How to Know What Boundaries to Set

You agreed to Sunday dinner at their parents' house again, and by the drive home you were snapping at your partner about the radio. That flash of irritation over something small is often the first trace of a limit you haven't named. A boundary protects the connection you want to keep; you find yours by tracing the feeling to an area of your life and stating the limit as something you will do.

How to tell which boundaries you need: the short answer

If you're searching for how to know what boundaries to set in a relationship, start with a feeling rather than a rulebook. Notice the moment your jaw sets or your answer comes out as "sure" when you meant "no," and locate it in an area of your life: your time, your body, your money, your phone, your opinions. Then put the limit into a sentence about your own behavior: "When this happens, I will do that."

That sequence keeps a boundary on the side of the relationship. Family-systems researchers call the ability to stay close to someone while keeping a clear sense of where you end "differentiation of self." In plain terms, it's being able to say "I disagree" without panicking and "I need you" without disappearing. A 2021 scoping review in Clinical Psychology Review by Calatrava and colleagues found that 39 of 56 studies testing differentiation against couple functioning supported the idea that more of it goes with better relationships; 12 studies reported mixed findings and 5 found no link. The same review notes that 250 of the 295 studies it screened were cross-sectional snapshots, and that no meta-analysis on the topic had been done. So researchers found an association among groups of couples, but they can't tell you that a specific boundary will change your specific relationship.

The next step is learning which feelings tend to show up before a boundary is missing.

The feelings that flag a missing boundary

Resentment can be one useful flag, and it rarely points at one event. A peer-reviewed validation study of the Resentment Rating Scale for Couples (844 participants) describes resentment as "not based on a present event but on several past events" and as "reliving a painful experience again and again." Psychotherapy Networker makes a similar point about chronic resentment: it tends to involve more than one event. If you keep replaying the same Sunday, the boundary may be about the pattern of Sundays.

Other feelings may be worth tracing to a possible limit:

Dread: Not looking forward to seeing your partner, or bracing before their car pulls in, appears in Ayala Malach Pines's description of couple burnout. She defines couple burnout as exhaustion "caused by a chronic discrepancy between expectations and reality." Could the dread relate to an expectation you haven't discussed?
Exhaustion: Fatigue that sleep doesn't fix, headaches, and stomachaches sit on Pines's list of physical markers. A study using 10 days of daily-experience data from 756 married individuals found that researchers associated providing emotion work, the labor of managing a partner's moods, with lower well-being for the person doing it, especially when the spouse had elevated depressive symptoms. Exhaustion may point to an emotional boundary: are you carrying feelings that aren't yours to carry alone?
Anger: Anger about a single incident is ordinary. Anger that arrives fast and out of proportion may have a backlog behind it. Terri Cole, LCSW, describes self-abandonment this way: "As we self-abandon, we build up resentment for ourselves and the other person."
The urge to lie or avoid: Hiding a purchase, deleting a text thread, or going quiet for two days rather than having the conversation may be attempts to protect a limit you haven't stated. Nedra Glover Tawwab, LCSW, adds guilt and apologizing after saying no to the list; if you say sorry every time you decline, the "no" itself may feel like a violation in your own mind.

These feelings show up in the body over time. A 20-year longitudinal study of married couples found that husbands' anger behaviors predicted increases in cardiovascular symptoms, while stonewalling (shutting down and withdrawing) predicted increases in musculoskeletal symptoms, after controlling for age, education, exercise, smoking, and alcohol. That is one cohort and one set of measures, so treat it as a reason to take the pattern seriously rather than a forecast for your own back.

Before you move on, you might pick the feeling from this list that showed up most this week. Ask yourself: what happened right before it, and was that the first time?

Run the seven domains: a boundary self-audit

The seven-domain list is a practitioner convention, and the research behind each domain is uneven. Tawwab's Set Boundaries, Find Peace names six types: physical, sexual, intellectual, emotional, material, and time. Cloud and Townsend's Boundaries lists four. "Digital" arrived later; the earliest Psychology Today piece in the research to include it is Marlynn Wei's from September 2024. A 2024 dissertation on relationship boundaries states plainly that researchers have no validated measures that directly assess relationship boundaries. The list works as a checklist for your own attention, and it isn't a diagnostic instrument.

