Bruce W. Livingston appeals from an amended judgment awarding custody of the parties’ children to his former wife, Angela M. Schneider. We affirm.
After over eleven years of marriage, Bruce and Angela divorced on February 22, 1992. By stipulation, they agreed to joint legal custody and shared physical custody of their three children: Nathan, age 10, Jacob, age 9, and Samuel, age 5. The shared physical custody consisted of each party having the children for three and one-half days per week, eventually evolving into Angela having the children from Wednesday morning until Saturday morning and Bruce having the children from Saturday morning until Wednesday morning.
In March 1994, Angela moved for an increase in child support. Bruce brought a counter motion, seeking sole physical custody of the children. Bruce based his counter motion on, among other things, the childrens’ maladjustment to the shared custody arrangement of continual shuffling back and forth between homes. Angela agreed the custody arrangement was not completely satisfactory. After the motions were heard on September 1, 1994, the trial court awarded custody of the children to Angela with liberal visitation privileges to Bruce.
Bruce disputes the trial court’s custody determination, arguing several of the court’s findings are clearly erroneous because they are not supported by the evidence. He challenges the findings addressing the best interests of the child factors (b), (d), (i), and (k) of N.D.C.C. § 14-09-06.2(1). Bruce also argues the trial court failed to consider properly the guardian ad litem’s testimony and recommendation on child custody.
This court treats a trial court’s child custody determinations as findings of fact subject to the clearly erroneous standard of review. Rule 52(a), N.D.R.Civ.P.; Klose v. Klose, 524 N.W.2d 94, 96 (N.D.1994). A finding of fact is clearly erroneous if it is induced by an erroneous view of the law, if no evidence exists to support it, or if the reviewing court, on the entire evidence, has a definite and firm conviction that the trial court made a mistake. Id.
The trial court found a significant change in circumstances had occurred since Bruce and Angela’s stipulated agreement. The court also stated Bruce and Angela essentially agreed a change in custody was necessary. Citing Dalin v. Dalin, 512 N.W.2d 685 (N.D.1994), the court made its custody determination, treating the proceedings as an original custody matter and applying the best interests of the child factors of section 14-09-06.2. We limit our review to the trial court’s determination of the children’s best interests.
Subsection (b) of section 14-09-06.2(1) requires a trial court to consider and evaluate “[t]he capacity and disposition of the parents to give the child love, affection, and guidance and to continue the education of the child.” N.D.C.C. § 14-09-06.2(l)(b). The trial court found this factor in favor of Angela because of her role as the children’s primary caretaker and her educational background. Bruce disputes the finding.
Although the primary caretaker rule has never gained a presumptive status in North Dakota, it is a factor properly considered by the trial court. E.g., Foreng v. Foreng, 509 N.W.2d 38, 40 (N.D.1993). “The primary caretaker is the parent who provides the child with daily nurturanee, care and support.” Gravning v. Gravning, 389 N.W.2d 621, 624 n. 1 (N.D.1986) (Levine, J., dissenting). Primary caretaker tasks include, among others, feeding the child, bath*231ing, grooming and dressing the child, attending to and laundering the child’s clothing, arranging the child’s medical care, disciplining the child, and educating and teaching the child elementary skills. Id.
Bruce and Angela shared essentially equal physical custody, with each party having the children for three and one-half days per week. Given this arrangement, the trial court’s determination that Angela was the primary caretaker intuitively seems mistaken. But see Dalin, 512 N.W.2d at 688 (finding neither parent primary caretaker when both parents substantially shared custody). A review of the record, however, reveals evidence supporting the determination.
At the motions hearing, Bruce could not respond definitively when asked the names of his children’s teachers.
“Q. Who was Nathan’s teacher last year?
