Most of what I write about between-session work is about absence. The client leaves the room, the insight evaporates somewhere around Tuesday, and the next session starts cold — the whole problem with how therapy insight doesn't stick is that too little carries over.
EMDR inverts that. The risk isn't that nothing happens — it's that too much keeps happening. You ran a set, the material started to move, and the client walked out with a memory network that's still loosening. The Adaptive Information Processing model underneath EMDR holds that reprocessing keeps integrating a memory once it's been activated (American Psychological Association, EMDR clinical practice guideline). That's the goal in the room. Outside it, with no clinician tracking it and no dual attention keeping one foot in the present, the same momentum can become intrusive images, disrupted sleep, and a Tuesday-afternoon spike the client has no structure to hold.
So this is the opposite of my usual argument. I'm not trying to get more to carry over — I'm trying to make sure the right thing does: grounding, distance, the felt sense of "I can set this down," with the reprocessing itself staying inside the protocol.
Why EMDR clients destabilize between sessions
The destabilization isn't a side effect of doing EMDR wrong. It's a predictable feature of having activated a memory network and then ending the hour. Three mechanisms stack.
The network is still online. When reprocessing starts, you've opened something, and channels of association keep surfacing — that's how the method works. If the session ends before a channel resolves, the client leaves with material that's moving and no container around it. The body keeps doing the work the protocol was supposed to be metering.
The resources don't transfer cleanly. This is the part I want clinicians to sit with. You installed a Calm Place in your office — regulated, witnessed, with your voice cuing it and bilateral stimulation reinforcing it. That resource is partly bound to the state and context it was built in. Classic memory research is blunt: what you encode in one state and setting is harder to retrieve in a different one (Godden & Baddeley, 1975; Tulving & Thomson, 1973). So the Calm Place that landed on Monday is genuinely harder to reach on Wednesday — not because the client is non-compliant, but because the cue is fighting the context. The same problem I describe for skills generalization between sessions shows up here with higher stakes, because the thing failing to generalize is the client's brake.
The disturbance doesn't keep office hours. It peaks when the network gets touched by ordinary life — a smell, an argument, a song — which rarely happens during your session. It happens in the gap, where the client is asked to be their own clinician with a degrading resource and no one tracking the SUD.
Put those together and the between-session period for an active EMDR client is the highest-risk stretch of the week. That's exactly why it's the worst possible place to hand someone an AI chatbot — a point I'll come back to hard.
Complete vs. incomplete sessions: the line that governs the week
The single biggest determinant of how a client does between sessions is whether you closed cleanly. Francine Shapiro's protocol draws a sharp line between a complete and an incomplete session — not a bureaucratic one, but about what state the client carries out the door.
A complete session reached an adaptive resolution: disturbance came down (a SUD at or near 0 where clinically appropriate), the positive cognition strengthened, you ran a body scan. The network was activated and then brought to rest.
An incomplete session ends while material is still active — you ran out of time, the client hit a block, the disturbance is still up. Shapiro's protocol is explicit that incomplete sessions require a deliberate closure precisely so the client doesn't leave mid-reprocess (American Psychological Association, EMDR clinical practice guideline describes the eight-phase approach this closure sits within). The Container exercise lives here: you guide the client to imagine putting the unresolved material somewhere contained until you reopen it together next session.
Here's the read that matters for the week. An incomplete session that gets a real closure can be safer than a "complete" session you rushed. A clean Container plus a rehearsed Calm Place hands the client an explicit instruction to set the material down. A rushed completion hands them a network you called "done" while their body says otherwise — so they get ambushed Tuesday with no permission to set anything down.
| Rushed "complete" session | Properly closed incomplete session | |
|---|---|---|
| What the client carries out | A network still active, labeled finished | Active material, explicitly contained |
| The instruction they have | None — they think it's done | "Set it down, don't open it, we resume together" |
| What happens at the Tuesday spike | Ambush, no framework, self-blame | Expected, named, has a resource to meet it |
| Between-session resource state | Untested, assumed reliable | Rehearsed in session, given a job |
The practical move: protect your closure time like it's part of the protocol, because it is. Stopping reprocessing ten minutes early to close well beats squeezing one more set and sending someone out the door mid-channel.
Resourcing that survives the week
If the Calm Place degrades by context, the answer isn't to install it harder in your office. It's to build it so it travels. A few things I do differently for EMDR clients now.
- Over-rehearse the resource before you ever reprocess. Phase 2 — preparation — exists so the client has affect-management resources before you open anything (American Psychological Association, EMDR clinical practice guideline). I don't treat the Calm Place as a one-and-done install. I have the client access it cold, in different states, on different days, until the cue word pulls the felt state without my voice. A resource rehearsed once in a calm office won't be there at the Tuesday spike.
- Build the cue to work out of context. Because state and context shape retrieval, I attach the resource to something portable — a cue word plus a physical anchor (a stone, a breath pattern, a hand on the sternum) — so it has a route back that doesn't depend on being in my office or in a regulated state.
- Practice the Container as a skill, not a metaphor. Most clients can imagine a container once; fewer can actually use one when intrusive material shows up at dinner. I have them rehearse closing it in session on small stuff, so the move is in the body before they need it on big stuff.
- Give the resources an explicit job for the week. "If something comes up, use your Calm Place" is too loose. "When you notice the image: name it, Container, Calm Place — and if it's still above a 5 in twenty minutes, text me" is a protocol the nervous system can follow. The somatic version of holding that thread is the same logic — grounding has to be rehearsed and specified, not assumed.