Answer two questions per domain and note where you hesitate or feel a pull to defend your partner before finishing the sentence:

Domain	Ask yourself	What the research adds
Physical	When I want space after work, do I say so or wait for them to guess? After time together, does my body feel rested or braced?	Personal-space and consent research treats this as implicit; no couple-specific scale exists.
Sexual	Have I gone along with something in the last month I didn't want? Can I say "not tonight" without a repair conversation afterward?	Willis and Smith (2021) found 31% of participants reported at least one sexual behavior against their will; 47.9% of women versus 22.3% of men.
Emotional	Am I checking their mood before I check my own? Whose problem am I holding right now?	Cloud and Townsend name responsibility for a partner's feelings as a core boundary problem.
Intellectual	Can I disagree out loud without being mocked or talked over? Do I edit my opinions before sharing them?	Not identified as a distinct empirical construct; a 2026 review of 126 self-silencing studies found a meta-analytic r = .39 with depression.
Time	How many hours this week were mine to plan? When did I last cancel something of my own for them, and did I choose it?	Only a qualitative taxonomy (Kreiner et al., 2009) exists; no validated romantic time-boundary scale.
Financial	What on my statement would I rather they not see, and why? Do we have an agreed amount above which we talk first?	Dew, Britt, and Huston found financial disagreements were the strongest disagreement type predicting divorce among 4,574 couples.
Digital	Do I know what my partner assumes about phone access, or am I guessing? How often does a conversation stall because one of us is on a screen?	Pew (2020, 4,860 U.S. adults) found 51% of partnered adults said their partner was distracted by a phone during conversation.

You could print the table, or copy the two questions per domain into your notes, and revisit it once at the start and once at the end of a two-week resentment log. The log exists because the RRS-C research frames resentment as a stack of past events, and a stack is hard to see in the moment. Each time the flash arrives, write down:

The date and time
What happened in one sentence, without interpretation
The feeling, using one of the five words from the previous section
The domain from the table
Whether you said anything to your partner about it

After fourteen days, count entries per domain. Three entries in "time" and none anywhere else tells you where you might start. Tawwab's first instruction for boundary work is to react faster: "react a little faster because sometimes with boundaries, we'll let things go on for years." Using the log may help you notice the pattern in two weeks instead of two years.

Your non-negotiables versus your preferences

Sort the log entries into two piles before you raise anything with your partner. Isabelle Morley, PsyD, separates universal boundaries, which concern safety and apply in any relationship, from individual ones based on preference. On universal boundaries, she writes: "If someone crosses our boundary, and it's not a universal one, it's not a reason to sound the abuse alarm; it's a reason to talk about the boundary." Repeated violations of universal boundaries, she adds, indicate an abusive relationship. A systematic review and meta-analysis defines coercive control as "an enduring pattern of domination, intimidation, isolation, and control."

Many entries in a two-week log may belong in the second pile, and that pile still deserves sorting. Two questions help:

Would this be a limit with any partner, or does it belong to this relationship and its history?
If my partner crosses it once more, what happens to me: a bad evening, or a change in how I see us?

If a limit survives both questions, you might treat it as a non-negotiable. If a limit depends on who you're with, and costs you an evening rather than your sense of self, you might treat it as a preference you can trade.

Knowing your values makes the sorting faster. Parise and colleagues (2019) studied 202 dating couples cross-sectionally and 97 married couples over time and found that self-concept clarity, a stable sense of who you are and what you want, predicted both partners' relationship satisfaction, with a longitudinal indirect effect of β = .12. The samples were small and the effect modest, so read this as support for a values inventory, not a guarantee. Write down the five things you'd protect in any relationship, then check each log entry against the list. You might place entries that touch one of the five in the non-negotiable pile.

Boundary or ultimatum? A two-question test

A boundary describes what you will do; an ultimatum describes what your partner must do. Hailey Magee, writing for the Gottman Institute, distinguishes the two directly: "a boundary is about our own limits, whereas an ultimatum is specifically designed to control somebody else." Megan Harrison, LMFT, defines an ultimatum in Verywell Mind as "a warning or demand issued to one's partner to act in a specified way within a specified period or they risk losing the relationship."

Run any sentence you're planning to say through two questions. First, whose behavior does the sentence describe? Second, could you carry out the consequence today, on your own, without their cooperation? If the answer to the first is "theirs" or the answer to the second is "no," you have a demand dressed as a boundary. Mark Travers, PhD, warns in Psychology Today that people can misuse boundary language to issue a demand, an ultimatum, or a unilateral condition that their partner must obey.

