¶ 1. Billie Jo Lambert appeals from a judgment awarding primary physical placement of her son, Neven, to the boy's father, David Helling. She claims the trial court erroneously considered her non-marital relationship with a third party as a negative factor absent any showing that the relationship was harmful to the child; that the court was biased against her based on her living arrangements and pregnancy; that the court failed to give adequate consideration to the harm which could result from removing the child from his primary attachment; and that there was insufficient evidence to support the court's conclusion that placement with the father would be in the child's best interest. We conclude that the trial court's opinion of the stability of nonmarital relationships in general, which it stated was based in part upon its view of other paternity cases it had seen, was insufficient to support a factual finding that the mother's specific living situation in this case was unstable. Because the trial court *799identified its finding that the mother's living situation was unstable as one of the "main factors" supporting its decision, we reverse and remand for further proceedings consistent with this opinion.
BACKGROUND
¶ 2. Neven was born on March 1, 2000. Lambert and Helling were living together at that time, and Helling formally acknowledged his paternity of Neven after the child's birth.
¶ 3. Helling and Lambert's relationship deteriorated. Helling moved out of the parties' shared residence in October of 2000. The parties agreed between themselves that Helling would have Neven overnight on Tuesdays, Thursdays and Saturdays each week. That arrangement continued until August of 2001, when Lambert moved in with her new boyfriend, Scott Weber, whom she had been dating for over a year.
¶ 4. In August of 2001, Helling filed a family court action seeking primary physical placement of Neven. The trial court entered a temporary order placing Neven with Helling on Tuesday and Thursday evenings and alternate weekends.
¶ 5. By the time of the hearing, Lambert had been dating Weber for about two years, living with him for three months, and was expecting a child with him. Weber testified that Lambert was not officially on his lease, but was paying him $425 per month in rent, and they were splitting other expenses. Lambert testified that she and Weber had no current marriage plans, but that she hoped to build a permanent long-term relationship with him. While at work, Lambert placed Neven in daycare with a woman caring for four other children about Neven's age. Helling's sister provided daycare for Neven during Helling's periods of placements.
*800¶ 6. Consistent with the guardian ad litem's recommendation, the trial court decided to award Helling primary physical placement, primarily citing the instability of Lambert's living situation. The details of the trial court's reasoning will be discussed more fully below.
DISCUSSION
¶ 7. We review the trial court's placement decision under the erroneous exercise of discretion standard. Wiederholt v. Fischer, 169 Wis. 2d 524, 530, 485 N.W.2d 442 (Ct. App. 1992). To be sustained, a discretionary determination must be based upon the facts appearing in the record and in reliance on the appropriate and applicable law. Luciani v. Montemurro-Luciani, 199 Wis. 2d 280, 294, 544 N.W.2d 561 (1996). Thus, although a court has broad discretion in making placement decisions, its power is still limited to that provided by statute. Schwantes v. Schwantes, 121 Wis. 2d 607, 360 N.W.2d 69 (Ct. App. 1984).
¶ 8. Under Wis. Stat. § 767.24(4)(a)2 (2001-02),1 the trial court "shall set a placement schedule that allows the child to have regularly occurring, meaningful periods of physical placement with each parent and that maximizes the amount of time the child may spend with each parent, taking into account geographic separation and accommodations for different households." The court must consider a variety of factors relevant to the best interest of the child in making its determination, including the wishes of the child and parents; the *801interaction of the child with his parents or other persons who may significantly affect his or her best interests; the amount of time the child has spent with each parent; the child's developmental needs and adjustment; any physical or mental health issues; any abuse issues; the availability of child care; the ability of each parent to cooperate and facilitate the other parent's contact with the child; and any professional assessments. § 767.24(5). In addition, the court shall consider "the need for regularly occurring and meaningful periods of physical placement to provide predictability and stability for the child," and "[s]uch other factors as the court may in each individual case determine to be relevant." § 767.24(5)(em), (k). However, because freedom of association is constitutionally protected, a court may not base a placement decision on a parent's nonmarital sexual conduct or relationship with a third party absent specific evidence that the conduct or relationship in question has had or would have a significant adverse impact on the child. See Schwantes v. Schwantes, 121 Wis. 2d at 625-26.2
¶ 9. Here, the trial court found that both parents had spent substantial amounts of time with the child and interacted well with him; that the child was well adjusted; that there were no mental health or abuse issues; and that the parents were able to communicate reasonably well with one another. While "not denigrating" the mother's daycare services, the trial court fa*802vored the father on that factor due to the strong relationship the father's sister had with the child. The trial court acknowledged that the mother had been the primary caregiver, but explained that it was giving that factor less weight than it might have in other cases because the child had demonstrated an ability to adjust well to the changes he had already experienced. We see no misuse of discretion in the trial court's consideration of any of these factors.