A. Oh, God — names again — I know her.
Q. Who was Jacob’s teacher?
A. Am I going to be judged for my poor memory?
Q. Well ...
A. Mrs. Aueh was one of them.
Q. Who?
A. I think it was Mrs. Aueh, and Mrs. Landsiedel was Nathan’s.
Q. The reason I asked is because you named the wrong teacher in your deposition.”
Brace did indicate he helps his children with their homework. The guardian ad litem testified Angela took the lead in handling the children’s educational needs. For example, Angela arranged a tutor for Jacob, who was having difficulty in school, and also took him to the library and spent time helping him read. She also spends considerable time helping the children with their daily homework and other school projects. Brace testified he was aware of Angela’s efforts.
Bruce also testified he had taken his children to the doctor twice that he could recall. His children are ages 10, 9, and 5. Otherwise, Bruce sometimes attended doctor visits with Angela.
Angela testified, and Bruce did not dispute, that she made a list for Brace every week when he picked up the children detailing the children’s upcoming schedule. She also indicated, until recently, she was laundering all of the children’s clothing and packing their suitcases for their three and one-half days with Bruce. To eliminate the children’s living out of a suitcase, Angela organized two wardrobes, one for each parents’ home. Brace agreed this was the arrangement, but he also indicated he now does laundry since the children have a wardrobe at his home. Both parents cook and clean for their children while in each parent’s custody.
Evidence concerning parental roles during Bruce and Angela’s marriage also was received by the trial court. Undisputed testimony from Angela’s family indicated Angela was primarily responsible for the care of the children. Bruce indicated he was primarily responsible for the family’s income.
The evidence supporting the trial court’s finding about Angela’s greater capacity to continue the children’s education consisted of Angela’s educational background, which included a degree in elementary education, and her experience as a substitute teacher in the school system. Bruce indicated he had completed two years of college.
The foregoing is a sampling of the evidence before the trial court concerning parental roles and educational capacity. The evidence suggests, although Bruce and Angela shared essentially equal physical custody, they did not share equally in caretaking tasks, such as contributing to and fostering the children’s education. We cannot say the trial court was mistaken when it found Angela was the primary caretaker and had the greater capacity to continue the children’s education. The trial court having found such relevant evidence, we cannot conclude the trial court clearly erred when it found subsection (b) favorable to Angela.
Brace contends the trial court overlooked evidence concerning subsection (d), the stability and continuity factor. He claims the children spent, in quantitative terms, more time with him, apparently attempting to equate greater quantity with greater stability. The trial court found the shared physical *232custody arrangement greatly impeded both Bruce and Angela’s ability to provide a stable lifestyle for their children. Elaborating further in its memorandum and order, the trial court stated:
“The boys have been ‘living out of a suitcase’ for nearly three (3) years now, and the end result is that they have not been in either home long enough to develop a sense of belonging, a feeling of security or the assurance that there will always be stability in their lives — all of which are so crucial to the children’s normal growth and development.”
“This Court has often stressed the importance of continuity and stability in a child’s life,” Mansukhani v. Palling, 318 N.W.2d 748, 753 (N.D.1982), but “more than a mathematical computation of the days spent in each parent’s” custody is involved in determining continuity and stability. See Dalin, 512 N.W.2d at 688. Recognizing “it is difficult, if not impossible, for a child to live a normal, happy life where there is [divided]” custody, Lapp v. Lapp, 293 N.W.2d 121, 130 (N.D.1980) (quoting Annot., 92 A.L.R.2d 695 (1963)), family law professionals have suggested stability may “best be attained with such an undivided custody as will prevent the child from being shunted back and forth between homes.” Id. Given the foregoing, we cannot conclude the trial court clearly erred when it found the stability and continuity factor in favor of neither Bruce nor Angela.
Bruce argues the trial court failed to properly address subsection (i), the reasonable preference of the children. Bruce claims the children, through isolated, early post-divorce behavior and statements, demonstrated a preference to live with him. Other than Bruce’s claim, the trial court received no evidence about the children’s preference because of their young ages, and thus implicitly determined the children were not capable of intelligently choosing between their parents. Barstad v. Barstad, 499 N.W.2d 584, 588 (N.D.1993) (discussing general rules about custody preference of children); See Weber v. Weber, 512 N.W.2d 723, 727 (N.D.1994) (stating trial courts have substantial discretion in custody matters). The trial court addressed the factor, and its finding is not clearly erroneous.
