that Burchill’s license would be reinstated soon. The court attacked Burc-hill’s choice “to remain on the night shift [for his job at WalMart] as he thinks it provides more job security,” since then “Justin sleeps at his grandparent’s house” nearby. The court condemned Burchill’s job choice, saying his “priority is shift preference over a less-nomadic lifestyle for his son,” without recognition that the night job allowed Burchill the most possible daytime hours with his son. Thus, the trial court faulted Burchill’s work ethic and his closeness to his extended family, even though those qualities are usually considered exemplary in a single parent’s life. See also NDCC 14-09-06.1 (grandpa-rental rights). The trial court reasoned:
When one considers [Burchill’s] work schedule, Justin’s school schedule, the fact the grandparents provide transportation, etc., the focus on whom is providing the environment is on the grandparents. The fact they have a new home is commendable. The fact that they are the primary care givers of Justin is also commendable. However, this custody contest is not between the grandparents and [Ludwig]. It is between the parties and it seems that [Burchill] has transferred from being the lead actor to a supporting role.
Missing from this reasoning is evaluation of the evidence about how Justin has gotten along in the four years with his father.
The trial court stressed that, although Burchill had been ordered in the 1991 decree denying change of custody to “attend Alcoholics Anonymous at least twice per month and [to] secure a sponsor,” Burchill had quit attending A.A. in the fall of 1991. The trial court reasoned:
He based his disobedience of the Court’s Judgment on the fact that an addiction counselor, Kerry Wicks, had told him he need not attend A.A. [Burchill] has failed to cite any authority which provides that a counselor may act as an appellate court and review the order of a trial court. This Court is not impressed.
Omitted is the fact that custody was not explicitly, but only implicitly, conditioned on this order. The court went on:
[Burchill] opted to continue to drink alcohol and risk losing custody of his son. Not a wise choice.
The psychologist, Dr. Neil Clark, testified that a second DUI conviction is a “hard indicator” of alcohol dependence. He believed that such dependency had a negative effect on parenting skills such as driving with the child and the known difficulty *680that children have growing up in alcohol dependent homes.
The court did not find that Justin was having difficulty at home.
The court also compared Burchill’s conduct with Ludwig’s, who also testified that she drank, usually “between three and four” beers, “[o]nce, twice a month.”
A change on behalf of [Ludwig] is that she and her husband have discontinued hosting any alcoholic parties that the Court was concerned with in 1991.
In 1991, [Ludwig] had a slight edge when it came to these two factors. That edge has widened into a pivotal lead.
Missing, again, is any evaluation of how Justin fared in his father’s custody for four years.
Justice Levine relies on Gravning v. Gravning, 389 N.W.2d 621, 623-24 (N.D.1986), to justify the trial court’s change of custody for breach of behavioral conditions. But Gravning was an original custody placement, not a change in custody. In Gravning, both parents appealed and complained about the separate custody of each of two children being “expressly conditioned” on certain behavior by each parent. The express conditions required the mother to get “formal education and/or gainful employment,” and required the father to abstain from “controlled substances including alcohol,” get gainful employment, and continue to reside with his parents. Gravning, 389 N.W.2d at 622. While we affirmed, we cautioned:
[T]he critical concern must be whether the effect of conditioning custody risks making the child the innocent victim of a parent’s misbehavior. Behavioral conditions directed to the conduct of the parents, like those used here, carry such a risk, particularly where the failure to live up to the condition does not clearly and directly affect the child’s welfare. Whether that risk is acceptable, when weighed with the utility of promoting compliance with the custodial decree, can probably be determined better with experience. Therefore, we leave full resolution of that issue to another day, while making several observations about the subject.
Gravning, 389 N.W.2d at 623. In our observations, we stressed:
[I]t may be better to consider the advantages and drawbacks of a particular condition when, and if, a trial court has determined that a breach of the expressed condition has taken place and has assessed the impact of the breach on the child involved.