¶ 10. The trial court then focused on the factor enumerated in Wis. Stat. § 767.24(5)(em), the "need for regularly occurring and meaningful periods of physical placement to provide predictability and stability for the child." Lambert argues that the trial court erred in interpreting this provision as a general requirement for predictability and stability in the child's life, rather than a preference for regularly occurring periods of physical placement. Even if we were to read that factor as Lambert urges, however, we are persuaded that predictability and stability would still be permissible factors for a court to consider as relevant to the child's best interest under the catchall provision in § 767.24(5)(k).
¶ 11. The main problem with the trial court's analysis, in our view, is not its consideration of stability as a proper factor, but rather its factual basis for finding that Lambert was "not in a predictable and stable situation." The trial court specifically found that Lambert's situation was" not stable "because ... [she had] recently established a relationship with Mr. Weber." It went on to state:
That relationship, if I calculate it correctly, has existed where you live together for at the most three months, and it's the kind of relationship that can be here today and gone tomorrow because when people do *803not get married, that relationship does not have the stability of law. It can end tomorrow. That concerns me.
It also concerns me, and, you know, I observed Mr. Weber up here. I observed Ms. Helling. I didn't hear any testimony about how that child gets along with Mr. Weber. Here we have got somebody that's in that child's life and is going to be in that child's life on a fairly consistent basis, and I'm concerned about that.
And I'm concerned by the new relationship. You don't own the home, that you are there at Mr. Weber's largess, and that is one of the most important things I think in this case.
And I also agree with Mr. Daniel, I think it shows poor judgment on your part that you entered into that relationship, that you will have another child out of that relationship with no predictability or stability in it.
And I guess in part my attitude about this type of thing is based on the fact that I have probably 40 to 50 paternity cases a month which are brand new, and I don't think that's right. And I don't think those are predictable or stable relationships, and I guess one of the criterion that I based that decision on, I see it 40 to 50 to 60 times a month. I see what happens. That concerns me.
(Paragraph divisions revised.) We see several flaws in the trial court's reasoning.
¶ 12. First, although it is true that Lambert and Weber had only been living together for three months by the time of the hearing, it was undisputed that they had been dating for two years. We therefore question the characterization of their relationship as having been "recently established."
*804¶ 13. Second, the trial court's assertion that Lambert and Weber's relationship is the kind that "can be here today and gone tomorrow" appears to have been based solely on the fact that the relationship was nonmarital in nature. The trial court's generalized assumption about the stability of nonmarital relationships ignores the reality that marital relationships can also end in divorce. Moreover, it violates the maxim that "[e]ach custody case must turn on its own facts and circumstances." Wendland v. Wendland, 29 Wis. 2d 145, 149, 138 N.W.2d 185 (1965). Lambert testified that she hoped to have a "permanent, long-term relationship" with Weber. We see nothing in the record which would undermine Lambert's testimony or indicate that Lambert and Weber's relationship was in any particular jeopardy. Indeed, the trial court's own comment that Weber was "going to be in that child's life on a fairly consistent basis" suggests a likelihood that Lambert and Weber's relationship would continue.