Subsection (k) requires the trial court to consider and evaluate the interaction and interrelationship of the child with a non-parent, such as a grandparent. N.D.C.C. § 14-09-06.2(l)(k). Bruce claims the trial court was controlled by Angela’s parents’ feelings for their grandchildren when it found this factor in favor of Angela. He dismisses the court’s findings about the grandparents’ considerable love and affection for their grandchildren with the phrase “so what,” and asserts the court ignored evidence of the grandfather’s alleged verbal abuse toward the children.
The relationship between the grandparents and the children clearly is an applicable factor for consideration and evaluation by the trial court. The record indicates these grandparents have fairly extensive involvement in the lives of their grandchildren, such as weekly caretaking. Given this evidence, the trial court was required by law to consider this factor. Mertz v. Mertz, 439 N.W.2d 94, 97 (N.D.1989) (stating trial court must consider applicable factors).
The trial court received evidence about the grandfather’s disciplinary style. Bruce described it as “verbal abuse” toward his children; Angela, members of her family and the guardian ad litem, based upon the children’s interviews, described it as “yelling.” In its finding, the trial court did not adopt either expression but did convey mild concern about the grandfather’s disciplinary style. Thus, contrary to Bruce’s assertion, the court did not ignore this negative evidence but used its discretion in how it addressed the evidence. See Catlin v. Catlin, 494 N.W.2d 581, 591 (N.D.1992); see also Freed v. Freed, 454 N.W.2d 516, 519 (N.D.1990). The court’s findings are supported by the evidence, and thus, are not clearly erroneous.
Bruce argues the trial court erred when it failed to consider properly the guardian ad litem’s testimony and recommendation about child custody. The trial court appointed a guardian ad litem for Bruce and Angela’s children. After interviewing Bruce, Angela, and the children, the guardian ad litem pre*233pared a report evaluating the best interests of the children and recommending custody be awarded to Bruce. The guardian ad litem testified her recommendation was based primarily on Bruce’s more appropriate disciplinary style and greater ability to put aside his needs for those of his children.
The trial court is statutorily vested with the responsibility of awarding custody of a child to the parent who will promote the best interests and welfare of the child. N.D.C.C. § 14-09-06.1; McAdams v. McAdams, 530 N.W.2d 647, 650 (N.D.1995). The court cannot delegate this responsibility to a guardian ad litem or other independent investigator. See Owan v. Owan, 541 N.W.2d 719 (N.D.1996). The weight assigned to a guardian ad litem’s testimony and recommendation is within the trial court’s discretion, and the court does not have to, nor should it, regard a guardian ad litem’s testimony and recommendation as conclusive. See McAdams, 530 N.W.2d at 650 (involving custody investigator’s report); see also Owan, 541 N.W.2d at 721 (involving social worker’s testimony).
Bruce’s argument is meritless. Evidence that the trial court considered the guardian ad litem’s testimony and recommendation is found in the court’s memorandum and order in which it made several references to the guardian ad litem’s report. The references included both agreement and disagreement with the guardian ad litem’s evaluation, findings independent of the guardian ad litem, and comments addressing the guardian ad litem’s primary parenting concerns. The trial court exercised its discretion in considering the guardian ad litem’s testimony and recommendation. We are not convinced the court erred simply because it did not agree with the guardian ad litem’s recommended award of custody to Bruce.
We conclude there is evidence to support each of the challenged findings. The trial court’s child custody determination is affirmed.
LEVINE and SANDSTROM, JJ., and RALPH J. ERICKSTAD, Surrogate Judge, concur. RALPH J. ERICKSTAD, Surrogate Judge, sitting in place of MESCHKE, J., disqualified.