None of this is novel to experienced EMDR clinicians. I spell it out because the between-session failure point is almost always under-rehearsed resourcing — a resource that worked in the room and quietly fell apart in the gap.
Should clients do any reprocessing on their own?
No. This is the bright line of the whole post.
Self-directed bilateral stimulation, "running the memory" at home, buzzers or eye movements without a clinician tracking dual attention and disturbance — these are not between-session homework. Reprocessing belongs inside the protocol, with someone watching for the channel that opens too wide, the abreaction that needs a hand, the dissociative slide that needs you to stop. The client's between-session job is the opposite: containment and grounding. Set it down. Stay in the present. Reach for the resource. Bring the material back to the room.
This is where EMDR and the generalization problem in other modalities point in opposite directions. In DBT or CBT you want the client practicing the skill in the wild. In active EMDR you want them not practicing the intervention in the wild — holding the line until they're back with you. Telling those two apart is a clinical judgment, not a feature you can automate.
How to tell a client is over-accessing midweek
You won't always catch it in the room — by your next session the client may have white-knuckled through it and minimized it. Worth asking about directly. The signals I track:
- Intrusions that escalate rather than settle. A few stray associations that fade is the work integrating. Images that intensify across days, or new disturbance the original target didn't hold, is the network opening wider than the closure contained.
- Sleep going first. Disrupted sleep, vivid distressing dreams, or 3 a.m. activation is often the earliest tell that reprocessing is continuing off-protocol.
- The resource stops working. "I tried my Calm Place and it did nothing" usually means the disturbance has outrun the resource — the closure didn't hold, not that the client failed.
- Functional pullback. Skipping work, canceling plans, isolating. Destabilization that lasts several days without easing warrants a between-session check-in, not a "we'll talk Thursday."
- The skipped session. A no-show after an activating session is frequently an avoidance spike. I treat it as clinical data, not a billing event (more on reading no-shows clinically).
The throughline: you can only respond to what you can see, and the gap is where you can't. Anything that gives you a low-friction read on how the week is actually going — without the client waiting until Thursday and without you guessing — closes some of that blind spot.
Where a between-session mirror fits — and where it absolutely does not
I'll be plain, because a destabilized EMDR client is the single worst place to get this wrong. I am not describing a chatbot you hand a client so it can do therapy between sessions. A bot that pretends to be the clinician — interpreting, "helping" a trauma client work through an intrusive image at midnight — is dangerous, and EMDR clients are exactly the population where it's most dangerous. The evidence backs the caution: licensed clinicians respond appropriately in roughly 93% of clinical situations, while consumer AI bots fall under 60% in the same comparisons (Stanford HAI / ACM FAccT, 2025). I drew that line in full in AI as a clinical tool, not a replacement, and it holds hardest right here.
A between-session mirror is a different object. It does not reprocess, interpret, advise, or stand in for you. It reflects the client's own state and pattern back to them — in their language, tied to the resources you built in the room — so they're cued toward grounding and containment, not toward opening anything. For an EMDR client that's a nudge back to the Calm Place sequence, a prompt to use the Container, a structured place to note "the image came up Tuesday at an 8" so it travels back to you — and a midweek read you otherwise don't have, a way to see the spike before Thursday. That's the read behind why I built VibeCheck for my own caseload: to hold the thread in the hours between sessions without ever crossing into doing the clinical work outside the room. It reinforces the brake. It never touches the gas.
FAQ
Why do EMDR clients destabilize between sessions and what helps?
Because reprocessing activates a memory network that can keep firing after the session ends, while the Calm Place and Container resources installed in the office degrade in a different context and state (Godden & Baddeley, 1975). What helps: a clean closure, resources over-rehearsed until they're portable, an explicit protocol for meeting a spike, and a clear instruction to contain rather than reprocess.
How do I help a client use their Calm Place or Container resource at home?
Build the resource to travel before you need it. Rehearse the Calm Place cold, in different states and on different days, until the cue word pulls the felt sense without your voice; attach a portable physical anchor; practice closing the Container on small material in session. Then give it a specific job — an if-this-then-that sequence — not a vague "use it if something comes up."
What is the difference between a complete and incomplete EMDR session for between-session safety?
A complete session reaches adaptive resolution — disturbance down, positive cognition strengthened, body scan run. An incomplete session ends with material still active and requires a deliberate closure (typically the Container) so the client doesn't leave mid-reprocess. For the week ahead, a properly closed incomplete session is often safer than a rushed "complete" one, because the client carries an explicit instruction to set the material down.
Should clients do any reprocessing on their own between EMDR sessions?
No. Self-directed bilateral stimulation or running the memory at home is not homework — reprocessing belongs inside the protocol with a clinician tracking dual attention and disturbance. The client's between-session job is the opposite: containment and grounding, staying present, and bringing the material back to the room.
Can an app safely support an EMDR client between sessions without replacing the clinician?
Only if it stays a mirror, not a therapist. A safe tool reflects the client's state and cues them toward the resources you already built — it does not interpret, advise, or attempt reprocessing. Anything that tries to do clinical work with a destabilized trauma client between sessions is unsafe; clinicians respond appropriately far more often than consumer bots do (Stanford HAI / ACM FAccT, 2025).
How do I tell if a client is over-accessing disturbing material midweek?
Watch for intrusions that escalate rather than settle, sleep disruption and distressing dreams, a Calm Place that suddenly "doesn't work," functional pullback or isolation, and avoidance no-shows after an activating session. Destabilization lasting several days without easing warrants a between-session check-in, not waiting for the next scheduled hour.