¶ 14. The trial court could reasonably take into consideration the nature of Weber's relationship with the child. However, Weber had been dating Lambert for two of the two-and-a-half years of the child's life, and the trial court found that the child was well adjusted. Absent any specific evidence of problems, we see no factual basis for the trial court to draw any unfavorable inferences about Weber's relationship with the child.
¶ 15. The trial court could also reasonably consider the stability of each parent's living situation. Again, however, there was nothing in the record to suggest that Lambert and Weber would not be living together for an indefinite time into the future, or that Lambert was any more likely to move again sooner than Helling, who had himself moved twice after Neven's birth. Nor could the trial court reasonably infer that *805Lambert's past relationship history made her more likely than Helling to fail in future relationships, when the only failed relationship in evidence was that of both Lambert and Helling. While the fact that Lambert was not formally listed on Weber's lease might legally allow Weber to ask Lambert to leave at any time, there was absolutely nothing in the record to suggest any reason why he would be likely to do so.
¶ 16. We next consider the trial court's comments that Lambert's entering into a relationship with Weber, and having a child with him, showed "poor judgment," and that the court did not think it was "right" that so many paternity cases were being filed. It again appears that the trial court was relying in part upon evidence outside of the record, since the "40 to 50 to 60" monthly paternity cases upon which the trial court claimed to rely for its finding that "those" relationships (which we understand to mean between unmarried parents) were not predictable or stable were not before the court in this case.
¶ 17. In sum, we are persuaded that the trial court's finding that Lambert's living situation was "unstable" was based primarily not upon evidence in the record, but rather upon the trial court's negative view of her unmarried status. We further agree with Lambert that the trial court's comments reflect an impermissible consideration of her nonmarital relationship with a third party as a negative factor absent any showing that the relationship was harmful to the child, in violation of her associational rights. Schwantes, 121 Wis. 2d at 625.
¶ 18. The dissent correctly notes that, unlike in Schwantes, the trial court here did not explicitly condition the mother's opportunity for continued physical placement upon the termination of her third-party *806relationship. We see little practical distinction, however, between an order which forces a parent to choose between a third-party relationship or continued periods of physical placement and one which denies equal or primary physical placement based upon an existing third-party relationship. In each instance, the parent's choice to associate with a third-party ultimately affects the amount of time the parent will have with his or her child. Accordingly, we believe that showing a significant adverse affect upon the child is a necessary prerequisite to adverse consideration of a third-party relationship. This is required to protect the same associational rights identified in Schwantes.
¶ 19. The dissent also correctly notes that the trial court's view of Lambert's living situation was not the only factor in its decision. If the trial court's comments in this regard had merely been made in passing, we would agree that the rest of the trial court's discussion would provide a reasonable basis for its decision. But the trial court itself noted that many of the other factors it considered "really don't cut either way." Lambert's living situation was one of the few main factors the court focused on in making its decision. It is impossible to know what conclusion the trial court would have reached had it not relied heavily on Lambert's living situation to reach that conclusion. Because the trial court lacked a sufficient factual basis for a finding upon which it relied heavily in making its placement decision, we conclude that it erroneously exercised its discretion.
¶ 20. We emphasize that we are not holding that a trial court cannot consider whether a parent's particular lifestyle choices have an impact on the best interests of a specific child. Findings regarding instability in living conditions must, however, be based upon evi*807dence specific to the individual case, not generalizations. There was insufficient evidence here from which the trial court could have made a finding that Lambert's relationship was so unstable as to adversely impact the child. We therefore remand to have the trial court reconsider the placement decision in this case without treating Lambert’s relationship with Weber as a negative factor.
By the Court. — Judgment reversed and cause remanded with directions.
All references to the Wisconsin Statutes are to the 2001-02 version unless otherwise noted.
Helling contends that Schwantes is inapplicable here because that case involved the modification of an initial placement. However, because the modification in that case was based upon the violation of a condition set forth in the initial placement decision, the court did in fact analyze the topic of third-party relationships in the context of initial placement decisions.