Three rewrites show the test in action; clinicians supplied the boundary scripts:

Situation	Sounds like an ultimatum	Rewritten as a boundary
Raised voices during arguments	"Yell at me one more time and I'm leaving you."	Travers: "If voices are raised during arguments, I'm going to take a break and revisit the conversation later."
Private details shared with others	"Tell your sister about us again and we're done."	Tawwab: On sharing private details, "I don't want you to share my personal business with others. If you do, I'll stop telling you things."
Personal attacks mid-conversation	"Stop attacking me or this conversation is over for good."	Gottman Institute: In a script about personal attacks, "I'd like to have this conversation with you, but only if you are willing not to make personal attacks. If you continue to do so, I will walk away from this conversation."

The Gottman Institute concedes the line isn't always crisp. Its guide to boundaries and ultimatums says: "the area between boundaries and ultimatums can be quite gray." You can distinguish them by examining your intention and delivery. Practitioners use this distinction as a convention; researchers have not formalized it within a therapy model. Tawwab herself uses the phrase "healthy ultimatums" for reasonable follow-through. What matters in practice is whether you're moving your own feet or trying to move theirs.

Signs of unhealthy boundaries: rigid, porous, and enmeshed

Salvador Minuchin's structural family therapy (1974) is the origin of the rigid, clear, and diffuse typology that most boundary-style quizzes borrow from. Rigid boundaries are "overly restrictive and permit little contact with outside subsystems"; diffuse boundaries are "so easily crossed that members have difficulty knowing where one member stops and the other begins." The APA Dictionary defines enmeshment as involvement in each other's lives "to an excessive degree, thus limiting or precluding healthy interaction and compromising individual autonomy and identity."

Read each row and notice which one you'd reach for under stress, since most people default to a style rather than living in it full-time:

Style	What it sounds like	What it tends to cost
Clear (the healthy range)	"I need an hour alone tonight, and I want to hear about your day after."	Requires saying uncomfortable things out loud.
Rigid	Silence for three days after a hurt; "I don't need anything from anyone."	Terri Cole describes withdrawal or punishment after being hurt and inflexibility; connection thins.
Porous	Automatic yes when you mean no; apologizing for having a preference.	Cole ties this to resentment build-up; Tawwab links it to exhaustion and guilt.
Enmeshed (diffuse)	Not knowing what you want until you've checked what they want; their bad day becomes your bad day.	Autonomy and identity erode, per the APA definition.

A distinction inside the rigid column matters. Murray Bowen's "emotional cutoff" is a defensive escape from unresolved attachment, not healthy autonomy. A 2022 cross-sectional study of 140 Spanish adults found differentiation positively associated with marital satisfaction (r = .32) and emotional cutoff negatively associated (r = −.43). Cutting someone off can look like a strong boundary from the outside and feel like one from the inside; the correlation runs the other way.

The romantic-relationship evidence on styles is thinner than the family-systems evidence. A 2024 dissertation (unpublished, not peer-reviewed) found clear boundaries associated with higher satisfaction and porous and controlling boundaries with lower satisfaction and higher odds of dating violence. And Minuchin's categories carry a cultural lens; a clinical text notes that his perspective reflected Western culture and that the categories are not universal. If your family shares finances across three generations and talks daily, that isn't enmeshment by default. The question is whether you can still locate what you want.

How to say it: scripts for each boundary type

Say what you need and what you'll do, in that order, in the first person. Tawwab's instruction is to state what you need, rather than naming only the problem. Thomas Gordon coined the "I-statement" in 1970, and a 2018 study of 253 students rating hypothetical statements found I-language less likely to provoke defensiveness, with a much larger advantage when the speaker also explained their perspective. The evidence has limits: in Korobov's recordings of 20 real couples, over 61% of arguments contained moments where I-language got deflected back onto the speaker. A script starts a conversation. It doesn't finish one.

Each script below pairs the need with the action you control, and you can swap in your own details:

Physical: "I want a few minutes to land when I get home before we talk. I'll come find you when I've changed."
Sexual: "I'm not up for sex tonight, and I'd like to stay close on the couch. If I feel differently later, I'll say so."
Emotional: "I want to hear about work, and I'm at capacity right now. Can we talk after dinner? I'll bring it up if you forget."
Intellectual: "I see this differently, and I'd like to finish my thought before you respond. If I get interrupted, I'll pause and come back to it later."
Time: "Thursday evenings are mine. I'll be home by nine and I'll text if that changes."
Financial: "I'd like us to talk before either of us spends more than $200 outside the plan. I'll do the same on my side."
Digital: "I don't read your messages and I'd like the same. If I need something from your phone, I'll ask first."