Id. at 624. An assessment of the impact of Burchill’s breach of the order on the child is missing here, both in the evidence and in the findings. In spite of the Gravning concerns, the trial court’s findings and Justice Levine’s opinion have not “assessed the impact of [Burchill’s] breach on the child involved,” Justin, nor have they analyzed whether “the failure to live up to the condition [has] clearly and directly affect[ed] the child’s welfare.”
Both the trial court and Justice Levine express fear about the future of a child growing up with an alcohol-dependent person, but neither identify any present impact on Justin. This might be a satisfactory analysis for initial placement of custody, but it is an unsatisfactory analysis for a change of custody based on a breach of condition, where it is essential to identify the present adverse effects on the child that would compel the change.
The evidence does not include any expert opinion about an existing effect on Justin. James Monson, a chemical dependency counselor for Ludwig, diagnosed Burchill with “alcohol dependence mild,” but did not evaluate Justin. Monson testified: “I’m not recommending treatment right now because his drinking pattern isn’t serious enough to warrant a full outpatient or inpatient program.” When asked, “[d]oes [being an alcoholic] affect your relationship with your children?”, Monson answered: “Well, it certainly can, but I believe my testimony is' just to my evaluation, isn’t it? ... I’m not real comfortable answering that question. My role here is to substantiate my diagnosis, not to make a determination as to his fitness as a parent.”
Dr. Neil Clark, a licensed psychologist for Ludwig, also evaluated only Burchill. Dr. Clark used Monson’s chemical dependency assessment because “he has a lot more depth and breadth of experience in that particular problem presentation.” Dr. Clark testified *681that BurehiU “does report ... continuing to drink periodicaUy, I believe two to three times a month, in which he wül have two to three beers. And that continuing pattern of drinking could result in further problems, especiaUy in the DUI context.” Dr. Clark opined that aleohoUsm
certainly can affect one’s parenting skiUs. You know, the first concern is, of course, whether or not there’s any transportation of the chüd while drinking which is a safety concern. Other impacts depend upon the severity of the degree of alcohol abuse or dependence.
This is a generalization, not an evaluation of the effect on Justin. In Weber v. Weber, 512 N.W.2d 723, 727 (N.D.1994), we recently rejected similar generalizations as inappropriate criteria for deciding chüd custody. “[I]n the absence of a complete study of all of the parties, there is logical frailty in applying a general premise to a specific case.”
Kerry Wicks, a Ucensed addiction counsel- or for BurehiU, testified that his evaluation of BurchUl “didn’t find enough evidence to warrant an addiction diagnosis.” Wicks added:
I didn’t find enough evidence to warrant an addiction diagnosis. I did comment that his drinking and driving is obviously a use of poor judgment and that his drinking is starting to cause him some problems and I recommended that he attend 20 hours of alcohol education, but there was no — no actual DSM-III-R diagnosis.
Whfie the trial court faulted BurehiU for fafiing to attend A.A., the court did not find that BurehiU was an alcohohc, nor could it have done so on this record.
More importantly, the trial court could not fairly find from the evidence that Justin had been or was being harmed by BurchiU’s use of alcohol. Even Justin’s mother did not testify that Justin was not doing weU with BurchUl. Her strongest criticism was that “[h]e’s being raised by his grandparents,” without identifying any particular shortcomings. The trial court was indignant about BurchiU’s behavior, but faüed to determine whether his behavior adversely affected Justin. BurehiU is not a perfect parent; few are. But he is a fit parent who has done a reasonable job of raising Justin.
In Hagel, this court unanimously adopted Justice Levine’s formulation that, in a modification proceeding, the “chüd is presumed to be better off with the custodial parent.” Hagel, 512 at 468, citing Delzer v. Winn, 491 N.W.2d 741, 747 (N.D.1992) (Levine, specially concurring). Here, Justice Levine rebuts this presumption with impheations that are based on BurchiU’s behavior instead of any adverse effect his behavior has had on Justin.
Because I beheve that the trial court misunderstood the law for change of custody, misappUed the law of behavioral conditions on custody, and mistakenly changed custody of Justin from his fit father, I respectfully dissent.