Saying no follows the same shape and stays short: "No, I can't do Sunday. Saturday afternoon works for me." The DBT interpersonal-effectiveness skill DEAR MAN puts the "Assert" step this way. In its guidance on saying no clearly, it says: "Ask for what you want or say no clearly. Do not assume that others will figure out what you want." One caution from a study of 59 couples: first-person plural pronouns ("we," "us") went with more positive problem solutions, while second-person pronouns went with more negative interactions. If your draft has three "you"s in it, consider rewriting before you speak.

You might rehearse before you say it. Read the script aloud once, alone. You could strike "always" and "never," then run the two-question test from the ultimatum section. Decide which consequence you can carry out if your partner crosses the boundary, and, if possible, choose a moment when neither of you is depleted: in a study of couples' conflict interactions, partners who had both slept less interacted more hostilely, and sleeping one hour less than average was associated with 6% higher inflammatory markers after conflict. Wednesday after dinner may give you more room than Sunday at midnight.

When your partner pushes back or keeps crossing the line

A boundary has little effect if you don't act on it. The Gottman Institute is blunt about enforcing a boundary: enforcement means leaving or ending the conversation when the other person raises their voice. April Eldemire, LMFT, explains the cost of not following through in a discussion of respecting your boundaries: your partner may see that you do not respect the limit yourself.

The ladder below is a way to respond in proportion, drawn from the clinicians in the research:

First crossing: Name it once, in the moment, using the script you rehearsed. Seth Gillihan, PhD, notes that consequences work best when they follow in an obvious way from the undesired behavior, so if you said you'd leave a shouting match, leave the room, and say you'll be back in twenty minutes.
Second crossing: Restate the boundary and carry out the consequence again without escalating it. Terri Cole's two rules: Consequences should be proportionate, and you cannot threaten something you are unwilling to do. Ask your partner directly what got in the way of the change they agreed to.
A pattern: Tawwab's definition of a repeat offender is "someone who violates a boundary you have clearly expressed, says they won't do it again and does." Her graduated response sequence runs from changing how you engage, to withholding information the person has shown they cannot keep private, to reducing contact, to modifying expectations, and finally to severing ties. Most preference-based boundaries never need the last three rungs.

No source in the research gives a number of violations that converts a pattern into a decision about the relationship. Gabriella Giachin, LCSW, frames the shift as evaluation rather than verdict. When a pattern emerges, she suggests discussing what you want from the relationship and evaluating whether your partner can respect your boundaries. Ellyn Bader, PhD, lists the conditions under which confrontation becomes necessary rather than optional: chronic failure to take responsibility, repeated deception or betrayal, significant deficits in conflict skills, and persistent denial about impact.

What changes after you talk is the most useful data you have. A 2024 study of 217 cohabiting couples found that daily attentive listening predicted more helpful reconciliation attempts, while contemptuous communication and habitual demand-withdraw did not. So after the second conversation, ask: did they listen, did anything shift for a week, and did they come back to it without being prompted?

Even then, you make the decision. Tawwab says there are times people aren't ready to leave a relationship, and she does not support leaving before they are ready. When one partner is ambivalent about staying, a structured protocol called Discernment Counseling exists; in an analysis of 100 consecutive cases, 47% chose six months of couples therapy, 41% chose separation or divorce, and 12% chose the status quo. That was an uncontrolled observational sample, and the protocol excludes cases involving intimate partner violence. If you are in an emotional or mental-health crisis in the United States, call or text 988. If you are in immediate danger, call 911. If abuse is present outside an immediate crisis, seek individual support and make a safety plan before an exit; Verywell Mind notes that risk after leaving may increase.

When your boundaries and theirs conflict

Two legitimate boundaries can point in opposite directions, and phone privacy is the cleanest example. You want your messages to stay yours; your partner wants shared access as a sign of trust. Neither of you is issuing an ultimatum. You're each protecting a different need.

Pew Research Center's 2020 survey of 4,860 U.S. adults shows how split the norms are: 75% of partnered adults had given their partner their phone password, 34% had looked through a partner's phone without their knowledge, and 70% said looking without knowledge was rarely or never acceptable. A 2024 study of 400 Black and Latinx women (a non-representative, more educated sample) found 56% opposed partner access to devices, 24% favored it, and 20% said it should be negotiated situationally. A 2026 CHI study identified five unspoken-boundary motivations: seeing privacy as unnecessary in intimacy, assuming implicit respect, signaling trust, avoiding conflict, and responding to societal expectations. If you and your partner have never said your assumptions out loud, you're probably running on two different ones.

Negotiating starts with the need under each position. Ask: what does access give you, and what does privacy protect for me? Often the answer on the access side is reassurance, and the research suggests that's the piece worth working on directly. A two-year longitudinal study of 322 young adults found that social-media jealousy predicted more electronic surveillance and lower satisfaction a year later, while surveillance itself did not predict later satisfaction. Reading the phone doesn't seem to settle the jealousy; something else has to.

A workable middle for this example might be: passcodes stay private, either partner can ask to see something specific and expect a yes, and the partner who wants reassurance names what would help beyond the screen. Researchers in two studies examined approaches that may help such conversations go better. Finkel and colleagues had 120 couples spend 21 minutes writing about a recent conflict from the perspective of a neutral third party "who wants the best for all involved"; over two years, the exercise eliminated the decline in marital quality seen in the control group. Sisson and colleagues (2022), across a lab study of 111 couples and 2,178 weekly reports from 151 couples, found that reappraisal rather than suppression was linked to better partner-change outcomes as rated by both people.

Money conflicts follow the same logic and resolve more slowly. Papp, Cummings, and Goeke-Morey analyzed 748 conflict instances from 100 couples and found money conflicts "more pervasive, problematic, and recurrent" than other kinds. Joseph and Peetz (2025) followed 124 couples and found those who hid more financial information reported less satisfaction at the study's end. If your conflict is "I want a private account" versus "I want to see everything," autonomy and security may be underlying needs, and a shared threshold for discussed purchases gives each side something concrete.

Where your boundary blueprint came from: attachment, family, and culture

The same limit can feel easy to one person and like a threat to the relationship to another, and part of the difference is learned. Attachment research describes tendencies here, with a firm caveat: adult attachment anxiety and avoidance are dimensional, continuously distributed variables, not boxes, and they vary by relationship and context. Nothing below is a diagnosis of you.

Attachment and autonomy: Anxiety may shape fears about conflict or rejection, while avoidance may shape withdrawal. Neither tendency tells you what one person will do.
Family and culture: Early roles and cultural norms may shape how people express needs, but group patterns do not define an individual relationship.

At the group level, people higher in attachment anxiety tend to perceive more conflict and escalate it; Campbell and colleagues (2005) found this in observer-rated videotaped discussions. They also tend to tolerate more. In a study of 158 heterosexual women, higher attachment anxiety went with more sexual compliance (b = .47) and less endorsement of confronting a partner. If saying "not tonight" feels like risking the relationship, that fear may have a history worth examining.

People higher in avoidance tend to withdraw. In a study of 175 couples, avoidant attachment predicted a person's own withdrawal (β = 0.41) more strongly than demand or aggression. But a rigid boundary and a real need for autonomy are different things. Brassard and colleagues (2025), working from 21-day diaries with 121 and then 154 couples, define autonomy as feeling free to act according to one's true self. Their data linked higher avoidance to lower daily autonomy and, through that, lower satisfaction at three months.

Security runs the other direction from what people expect. Feeney (2007) tested the "dependency paradox" with a six-month follow-up and found that a partner's acceptance of dependence and sensitive responsiveness went with the supported person's increased autonomy. Being able to lean on someone appears to make it easier to stand apart from them, which is another way of saying boundaries and closeness aren't a trade.

Family of origin sets an early template. Skowron (2000) found that couple differentiation scores explained 74% of the variance in husbands' marital adjustment and 61% in wives'. Children who took on adult emotional roles carry a specific pattern forward: in a study of 225 Israeli young women, those who felt parentified tended to have negative expectations of mutuality and difficulty disclosing their needs and wishes to a partner. If you were the kid who managed a parent's moods, "whose problem am I holding right now?" may be the hardest question in the audit.

Culture shapes what counts as a boundary at all. Greater Good Magazine notes that enmeshment can be protective where it is a cultural norm. A study of 277 South Korean and White American participants found the White Americans scored higher on differentiation, and the authors cautioned that the instrument may not accurately measure enmeshment or rigidity in Korean families. Self-silencing findings conflict outright: one U.S. college sample (N = 604) found Asian participants reported the highest levels, while a comparison of Iranian-Australian and European-Australian adults found no group differences in culture or gender. Group patterns don't predict you.

A therapist becomes worth considering when the audit keeps returning the same domain and the conversations keep ending the same way, or when the history behind a boundary is heavier than the boundary itself. Bader's list of conditions for necessary confrontation is a reasonable threshold. One safety note: clinical guidance advises against conjoint therapy when intimate partner violence is present, so individual support comes first in that situation. Esther Perel describes the therapist's role as accompanying people wherever they choose to go because they will live with the consequences of their decisions.

Turning reflection into clarity

Reflection and rumination can feel identical at 1 a.m. Rumination replays the Sunday dinner and arrives back where it started. Reflection asks a question the loop hasn't asked yet: which domain is this, is it a preference or a non-negotiable, and what will I do the next time it happens? The audit and the log in this article are structured reflection on paper. Some people want the same structure with something that talks back.

Mutually is a solo-user, AI-guided space for that. You say what happened, and it asks open-ended reflective questions about your side of the relationship rather than telling you what to do. Sessions are discrete, with a summary and takeaways you can reread later, so you can check whether your thinking about a boundary has moved between week one and week three. You can start as a guest without an email, name, or payment; before the first exchange, Mutually asks you to consent to sharing your messages with Anthropic's AI to generate replies and summaries, and you can revoke that consent in Settings or delete any conversation. The first session is free; after that it costs $9.99 per month or $59.99 billed annually. AI-guided. Not therapy. It doesn't diagnose, and it doesn't decide for you.

If you want to think through your own situation, start a free session with Mutually. Say what happened, then follow the questions wherever they lead.

FAQ

These answers condense the sections above for quick reference.

What are the seven types of boundaries in a relationship?

The list most articles use is physical, sexual, emotional, intellectual, time, financial, and digital. It's a practitioner convention rather than a validated framework; Tawwab's book names six without digital, and Cloud and Townsend's names four. The evidence base differs by domain, with sexual, financial, and digital boundaries having the strongest research behind them and intellectual boundaries having none as a distinct construct.

What feelings signal that a boundary has been crossed?

Resentment that replays several past events rather than one, dread before seeing your partner, exhaustion that sleep doesn't touch, anger out of proportion to the trigger, and an urge to hide or avoid rather than talk may all be worth examining. Guilt and reflexive apologizing after saying no belong on the list too. One incident that produces any of these is information. Repeated feelings over time may signal a pattern worth examining, especially when they last, affect you strongly, or continue after you talk about what happened.

What are the 3-6-9 rule and the 7-7-7 rule, and do they help with time boundaries?

The 3-6-9 rule divides a new relationship into a honeymoon stage (months 0–3), a conflict stage (months 4–6), and a decision stage (months 7–9); it traces most clearly to TikToker Annabelle Gesson. Couples therapist Daniel Dashnaw rejects an exact timeline, noting that no single peer-reviewed study validates it. A 2023 study of 1,345 dating young adults found relationship quality outcomes peaking at 6–12 months for sexual activity, 1–2 years for companionship, and 2–3 years for commitment, none of which lines up with the rule.

The 7-7-7 rule prescribes a date every 7 days, a night away every 7 weeks, and a vacation every 7 months; writers attribute it to Steffo Shambo. Suzanne Wallach, PsyD, LMFT, says she knows of no supporting research showing that doing things in sevens improves relationships. What the research does support is shared time that feels new: Aron and colleagues (2000) found that novel and arousing activities predicted greater increases in relationship quality, and a 2023 National Marriage Project report of 2,000 married adults found 83% of wives and 84% of husbands with regular date nights reported being very happy versus 68% and 70% without, while noting the causal arrow may point the other way.

Both rules are schedules for time you spend together. A time boundary works in the other direction: the hours you protect for yourself, stated as something you'll do ("Thursday evenings are mine"). If a schedule rule helps you notice you haven't had a real evening together in two months, use it as a prompt. It won't tell you which hours you need alone.

How do I say a boundary without starting a fight?

Lead with the need and what you'll do, in the first person, and pick a time when you've both slept. Cut "always" and "never." Check whether the sentence describes your behavior or theirs; if it's theirs, rewrite. Expect that a first attempt may get deflected, since Korobov's recordings found this in over 61% of real arguments, and plan to restate rather than escalate.

What if our needs are incompatible?

Find the need under each position before deciding they can't coexist. Wanting phone access may be a request for reassurance; wanting privacy may be a request for autonomy, and the two can sometimes be met in different ways at once. If, after several conversations, a non-negotiable of yours keeps being crossed and nothing changes after you talk, you may shift your question from "how do we negotiate this" to "can this person respect this limit?" You can answer that question at your own pace.